Single-Cell Technologies for Monitoring
Interactions Between Immune Cells
by
Yvonne J. Yamanaka
B.S.E., Biomedical Engineering, Duke University (2008)
Submitted to the Department of Biological Engineering
in partial fulfillment of the requirements for the degree of
Doctor of Philosophy
at the
MASSACHUSETTS INSTITUTE OF TECHNOLOGY
June 2014
© Massachusetts Institute of Technology 2014. All rights reserved.
Author . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Department of Biological Engineering
May 12, 2014
Certified by . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
J. Christopher Love
Associate Professor of Chemical Engineering
Thesis Supervisor
Certified by . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Darrell J. Irvine
Professor of Materials Science and Biological Engineering
Thesis Supervisor
Accepted by . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Forest M. White
Professor of Biological Engineering
Chair, Graduate Program Committee
2
Single-Cell Technologies for Monitoring Interactions
Between Immune Cells
by
Yvonne J. Yamanaka
Submitted to the Department of Biological Engineering
on May 12, 2014, in partial fulfillment of the
requirements for the degree of
Doctor of Philosophy
Abstract
Immune cells participate in dynamic cellular interactions that play a critical role
in the defense against pathogens and the destruction of malignant cells. The vast
heterogeneity of immune cells motivates the study of these interactions at the singlecell level. In this thesis, we present new tools to characterize how individual immune
cells interact with each other and with diseased cells.
We develop a nanowell-based platform to investigate how natural killer (NK)
cells interrogate and attack diseased target cells. This platform enables integrated
analysis of cytolytic activity, secretory activity, receptor expression, and dynamic
parameters of interactions between thousands of individual NK cells and target cells.
Using this platform, we show that NK cells operate independently when lysing a
single target and that motility during contact is associated with the secretion of
certain cytokines. Extending the platform, we investigate how contact with a target
induces the shedding of CD16 from the surface of an NK cell. NK cells use CD16
to recognize antibody-coated target cells, and thus the loss of CD16 is of clinical
interest. We show that the loss of CD16 is correlated to the length of time that the
NK cell spends in contact with a target but that not all NK cells that shed CD16
exert common effector functions. In wells with multiple NK cells, shedding occurs
in a more coordinated manner than would be expected by chance alone. Next, we
compare the functional properties of NK cells with distinct repertoires of inhibitory
receptors. Inhibitory receptors prevent NK cells from attacking healthy cells, and
their expression can confer NK cells with increased functional activity in a process
known as “licensing”. We show that despite forming prolonged contacts with target
cells, unlicensed NK cells are less likely than licensed NK cells to secrete cytokines.
Finally, we present tools to study other modes of interaction between immune cells.
Towards this end, we develop and apply fluorescent cellular barcoding strategies to
efficiently analyze the secretory properties of individual immune cells from different
populations. Altogether, this thesis contributes new tools for single-cell analysis and
applies them to reveal new insights about intercellular interactions.
3
Thesis Supervisor: J. Christopher Love
Title: Associate Professor of Chemical Engineering
Thesis Supervisor: Darrell J. Irvine
Title: Professor of Materials Science and Biological Engineering
4
Acknowledgments
I have thoroughly enjoyed my time as a graduate student at MIT—both inside and
outside of the lab—and have many people to thank for their guidance, support, and
friendship along the way.
My advisors, J. Christopher Love and Darrell Irvine, are excellent mentors and
inspiring scientists to work with. I am especially grateful to Chris for the wonderful combination of freedom and insightful direction that he has provided. He has
constantly encouraged me to pursue my interests and supported every project that
I took on. It has been a great experience training with mentors as open-minded
and creative as Chris and Darrell, and I hope that I can propagate the examples
they have set. I have learned a great deal from the discussions I have had with my
thesis committee, and I would like to thank each member for sharing their advice,
ideas, and time so generously. Special thanks to Marcus Altfeld for contributing his
knowledge about NK cells and helping me connect with other NK cell biologists in
the area, and to Douglas Lauffenburger for his guidance as the chair of the committee.
The Love Lab was just over a year old when I first joined the group. Now, one
building, countless gallons of PDMS, >38 TB of data, and 5.5 years later, it has gone
through many exciting transformations. It has been a real pleasure to be a part of
these changes and grow up with the lab, and I look forward to seeing what new areas
the group will tackle next.
Many thanks to each member of the lab for being a great colleague and friend.
In particular, I would like to thank Todd Gierahn for being like a third advisor, continually sharing his technical expertise and encyclopedic knowledge of science, and
always being up for having an interesting conversation about new ideas; Denis Loginov for keeping up the lab servers and for writing codes that were instrumental not
only for this thesis but for the lab as a whole; Qing Han, Adebola Ogunniyi, Yuan
Gong, Navin Varadarajan, and Kerry Love for training me when I first joined the
lab; Li-Lun Ho for teaching me about RNA-Seq; Rita Lucia Contento for working
together on several different T cell projects; Alan Stockdale for his work on the clear
clamp; Viktor Adalsteinsson for sharing his expertise on sequencing and the CellCelector; Rachel Barry for keeping the lab well-stocked with stamps and slides and for
helping with the CellCelector; Ross Zimnisky and Kartik Shah for helping with the
biosensor project; Brittany Thomas, John Ballew, Chris Dupont, Bin Jia, Andy Tu,
Konstantinos Tsioris, Lionel Lam, Tom Douce, Nick Mozdzierz, John Clark, Rachel
Leeson, and Brinda Monian for all of the interesting conversations about immunology, biotechnology, and the world outside of lab; Ayca Yalcin Ozkumur, Alex Torres,
Tim Politano, Sangram Bagh, Jonghoon Choi, Vasiliki Panagiotou, Joe Couto, Sarah
Schrier, Abby Hill, and Narmin Tahirova for their help and friendship during the time
that we overlapped; and Mariann Murray for keeping the lab running smoothly.
5
I have had the good fortune to work with and learn from many wonderful collaborators:
At the Ragon Institute, I would like to thank Galit Alter and members of her lab
for guiding and supporting the work on NK cells. Magdalena Sips, Christoph Berger,
and Cormac Cosgrove from the Alter Lab have each been fantastic collaborators to
work with. I would also like to thank Marcus Altfeld and the members of his lab,
particularly Christian Korner, for sharing their NK cell expertise, and Mary Carrington and her team for sharing their list of KIR- and HLA-typed blood donors and for
their excellent suggestions about how to study licensing in NK cells.
From the Brigham and Women’s Hospital, Philip De Jager, Laura Glick, and others involved in the PhenoGenetic cohort have been extremely helpful in arranging
access to extensively characterized blood donors. Cells from these donors were a key
element in the work on NK cell licensing and inhibitory receptors.
At the Broad Institute, I have truly enjoyed working with Arnon Arazi and Chloe
Villani from Nir Hacohen’s lab and Alex Shalek and Rahul Satija from Aviv Regev’s
lab. They have each taught me a great deal about single-cell analysis, computational
biology, and how to design experiments that are both innovative and useful. I am especially grateful to Chloe and Nir for so readily extending their platform for single-cell
RNA-Seq to address questions in NK cell biology, and to Arnon for working together
on several different projects and teaching me about statistics, lupus, and even some
Hebrew along the way.
The Koch Institute has been a fantastic home-base, and I am lucky to share a
floor with many collaborators and friends. From the Irvine Lab, Greg Szeto and
Talitha Forcier were a great team to work with on the barcoding project. I would
also like to thank Greg, Prabhani Atukorale, and Maria Foley for sharing interesting
discussions and ideas about immunology and biomedical research. From the Wittrup
Lab, I have had a lot of fun working with Xiaosai Yao on cancer-related projects, and
I always look forward to learning new things about immunotherapy, protein engineering, and in vivo work from Tiffany Chen. I would also like to say a special thanks to
Ermelinda, who not only keeps all the labs spick-and-span, but also keeps all of us
happy by sharing her friendship and her homemade Portuguese treats.
The scientists at the core facilities have contributed to many aspects of the work
presented in this thesis. From the Swanson Biotechnology Center, I would like to
thank Glenn Paradis and the other members of the Flow Cytometry Core, and Eliza
Vasile from the Microscopy Core, for their help and training. From the MIT BioMicro Center, I would like to thank Stuart Levine, each of the staff members who have
helped with sequencing, and Vincent Butty for help with data analysis and for discussions around our mutual fascination with KIR and HLA.
From UC San Diego, I would like to thank Yingxiao Wang and Kathy Lu for shar6
ing their SHP2 biosensor and discussing ideas about how biosensors could be used to
study signaling in immune cells.
It has been a privilege to be a part of the Biological Engineering community here
at MIT. My BE classmates have been a constant source of fun and inspiration. Working in the Love Lab has also given me the chance to join the Chemical Engineering
student community, and I appreciate how welcoming they have been. Thanks also
to the many professors across MIT and Harvard who have taught me, and especially
to Douglas Lauffenburger, K. Dane Wittrup, and Eric Alm for their discussions and
mentorship.
I have been lucky to work with outstanding teachers and mentors before coming
to MIT. Many thanks to Kam Leong, Stephen Craig, Robert Cramer, and Yihua Loo
from Duke University; Mingdi Yan and Carl Wamser from Portland State University;
and Rosa Hemphill, Bill Lamb, David Bermudez, and Angela Hancock from Oregon
Episcopal School.
I am grateful to the NSF Graduate Research Fellowship, the Collamore-Rogers
Fellowship, the Siebel Scholars program, the Merck Innovation Cup, and the MIT
Office of the Dean for Graduate Education for providing funding and/or exciting educational opportunities throughout my time as a graduate student.
The last few years have flown by thanks to all of the fun times I have had the pleasure of sharing with wonderful friends, especially Tiffany Chen, Xiaosai Yao, Amneet
Gulati, Yuan Gong, Rita Lucia Contento, Adebola Ogunniyi, Denis Loginov, Audrey
Fan, Lori Ling, Yuchen Yu, Lihua Zou, Tiky Luo, Annie Gai, Harvey Xiao, Nicole
Casasnovas, Chelsea He, Timothy Curran, the Sidney-Pacific Graduate Community,
and running teammates at the Cambridge Sports Union.
And most of all, to my parents, Susan and Masatoshi, to Grandma Mary, and to
James—thank you for bringing me from the test tube to the test tubes and making
me look forward to every new day.
7
8
Contents
1 Introduction: Single-Cell Measurements in Immunology
1.1
1.2
1.3
1.4
21
Importance of single-cell measurements in understanding immune responses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
22
Technologies for analyzing individual immune cells . . . . . . . . . . .
25
1.2.1
Flow cytometry . . . . . . . . . . . . . . . . . . . . . . . . . .
26
1.2.2
Cytometry by time-of-flight (CyTOF) . . . . . . . . . . . . . .
28
1.2.3
Single-cell secretory measurements . . . . . . . . . . . . . . .
29
1.2.4
Single-cell transcriptional profiling . . . . . . . . . . . . . . . .
31
Technologies for monitoring interactions between immune cells . . . .
32
1.3.1
Intravital microscopy . . . . . . . . . . . . . . . . . . . . . . .
32
1.3.2
In vitro microscopy for monitoring cytolytic interactions . . .
34
Overview of thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
36
2 A Nanowell-Based Platform for Analyzing Interactions Between Individual Natural Killer Cells and Target Cells
39
2.1
Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
40
2.2
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
40
2.3
Materials & Methods . . . . . . . . . . . . . . . . . . . . . . . . . . .
42
2.3.1
Cells and stimulations . . . . . . . . . . . . . . . . . . . . . .
42
2.3.2
Fabrication of nanowell arrays . . . . . . . . . . . . . . . . . .
43
2.3.3
Single-cell cytolysis assay and video microscopy . . . . . . . .
43
2.3.4
Microengraving to detect secreted proteins . . . . . . . . . . .
44
2.3.5
Functional and phenotypic flow cytometry . . . . . . . . . . .
44
9
2.3.6
2.4
Data analysis . . . . . . . . . . . . . . . . . . . . . . . . . . .
45
Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
46
2.4.1
Individual NK cells lyse target cells when co-incubated in nanowells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.4.2
Small groups of NK cells do not cooperate to kill single target
cells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.4.3
2.5
51
Microengraving measures the short-term secretory activity of
individual NK cells . . . . . . . . . . . . . . . . . . . . . . . .
2.4.5
49
IL-2-activated NK cells lyse target cells with heterogeneous dynamics and in an order-dependent manner . . . . . . . . . . .
2.4.4
46
52
Secretion of IFN-γ from IL-2-activated NK cells is associated
with reduced motility during contact with target cells . . . . .
55
Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
61
3 Dynamic and Functional Correlates of Activation-Induced Shedding
of CD16 from Natural Killer Cells after Discrete Interactions with
Target Cells
65
3.1
Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
66
3.2
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
67
3.3
Materials & Methods . . . . . . . . . . . . . . . . . . . . . . . . . . .
70
3.3.1
Cells and stimulations . . . . . . . . . . . . . . . . . . . . . .
70
3.3.2
Measurement of surface-expressed CD16 and NKG2D by flow
cytometry after bulk stimulation with target cells . . . . . . .
3.3.3
Nanowell-based assays to measure the shedding of CD16 and
functional activity of individual NK cells . . . . . . . . . . . .
71
Data analysis . . . . . . . . . . . . . . . . . . . . . . . . . . .
73
Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
74
3.3.4
3.4
71
3.4.1
CD16 is shed from the surface of NK cells after bulk stimulation
with target cells . . . . . . . . . . . . . . . . . . . . . . . . . .
10
74
3.4.2
CD16 is shed from the surface of NK cells after stimulation with
single target cells in nanowells . . . . . . . . . . . . . . . . . .
3.4.3
77
CD16 is shed with a probability that depends on the duration
of contact with a target cell and the baseline activation state of
the NK cell . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.4.4
Bystander NK cells lose CD16 more frequently than expected
by chance . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.4.5
80
82
NK cells that perform effector functions shed CD16, but many
cells that shed CD16 do not perform common effector functions 84
3.5
Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
88
4 Single-Cell Analysis of the Dynamic and Functional Properties of
Licensed and Unlicensed Natural Killer Cells
95
4.1
Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
96
4.2
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
96
4.3
Materials & Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
4.4
4.3.1
Nanowell-based assays . . . . . . . . . . . . . . . . . . . . . . 101
4.3.2
Population-level RNA-Seq . . . . . . . . . . . . . . . . . . . . 102
Results & Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
4.4.1
An integrated platform to correlate the expression of inhibitory
receptors (IRs) with functional activity and dynamic properties
at the single-cell level . . . . . . . . . . . . . . . . . . . . . . . 105
4.4.2
NK cells that express IRs are more functionally active against
target cells in nanowells than those that do not . . . . . . . . 107
4.4.3
Functional differences between IR+ and IR– NK cells responding to target cells are maintained across different durations of
contact . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
4.4.4
4.5
Transcriptional profiling of licensed and unlicensed NK cells . 114
Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
5 Cellular Barcodes for Efficiently Profiling Single-Cell Secretory Re11
sponses by Microengraving
121
5.1
Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
5.2
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
5.3
Materials & Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
5.4
5.3.1
Fabrication of arrays of nanowells . . . . . . . . . . . . . . . . 124
5.3.2
Cells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
5.3.3
Stimulations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
5.3.4
Cellular barcoding . . . . . . . . . . . . . . . . . . . . . . . . 125
5.3.5
Loading of cells onto arrays of nanowells . . . . . . . . . . . . 126
5.3.6
Detection of secreted proteins by microengraving . . . . . . . 126
5.3.7
Staining for viability and surface marker expression . . . . . . 126
5.3.8
Imaging cytometry . . . . . . . . . . . . . . . . . . . . . . . . 127
5.3.9
Data analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
Results & Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
5.4.1
Overview of cellular barcoding applied to multiplex single-cell
secretory measurements . . . . . . . . . . . . . . . . . . . . . 127
5.4.2
Development and validation of cellular barcodes . . . . . . . . 128
5.4.3
Application 1: Efficient construction of dose-response curves . 131
5.4.4
Application 2: Profiles of secretory responses induced by diverse
stimuli . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
5.4.5
5.5
Application 3: Profiles of lineage-dependent secretory responses 136
Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
6 Summary & Future Directions
139
A Supplementary Information for Chapter 2
145
A.1 Supplementary Methods . . . . . . . . . . . . . . . . . . . . . . . . . 145
A.1.1 Fabrication of arrays of nanowells . . . . . . . . . . . . . . . . 145
A.1.2 Microengraving to detect secreted proteins . . . . . . . . . . . 145
A.2 Supplementary Figures . . . . . . . . . . . . . . . . . . . . . . . . . . 147
12
B Supplementary Information for Chapter 3
151
B.1 Supplementary Methods . . . . . . . . . . . . . . . . . . . . . . . . . 151
B.1.1 Antibodies for flow and imaging cytometry . . . . . . . . . . . 151
B.2 Supplementary Figures . . . . . . . . . . . . . . . . . . . . . . . . . . 152
C Supplementary Information for Chapter 4
155
C.1 Supplementary Methods . . . . . . . . . . . . . . . . . . . . . . . . . 155
C.1.1 Antibodies for flow and imaging cytometry . . . . . . . . . . . 155
C.2 Supplementary Figures . . . . . . . . . . . . . . . . . . . . . . . . . . 155
D Supplementary Information for Chapter 6
167
D.1 Supplementary Methods . . . . . . . . . . . . . . . . . . . . . . . . . 167
D.1.1 Fabrication of arrays of nanowells . . . . . . . . . . . . . . . . 167
D.1.2 T helper (Th) cell biasing . . . . . . . . . . . . . . . . . . . . 168
D.1.3 Detection of secreted proteins by microengraving . . . . . . . 168
D.1.4 Calculation of the classification accuracy of cellular barcoding
170
D.2 Supplementary Figures & Tables . . . . . . . . . . . . . . . . . . . . 172
13
14
List of Figures
2-1 Schematic of the single-cell cytolysis (SCC) assay . . . . . . . . . . .
47
2-2 NK cells interrogate and lyse K562 target cells when co-incubated in
nanowells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
48
2-3 NK cells act independently when lysing a single target cell in a nanowell 50
2-4 Automated tracking of interactions between NK cells and target cells
53
2-5 Dynamics of cytolytic interactions between NK cells and target cells .
54
2-6 Schematic of microengraving to detect secreted cytokines and chemokines 56
2-7 Microengraving measures the short-term secretory activity of individual NK cells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
57
2-8 Correlated measurement of the dynamics of NK cell-target cell interactions and secreted products . . . . . . . . . . . . . . . . . . . . . .
59
2-9 Secretion of IFN-γ is associated with the velocity of NK cells during
contact with target cells . . . . . . . . . . . . . . . . . . . . . . . . .
60
2-10 Cell tracks segregated by functional outcome . . . . . . . . . . . . . .
61
3-1 Expression of CD16 and NKG2D on CD56dim NK cells after stimulation
with target cells in bulk . . . . . . . . . . . . . . . . . . . . . . . . .
76
3-2 Expression of CD16 on IL-2-activated CD56dim NK cells after stimulation with target cells in nanowells . . . . . . . . . . . . . . . . . . . .
78
3-3 The majority of NK cells shed CD16 when co-incubated with target
cells in nanowells . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
79
3-4 Distribution of contact durations between NK cells and target cells .
80
15
3-5 Amount of CD16 remaining on the surface of NK cells after varying
durations of contact with target cells. . . . . . . . . . . . . . . . . . .
81
3-6 NK cells shed CD16 with a probability that depends on their longest
duration of contact with a target cell and their activation state . . . .
83
3-7 Bystander NK cells shed CD16 more often than expected by chance .
85
3-8 Functionally active NK cells shed CD16 . . . . . . . . . . . . . . . . .
87
3-9 Not all NK cells that shed CD16 perform common effector functions .
89
4-1 Expression of inhibitory receptors on NK cells can be quantified by
on-chip staining and imaging cytometry . . . . . . . . . . . . . . . . . 106
4-2 Detection of secreted cytokines by microengraving concurrently with
time-lapse microscopy
. . . . . . . . . . . . . . . . . . . . . . . . . . 108
4-3 Secretory activity of IR+ and IR– NK cells measured by microengraving109
4-4 IR+ NK cells secrete cytokines more frequently than IR– NK cells do
upon interacting with target cells . . . . . . . . . . . . . . . . . . . . 110
4-5 Distribution of contact durations between NK cells and target cells,
segregated by the expression of IRs on NK cells . . . . . . . . . . . . 113
4-6 NK cells secrete cytokines with a probability that depends on their
longest duration of contact with a target cell . . . . . . . . . . . . . . 115
4-7 IR– NK cells are less likely than IR+ cells to secrete cytokines, even
after prolonged contact with a target . . . . . . . . . . . . . . . . . . 116
5-1 Schematic for using cellular barcodes to increase the throughput of
secretory measurements from single cells . . . . . . . . . . . . . . . . 129
5-2 Matched imaging cytometry and microengraving . . . . . . . . . . . . 130
5-3 Barcoded T cells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
5-4 Classification accuracy . . . . . . . . . . . . . . . . . . . . . . . . . . 132
5-5 Dye-rotation experiments . . . . . . . . . . . . . . . . . . . . . . . . . 132
5-6 Application of cellular barcodes to construct dose-response curves . . 134
5-7 Application of cellular barcodes to measure secretion from PBMCs
responding to mechanistically distinct stimuli . . . . . . . . . . . . . 135
16
5-8 Application of cellular barcodes to profile lineage-dependent secretory
responses from CD4+ Th cells . . . . . . . . . . . . . . . . . . . . . . 137
A-1 Flow cytometry gating scheme for NK cells . . . . . . . . . . . . . . . 147
A-2 Surface expression of NKG2D ligands on K562 target cells . . . . . . 147
A-3 Distribution of NK cells and K562 target cells in arrayed nanowells . 148
A-4 Phenotypes of NK cells after 44 h stimulation with IL-2 . . . . . . . . 149
B-1 Expression of CD16 on CD56dim resting NK cells after stimulation with
target cells in bulk . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
B-2 NK cells shed CD16 with a probability that depends on their cumulative duration of contact with a target cell and their activation state . 153
B-3 NK cells act independently when lysing antibody-coated P815 target
cells in nanowells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
B-4 Not all NK cells that shed CD16 lyse target cells . . . . . . . . . . . . 154
C-1 Chilling the array of nanowells improves the retention of NK cells . . 157
C-2 Functional activity of IR+ and IR– NK cells against K562 target cells
158
C-3 Distribution of secretion intensities as a function of contact time parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
C-4 Frequency of secretion as a function of contact time parameters . . . 160
C-5 Median intensity of positive secretion events as a function of contact
time parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
C-6 Frequency of cytolysis as a function of contact time parameters
. . . 162
C-7 Frequency of secretion as a function of contact time parameters, segregated by the expression of IRs on NK cells . . . . . . . . . . . . . . 163
C-8 Median intensity of positive secretion events as a function of contact
time parameters, segregated by the expression of IRs on NK cells . . 164
C-9 Frequency of cytolysis as a function of contact time parameters, segregated by the expression of IRs on NK cells . . . . . . . . . . . . . . . 165
D-1 Automated counting of viable cells . . . . . . . . . . . . . . . . . . . 172
17
D-2 Intensities of secretion in dye-rotation experiments . . . . . . . . . . . 174
18
List of Tables
B.1 Antibodies for flow and imaging cytometry . . . . . . . . . . . . . . . 151
C.1 Antibodies for flow and imaging cytometry . . . . . . . . . . . . . . . 156
D.1 Classification accuracy for double-positive cells . . . . . . . . . . . . . 173
19
20
Chapter 1
Introduction: Single-Cell
Measurements in Immunology
Portions of this chapter are adapted from Ref. 1 and Ref. 2.
Yamanaka, Y.J., Gierahn, T.M., Love, J.C. The dynamic lives of T cells: new
approaches and themes. Trends in Immunology, 2013, 34(2):59-66.
Yalcin A., Yamanaka Y.J., Love J.C. Analytical technologies for integrated
single-cell analysis of human immune responses. Methods in Molecular Biology, 2012, 853:211-235.
21
1.1
Importance of single-cell measurements in understanding immune responses
Attempting to dissect cellular immune responses presents a conundrum: On the one
hand, the vast heterogeneity of immune cells means that we need to characterize each
individual cell with a high level of detail; on the other hand, the dense interconnectedness of the immune network means that we must also assess how cells interact with
and are influenced by the cellular community as a whole. In this thesis, we develop
and apply tools to characterize natural killer (NK) cells and T cells, which are two
types of immune cell that epitomize this conundrum due to the fact that they are
extremely heterogeneous, polyfunctional, dynamic, and involved in many types of
cellular interactions.
Heterogeneity
One of the most striking examples of heterogeneity in the immune system comes
from the rearrangement of T cell receptor (TCR) gene segments to generate receptors
capable of recognizing a diverse assortment of antigens. It is estimated that the lower
limit of each person’s TCR repertoire is on the order of 106 clonotypes, 3 which means
that typically only one in 103 – 105 T cells will recognize any given antigen. 4 Similarly,
although NK cells do not express antigen-specific receptors that are generated by
gene rearrangement, they express a variety of germline-encoded surface receptors that
allow them to differentiate between diseased target cells and healthy host cells. 5 The
expression of these receptors varies not only between individuals, but also between
NK cells within the same individual. This variability is partly due to the stochastic
gene expression of certain receptors. 6
In addition, a substantial amount of heterogeneity in populations of T cells and NK
cells exists even when specific gene rearrangements or stochastic receptor expression
are ignored. Both T cells and NK cells can be subdivided into many different—but
often overlapping—classes that reflect differences in function, prior history, tissue
localization, and state of differentiation and activation. For example, studies char22
acterizing a large number of surface receptors and cytokines expressed by individual
cells have shown that CD8+ T cells populate a continuum of intermediary states between major states, such as naı̈ve and memory, and that even highly specified subsets
of T cells exhibit extensive combinatorial diversity in their expression of cytokines. 7
Multiparameter characterizations of receptors on NK cells have demonstrated that
these cells also exhibit extensive combinatorial diversity—currently, it is estimated
that each person carries 6,000 – 30,000 distinct subsets of NK cells. 8 Thus, technologies are needed that enable the analysis of large numbers of individual cells in an
efficient manner.
Polyfunctionality
Immune cells are multitalented—like precocious students, they are able to participate in many different activities, often at the same time. One form of polyfunctionality arises when a single cell performs two (or more) distinct types of function.
For example, CD8+ cytotoxic T lymphocytes (CTLs) and NK cells mediate two distinct, hallmark functions: the cytolysis of target cells and the secretion of cytokines.
Depending on the setting and the specific cells and cytokines analyzed, these two
functions can occur discordantly 9 or with varying degrees of overlap 10–12 within each
cell. Thus, technologies are needed that enable the integrated analysis of multiple
types of functions performed by the same individual cell.
Polyfunctionality can also arise due to the fact that within one class of function,
it is possible to have many different “flavors”. For example, a single T cell or NK cell
can secrete a wide range of different cytokines, each with different downstream effects. In T cells, this form of secretory polyfunctionality is associated with protective
immune responses against pathogens (discussed in Section 1.2.3). Thus, technologies
are needed that enable multiplexed analysis of cellular functions.
Dynamic properties
Both T cells and NK cells transition through a continuum of different phenotypic
and functional states as they develop, encounter antigen or other stimuli, prolifer23
ate, home to sites in the periphery, mediate effector functions, and possibly form
long-lived populations of memory1 cells. In addition, even cellular subsets that were
initially thought to represent stable lineages (e.g., classes of CD4+ T helper cells)
are increasingly being recognized as having plastic, flexible characteristics. 14 These
slowly evolving transitions, which occur over days or longer, highlight the need for
technologies to monitor—or infer—cellular histories spanning long periods of time.
Dynamic events that occur over shorter time frames are also of great importance
in immunology. Upon discrete stimulation (for example, a T cell meeting an antigenpresenting cell (APC), or an NK cell meeting a target cell), intracellular signaling
responses can propagate in seconds, triggering dynamic outcomes that evolve over
minutes to hours. These outcomes include changes in motility, the formation of a
synapse, cytolysis, and the secretion of cytokines. Thus, technologies are needed that
monitor dynamic properties of single cells with a temporal resolution that is sufficient
to capture transitions of interest.
Cellular interactions
Immune cells do not lead hermetic lives; rather, they interact with each other and
with other types of cells. These interactions can take place through direct contact or
through intermediaries such as secreted cytokines.
Both T cells and NK cells are highly motile. They migrate around lymphoid
and peripheral tissues collecting and interpreting signals that are presented on the
surfaces of other cells. This process is mediated by direct cell-cell contacts that can
be either transient (immunological kinapses) or stable (immunological synapses). 15 In
the case of a naı̈ve T cell interacting with an APC displaying cognate peptide on major
histocompatibility (MHC) complexes, the signaling that results from the interaction
can activate the T cell and induce proliferation, synthesis of cytokines and effector
molecules, and differentiation into effector or memory subsets. In the case of a CTL
or NK cell interacting with a target cell, the synapse or kinapse can serve as a conduit
through which cytotoxic granules containing perforin and granzyme are transferred
1
Certainly in the case of T cells, and possibly for NK cells 13 as well.
24
from the CTL or NK cell to the target cell. 16 Thus, to investigate processes such as
the activation of antigen-specific T cells or the cytolysis of target cells, technologies
are needed that monitor how cells respond to discrete, contact-mediated interactions
with other cells.
In addition to cellular interactions arising from direct physical contact, immune
cells communicate with each other and with non-immune cells by secreting soluble factors such as cytokines. This route of communication forms a dense network
(where cells are nodes and cytokines are edges) that has similarities to human social networks. 17 T cells and NK cells use secreted proteins to attract other cells to
the site (by the secretion of small cytokines known as chemokines), bias CD4+ T
helper cells towards different lineages, induce or inhibit proliferation, and control the
balance between inflammation and immunosuppression. Secreted cytokines can also
directly modulate the susceptibility of cells to pathogens. For example, interferon-γ
(IFN-γ) inhibits viral replication in infected cells and enhances the microbicidal activity of macrophages, 18 and macrophage inflammatory protein-1β (MIP-1β, or CC
chemokine ligand 4 (CCL4)) suppresses human immunodeficiency virus (HIV) infection by inhibiting the binding of HIV to a co-receptor on susceptible cells. 19,20 These
examples highlight just a few of the ways that immune cells use secreted factors to
influence each other and to drive disease outcomes. Because immune cells display heterogeneous and dynamic secretory properties, technologies are needed that measure
the secretory profiles of individual cells (or small groups of cells) over time.
1.2
Technologies for analyzing individual immune
cells
Here, we discuss both established and recently developed techniques for analyzing the
phenotypic, functional, and transcriptional properties of large numbers of individual
immune cells, with a focus on T cells and NK cells.
25
1.2.1
Flow cytometry
To date, flow cytometry has been the method of choice for creating single-cell phenotypic profiles of immune cell populations. Flow cytometry enables rapid quantification of ∼17 analytes (surface markers, intracellular proteins, labeling dyes, etc.)
on or in single cells. 21 The surface-marker phenotypes of immune cells are commonly
measured by flow cytometry after staining the cells with fluorophore-conjugated antibodies that are specific for the markers of interest. Intracellular proteins can be
assessed by fixing and permeabalizing cells prior to staining and analysis. This technique has been applied to measure the phosphorylation status of intracellular proteins
(“phospho-flow”), 22 providing rich sets of data with which to construct intracellular
signaling networks in T cells. 23 Cellular production of cytokines can also be assessed
by flow cytometry using a technique, termed intracellular cytokine staining (ICS), in
which cells are treated with pharmacological inhibitors of secretion so that cytokines
that would normally be secreted are instead associated with the cell (and hence with
the signal measured by the flow cytometer). 24 In addition to these applications, flow
cytometry is a useful tool for characterizing the proliferative and cytolytic capacities
of individual cells, discussed below.
Proliferation
The proliferation of individual immune cells can be characterized using a variety of flow cytometric approaches. 25 Flow cytometry-based proliferation assays can
be combined with simultaneous surface marker and intracellular cytokine staining to
generate integrated measurements of proliferation, phenotype, and cytokine production from single cells. A classic method for identification of actively proliferating cells
is incorporation of thymidine analogues such as 5-bromo-2’-deoxyuridine (BrDU) or
5-ethynyl-2’-deoxyuridine (EdU) into newly synthesized DNA. 26,27 The incorporated
analogues are stained with fluorescently labeled antibodies (for BrDU) or azides (for
EdU) and the resulting fluorescence signal is read out on a flow cytometer to identify cells with newly synthesized DNA. One drawback of BrDU and EdU labeling is
26
that cells must be fixed and permeabilized prior to data acquisition, and thus cannot
be repeatedly observed over multiple rounds of division or recovered for downstream
analysis.
Several types of stably incorporated live-cell fluorescent dyes have been developed
that allow proliferation measurements to be made on viable cells. A parent population of cells is labeled with the dye, and upon each round of division the dye content
is split equally between the two daughter cells. The division history of individual cells
is inferred based on the relative intensity of dye staining measured by flow cytometry. Cells from the parent population are brightest, whereas cells from subsequent
generations are dimmer due to the dilution of dye over successive rounds of division.
Carboxyfluorescein diacetate succinimidyl ester (CFSE) is a commonly used dye for
such assays, and can track up to eight rounds of division. 28
Cytolytic outcomes
The death of individual target cells can be identified with flow cytometry using
fluorescent stains against different markers of cell death. Cells undergoing apoptosis
are identified by annexin V, which labels phosphatidylserine that translocates to the
outer cell membrane early during an apoptotic event. 29 Cells that have undergone nuclear fragmentation are identified by membrane-impermeant DNA intercalating dyes
such as propidium iodide (PI), 29 7-Amino-Actinomycin D (7-AAD), 30 or SYTOX. 31
One limitation of traditional dead cell markers is that they do not distinguish between
target cells that died in a cytolytic interaction and target cells that died from other
causes. This limitation can be addressed by using fluorogenic caspase 32 or granzyme
B 33 substrates to more specifically label target cells that have received a lytic hit
from an effector cell.
The lytic potential of individual effector cells is commonly monitored using indirect
readouts of cytolytic activity. Surface expression of the degranulation marker CD107a
on effector cells can be conveniently detected using flow cytometry and has been shown
to correlate with target cell death. 11 However, stimulated effector cells that have not
been involved in a cytolytic interaction can also degranulate and express surface
27
CD107a, limiting the specificity of the readout under certain conditions. Conjugates
between effector and target cells can be assessed by flow cytometry, 30 although this
strategy cannot be used to identify target and effector cells that interacted but then
dissociated prior to the data acquisition.
1.2.2
Cytometry by time-of-flight (CyTOF)
In traditional flow cytometry, spectral overlap between the fluorescent reagents that
are used to label cellular markers imposes an upper limit on the number of analytes
that can be simultaneously resolved (typically <20 analytes per cell. 21,34 To increase
the depth of profiling, a cytometry technique has been developed that uses mass
spectrometry to analyze single cells that have been labeled with antibodies that are
conjugated to elemental (metal) tags instead of fluorophores. 35,36 This method, called
mass cytometry or cytometry by time-of-flight (CyTOF), eliminates the problem of
spectral overlap and therefore enables the simultaneous measurement of >36 analytes
per cell (in theory, ∼100 analytes can be measured) (reviewed in Ref. 37). New computational approaches have been developed to analyze the large, high-dimensional
datasets generated by CyTOF. 38 The use of metal-labeled antibodies has recently
been extended to the analysis of tissue sections, 39,40 thus providing a highly multiplexed alternative to immunohistochemistry (IHC) in the same way that CyTOF has
provided a highly multiplexed alternative to flow cytometry.
Profiles of human CD8+ T cells collected using CyTOF have shown that T cells
populate a continuum of intermediary states between major states, such as naı̈ve and
memory, and that even highly specified subsets of T cells exhibit extensive combinatorial diversity in their expression of cytokines. 7 These findings highlight the complexity
and heterogeneity of the CD8+ T cell compartment and demonstrate the utility of
CyTOF for identifying cells with transitional phenotypes.
CyTOF has also been applied to map the phenotypic diversity of the human NK
cell population within and between individuals. 8 Measurements of the combinatorial
expression patterns of 28 NK cell receptors suggested that within a single individual,
there are 6,000 – 30,000 distinct phenotypes of NK cells (out of the total 228 =
28
268,435,456 combinations that could have been recovered by Boolean gating on the
28 measured surface markers). This analysis also found that even the 50 most common
phenotypes account for an average of only 7 – 24% of the NK cells in an individual.
1.2.3
Single-cell secretory measurements
There is considerable interest in understanding how cytokine profiles from individual
immune cells within a population correlate with outcomes of infection, vaccination,
and other immunological processes. Because cytokines are important mediators of
intercellular communication, single-cell secretory measurements shed light on how
cells in a population interact with each other even in the absence of direct physical
contact. Here, we discuss several different approaches for measuring the secretion of
cytokines from single cells.
Approaches based on flow cytometry
ICS combined with multiparameter flow cytometry has been used to efficiently
profile the production of cytokines from T cells and NK cells in a variety of experimental and clinical settings. 12,41 These studies have revealed many interesting
properties of cytokine secretion at the single-cell level. For example, ICS studies have
shown that the breadth and magnitude of cytokines secreted by stimulated T cells
shift globally over time, 42–44 and that antigen-specific T cells that produce multiple
cytokines (“polyfunctional” T cells) are associated with protective immune responses
to pathogens such as HIV, 45,46 vaccinia virus, 47 and Leishmania major. 48 Because
ICS requires cells to be fixed and permeabalized prior to analysis, it cannot be used
to track dynamic secretory patterns in the same individual cell over time or to recover
viable, cytokine-secreting cells.
An alternative, non-destructive flow cytometric method for analyzing the secretion of cytokines from single cells is to use bispecific antibodies to link cytokines to
the surface of the cell that secreted them (reviewed in Ref. 49). This method has
been applied to detect and isolate antigen-specific CD8+ T cells that secreted IFN-γ
or interleukin-4 (IL-4), 50 but it is not widely used due to issues with sensitivity and
29
false positives. 49
Approaches based on the capture of analytes on solid surfaces
Enzyme-linked immunosorbent spot (ELISpot) assays, in which cytokines secreted
by single cells are captured and subsequently detected on a membrane coated with
cytokine-specific antibodies, are widely used to assess the frequencies of cytokinesecreting cells in both clinical and research settings (reviewed in Ref. 51). In addition
to the detection of secreted cytokines, ELISpot assays have also been developed to
measure the release of cytotoxic molecules such as perforin 52 and granzyme 53 from
individual cytolytic effector cells. A dual Lysispot/ELISPot secretion-based assay
has also been developed to enable the concurrent detection of target cell lysis (Lysispot) and effector cell IFN-γ secretion (ELISpot) for individual cytolytic interactions
between effector and target cells. 54
ELISpot and similar methods based on fluorescent detection (Fluorospot) 55 can
analyze up to three cytokines in a single assay. To increase the number of secreted
analytes measured per cell, a platform known as the single-cell barcode chip (SCBC)
has been developed. 56 The SCBC operates by isolating single cells in microfluidic
chambers (0.54 – 3 nL in volume) that contain spatially barcoded arrays of capture
antibodies. The cells are incubated for 12 – 24 h in these chambers (>1000 per
device), and the cytokines they secrete are captured and subsequently detected on
the spatially barcoded array. 56,57 The dense patterning of the antibody barcodes in
each chamber enables detection of the secretion of 12 – 19 cytokines from a single
cell. The SCBC has been applied to analyze the secretory profiles of macrophages, 56
tumor cells, 57 and tumor-specific patient-derived T cells. 56,58
Our lab has developed an array-based method to enable the rapid and sensitive
detection of multiple secreted proteins from large numbers of single cells. In this process, known as microengraving, 59 cells are isolated in an array of subnanoliter wells
(nanowells) that are each ∼125 pL. To analyze secreted cytokines (up to four per
process), a glass slide coated with cytokine-specific antibodies is compressed on top
of the array to capture the cytokines secreted by the cell(s) residing in each well over
30
a short period of time (typically one to several hours), and the resulting microarray of
secreted proteins is registered back to the wells—and hence the cells—that produced
the cytokines. 9,60,61 Microengraving is non-destructive and can be repeatedly applied
to measure the secretory profiles of the same individual cells at multiple points in
time, thereby enabling the study of single-cell secretory dynamics and the recovery of
viable cytokine-secreting cells. Furthermore, because each array contains ∼80,000 to
∼250,000 wells, large numbers of single cells can efficiently be interrogated. Microengraving has been applied to measure the dynamics of cytokine secretion from T cells 62
and to recover viable T cells with defined secretory properties for downstream analysis. 60,61 In the work presented in this thesis, we demonstrate that microengraving
can be combined with simultaneous time-lapse microscopy (Chapter 3 and Chaper 4)
and introduce strategies to increase the efficiency of microengraving when analyzing
mixed populations of cells (Chapter 5).
1.2.4
Single-cell transcriptional profiling
Detailed profiles of variability in gene expression patterns are shedding light on the degree of diversity that is present among immune cells and the dynamic transitions that
they undergo. Transcriptional profiling of single murine T cells using a nanofluidic
platform analyzing the expression of 96 genes has shown that when different primeboost strategies are used to deliver the same antigen, the CD8+ memory T cells that
are elicited display distinct patterns of gene expression despite having equivalent frequencies in the T cell population and equivalent profiles of cytokine production. 63
These findings illustrate how cells that share common phenotypic properties can still
access a variety of unique transcriptional states dictated by their current and past
experiences.
Transcriptional profiling has been extended even further by the advancement of
single-cell RNA-Seq, which can be used to analyze the entire transcriptome (>10,000
genes) of single cells. 64 Improved methods 65–67 for the generation of cDNA and the
preparation of sequencing libraries are making this technique increasingly feasible—
from the standpoint of both cost and sensitivity—to use to profile dozens to hundreds,
31
and even thousands, of single cells. Applied to immune cells, single-cell RNA-Seq is
proving to be a valuable tool for characterizing heterogeneity in the transcriptional
response of bone-marrow-derived dendritic cells to stimulation 68 and for the unsupervised classification of diverse types of immune cells. 69 Furthermore, the recent
development of methods to sequence RNA in situ with subcellular resolution 70 offers the possibility of someday probing the transcriptional landscape of immune cells
while also preserving information about their spatial organization in lymphoid organs,
tumor tissues, or other sites of interest.
1.3
Technologies for monitoring interactions between immune cells
Many critical functions of immune cells are mediated by their contact-dependent
interactions with other cells. Analyzing the dynamics of these interactions provides
valuable insight into how immune cells collect and respond to signals from other cells
(e.g., the activation of T cells by APCs) and how they exert their effector functions
(e.g., the killing of tumor cells by NK cells). Here, we discuss how the properties
of dynamic interactions between individual immune cells can be monitored in living
organisms or in in vitro settings.
1.3.1
Intravital microscopy
Dynamics of T cell activation
Multiphoton microscopy is a powerful tool for analyzing dynamic interactions
among individual cells in intact tissues (reviewed in Ref. 71–75), and has been used
extensively to study the dynamics with which naı̈ve T cells are primed by antigenloaded dendritic cells (DCs) in lymph nodes. These experiments have shown that
under certain conditions, T cells go through an initial phase of transient, serial interactions (typically <10 min per interaction) with DCs before progressing to a phase
of long-lived (>1 h) interactions. 76,77 Under other conditions, however, T cells do
32
not go through a prolonged phase of transient interactions but instead rapidly form
stable interactions with DCs. 78,79 One proposed explanation for these dynamic differences is that T cells integrate the antigenic, costimulatory, and chemotactic signals
they collect from different DCs and use the cumulative signal to trigger functional responses. According to this model, a T cell may need to serially encounter many DCs
presenting weak signals before reaching the same level of activation as a T cell that
encounters fewer DCs that each present strong signals. 80 Indeed, several experiments
have suggested that T cells integrate signals from APCs over time, encounters, doses,
and potencies, and then use these cumulative signals to induce outcomes such as cell
division or the secretion of cytokines. 79–83 These studies illustrate the importance
of incorporating measurements of the dynamic properties and histories of cell-cell
interactions when assessing the functional states of individual cells.
The dynamics of T cell activation have been investigated in further detail using a
technique called dynamic in situ cytometry (DISC), which uses intravital multiphoton
microscopy to measure the dynamics of cellular interactions while simultaneously
monitoring changes in the expression of cellular surface markers. 84 DISC analysis of
the shedding of CD62L (indicative of signaling through the TCR) from splenic T
cells has shown that T cells shed CD62L both when they form stable immunological
synapses with APCs and when they form transient interactions (kinapses) with APCs,
although the shedding is more rapid when stable synapses are formed. 84
Investigations of the subcellular properties of T cell-APC interfaces have provided
additional details about the dynamics of T cell activation. In vivo visualization of
the dynamics of protein localization at the T cell-APC interface has been achieved by
using multiphoton microscopy to monitor enhanced green fluorescent protein (EGFP)
fusion proteins in T cells. This approach was used to demonstrate that downstream
activities associated with antigen recognition can be induced either by transient or
stable clustering of TCR molecules at the T cell-APC interface, 85 and that migratory
and synaptic dynamics are influenced by the history of the T cell (e.g., naı̈ve versus
recently activated) and the characteristics of the APC. 86
33
Dynamics of interactions between NK cells and target cells
The in vivo interactions between adoptively transferred murine NK cells and target cells have been characterized by two-photon intravital microscopy. 87 NK cells
residing in the lymph node migrate along collagen fibers with a mean velocity of
6.7 µm/min, but upon contacting target cells (allogeneic B cells) they form conjugates lasting from 1 min to >50 min and routinely display cytolytic activity. 88 In this
system, the majority of target cell interactions are monogamous. However, in some
cases the contact between a target cell and an NK cell results in the recruitment of
additional NK cells, suggesting that NK-to-NK communication plays a role in target
cell interactions. NK cells residing in a tumor that expressed a ligand for the activating NK cell receptor NKG2D were also found to dynamically patrol the area (mean
velocity 5.2 µm/min) and interrogate target cells. 89 Compared with CTLs, NK cells
contacted tumor target cells for shorter periods of time, typically <10 min, but were
still capable of mediating cytolysis.
1.3.2
In vitro microscopy for monitoring cytolytic interactions
In vitro video microscopy is a useful approach for precisely tracking the trajectories,
contact histories, and certain functional outcomes of cell-cell interactions over long
periods of time. This approach has been used by several groups to characterize the
dynamics of cytolytic interactions between effector cells and their targets. For example, in the case of HIV-specific CTL clones, video microscopy studies have revealed
that after killing an HIV-infected target cell, CTLs undergo a prolonged period of
arrest. 90 During this time, they remain attached to the killed target cell and fail to
engage the majority of other target cells that migrate past them.
Video microscopy studies have also been used to examine how NK cells interrogate
and lyse their targets. Studies of interactions between NK cells and a variety of
target cells have demonstrated that the contact behavior of NK cells is dynamic,
heterogeneous, and tightly regulated by the triggering of NK cell receptors. 91–94 One
34
example of the heterogeneity of dynamic behaviors has come from a study of the
serial killing activity of NK cells. 93 This study demonstrated that over a 16 h period,
human NK cells that had been expanded in IL-2 could kill up to six target cells
(722.221 MHC Class I deficient cell line) in a serial manner. However, some NK cells
did not kill any targets despite forming a conjugate with a target cell, illustrating
the large amount of cell-to-cell variability in target cell interaction and killing ability
within the NK cell population. The variable killing pattern that NK cells display
when they interact with multiple target cells in series has been proposed as a metric
with which to define distinct classes of NK cells. 95
One source of NK cell variability in target interaction behavior stems from the
different levels and combinations of activating and inhibitory receptors expressed on
NK cells. High-resolution microscopy studies have shown that activating signaling
induced by contact with a target cell causes the NK cell to halt its motility and initiate
the formation of a lytic synapse with the target cell, whereas inhibitory signaling
limits the duration of time that the NK cell spends in contact with the target. 94
The baseline activation state of the NK cell prior to encounter with a target can also
significantly affect both the dynamics and outcome of the encounter. For example,
NK cells that have been activated with IL-2 form longer lasting contacts with target
cells and are also much more likely to mediate cytolysis compared with unactivated
NK cells. 96 Even NK cells that share similar dynamic properties, however, can exhibit
heterogeneous killing behaviors. For example, in an experiment using IL-2-activated
NK cells and 293T target cells, only 40% of the NK cells that arrested upon contacting
a target successfully lysed the target. 97
Collective interactions between groups of NK cells and target cells have been
examined by in vitro microscopy in the context of chemokine-induced NK cell migratory behavior. 98 NK cells treated with monocyte chemoattractant protein-1 (MCP-1)
migrated to contact K562 target cells via their leading edge while simultaneously
adhering to untreated NK cells via their uropod. This tethering behavior resulted
in an increased local concentration of NK cells around target cells, suggesting that
chemokine-induced NK-NK adhesion may lead to cooperative behavior amongst NK
35
cells when responding to target cells.
1.4
Overview of thesis
This thesis presents new sets of tools and approaches to characterize how individual
immune cells interact with each other and with diseased cells.
Chapter 2 to Chaper 4 focus on the development of tools to understand how NK
cells recognize, interrogate, and attack diseased target cells. The functional outcomes
of interactions between NK cells and target cells are dictated by the balance between
activating and inhibitory signals that NK cells receive from receptors recognizing
ligands on target cells and on healthy host cells. We explore how both activating
(CD16) and inhibitory (KIR and NKG2A) receptors on individual NK cells control
the responses that each NK cell mounts against a target, as well as how the receptors
themselves can be modulated by interactions with a target cell.
In Chapter 2, we introduce a nanowell-based platform for measuring interactions
between individual NK cells and target cells. This platform enables the integrated
analysis of cytolytic activity, secretory activity, and dynamic parameters of interactions between cells. Using this platform, we show that NK cells operate independently
when lysing a single target cell and that lysis is most probable during an NK cell’s
first encounter with a target. Furthermore, we demonstrate that the secretion of
interferon-γ (IFN-γ) occurs most often among NK cells that become the least motile
upon contacting a target cell but is largely independent of cytolysis.
In Chapter 3, we extend the platform to investigate how discrete contacts with
individual target cells induce the shedding of CD16 (FcγRIIIA) from NK cells. NK
cells use CD16 to recognize and attack antibody-coated target cells, and thus the loss
of surface-expressed CD16 from NK cells is of clinical interest in the context of viral
infection and antibody-based cancer therapies. We show that the loss of CD16 from
an NK cell is correlated to the length of time that the NK cell spends in contact with a
target cell, and that NK cells that contact antibody-coated target cells for a prolonged
time almost completely shed CD16. Unlike cytolytic activity, CD16 shedding occurs
36
more cooperatively than would be expected by chance. In addition, we demonstrate
that although the loss of CD16 is associated with cytolytic activity and secretion of
IFN-γ and CCL4, a significant fraction of NK cells that shed CD16 do not exert these
common effector functions.
In addition to activating receptors such as CD16, NK cells also express a diverse
range of inhibitory receptors that recognize HLA Class I molecules and prevent NK
cells from attacking healthy host cells. Expression of these inhibitory receptors can
confer NK cells with increased functional activity against HLA-deficient target cells in
a process known as “licensing”. In Chapter 4, we apply the nanowell-based platform
to study the dynamic and functional properties of individual licensed and unlicensed
NK cells. We show that all NK cells secrete cytokines with a frequency that increases
as they spend more time in contact with a target cell, but that unlicensed NK cells
are less likely than licensed NK cells to secrete cytokines, even after prolonged contact
with a target. In ongoing work, we further investigate licensing and responses to target
cells by comparing the transcriptional profiles of NK cells with distinct repertoires of
inhibitory receptors.
In addition to direct contact-mediated interactions, immune cells communicate
with each other and with their surrounding cellular environment by secreting soluble
factors such as cytokines. In Chapter 5, we develop strategies to fluorescently barcode multiple populations of immune cells and then quantify the secretory profiles
of each individual cell in parallel. We apply this approach to efficiently analyze the
secretory properties of immune cells treated with different doses and classes of stimuli
and to compare the secretory behaviors of different lineages of CD4+ T cells.
37
38
Chapter 2
A Nanowell-Based Platform for
Analyzing Interactions Between
Individual Natural Killer Cells and
Target Cells
This chapter is presented as it appeared in Ref. 99, with minor modifications.
Yamanaka, Y.J., Berger, C.T., Sips, M., Cheney, P.C., Alter, G., Love, J.C. Singlecell analysis of the dynamics and functional outcomes of interactions between human natural killer cells and target cells. Integrative Biology, 2012,
4(10):1145-1312.
39
2.1
Abstract
Natural killer (NK) cells are a subset of innate immune lymphocytes that interrogate
potential target cells and rapidly respond by lysing them or secreting inflammatory
immunomodulators. Productive interactions between NK cells and targets such as
tumor cells or virally infected cells are critical for immunological control of malignancies and infections. For individual NK cells, however, the relationship between the
characteristics of these cell-cell interactions, cytolysis, and secretory activity is not
well understood. Here, we used arrays of subnanoliter wells (nanowells) to monitor
individual NK cell-target cell interactions and quantify the resulting cytolytic and
secretory responses. We show that NK cells operate independently when lysing a
single target cell and that lysis is most probable during an NK cell’s first encounter
with a target. Furthermore, we demonstrate that the secretion of interferon-γ (IFNγ) occurs most often among NK cells that become the least motile upon contacting
a target cell but is largely independent of cytolysis. Our findings demonstrate that
integrated analysis of the cell-cell interaction parameters, cytolytic activity, and secretory activity of single NK cells can reveal new insights into how these complex
functions are related within individual cells.
2.2
Introduction
Natural killer (NK) cells are granular lymphocytes classically defined by their capacity to recognize and kill tumor or virally infected cells without the need for antigen
sensitization. They can also secrete cytokines and chemokines that promote the induction of a robust immune response. 100,101 NK cells identify target cells using a
variety of activating and inhibitory germline-encoded surface receptors. 5,102 The integration of signals from these receptors regulates the cells’ cytolytic and secretory responses. 12,103–106 In addition, cytokines such as interleukin-2 (IL-2) from T cells, 107–110
interferon-α (IFN-α) from plasmacytoid dendritic cells (pDCs), 111 and IL-12, IL-15,
and IL-18 from macrophages and DCs 112–116 can significantly enhance cytotoxicity
40
and secretion of cytokines from NK cells.
Recent data have shown that NK cells perform several contact-dependent immunomodulatory functions in addition to directly eliminating target cells. 101 For example, NK cells can tune the development of antiviral adaptive immunity by eliminating virus-infected DCs, 117 by maturing and ‘editing’ pools of DCs to optimize
antigen presentation, 118,119 or by lysing and suppressing activated T cells. 120,121 In
these regulatory interactions, as well as in direct interactions with target cells, the
efficacy of the NK cell-mediated immune response is determined by the cytolytic and
secretory decisions that each NK cell makes upon encountering another cell. The
direct relationship between cell-contact parameters, cytolysis, and the secretion of
cytokines is challenging to study, however, due to the large cell-to-cell variability in
both the functional capacities of NK cells 12,100,112,122–124 and the dynamics of NK cell
interactions (e.g., duration of conjugation, time to lysis, motility). 91,93,94,97,125
This functional diversity motivates the study of NK cell-mediated cytolysis, secretion, and interaction dynamics for individual NK cells. Conventional technologies,
however, cannot efficiently match observations of these traits with single-cell resolution. Standard microscopy-based approaches offer sub-cellular imaging resolution of
cell-cell interactions but are limited in throughput and are not amenable to measuring
cytokine production from untransfected human cells. Lysispot/ELISpot assays enable
the concurrent detection of cytolysis and secretion 54 but are not conducive to monitoring interactions over time. Finally, flow cytometry, combined with intracellular
cytokine staining (ICS), provides robust measures of the phenotypic and functional
attributes of large numbers of single NK cells but de-couples the history of cell-cell
interactions from the functional readout.
Multiple functional properties of NK cells can be measured at the single-cell level
by isolating and monitoring individual NK cells in submicroliter wells (microwells) or
subnanoliter wells (nanowells). This approach has been used to demonstrate that the
lysis of 293T target cells colocalized with IL-2-activated primary human NK cells in
nanowells occurs with diverse dynamics—some targets display a rapid loss of calcein (a
live-cell dye) when they are lysed by an NK cell, some display a slow loss of calcein, and
41
others remain unlysed despite being contacted by an NK cell. 91 Analyses in microwells
have also shown that individual primary human NK cells undergo dynamic changes
in their migratory behavior by transitioning between directed migration, random
migration, and periods of transient arrests in migration. 97
Microscopy-based assays using micro- and nanowells are well suited for monitoring
cytolytic activity and motility, but to date there has not been an analytical process
that directly integrates the measurement of these dynamic behaviors with other functional outcomes, such as secretory activity, at the single-cell level. Because one of
the primary functions of NK cells is to secrete cytokines that orchestrate immune
responses to pathogens and malignant cells, 100,101 we developed a nanowell-based approach to concomitantly relate the secretory activity of individual NK cells to the
dynamics of their interactions with target cells. Using this integrated process, we
monitored hundreds of isolated NK cell-target cell interactions in parallel and analyzed the relationship between cytolytic activity, secretory activity, and motility. We
found that the acute secretion of IFN-γ from an NK cell is associated with its motility
during contact with the target cell but is not associated with the cytolytic outcome
of the encounter.
2.3
2.3.1
Materials & Methods
Cells and stimulations
Human peripheral blood mononuclear cells (PBMCs) were isolated from whole blood
by Ficoll-Hypaque (Sigma-Aldrich) gradient centrifugation.
Study subjects were
drawn from a healthy donor cohort at the Massachusetts General Hospital in Boston.
The Partners Healthcare Institutional Review Board and the Massachusetts Institute
of Technology Committee on the Use of Humans as Experimental Subjects approved
the study, and each subject gave written informed consent. NK cells were isolated
from PBMCs by negative selection (EasySep Human NK Cell Enrichment Kit; STEMCELL Technologies) and maintained in complete media (RPMI 1640 (Mediatech) sup42
plemented with 10% heat-inactivated fetal bovine serum (PAA), 2 mM L-glutamine,
100 U/mL penicillin, 100 µg/mL streptomycin, and 10 mM HEPES buffer (all from
Mediatech)). Where indicated, NK cells were stimulated with IL-2 (50 U/mL; NIH
AIDS Research & Reference Reagent Program), IFN-α (10 ng/mL; PBL InterferonSource), IL-12 (10 ng/mL), IL-15 (100 ng/mL), or IL-18 (100 ng/mL) (all from R&D
Systems). The major histocompatibility complex (MHC) class I-deficient K562 cell
line (ATCC) was maintained in complete media and used as a model target cell.
2.3.2
Fabrication of nanowell arrays
Arrays of poly(dimethylsiloxane) (PDMS; Dow Corning) nanowells containing either
30 µm or 50 µm cubic wells were prepared by injection molding, as described previously 9,60,62 and in the Supplementary Methods (Appendix A.1.1).
2.3.3
Single-cell cytolysis assay and video microscopy
The single-cell cytolysis (SCC) assay was adapted from an assay we previously developed to identify antigen-specific cytolytic CD8+ T cells. 9 The assay operates by
loading small groups (∼1–5 cells/well) of effector and target cells on an array of
nanowells and observing the cytolytic encounters that occur in each well (Figure 21). Fluorescently labeled NK cells (calcein violet, 1 µM; Invitrogen, or for video
microscopy, eFluor 670, 0.5 µM; eBioscience) and target cells (CellTracker Red, 2.5
µM; Invitrogen) were adjusted to a density of 5 × 105 cell/mL and loaded onto the
arrays. Arrays with 30 µm or 50 µm cubic wells were used to favor the loading of
single or multiple targets per well, respectively. Media containing SYTOX green (nucleic acid stain to identify dead cells, 0.5 µM; Invitrogen) was applied and the arrays
were covered with lifter slips (Electron Microscopy Sciences) or glass slides.
Arrays were imaged using an automated inverted epifluorescence microscope (Axio
Observer; Carl Zeiss, 10×/0.3 objective) fitted with an EM-CCD camera (ImagEM;
Hamamatsu). For the SCC assay, images were collected at 0 h to determine the
initial occupancy of each well and at 4 h to validate the occupancy and determine
43
the acquisition of SYTOX by targets. Arrays were incubated at 37°C and 5% CO2
between the initial and final images. Video microscopy was performed in a similar
manner by imaging a subsection of the array at 8 min intervals over a total of 4 h. A
custom-written script 60 was used to process the images.
2.3.4
Microengraving to detect secreted proteins
The secretion of cytokines and chemokines from the cells residing in each nanowell was measured by microengraving using previously reported protocols. 9,59–62,126
Microengraving measures secretory events from both single- and multi-celled wells,
but it does not directly attribute secretions from multi-celled wells to specific cells
within the well. (The cells’ respective contributions may be deconvolved based on
independently determined knowledge of cellular properties (e.g., certain cells do not
secrete specific analytes under certain conditions).) Secreted macrophage inflammatory protein-1β (MIP-1β, or CC chemokine ligand 4 (CCL4)) and IFN-γ from each
well were captured for 1 h (30 µm wells) or 2 h (50 µm wells) on a glass slide coated
with the corresponding capture antibodies (Figure 2-6). The resulting microarray of
captured proteins was then probed with fluorescent detection antibodies and imaged.
See Supplementary Methods (Appendix A.1.2) for additional details.
2.3.5
Functional and phenotypic flow cytometry
For the functional flow cytometric assays, 106 PBMCs were pre-incubated in the
presence or absence of stimulatory cytokines (as indicated in the text), and then
incubated with K562 target cells at an effector:target (PBMC:K562) ratio of 10:1
in the presence of 0.5 µg/mL Brefeldin A (Sigma-Aldrich), 0.3 µg/mL Monensin
(Golgi-Stop, BD Biosciences), and anti-CD107a-PE-Cy5 antibody (BD Biosciences).
After 4 h of incubation, cells were stained using the LIVE/DEAD Fixable Blue Dead
Cell Stain (Invitrogen), washed, and then incubated for 15 min with the following
antibodies: anti-CD3-Pacific Blue (PB), anti-CD14-PB, anti-CD19-PB, anti-CD16allophycocyanin (APC)-Cy7, and anti-CD56-phycoerythrin (PE)-Cy7 (all BD Bio44
sciences). Cells were then fixed and permeabilized using the Cytofix/Cytoperm solution kit (BD Biosciences) according to the manufacturer’s instructions. Intracellular
cytokine staining was performed for 30 min with anti-IFN-γ-fluorescein isothiocyanate
(FITC) and anti-MIP-1β-PE (both BD Biosciences).
To assess how the different stimulation conditions might affect the phenotypes and
activation states of NK cells, PBMCs were separately stained with the immunophenotypic markers anti-NKG2D-APC, anti-NKp46-PE (both BD Biosciences), antiCD69-FITC, and anti-perforin-peridinin chlorophyll protein (PerCP)-Cy5.5 (both
eBioscience). In all experiments, NK cells were defined as live CD3– CD14– CD19–
lymphocytes expressing CD56 and/or CD16 (Figure A-1). In separate experiments,
surface expression of NKG2D ligands on K562 cells was confirmed using anti-UL16
binding protein (ULBP)-1-PE, anti-ULBP-2-PE, anti-ULBP-3 (primary) (all from
R&D Systems) with goat anti-mouse-APC (secondary), and MHC class I chain-related
protein A/B (MICA/B)-PE (both BD Biosciences) (Figure A-2). Data were acquired
on a BD LSRII (BD Biosciences) flow cytometer and analyzed using FlowJo software
(V9.3.1; Tree Star).
2.3.6
Data analysis
Raw data were processed as described above and in the Supplementary Methods
(Appendix A.1.2). In all experiments, the wells that were located on the edge of the
array were excluded from analysis to minimize any potential edge effects. To analyze
the video microscopy data, a custom-written MATLAB (R2010b; MathWorks) script
was used to calculate parameters describing the interactions of single NK cells with
up to four target cells. Intervals of NK cell-target cell contact were manually verified.
Custom-written MATLAB scripts were used to correlate data from the SCC assay
or video microscopy with data from microengraving on a per-well basis. Statistical
analyses were performed with Prism 5 (GraphPad Software); specific statistical tests
are indicated in the text.
45
2.4
2.4.1
Results
Individual NK cells lyse target cells when co-incubated
in nanowells
We developed a single-cell cytolysis (SCC) assay to directly measure the cytolytic
behavior of thousands of individual NK cells (Figure 2-1). NK cells and K562 target
cells were co-deposited onto an array of 30 µm cubic nanowells to obtain small, isolated groups of cells. The stochastic loading procedure resulted in the majority of the
nanowells being filled with 0–3 cells of each type (Figure A-3). For the SCC assay,
analysis was restricted to the 1,000–5,000 wells per array that contained exactly one
NK cell and one target cell. After 4 h of co-incubation in the nanowells, productive
cytolytic interactions were identified by the appearance of SYTOX+ target cells (Figures 2-1, 2-2A). These initial experiments established that thousands of cytolytic and
non-cytolytic interactions between individual NK cells and targets could be observed
in parallel.
The cytolytic potential of NK cells is strongly modulated by cytokines that are
released locally during an immune response. 108–111,114–116 We therefore assessed how
stimulation with exogenous cytokines affected the cytolysis measured in the SCC
assay (Figure 2-2B). Prior to the assay, NK cells were incubated for 20 h with media,
IL-2, IFN-α plus IL-2, or a combination of IL-12, IL-15, and IL-18. In the absence of
cytokine stimulation, 7% ± 6% of NK cells in nanowells containing a single NK cell
and a single target lysed the target. In each experiment, this rate was significantly
above the 2% ± 1% background death rate of single target cells on the same array (P
< .0001, Chi-squared with Yate’s correction). Activation of NK cells with exogenous
cytokines prior to the assay resulted in a significant increase in the frequency of target
cell death (23–26%) compared with unstimulated NK cells (P < .05, one-way analysis
of variance (ANOVA) followed by Tukey’s post-test) (Figure 2-2B).
The cytolytic potential of NK cells 11 and T cells 10 is associated with target cellinduced surface expression of CD107a, a degranulation marker. Assays employing
46
Image 1 (t = 0 h)
Image 2 (t = 4 h)
0.5 ȝM SYTOX
in media PDMS
Live NK cell
Live target cell
30 µm
Dead target (SYTOX+)
10x
10x
Figure 2-1: Schematic of the single-cell cytolysis (SCC) assay. NK cells and
K562 target cells are loaded onto an array of nanowells. An initial set of images is
acquired to determine the number of cells per well; after 4 h of incubation, a second
set of images is acquired to identify wells in which a killing event occurred (SYTOX+
target cell).
flow cytometry commonly use CD107a as a surrogate measurement for the cytolytic
activity of single NK cells. 11 We therefore compared the frequency of single-cell cytolysis (measured by the SCC assay) with the frequency of NK cells expressing CD107a
after 4 h of co-incubation with targets (measured by flow cytometry) (Figure 2-2B).
For each stimulation condition, the frequency of NK cells that upregulated CD107a
expression was higher than the frequency of cytolytic NK cells observed in the SCC
assay, but the two measurements were significantly correlated (R2 = 0.89, P < .05,
Pearson correlation). The discrepancy in the absolute frequencies may be explained
by differences in bulk compared to single-cell stimulation, by the fact that CD107a can
be expressed on the surface of NK cells that have been activated but have not lysed a
target, 11 or by the fact that the acquisition of SYTOX signal marks the end-stages of
target cell death (membrane permeabilization). Overall, these results demonstrated
that the SCC assay monitors the cytolytic outcome of thousands of individual NK
cell-target cell pairs, and that the cytolytic frequencies measured in this assay reflect
the cytolytic activity as measured in bulk cultures.
47
SCC
80 Pearson
R2 = 0.89
60
**
*
40
*
0
Flow cytometry
**
80
***
**
60
40
###
20
Media
IL-2
IFN- + IL-2
IL-12/15/18
20
No NK
Media
IL-2
IFN- + IL-2
IL-12/15/18
10 µm t = 4 h
Cytolysis (%)
Lysis
t = 0 h
B
CD107a (%)
No lysis
A
0
Figure 2-2: NK cells interrogate and lyse K562 target cells when coincubated in nanowells. (A) Representative images of NK cells (blue) co-incubated
with target cells (red) in the nanowells. Lysed target cells become positive for SYTOX
(green). (B) Frequencies of cytolytic events measured by the SCC assay and of degranulated NK cells measured by flow cytometry after 4 h of NK cell-target cell
interactions. NK cells were incubated for 20 h with the indicated stimulation conditions prior to either assay. Data from the SCC assay were gated to include wells
containing a single NK cell and a single target. Target cell-induced degranulation is
expressed as ∆CD107a, which indicates the difference in expression of CD107a on the
surface of NK cells that were or were not exposed to target cells in bulk. Mean and
standard deviation (SD) for at least three donors per condition are shown. ###P
< .0001, Chi-squared with Yate’s correction comparison of lysis rates of target cells in
wells on the same array containing zero NK cells (background death) or one NK cell.
*P < .05, **P < .01, ***P < .001, one-way ANOVA followed by Tukey’s post-test.
48
2.4.2
Small groups of NK cells do not cooperate to kill single
target cells
We next quantified the frequency of cytolysis in nanowells containing different numbers of NK cells (zero, one, two, or three NK cells) in the presence of a single target
to determine whether NK cells kill more efficiently in groups than in isolation (Figure
2-3A). As expected, increasing the number of NK cells in the well led to increased
frequencies of cytolysis. To determine whether this increase occurred in a cooperative
or independent manner, we tested if the observed frequency of lysis in wells containing one NK cell could predict the frequency of lysis in wells containing multiple NK
cells. If NK cells act independently when killing a single target, then the predicted
frequency of dead targets in wells with n NK cells is
1 − (1 − Pdeath )n
(2.1)
where Pdeath is the frequency (probability) of the target cell being lysed when colocated with a single NK cell. We found that the observed frequencies of lysis in wells
with multiple NK cells were not significantly different from the predicted frequencies
(P > .05 for 23 of the 24 cases tested, Fisher’s exact test) (Figure 2-3B). This result
suggested that NK cells residing in small groups in nanowells operate independently to
mediate the acute lysis of a single target cell; in other words, they do not attack more
effectively as a group than as individuals (e.g., by enhancing the cytolytic behavior
of neighboring NK cells or by delivering partial lytic hits that cumulatively, but not
individually, result in lysis). We note that cooperative effects over larger length scales
in vivo could occur by the selective recruitment of specific subsets of NK cells to the
site of infection or tumorigenesis.
49
CellTracker Red (MFI)
A
10
4
10
3
10
2
0 NK cells
1%
10
2
10
4
10
3
10
2
1 NK cell
11%
10
2
10
4
10
3
10
2
2 NK cells
20%
10
2
10
4
10
3
10
2
3 NK cells
34%
10
2
Target cell SYTOX at t = 4 h (MFI)
B
Cytolysis (%)
80
80
Media
80
IL-2
80
IFN- + IL-2
60
60
60
60
40
40
40
40
20
20
20
20
0
0
1
2
NK cells (#)
3
0
0
1
2
3
NK cells (#)
0
0
1
2
NK cells (#)
3
0
IL-12/15/18
0
1
2
3
NK cells (#)
Figure 2-3: NK cells act independently when lysing a single target cell in a
nanowell. (A) Representative distribution of SYTOX intensity in target cells after
being co-incubated for 4 h in nanowells with one, two, or three NK cells. Data were
gated to include wells containing a single target cell. Example images show the initial
occupancy of the nanowells for each combination. The percentage of lysed target
cells (double-positive for SYTOX and CellTracker Red) is indicated. In this example,
the NK cells were obtained from the donor depicted in gold in the plots in Panel
B and were activated with IFN-α plus IL-2. MFI, median fluorescence intensity.
(B) Observed frequencies of cytolysis (filled circles; colors indicate unique donors)
in nanowells containing 0–3 NK cells. For each donor, the frequencies of cytolysis
predicted from the frequency observed in wells containing a single NK cell are shown
as dotted lines. Prior to the assay, NK cells were stimulated with the indicated
conditions for 20 h.
50
2.4.3
IL-2-activated NK cells lyse target cells with heterogeneous dynamics and in an order-dependent manner
The induction of functional responses in NK cells depends heavily on the dynamics of
contact between NK cells and the target cells that they interrogate. 94,125 To characterize the dynamics of contact, we performed time-lapse microscopy on a subsection
of the array to collect movies of single NK cells interrogating 0–4 target cells in 50
µm cubic wells (Figure 2-4). NK cells were stimulated for 44 h in 50 U/mL IL-2 prior
to the assay to ensure a robust functional response. This stimulation activated NK
cells (47% ± 9% CD69+ ) but did not change the expression of NKp46, NKG2D, or
perforin compared with unstimulated NK cells (Figure A-4).
NK cells patrolled the nanowells and formed multiple contacts with targets in the
wells (Figure 2-5A). NK cells that formed contacts with targets mediated the cytolysis
of one or more targets in 25% of the cases (81 out of 324, compiled from two donors).
Among cytolytic NK cells, the duration of contact with a target prior to lysis (marked
by membrane permeabilization) ranged from less than 8 min to over 200 min (Figure
2-5B). This wide distribution in times implies that the initiation and execution of
cytolytic activity following contact with a target occurs with heterogeneous dynamics.
The probability of a target cell being lysed depended upon the order in which it
was contacted by the NK cell—the first target that an NK cell encountered was more
likely to be lysed than targets that were encountered subsequently (Figure 2-5C). A
decrease in cytolytic activity over sequential encounters with target cells could occur
if the cytolytic potential of an NK cell is diminished after the delivery of a lytic hit.
To test this hypothesis, we examined whether NK cells that lysed the first target they
encountered were able to lyse other target cells upon subsequent contact. Among the
NK cells that contacted 2–4 target cells, 27% (37 out of 135) lysed the first target
that they encountered. Within the group of cells that exhibited cytolytic activity
upon the first encounter, 19% (7 out of 37) also lysed additional targets that they
encountered later. This result suggested that NK cells retain their cytolytic potential
after delivering a lytic hit (consistent with previous reports of serial killing 93 and
51
bulk studies demonstrating that NK cells retain perforin and granzyme loading after
target cell-induced degranulation 127 ), but that only a subset of the initially cytolytic
NK cells proceed to serially lyse multiple targets. In addition, among the group of
NK cells that did not lyse the first target, 93% (92 out of 98) also did not lyse any
of the other targets that they subsequently encountered. This result demonstrated
that NK cells that do not lyse the first target they encounter are unlikely to exhibit
cytolytic activity upon subsequent encounters with similar target cells within short
periods (4 h). Together, these results suggest that the decrease in cytolytic activity
over sequential encounters with target cells occurs because (1) only a fraction of NK
cells that lyse the first target they encounter continue on to lyse additional targets,
and (2) NK cells that do not lyse the first target they encounter are unlikely to lyse
targets that they encounter later.
2.4.4
Microengraving measures the short-term secretory activity of individual NK cells
In addition to their cytolytic activity, NK cells also contribute to immune control by
secreting cytokines and chemokines. We used microengraving, 9,59–62,126 a technique to
detect the proteins that are secreted by the cells residing within each nanowell on the
array (Figure 2-6), to measure the short-term secretory profiles of individual NK cells
and match them to the corresponding observations of the NK cell-target cell interactions (Figure 2-7A). We focused our analysis on the secretion of MIP-1β and IFN-γ.
These proteins are secreted by NK cells and play important roles in coordinating the
immune response. MIP-1β is a chemokine that can attract NK cells 128 and other
immune cells; it also suppresses human immunodeficiency virus (HIV) infection by
inhibiting the binding of HIV to CC chemokine receptor type 5 (CCR5). 19,20 IFN-γ
is a cytokine that inhibits viral replication, increases the presentation of antigen, and
polarizes the differentiation of T helper cells to the Th1 lineage. 129
We first validated that microengraving could detect characteristic patterns of secretory activity from NK cells that interacted with target cells. Prior to microen52
16 min
64 min
120 min
176 min
Target cell
log(SYTOX) (MFI)
Center-to-center
distance (µm)
20 µm 40
20
0
5
3
1
0
40
80
120
160
Time (min)
200
240
Figure 2-4: Automated tracking of interactions between NK cells and target
cells. Representative composite micrographs of a single IL-2-activated (50 U/mL, 44
h) NK cell (blue) interacting with a target cell (red) and inducing cytolysis (green;
SYTOX). Images were acquired every 8 min for 4 h. For each target cell, the SYTOX
intensity and the distance to the NK cell were calculated using automated tracking
scripts.
53
Frequency (%)
A
100
Contacted
targets (#)
80
0
1
2
3
4
60
40
20
0
1
2
3
4
B
C
Lytic events (#)
15
First contact
After t0
Before
or at t0
10
5
0
8
56 104 152 200
Rel. probability of lysis
Targets in well (#)
1.0
Contacted
targets (#)
1
2
3
4
0.8
0.6
0.4
0.2
0.0
1
2
3
4
Order of target contact
Contact before lysis (min)
Figure 2-5: Dynamics of cytolytic interactions between NK cells and target
cells. (A) Distribution of the number of contacts formed by single NK cells with
distinct target cells over 4 h as a function of the total number of target cells residing
in each nanowell. (B) Histogram of the total NK cell-target cell contact time required
for cytolytic NK cells to induce the permeabilization of the membrane of the target
cell. Data are segregated into pairs of NK cells and target cells that were not yet
in contact in the first image (dark grey bars) or were already in contact in the first
image (light grey bars). (C) Relative probability of an NK cell lysing the first, second,
third, or fourth target cell that it encountered. Groups are segregated based on the
total number of distinct target cells that were encountered, and relative probabilities
were calculated as the fraction of lysed target cells in each group that were the nth
target cell to be contacted by the NK cell. Panels A and B are representative of two
independent donors; Panel C shows combined data from both donors.
54
graving, NK cells were incubated for 20 h in either media alone or a combination of
IL-12, IL-15, and IL-18. NK cells were then loaded with target cells into 30 µm cubic
nanowells and incubated for 4 h to allow interactions with targets to take place. Following incubation, microengraving was performed for 1 h to capture secreted MIP-1β
and IFN-γ, and wells that contained a single NK cell and multiple target cells were
analyzed for secretion (Figure 2-7B). Previous reports have shown that, in the absence
of activation with exogenous cytokines, MIP-1β is the dominant secretory product
from NK cells that interact with K562 target cells. 12 Similarly, we found that MIP1β was secreted by a small but consistently detectable fraction of NK cells that were
cultured in media alone prior to being introduced to the target cells in the nanowells.
NK cells that were activated with IL-12, IL-15, and IL-18 prior to the assay exhibited increased secretion of MIP-1β and strong secretion of IFN-γ. This finding was
consistent with previous reports showing that IL-12, IL-15, and IL-18 activate NK
cells and induce the secretion of cytokines, especially IFN-γ. 12,112,113 Analogous assays
performed using ICS in combination with flow cytometry produced similar trends in
relative secretory activity (Figure 2-7B). Together, these experiments confirmed that
microengraving can be used to profile the secretion of MIP-1β and IFN-γ from NK
cells interacting with target cells in the array of nanowells.
2.4.5
Secretion of IFN-γ from IL-2-activated NK cells is associated with reduced motility during contact with target
cells
As a population, NK cells patrol their local microenvironment (e.g., lymph node or
tissue), interrogate target cells, lyse target cells, and secrete cytokines. How these
diverse functional parameters are related in each individual NK cell, however, has
not been well characterized. To address this question, we matched observations of
interactions (determined by time-lapse microscopy) between single IL-2-activated (50
U/mL, 44 h) NK cells and target cells in 50 µm cubic wells to measurements of secreted
MIP-1β and IFN-γ collected via microengraving from the same nanowells (Figure 255
Secreted cytokine
Capture antibodies
Glass
Separate and probe with
detection antibodies
PDMS
Detection antibodies Secreted cytokine
Capture antibodies
Microarray of secreted products
Figure 2-6: Schematic of microengraving to detect secreted cytokines and
chemokines. A capture antibody-coated glass slide is placed in contact and incubated with the array of nanowells; this process creates a matched microarray of
secreted proteins that is subsequently detected with fluorescently labeled antibodies,
imaged, and analyzed to correlate secretion events to the nanowell of origin.
56
A
t = 0 h
t = 4 h
MIP-­1ȕ
IFN-­Ȗ
10 µm ICS
20
IL-12/15/18
Media
0
IL-12/15/18
10
80
60
40
20
0
IL-12/15/18
30
100
Media
40
ICS+ NK cells (%)
MIP-1
IFN-
Media
50
Media
Secreting single
NK cells (%)
Microengraving
IL-12/15/18
B
Figure 2-7: Microengraving measures the short-term secretory activity of
individual NK cells. (A) Representative matched images measuring cytolytic activity and the secretion of proteins from the same NK cell. Cells are labeled as
described in Figure 2-2A. (B) Production of cytokines and chemokines measured by
microengraving or intracellular cytokine staining (ICS). NK cells were incubated for
20 h with the indicated stimulation conditions prior to either assay and then incubated with target cells in the nanowells (microengraving assay) or in bulk (ICS assay)
for 4 h. The microengraving data were gated to include wells containing a single NK
cell and multiple target cells. Mean and SD for at least three donors per condition
are shown. Note difference in scale between the two graphs.
57
8). The resulting collection of functional profiles allowed us to quantitatively analyze
cytolysis, secretion, and the dynamics of interaction in individual NK cells.
We first examined whether target cell-induced secretory activity and cytolytic
activity were associated for individual NK cells. NK cells that contacted at least one
target cell displayed robust secretion of MIP-1β and IFN-γ (Figure 2-9A). Secretion
was not observed in the absence of contact with a target, even if a target cell was
present in the well. Among NK cells that contacted at least one target, however,
there was no significant difference in the secretion of MIP-1β or IFN-γ between the
cytolytic and non-cytolytic groups of NK cells (Figure 2-9A). These results suggest
that acute, target cell-induced secretory activity is not closely correlated to cytolytic
activity in individual IL-2-activated NK cells.
To investigate how the dynamics of contact with target cells may regulate cytolysis
and secretion in single NK cells, we next examined whether the motility of NK cells
during contact with targets was associated with the subsequent cytolytic or secretory
outcome of the encounter. The velocities of NK cells during periods of contact with
target cells were reduced compared with the velocities in the absence of targets or after
disengagement from a target (Figure 2-9B; P < .001, Kruskal-Wallis test followed by
Dunn’s post-test). When the velocities of NK cells during contact with targets were
further segregated based on the functional response (MIP-1β –/+ , IFN-γ –/+ , lysis–/+ ),
we found that cells that secreted IFN-γ had a significantly reduced velocity during
contact with targets compared with cells that did not secrete IFN-γ (Figure 2-9C; P
< .05, Mann-Whitney test), and that these cells remained more arrested at the initial
point of contact with the target (Figure 2-10). There was no significant difference
between the during-contact velocities of NK cells that were segregated by the secretion
of MIP-1β or cytolytic outcome (Figure 2-9C). Together, these results suggest that
NK cells receive stop signals of varying strengths after contacting a target, and that
the degree of arrest in motility is related to the subsequent secretion of IFN-γ.
58
Secreted factors
Time-­lapse imaging
8 min
48 96 144
200
240
MIP-­1ȕ IFN-­Ȗ
20 µm Figure 2-8: Correlated measurement of the dynamics of NK cell-target cell
interactions and secreted products. Representative composite micrographs of
IL-2-activated (50 U/ml, 44 h) NK cell-target cell interactions and corresponding
measurements of secreted chemokines and cytokines. Interactions were tracked for 4
h at 8 min intervals using time-lapse microscopy (labeling as described in Figure 2-4);
microengraving was performed immediately afterward for 2 h to capture chemokines
and cytokines secreted by each NK cell.
59
Secreted protein (MFI)
MIP-1
Secreted protein (MFI)
A
n.s.
***
103
- +
Contact
- +
Lysis
*** ***
6
4
C
After
During
2
0
102
Velocity (µm/min)
8
**
n.s.
- +
Contact
- +
Lysis
103
0 K562
B
Velocity (µm/min)
102
IFN-
3
n.s.
*
n.s.
- +
- +
- +
2
1
0
MIP-1 IFN- Lysis
During contact
Contact
Figure 2-9: Secretion of IFN-γ is associated with the velocity of NK cells
during contact with target cells. (A) Secretion of MIP-1β and IFN-γ from single
NK cells. NK cells are segregated based on whether they contacted a target cell, and,
if they did contact at least one target cell, whether they exhibited cytolytic activity.
Dashed lines indicate the threshold for positive secretion events (background + 2
SD). **P < .01, ***P < .001, n.s., not significant, Mann-Whitney test. (B) Mean
velocities of NK cells in nanowells in the absence of target cells (0 K562), during
contact with target cells, and after contact with target cells. ***P < .001, KruskalWallis test followed by Dunn’s post-test. (C) Mean velocities of NK cells in nanowells
during contact with target cells. Data are segregated by functional outcome. *P
< .05, n.s., not significant, Mann-Whitney test. The mean and standard error of
the mean (SEM) are indicated in red. Results are representative of two independent
donors.
60
20 ȝm
0 K562
-­
IFN-­Ȗ
+
IFN-­Ȗ
During contact
Figure 2-10: Cell tracks segregated by functional outcome. Tracks of NK cells
in the absence of target cells (left) or during contact with a target cell (right). The
tracks during contact are segregated by secretory outcome. Each plot displays n =
15 tracks that were randomly selected from the pool of all eligible cell tracks. Results
are representative of two independent donors.
2.5
Discussion
NK cells induce cytolysis and secrete chemokines and cytokines in response to malignant or infected cells, and thus play a critical role in the immune control of cancer
and infectious diseases. 101,130 Characterizing how individual NK cells interact with,
and subsequently respond to, potential target cells is important for understanding the
mechanisms by which they eliminate targets but leave healthy cells unharmed. Our
group has previously described the use of arrays of nanowells to screen for HIV-specific
CD8+ T cells. 9 Here, we extended this approach to monitor interactions between individual NK cells and K562 target cells and then correlated these interactions to
linked measurements of the cytolytic and secretory outcomes from the individual NK
cells. The results of these dynamic, single-cell measurements begin to define the rules
of engagement that modulate the functionality of NK cells during their response to
target cells.
The unique ability of the SCC assay to precisely count the number of NK cells
and targets in each nanowell allowed us to quantitatively investigate whether small
61
groups of NK cells act cooperatively when killing an isolated target cell. Our results
were consistent with a model in which each NK cell operates as a cytolytic agent that
is independent of other NK cells present in the well. The lack of cooperative cytolytic
behavior suggests that local paracrine signaling (over the sub-nanoliter volume of the
well) between NK cells does not induce them to work synergistically to kill single
target cells within a 4 h interval. This finding, however, does not eliminate the possibility for cooperative killing behavior among NK cells when killing multiple targets,
killing resistant targets, or operating over larger length scales or longer intervals of
time. Indeed, the bulk self-recruitment of NK cells into regions containing a high
density of target cells is commonly observed. 89,98 In contrast, the self-recruitment of
individual NK cells to aid in the acute elimination of a single target cell has been observed only on occasion, 88 thus supporting our finding that local cooperative behavior
among NK cells is not a major mechanism by which the lysis of single, susceptible
targets is achieved.
The variability in time to lysis that we observed may involve mechanisms by
which NK cells progress through cytoskeletal polarization checkpoints before delivering a successful lytic hit 125 or cell-to-cell differences in the kinetics of apoptosis in
targets. 131 NK cells lyse K562 target cells primarily by delivering perforin/granzymeloaded cytolytic granules into the lytic synapse 132,133 rather than by triggering death
receptor-mediated pathways of apoptosis induced by Fas ligand 134–136 (FasL) or tumor necrosis factor (TNF)-related apoptosis-inducing ligand 137 (TRAIL). Some of the
heterogeneity in lysis times and behaviors, however, could result from NK cells inducing the death of target cells by other mechanisms—for example, by using membrane
nanotubes to facilitate cytolytic interactions with target cells. 138
Our finding that cytolytic NK cells are most likely to lyse the first target cell they
encounter suggests that they do not need to integrate activating signals collected during prior encounters with targets before overcoming the signaling threshold required
for the delivery of a lytic hit; instead, they exert their cytolytic activity at the first
opportunity. This pattern of activation is different from the pattern followed by T
cells during priming in the lymph node under certain conditions, where it has been
62
shown that the signals received through the T cell receptor are integrated over serial
encounters with antigen-presenting cells until a threshold of activation is reached. 80
Because both the density and the type of ligands displayed by target cells have a
large effect on the functional response of NK cells 12,103,104 it will be interesting to test
whether different target cells produce different patterns of order-dependent cytolytic
activity.
Cytolysis and secretion are the primary functions of NK cells during an immune
response. By collecting linked measurements of the cytolytic and secretory activities
of individual IL-2-activated NK cells upon exposure to target cells, we demonstrated
that the acute secretion of IFN-γ and MIP-1β is not associated with the cytolytic
status of NK cells. Studies using bulk mixtures of NK cells and target cells have
also suggested that cytolysis and secretion may be differentially regulated or temporally distinct. 12,139–142 Furthermore, cytokines (e.g., IFN-γ, TNF) and perforin-loaded
cytolytic granules are sorted into separate endosomes and released in distinct fashions from NK cells, 139 providing mechanistic support to the idea that cytolysis and
secretion can be controlled independently within individual cells.
The relationship between the motility of NK cells and their secretory responses to
target cells has not previously been analyzed at the single-cell level. Using integrated
measurements of single, IL-2-activated NK cells, we found that target cell-induced secretion of IFN-γ was associated with the motility of the NK cells during contact with
targets. The during-contact velocity of NK cells that secreted IFN-γ was significantly
lower than that of NK cells that did not secrete IFN-γ. Previous reports have demonstrated that NK cells display a dose-dependent reduction in velocity when exposed
to increasing concentrations of surface-bound MICA, an NKG2D ligand expressed
on K562 target cells, and that these activating signals induce NK cells to arrest and
form stable synapses with target cells. 94 It has also been shown that the level of
activation that is necessary to induce the secretion of IFN-γ is higher than that necessary to induce degranulation or the secretion of MIP-1β. 12 Our findings, therefore,
are consistent with a model in which NK cells that receive the strongest activating
signals upon encounter with a target have the lowest during-contact velocity (due to
63
receiving a strong stop signal) and are also the most likely to surpass the signaling
threshold required to induce the secretion of IFN-γ. A similar association between
antigen-induced arrest and the secretion of IFN-γ has also been observed at the population level in mouse CD4+ T cells. 143 Interestingly, we did not detect a significant
difference in the during-contact motility of cells that did or did not secrete MIP-1β
(a chemokine), although in other systems of interacting lymphocytes the secretion of
chemokines has been shown to affect the characteristics of the immune synapse. 144
In general, the association between the during-contact velocity and the secretion of
IFN-γ suggests that, for NK cells, there is overlap in the target cell-triggered signaling
pathways that control motility and those that control the secretion of cytokines.
In summary, we have developed and validated a nanowell-based platform to quantitatively monitor and correlate the contact dynamics and functional outcomes (cytolysis, secretion) of interactions between individual NK cells and target cells. Integrated measurements such as these have not previously been possible to perform at
the single-cell level due to the constraints of conventional assays. In the present study,
we analyzed interactions between NK cells and tumor target cells, but the nanowellbased platform presented here could also be used to examine cross-regulation between
NK cells and virally infected cells, DCs, or T cells with single-cell resolution. Moreover, the platform is suitable for use with samples containing a limited number of cells
(104 –105 ), and thus will enable unprecedented functional analysis of clinical biopsy
specimens. Such applications could include the analysis of mucosal NK cells in HIV infection 145 or of intrahepatic NK cells in viral hepatitis. 146 In the current study, which
focused on interactions between peripheral blood NK cells and K562 tumor target
cells, we demonstrated that local cooperativity among NK cells does not contribute
significantly to the lysis of single target cells, and that short-term secretory responses
arising from individual NK cells are correlated with the dynamics of NK cell-target
cell interactions but not cytolysis. Together, these results illustrate new properties by
which NK cells perform immune surveillance and mediate the elimination of target
cells.
64
Chapter 3
Dynamic and Functional
Correlates of Activation-Induced
Shedding of CD16 from Natural
Killer Cells after Discrete
Interactions with Target Cells
65
3.1
Abstract
FcγRIIIA (CD16) is an activating receptor that is expressed on ∼90% of resting
natural killer (NK) cells. CD16 mediates the recognition and cytolysis of antibodycoated target cells by NK cells, but it is proteolytically shed upon stimulation by
target cells. This shedding renders NK cells deficient in responding to subsequently
encountered antibody-coated target cells. The dynamics by which and extent to
which CD16 is shed from an NK cell during a discrete interaction with a single target
cell are unknown. Here, we addressed this question by using video microscopy to
track the dynamic interactions between individual NK cells and target cells and then
measuring the amount of CD16 that remained on the surface of the NK cells at the
conclusion of the interaction. By profiling thousands of pairs of interacting NK cells
and target cells with a distribution of contact times, we found that the loss of CD16
from an NK cell was directly correlated to the length of time that the NK cell spent
in contact with a target cell. Contact with a single antibody-coated target cell was
sufficient to induce the loss of CD16. The length of contact time required to induce
the loss of CD16 was shorter in IL-2-activated NK cells than in resting NK cells. Loss
of CD16 was associated with the performance of effector functions such as lysis of
the target cell and secretion of cytokines and chemokines, but a significant fraction of
NK cells that shed CD16 did not exert any of the effector functions that we assessed
(killing, secretion of IFN-γ, secretion of CCL4). Furthermore, confined groups of two
or three NK cells (with at least one target cell) were more likely to all shed CD16 than
would be expected by chance, which suggests that shedding occurs in a cooperative
fashion. Together, these results demonstrate that prolonged engagement of a single
antibody-coated target cell is sufficient to induce the shedding of CD16 but that NK
cells that shed CD16 do not necessarily perform effector functions such as cytolysis
or the secretion of IFN-γ or CCL4.
66
3.2
Introduction
Antibodies that recognize antigens on diseased target cells can be generated naturally by humoral immune responses (e.g., non-neutralizing antibodies against viral
pathogens) or administered as part of treatment regimes (e.g., tumor-targeting monoclonal antibodies). The Fab (Fragment antigen binding) regions of these antibodies
bind to the antigens expressed on the target cell, while the Fc (Fragment crystallizable) regions are left exposed on the surface of the cell. Innate immune cells, including
natural killer (NK) cells, macrophages, monocytes, and granulocytes, are equipped
with receptors that recognize the Fc regions of antibodies and induce effector mechanisms that lead to the destruction of antibody-coated target cells. 147
The majority (∼90%) of CD56dim NK cells express the receptor FcγRIIIA (CD16),
which binds with low affinity to the Fc region of certain subclasses of immunoglobulin G (IgG). 100,147–149 Upon encounter with an antibody-coated target cell, the CD16
receptors on NK cells bind to the Fc regions of the coating antibodies. The resulting
engagement of CD16 induces the phosphorylation of immunoreceptor tyrosine-based
activation motif (ITAM)-bearing adaptor molecules including CD3ζ and Fc�RIγ,
which trigger a cascade of activating signaling. 106,150,151 These activating signals can
stimulate NK cells to mediate antibody-dependent cellular cytotoxicity (ADCC) by
directing cytolytic granules into the target cells. In addition to killing target cells via
ADCC, stimulation through CD16 also induces NK cells to secrete anti-viral and immunostimulatory cytokines and chemokines such as interferon-γ (IFN-γ), macrophage
inflammatory protein-1β (MIP-1β, or CC chemokine ligand 4 (CCL4)), and tumor
necrosis factor-α (TNF). 12 Unlike other activating receptors on NK cells, which only
induce robust cytolytic or secretory responses when they are synergistically triggered
in combination, engagement of CD16 alone is sufficient to induce strong effector responses from NK cells. 103,104
The activation of NK cells by a variety of stimuli results in the loss of CD16
from the cell surface. Expression of CD16 is significantly reduced on NK cells that
have been treated with phorbol 12-myristate 13-acetate (PMA), 152–154 which potently
67
stimulates immune cells by binding and activating protein kinase C (PKC). Surfaceexpressed CD16 is also reduced on NK cells after exposure to activating cytokines such
as interleukin-2 (IL-2), IL-12, IL-15, and IL-18. 153,154 The engagement of activating
receptors on NK cells (CD16, NKG2D, etc.), either by antibody-coated plates or by
target cells, also leads to loss of CD16. 153,155–157 Interestingly, one of the most potent
inducers of the loss of CD16 is the engagement of and stimulation through CD16
itself. 153
Receptors such as CD16 can be removed from the surface of cells by internalization or shedding. Although the application of soluble anti-CD16 to NK cells induces
the internalization of CD16, 158 several lines of evidence suggest that metalloproteaseinduced shedding drives the loss of CD16 following other routes of activation. After
activation with PMA or IL-12 and IL-18, soluble CD16 fragments can be detected
in the supernatant, suggesting that CD16 is cleaved from the surface of activated
NK cells rather than (or perhaps in addition to) being internalized. 152,153 Furthermore, stimulation of NK cells in the presence of phenylarsine oxide (PAO), which
disrupts endocytosis, does not stop the loss of CD16 from the surface of the cells. 154
In contrast, retention of surface-expressed CD16 on activated NK cells is significantly
increased when the cells are treated with broad-spectrum inhibitors of proteases or
more selective inhibitors of specific matrix metalloproteinases (MMPs) or proteases
from the ADAM (a disintegrin and metalloprotease) family. 152–157 In support of these
findings, ADAM17 and MMP25 have recently been shown to cleave CD16 from the
surface of activated NK cells. 153,154,157
The shedding of CD16 from the surface of NK cells after encounters with target
cells is of clinical interest, because the loss of this receptor renders NK cells deficient
in their ability to mediate ADCC or to mount other effector functions in response to
antibody-coated target cells. 156 In many disease settings, target cells are a minority
population amidst a sea of healthy cells (e.g., antibody-coated HIV-infected CD4+ T
cells in the blood, or malignant NKG2D+ cells in precancerous tissue). Experimentally, however, it is difficult to quantify on a per-cell basis the relationship between
the stimulation that an NK cell receives from individual encounters with target cell(s)
68
and the loss of CD16 (or other functional outcomes of interest). This difficulty stems
from the technical challenge of correlating the effective “dose” of target cells that any
given NK cell is exposed to with the subsequent shedding of CD16 from the NK cell.
Standard experimental protocols commonly modulate the global amount or type
of stimulation that is presented by the population of target cells. Global modulation
can be achieved by varying the coating density of activating ligand (e.g., IgG) on the
surface of the target cells 12 or by using target cells that stimulate through different
activation receptors on NK cells. 153 Although this type of modulation provides useful
information about stimulatory outcomes, it does not enable control over the local
concentration of target cells that each NK cell is exposed to. For example, in a bulk
mixture, an NK cell may have the opportunity to interact with tens or hundreds of
target cells over the course of the assay.
One approach to limit the number of target cells encountered by an NK cell is to
use vastly more non-target cells than target cells in the cellular mixture. Although
this approach would ensure that the majority of NK cells would encounter at most
a single target, it would not answer the question of whether an NK cell that retains
CD16 on its surface does so because it did not encounter a target cell (and hence
was not stimulated) or because it encountered a target cell but did not shed CD16.
Target-induced stimulation can also be controlled by restricting the co-incubation
time; however, this approach precludes the analysis of responses that are induced by
an interaction with a single target cell but that require time to develop. Instead, it
would be desirable to measure the phenotypic and functional responses (such as the
shedding of CD16) of NK cells in a way that links each cell’s response to the duration
of stimulation that it received from a discrete encounter with a target cell.
Here, we extend the nanowell-based platform introduced in Chapter 2 to enable
measurement of the shedding of CD16 from NK cells that engaged a single or small
number of target cells for varying durations of time. We also correlate the activationinduced shedding of CD16 from individual NK cells with the functional properties
(cytolysis and secretion of cytokines) of each cell.
69
3.3
3.3.1
Materials & Methods
Cells and stimulations
Peripheral blood mononuclear cells (PBMCs) were isolated by density centrifugation
using Ficoll-Paque PLUS (GE Healthcare) from the whole blood of healthy donors
(Research Blood Components, Boston, MA; or the Brigham & Women’s Hospital
(BWH) PhenoGenetic Project, Boston, MA) and were cryopreserved in 90% heat
inactivated fetal bovine serum (FBS; PAA Laboratories) with 10% dimethyl sulfoxide (DMSO; Electron Microscopy Science). Before use, cryopreserved PBMCs were
thawed, washed with complete media (RPMI-1640 (Mediatech) supplemented with
2 mM l-glutamine, 100 U/mL penicillin, 100 µg/mL streptomycin, 10 mM HEPES
buffer (all from Mediatech), and 10% FBS), and then incubated (37°C, 5% CO2 ) in
complete media. Cells were either rested for 1 h in media only or were activated for
20 h with 100 U/mL of recombinant human IL-2 (PeproTech).
P815 and K562 cells were used as target cells. P815 cells are a murine mastocytoma cell line, and K562 cells are a major histocompatibility complex (MHC) class
I-deficient chronic myelogenous leukemia cell line. Both cell lines were purchased
from ATCC and maintained in complete media without antibiotics. Target cells were
split the day before the assay. Immediately before preparing the target cells for the
single-cell assays, dead target cells and cellular debris were removed from the cultures
by density centrifugation using Ficoll-Paque PLUS. To stimulate NK cells through
the CD16 receptor and induce ADCC, P815 target cells were coated with rabbit antimouse lymphocyte antibody (1 µL per 100,000 cells; Accurate Chemical & Scientific
Corporation) for 30 min and then washed with complete media.
70
3.3.2
Measurement of surface-expressed CD16 and NKG2D
by flow cytometry after bulk stimulation with target
cells
PBMCs (106 /well) and target cells (105 /well) were added in a total of 200 uL to a
U-bottom 96-well plate. The plate was centrifuged for 1 min at 10 g to bring the
cells to the bottom. Cells were incubated for the indicated periods of time (37°C, 5%
CO2 ) and then stained with antibodies against surface markers or isotype controls
(anti-CD3-FITC, anti-CD14-FITC, anti-CD19-FITC, anti-CD56-PerCPCy5.5, antiCD16-BV421, anti-NKG2D-PE, BV421 IgG1 κ isotype control, PE IgG1 κ isotype
control). Antibodies are described in further detail in the Supplementary Methods
(Appendix B.1.1). Staining was performed in media at 4°C for 30 min. Viability was
assessed using the LIVE/DEAD Fixable Aqua Stain (Life Technologies). Cells were
then washed in media, suspended in fixation buffer (1% paraformaldehyde (Electron
Microscopy Sciences) in phosphate-buffered saline (PBS)), and acquired on a BD LSR
Fortessa (BD Biosciences) flow cytometer. Compensation was performed using singlestained beads (Quantum Simply Cellular anti-mouse IgG beads; Bangs Laboratories).
CD56dim NK cells were gated as dead– CD3– CD14– CD19– CD56dim lymphocytes.
3.3.3
Nanowell-based assays to measure the shedding of CD16
and functional activity of individual NK cells
Preparation of cells
CD56dim NK cells were isolated from PBMCs by flow sorting. PBMCs were stained
with anti-CD3-FITC, anti-CD14-FITC, anti-CD19-FITC, and anti-CD56-APC in media at 4°C for 30 min. Antibodies are described in further detail in the Supplementary
Methods (Appendix B.1.1). After washing, cells were resuspended in PBS with 5%
FBS, filtered, and sorted by gating on CD3– CD14– CD19– CD56dim lymphocytes (BD
FACSAria III, BD Biosciences). Compensation was performed using single-stained
beads. Sorted CD56dim NK cells were rested in media for 1 h at 37°C and then stained
71
with the proliferation dye eFluor 670 (2 µM in PBS for 5 min at 37°C, eBioscience).
Target cells were labeled with either CellTrace Violet (CTV; 2 µM in PBS for
20 min at 37°C) or CellTracker Red (CTR; 2.5 µM in media for 20 min at 37°C)
(both from Life Technologies) as indicated in the text. For experiments in which
both antibody-coated P815 and K562 cells were used on the same array, the P815
cells were labeled with CTV and the K562 cells were labeled with CTR. For experiments with a single type of target on the array, both targets were labeled with CTV.
Labeling was performed prior to density centrifugation using Ficoll-Paque PLUS and
(in the case of the P815 cells) coating with antibody.
Video microscopy and microengraving
Arrays of 50 µm cubic nanowells were prepared as described in Appendix D.1.1,
with the following modifications: First, arrays were designed to include 121 wells per
field of view (as opposed to 49); the resulting arrays contained 83,490 wells. Second,
the poly(dimethylsiloxane) (PDMS) was cured overnight. Third, the arrays were
collected into media immediately after plasma treatment.
Labeled NK cells and target cells were adjusted to a concentration of 106 /mL,
combined and mixed, and then deposited onto the array (∼300 µL). After letting the
cells settle by gravity for 5 min, arrays were gently washed with media containing
SYTOX green (nucleic acid stain to identify dead cells, 1 µM; Life Technologies) and
0.02% human serum (containing IgG; to facilitate registration of the microarray of
secreted proteins). The array was then sealed with a glass slide coated with capture antibodies against IgG (Life Technologies / ZyMAX), IFN-γ (Mabtech), and
CCL4 (R&D Systems) (prepared as described in Appendix D.1.3) and clamped in a
microarray hybridization chamber (Agilent).
The sealed array was removed from the chamber after 20 min at room temperature
and transferred to a stage-top incubator (TOKAI HIT) mounted on an automated inverted epifluorescence microscope (Axio Observer; Carl Zeiss, 10×/0.3 objective with
a 0.63× demagnifying lens) fitted with an EM-CCD camera (ImagEM, Hamamatsu).
A subsection of the array (>35,000 wells) was imaged at 8 min intervals for 6 h. The
72
array remained sealed with the glass slide during the imaging, allowing the collection
of secreted proteins (microengraving) and also ensuring that the contents of each well
(both cells and secreted molecules) were physically isolated from neighboring wells.
Immediately after the last image was collected, the sealed array was removed
from the stage-top incubator and placed in a 4-well dish on ice for 1 h. This step was
necessary to reduce the number of NK cells that were lost when the array was subsequently unsealed. The 4-well dish was then centrifuged (340 g, 1 min),1 and the glass
slide (containing the microengraved proteins) was carefully removed from the array
of nanowells under ice-cold media. The microengraved slide was then processed with
detection antibodies (anti-IgG-AF700, Jackson Immunoresearch; anti-IFN-γ-AF594,
Mabtech; anti-CCL4-AF647, R&D Systems) as described in Appendix D.1.3 and imaged on a commercial microarray scanner (GenePix 4400A, Molecular Devices).
On-chip staining for CD16 and imaging cytometry
On-chip staining for CD16 was performed by gently aspirating excess media from
the array of nanowells, applying staining solution (1 test volume of anti-CD16-PerCP
(Biolegend, clone 3G8) in 200 µL media), covering the array with a lifter slip (Electron Microscopy Sciences), and then incubating for 1–2 h at 4°C. The lifter slip was
then removed under ice-cold PBS; the array was left in the PBS for 15–20 min to
allow unbound antibody to diffuse away. Excess PBS was aspirated and the array was
covered with a lifter slip. The array was then imaged on the microscope. Compensation was performed by imaging single-stained cells and beads (in a 96-well plate) in
each of the relevant fluorescent channels.
3.3.4
Data analysis
Flow cytometry data were analyzed with FlowJo (Tree Star). Statistical analyses
were performed with Prism 5 (GraphPad Software) or R (version 3.0.2). The protein
1
NK cells sometimes crawl onto the surface of the glass slide during the course of the incubation.
This step was performed in an attempt to bring these cells down back down into the wells.
73
arrays produced by microengraving were analyzed using commercial image processing
software (GenePix Pro 7, Molecular Devices).2 The microscopy images were processed
with a custom-written script based on ImageJ to segment the cells and extract the
fluorescence intensities. After this initial step of data extraction, the data from the
video microscopy portion of the assay were analyzed using custom-written MATLAB
(R2013b; MathWorks) scripts. The surface marker data (from the imaging cytometry performed at the conclusion of the nanowell-based assay) were analyzed using a
custom-written script based on MATLAB and R.
3.4
3.4.1
Results
CD16 is shed from the surface of NK cells after bulk
stimulation with target cells
To assess the magnitude and kinetics of the shedding of CD16 from the surface of
NK cells, we incubated IL-2-activated (100 U/mL, 20 h) PBMCs with target cells for
0.5–4 h and then quantified the amount of CD16 remaining on the surface of CD56dim
NK cells by flow cytometry. We used antibody-coated P815 target cells to stimulate
NK cells through CD16, and K562 target cells, which express NKG2D ligands, to
stimulate NK cells through NKG2D. Stimulation of NK cells with antibody-coated
P815 target cells resulted in a significant and rapid decrease of surface-expressed
CD16 (Figure 3-1A). In the absence of stimulation, 80% of NK cells were CD16hi .3
After 30 min of stimulation with antibody-coated P815 target cells, only 9% of the
NK cells remained CD16hi , and by 4 h, <1% remained CD16hi . Stimulation with
K562 target cells also resulted in a rapid decrease in surface-expressed CD16, but
2
The automated image processing software (Crossword) that was recently developed by our lab 159
for the analysis of protein microarrays could not be used here. Crossword relies heavily on the
presence of a clearly defined signal in the background channel. Microengraving without constant
clamping produces a clearly defined signal in the background channel, but this signal is often slightly
shifted (appearing as a “double-print”), making it difficult for Crossword to function properly. The
shift does not occur in the cytokine channels, which makes it straightforward to semi-automatically
align the analysis grid in GenePix Pro 7.
3
Including CD16med cells, 89% were CD16+ .
74
to a much smaller extent than that observed after stimulation with antibody-coated
P815 target cells: 57% remained CD16hi after 4 h (Figure 3-1B).
We also assessed changes in the surface expression of NKG2D following stimulation
with target cells. In the absence of stimulation, 99% of CD56dim NK cells were
NKG2D+ . The expression of NKG2D remained unchanged following stimulation with
antibody-coated P815 cells (Figure 3-1C) but decreased gradually upon stimulation
with K562 cells (Figure 3-1D). After 30 min of stimulation with K562 target cells, 96%
of the NK cells remained NKG2D+ , and by 4 h, 84% still remained NKG2D+ . These
results suggest that “cis” downregulation was greater than “trans” downregulation:
CD16 was downregulated the most by stimulation through CD16 (antibody-coated
P815 target cells) and NKG2D was downregulated the most by stimulation through
NKG2D (K562 target cells).
It has previously been shown that IL-2 induces the shedding of CD16 from NK
cells when high doses (400–500 U/mL) and/or long activation times (up to 4 d) are
used. 153,154 At the dose and activation time used in our experiments (100 U/mL, 20
h), however, we did not observe a significant decrease in the levels of CD16 expressed
on the surface of NK cells in the absence of stimulation with target cells. Of the
CD56dim NK cells, 81% were CD16hi in the absence of activation with IL-2 and 80%
were CD16hi after activation with IL-2. After stimulation with antibody-coated P815
target cells, IL-2-activated CD56dim NK cells downmodulated CD16 more rapidly
than resting CD56dim NK cells (Figure B-1). After stimulation of resting NK cells
with K562 target cells, we observed a decrease in surface-expressed CD16 that was
similar to that observed in Ref. 153 using similar conditions.
Together, these results demonstrated that a large fraction of both resting and IL2-activated NK cells rapidly and completely shed CD16 from their surface in response
to bulk stimulation with antibody-coated P815 target cells, and that a smaller fraction
shed CD16 from their surface in response to bulk stimulation with K562 target cells.
75
CD16 -­ BV421
K562
No target
0.5 h
1 h
2 h
4 h
Isotype control
CD16 -­ BV421
D
K562
Frequency (% max)
P815
Frequency (% max)
C
B
Frequency (% max)
P815
Frequency (% max)
A
NKG2D -­ PE
NKG2D -­ PE
Figure 3-1: Expression of CD16 and NKG2D on CD56dim NK cells after
stimulation with target cells in bulk. IL-2-activated PBMCs were incubated
with (A, C) antibody-coated P815 target cells or (B, D) K562 target cells for 0, 0.5,
1, 2, or 4 h and then stained, fixed, and acquired by flow cytometry. CD56dim NK
cells were gated as CD3– CD14– CD19– CD56dim lymphocytes.
76
3.4.2
CD16 is shed from the surface of NK cells after stimulation with single target cells in nanowells
To precisely quantify and control the number of target cells that each NK cell was
exposed to, we used the platform that was introduced in Chapter 2 to isolate individual NK cells and target cells in nanowells. We flow sorted IL-2-activated CD56dim
NK cells, labeled them with the proliferation dye eFluor 670 to allow unambiguous
identification, and then co-loaded NK cells and target cells (CTV-labeled antibodycoated P815 cells and CTR-labeled K562 cells) onto the array of nanowells. The
array was sealed with a glass slide and incubated for 6 h, during which time the
cellular occupancy of each well was measured by imaging cytometry. To quantify the
amount of CD16 remaining on the surface of each NK cell at the conclusion of the
6 h incubation, we removed the glass slide, stained the array with CD16-PerCP, and
performed imaging cytometry.
Single NK cells residing in wells without any target cells maintained high levels
of CD16, but the expression of CD16 was drastically reduced on NK cells in wells
containing antibody-coated P815 target cells or both types of target cells (Figure 32A). NK cells in wells with K562 cells also exhibited a slight but significant decrease in
the expression of CD16. These results were consistent when the analysis was further
restricted to only include wells with exactly one target cell (Figure 3-2B). Across three
independent donors, 89% ± 3% of NK cells in wells with no target cells were CD16+
(consistent with the baseline frequency of CD16+ NK cells in the CD56dim subset),
whereas only 22% ± 4% of NK cells in wells with 1–3 antibody-coated P815 cells were
CD16+ at the end of the assay (Figure 3-3). Together, these results demonstrated
that engagement of a single antibody-coated target cell is sufficient to induce the
shedding of CD16 from the majority of CD56dim NK cells.
77
●
●
4.0
3.5
●
2.5
3.0
●
●
2.0
●
●
1.5
●
CD16 intensity (logicle)
1.5 2.0 2.5 3.0 3.5 4.0
B
CD16 intensity (logicle)
A
K562
K562
P815+Ab
P815+Ab
(0, 0)
(1+,
1 0) (0, 1+)
1
1
1
Figure 3-2: Expression of CD16 on IL-2-activated CD56dim NK cells after
stimulation with target cells in nanowells. NK cells were flow sorted from IL-2activated PBMCs (gating on CD3– CD14– CD19– CD56dim lymphocytes), labeled with
eFluor 670, and co-loaded onto an array of nanowells with a mixture of K562 and
antibody-coated P815 target cells. After 6 h, cells were stained on the array for CD16PerCP and imaged. (A) The expression of CD16 was analyzed on NK cells in wells
that contained one NK cell and at least one target cell of the indicated type. From
left to right, the number of single NK cells analyzed was: 2064 (no target), 958 (K562
only), 601 (P815 only), 550 (both targets). With the exception of the P815 v. both
targets comparison (no significant difference), the levels of CD16 were significantly
different in all comparisons (***P < 0.001, Kruskal-Wallis test followed by Dunn’s
post-test). (B) Events were further gated to include only single NK cells in wells with
exactly one target cell of the indicated type. From left to right, the number of single
NK cells analyzed was: 2064 (no target), 714 (one K562), 477 (one P815), 198 (one
of each target). With the exception of the empty v. K562 comparison (**P < 0.01),
significance levels were the same as in (A). Results are representative of 2–3 donors
and 2–5 experiments.
78
CD16+ NK cells (%)
***
100
80
60
40
20
0
No
target
1-3
1-3
K562 P815+Ab
Target cells in well
Figure 3-3: The majority of NK cells shed CD16 when co-incubated with
target cells in nanowells. IL-2-activated NK cells were co-incubated on arrays of
nanowells with either K562 cells or antibody-coated P815 cells for 6 h. The frequency
of CD16+ NK cells was analyzed by staining for CD16 and imaging the cells on the
array at the conclusion of the incubation. Only wells containing a single NK cell were
included in the analysis. Bars indicate the median of three donors. A single type of
target cell was used per array, resulting in six measurements of empty wells across
the three donors. ***P < .001, one-way analysis of variance (ANOVA) followed by
Tukey’s post-test.
79
Cumulative
contact
200
100
Time (min)
300
Longest
contact
●
●
●
0
●
P815
+Ab
K562
P815
+Ab
K562
Figure 3-4: Distribution of contact durations between NK cells and target
cells. Interactions between IL-2-activated NK cells and target cells were tracked
(8 min/frame) over the course of a 6 h co-incubation in nanowells. The longest
continuous period of contact (left) and cumulative time in contact (right) were scored
for single NK cells in wells with the indicated target cells. Only NK cells that had at
least one period of contact with a target cell were included in the analysis. Results
are representative of three donors.
3.4.3
CD16 is shed with a probability that depends on the
duration of contact with a target cell and the baseline
activation state of the NK cell
Given that a single engagement with a target cell could induce the shedding of CD16,
we next sought to understand the kinetics by which this process occurs. We used
video microscopy to track the interactions between IL-2-activated NK cells and target
cells on the array of nanowells during the 6 h incubation. This approach enabled us
to make correlated measurements of dynamic interaction parameters and end-point
expression of CD16 for thousands of single NK cells on each array. NK cells exhibited
a wide distribution of contact times (Figure 3-4). NK cells spent longer in contact
with antibody-coated P815 target cells than with K562 target cells, both in terms of
their longest continuous period of contact (Figure 3-4, left) and their cumulative time
in contact (Figure 3-4, right).
80
3.5
C D16 intensity (logicle)
B
P815+Ab
3.0
2.5
2.0
1.5
●
●●●●●●●
●
●● ●●● ● ● ●
●●
●●
●
●●
●
●●●
●
●●
●●
●●
●
●
●
●
●●● ●
●
●●
●●
●●
●
●● ●●●
●
●
●●
●
●●
●●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●●●
●
●●
●●
●
●
●
●● ●●●●● ●● ●
●
●
●●
●
●●
●
●●
●●●
●
●●
●
●●
●● ●●●
●
●
●
●●
●●
●●
●●●
●●
● ●
●●
●
●
●●●
●●
●
●
●
●
●
●●●
●●●
●● ● ●
●●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●●
●●●
●
●●●●
●●●
●
●●●
●
●
●
●●
●
●
●
●
● ● ●
●●
●●
●● ●
●●
●●●●
●●
●●
● ●●● ● ● ●
●●
●●
● ●
●●
●
●● ●●● ●
●
●●
●●●
●●
●● ● ● ● ●
●
●●
●
●●
●
●● ● ● ● ●●
●
●●●● ● ● ●
●
●●● ●●
●●●
●● ●
●
●
●
●●
●
●●
●
● ●● ●
●
●
●●●●●●
●●
●● ●● ●● ●●
●●
●●●
●●●●
●●●●
●●●
●●●
●●
●●
●● ●● ●● ●
●●
●
●●
●
●
●
●●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●
●
● ●●● ●
●
●●● ●●
● ●● ● ● ●
●
●● ●●
●
●●●●●
●●●
●●●● ●
●●
●
●
●
●●●●●●
●● ●●
● ●
● ● ●●●
●●●
●●
●●
●
●
●●
●●
●●
●●
●●
●●●
● ●●●●
●
●● ●● ●
●
●●
●●
●●
●
●● ●●
●●
●●
●●●
●
●●
●●●
●●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●● ●●●
●● ●●●
●●● ●
●
●
●●●
●●●
●●
●
●
●● ●
●●●
●●
●●
●●●
●
●●
●
●
●●
●●●
●
●
●●
●●
●
●
●
● ●
●●
●
●
●●
●●●
●
●●
●
●●
●
●
●●●●
●
●●
●●● ●●
●●● ●
●
●
●●
●●
●●●
●
●
●
●
●
●
●
●
●
●●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●●
●●●
●●
●
●
●● ●●●
●●
●●
●●
●●●
●●
●
●
●●
●●●
●●●
●●
●●●●●
●●
●●
●● ●
●●
●●
●
●
●
●●●
●
●
●
●●
●●
●
●
●●
●●●
●●
●●
●●●
●●
●●
●●
●●●
●●
●
●●
●●●
●●
●
●
●●
●●
●
●●
●●●
●●
●●
●●
●● ●
●●
●
●●
●●
●●
●●
●
●
●
●
●
●●
●
●●
●●
●●
●
●●
●●
●
●●
●●
●●
●●
●
●●●●
●●●●●●●
●●
●●●
●
●●●
●●
●●
●
●●
●●●
●
●●
●●
●●
●●
●●
●
●
● ●●
●
●
●●
●
●●
●●
●
●
●
●●
●●●●●●
●●
●
●●
●
●●
●●
●●
●●
●
●
●
●
●● ●
●● ●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
● ●●
●
●
●
●
●
●
●
●
●
●● ● ●
●
● ●
●
●●
●
●● ●●●●
0
100
200
CD16 intensity (logicle)
A
K562
3.5
3.0
2.5
2.0
1.5
300
●
●
●●
●
●●
●●
●●
●
●
●
●●
●●
●●
●
●● ●●●
●●● ● ● ●
●
●
●
●
●
●
●●
●
●
●●
●
●
●●
●
●
●
● ●●●●●● ●
●●
●
●
●
●●
●
●●
●
●●
●
●
●
●●
●
●●
●●
●
●●
● ● ● ●●●●●●●●● ●
●●●●
●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●●
●●
●●
●●
●●
●
●●●
●●●
●●
●
●
●
●
●
●●●
●
●
●●
●
●●
●
●
●● ●
● ● ●
●
●●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●●●
●●
●●●
●
●●
●
●
●
●
●
●●
●●●● ● ●● ●● ●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●● ●
●●
●
●●
●●
●●
●●
●● ●●
●●●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●● ● ●●
●●
●●●●●
● ●●
●
●
●●●
●
●●
●●
●
●●
●●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
● ●●● ● ● ● ●
●●
●
●●
●●
●
●
●●●●●
●
●
●●
●
●●
●●
●
●
●●
● ●●●
●
●●
●
●●
●●
●
●
●
●●
●● ● ●●
●●
●
●●
●●
●
● ● ●●
●●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ● ●● ●
● ●●●●●
●
●
●
●●
●●
●
●●●● ●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●●
●●●
●●
●● ●● ●
●●
●● ●●
●●
●●●
●●●● ●● ● ● ● ● ● ●
●
●●
●●
●
●
●
●●●●●● ● ●
● ● ●●● ● ● ●●
●
●●
●●
●●
●●●
●●
●●
●
● ● ● ● ●●●
●
●●●
●
●●
●●
● ●●●
●●●
●●
●● ●●
●
●●●
●
●●
●
● ●
●
●
● ●●●
●
●
●
●
●
●
●
●
●
●
●●● ●●
●
●
● ●●
● ●●● ●● ● ●
● ●● ●
●●●
●●●● ●●●
●
●● ●● ● ●● ● ●●●● ●●●
● ●●●
●●
●●●
●
●
●
●
●
●
●
●
●
●
● ●● ●● ●
● ●
●●
●
● ●●● ● ●
●●●●●●
●● ●●●● ●●●●●●● ●●
●● ●
● ●● ● ●●●
●
● ●●
●● ●● ● ●●
●●
●
●●●
●●●
●
● ● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●●
● ●●
●
●
●
● ●●
●● ●
●● ●
●●●●●
●●●●
●●
●●●
●●
● ● ●●● ●
●●
●●
●●
●●●● ●
●●●●
●●
●● ●●●
●● ●●
●
●●●
●
●●
●●●●●
●●●
●●
●● ● ●
●●
●●
●●
●
●●
●● ●
● ●●
●
●
●●
●
● ●●
● ●●●● ●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●
●● ●●●●●
●●
●●
●
●●
●●
●●●
● ●●●●●● ●●
●
● ●●
●
●●
●● ●●● ● ●● ●●●
●●
●●
●●
●●
●
●
●●
●●
●
● ●●
●●●
●● ● ●
● ●●●
●●●
● ●●
●●●
●●●
●●●●
●●●
●●
●●
●●
● ●
●
●●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●● ● ●● ●
● ●
●●●●
●●
●●●
●
●●●
●
●
●
●
● ●
●●
●
●
●●
● ●●
●●
●
●
●●●● ●● ●
●●
●
● ●● ● ●
●●
●● ●
● ● ●●●
●
●
●
0
Longest contact (min)
100
200
Point density
300
Longest contact (min)
Figure 3-5: Amount of CD16 remaining on the surface of NK cells after
varying durations of contact with target cells. Interactions between IL-2activated NK cells and target cells were tracked by video microscopy (8 min interval,
6 h duration) to measure the longest continuous period that each NK cell spent
in contact with either (A) antibody-coated P815 target cells or (B) K562 target
cells. Cells on the array were then stained for CD16 and imaged again. Each point
represents a single NK cell. Only NK cells that had at least one period of contact with
a target cell are shown (n = 2001 cells (A), 2596 cells (B)). Results are representative
of three donors.
The amount of CD16 remaining on the surface of NK cells after interactions with
antibody-coated P815 cells was inversely related to the amount of time that they
spent in contact with the target (Figure 3-5A). Cells that formed only short contacts
(<30 min) retained a high level of surface-expressed CD16, whereas the vast majority
of cells that engaged a target for a prolonged period of time (>120 min) displayed a
complete loss of CD16. NK cells interacting with K562 target cells were much less
likely to form prolonged contacts or to shed CD16 (Figure 3-5B).
To further investigate the correlation between contact time and the shedding of
CD16, we binned the NK cells by their longest continuous period of contact with a
target cell and then calculated the fraction of NK cells that retained expression of
CD16 (Figure 3-6). This analysis was performed for three donors using IL-2-activated
NK cells; for one of these donors, we also collected a matched dataset using resting
NK cells. When antibody-coated P815 cells were used as the targets, the fraction of
81
NK cells retaining expression of CD16 declined sharply as a function of the longest
duration of contact (Figure 3-6A). This decline was more rapid for IL-2-activated NK
cells than for resting NK cells. For example, of the NK cells with a longest period of
contact of 100 min, ∼50% of the resting NK cells expressed CD16 at the end-point
of the assay, whereas only ∼20% of the IL-2-activated NK cells did. When K562
cells were used as the targets, the fraction of NK cells retaining expression of CD16
declined gradually as a function of the longest duration of contact; again, the decline
was more rapid in IL-2-activated NK cells than in resting NK cells (Figure 3-6B).
When antibody-coated P815 cells were used as the targets, almost every NK cell
(both IL-2-activated and resting) that was in continuous contact with a target cell
for >300 min shed CD16. When K562 cells were used as the targets, however, the
fraction of NK cells that had shed CD16 did not reach a plateau within the duration
of the 6 h assay, suggesting that the likelihood of shedding CD16 would continue to
increase in NK cells that remained in contact with a K562 target for longer than 6
h. Similar trends were observed when cumulative contact time (rather than longest
continuous period of contact) was considered (Figure B-2). Together, these results
demonstrated that NK cells that spend a longer period of time in contact with a
target cell are more likely to shed CD16, and that for a given duration of contact,
shedding is more likely in NK cells that are IL-2-activated than in those that are
resting.
3.4.4
Bystander NK cells lose CD16 more frequently than
expected by chance
Monitoring single IL-2-activated NK cells in wells with antibody-coated P815 cells
demonstrated that NK cells shed CD16 with a probability that is well predicted by
their duration of contact with a target. We next sought to address whether CD16 is
also shed from “bystander” NK cells in the local vicinity. This task is experimentally
challenging, because in wells with multiple NK cells, the measurements of contact (by
video microscopy) and the measurements of CD16 can be correlated at the level of
82
$
%
1.00
&'
CD16+1.FHOOVIUDFWLRQ
NK cells (fraction)
&'
C D16+1.FHOOVIUDFWLRQ
N K cells (fraction)
3$E
0.75
0.50
0.25
0.00
0
100
200
.
1.00
0.75
IL−2
0.50
Resting
0.25
0.00
300
0
Longest contact (min)
100
200
300
Longest contact (min)
Figure 3-6: NK cells shed CD16 with a probability that depends on their
longest duration of contact with a target cell and their activation state.
The fraction of NK cells that retained expression of CD16 is plotted as a function of
the duration of their longest continuous period of contact with (A) antibody-coated
P815 target cells or (B) K562 target cells. NK cells were either IL-2-activated (orange,
combined data from three donors) or resting (green, data from one donor, who was
also included in the IL-2-activated data set). Trend lines were estimated by LOESS
regression; shading represents the 95% confidence interval.
each individual well but not at the level of each cell in the well.4 Furthermore, the
time-interval used in these experiments (8 min/frame) was not short enough to enable
the accurate tracking of multiple highly motile NK cells when their paths intersected.
We instead approached the question of CD16 shedding from bystander NK cells
by using a simple application of the binomial distribution. The binomial distribution
states that the probability of observing k successes in N independent trials, each of
which succeed with a probability of p, is:
�
�
N k
P (k) =
p (1 − p)N −k
k
(3.1)
Adapted to the question of CD16 shedding, each NK cell in a well can be treated as
a trial that succeeds if the NK cell is CD16+ at the end of the assay and fails if the
4
For example, consider a well with two NK cells, Cell A and Cell B. In the video microscopy
portion of the assay, we track the dynamic properties and Cell A and Cell B. We then stain for
CD16 and image the well again, but because the cells may have shifted between the last image of
the time-lapse sequence and this image, it is not possible to unambiguously tell which surface marker
phenotype corresponds to Cell A and which corresponds to Cell B.
83
NK cell is CD16– . We used the data from wells containing a single IL-2-activated NK
cell and 1–3 antibody-coated P815 target cells to calculate pCD16+ , the probability
of success for any individual NK cell. We then applied the binomial distribution
to calculate the predicted number of CD16+ NK cells (i.e., k successes) in wells
containing two or three NK cells (i.e., N = 2 or 3 trials) and 1–3 target cells.
The predicted frequencies of wells in which all NK cells remained CD16+ matched
the experimentally observed frequencies quite closely (Figure 3-7, gray dots). In these
wells, presumably none of the NK cells engaged a target long enough to induce the
shedding of CD16. In contrast, the observed frequencies of wells in which some or
all of the NK cells were CD16– deviated from the predictions. Wells in which all NK
cells were CD16– (Figure 3-7, red dots) were observed more frequently than expected,
whereas “mixed” wells containing some CD16+ NK cells and some CD16– NK cells
(Figure 3-7A, blue dots; Figure 3-7B, blue and green dots) were observed less frequently than expected. These deviations suggest that when one NK cell sheds CD16,
other nearby NK cells have an increased liklihood of shedding CD16 compared with
when they are in isolation. In contrast, killing did not occur in a cooperative manner
(Figure B-3), similar to our previous findings using K562 target cells. 99 Although the
frequencies of wells with mixed states of occupancy were lower than predicted, there
were still many wells that exhibited these mixed states (for example, 22% ± 6% of
the wells containing two NK cells had one CD16+ NK cell and one CD16– NK cell).
Together, these findings suggest that there is a mechanism by which nearby NK cells
shed CD16 in a coordinated manner, but that shedding of CD16 from neighboring
cells is not an all-or-none event, even when NK cells are confined in close proximity.
3.4.5
NK cells that perform effector functions shed CD16,
but many cells that shed CD16 do not perform common
effector functions
In addition to inducing the shedding of CD16, engagement with target cells can
stimulate NK cells to perform effector functions such as cytolysis and the secretion of
84
Experiment: 20140326, 20140328, 20140331 (IL-­2 stim)
A
B
Wells with 2 NK cells Wells with 3 NK cells + 1-­3 P815+Ab targets
CD16+ cells
Actual fraction
Actual fraction
0.8
+ 1-­3 P815+Ab targets
0 of 2
1 of 2
2 of 2
0.6
0.4
0.2
CD16+ cells
0.8
0 of 3
1 of 3
2 of 3
3 of 3
0.6
0.4
0.2
0.0
0.0 0.2 0.4 0.6 0.8
0.0
0.0 0.2 0.4 0.6 0.8
Predicted fraction
Predicted fraction
Figure 3-7: Bystander NK cells shed CD16 more often than expected by
chance. Wells containing (A) two or (B) three IL-2-activated NK cells and 1–3
antibody-coated P815 target cells were included in the analysis. The binomial distribution was used to predict the fractions of wells containing the indicated numbers of
CD16+ NK cells, as described in the text. Predicted fractions were then compared
with the experimentally observed fractions. Data from three donors are shown. In
(A), all three of the actual (observed) fractions differed significantly (at the α = 0.05
level) from the predicted fractions according to Pearson’s chi-squared test, corrected
for multiple comparisons with the Benjamini-Hochberg method of controlling the false
discovery rate (FDR). In (B), none of the comparisons met the significance threshold
after correcting for multiple comparisons. The total number of wells analyzed from
each donor/experiment was (A) 216, 401, and 403; (B) 19, 63, and 33.
85
cytokines. To characterize the relationship between target-induced shedding of CD16
and the functional activity of each cell, we made correlated measurements of cytolytic
activity (assessed during the video microscopy portion of the assay), the secretion of
IFN-γ and CCL4 (assessed by microengraving at the same time as performing video
microscopy), and the end-point expression of CD16. We found that the NK cells
that were CD16– at the conclusion of the co-incubation with target cells were almost
exclusively responsible for the performance of effector functions when either antibodycoated P815 (Figure 3-8A) or K562 (Figure 3-8B) cells were used as the targets. NK
cells that remained CD16+ were almost entirely devoid of functional activity when
antibody-coated P815 cells were used as the targets, suggesting that only NK cells
that shed5 CD16 were functionally active in this condition. In contrast, a small but
reproducible fraction of NK cells that remained CD16+ exhibited effector functions
when K562 cells were used as the targets, suggesting that for this stimulation, a
minority of NK cells could perform effector functions without the concurrent loss of
CD16.
Finally, we asked whether all NK cells that actively shed CD16 from their surface
in response to stimulation with a target cell also killed the target or secreted IFN-γ
or CCL4. This question cannot be answered directly from the observed data due
to the fact that ∼10% of the NK cells in wells with target cells will be CD16– at
baseline. Thus, from the end-point measurement of the expression of CD16, we
cannot definitively say whether any given CD16– cell in a well with target cells began
the assay CD16+ and then actively shed CD16, or whether it simply was CD16– to
begin with. However, for these wells, we can estimate the fraction of CD16– NK cells
that actively shed CD16 (call this fraction A) by using the observed baseline fraction
of CD16– NK cells on each array (B, from the wells containing an NK cell but no
target) and the observed fraction of single NK cells (in wells with targets) that were
5
Because CD16 mediates the recognition of antibody-coated P815 cells, we can infer in this case
that NK cells that were CD16– at the conclusion of the co-incubation, but that exhibited effector
functions, actively shed CD16 (as opposed to being from the ∼10% population of NK cells that were
CD16– at baseline).
86
Experiment: 20140326 (IL-­2 stim)
Cells with no contact not plotted
A
P815+Ab (22% CD16+ / 78% CD16-­)
P = .08
20
10
0
+
40
20
0
-
CD16
B
60
40
20
0
-
CD16
+
-
CD16
K562 (84% CD16+ / 16% CD16-­)
P = .1
IFN-+ (%)
4
3
2
1
0
+
-
CD16
*
50
40
30
20
10
0
+
*
60
CCL4 (%)
5
Lysis+ (%)
+
*
80
CCL4 (%)
30
*
60
IFN-+ (%)
Lysis+ (%)
40
-
CD16
40
20
0
+
-
CD16
Figure 3-8: Functionally active NK cells shed CD16. The cytolytic and secretory activity of single IL-2-activated NK cells in wells with 1–3 (A) antibody-coated
P815 or (B) K562 target cells were measured and segregated based on whether the
NK cell expressed CD16 at the conclusion of the 6 h co-incubation (mean percentages
for each group are indicated). Paired data from three donors are shown. The percentage of cells that were positive for each function were corrected for the background
frequencies, which were calculated from wells with 0 NK cells and 1–3 target cells (for
background lysis) or wells with an NK cell but no targets (for background secretion).
Note difference in scale between graphs. *P < .05, paired t-test.
87
CD16– at the conclusion of the assay (call this fraction C ):
A=1−
B
C
(3.2)
When we compared this estimated fraction of actively shedding CD16– NK cells to
the fraction of CD16– NK cells that exhibited any of the three effector functions
measured (cytolysis, secretion of IFN-γ or CCL4), we found that shedding occurred
more frequently than the performance of effector functions when antibody-coated
P815 cells were used as targets (Figure 3-9A), but that shedding and effector function
occurred at similar frequencies when K562 cells were used (Figure 3-9B). A significant
portion (28% ± 7%) of NK cells that shed CD16 in response to antibody-coated
target cells failed to demonstrate an effector function, whereas this portion was only
7% ± 4% when K562 cells were used as targets. Both types of target cell induced
significantly more NK cells to shed CD16 than to mediate cytolysis (Figure B-4).
Overall, these results demonstrated that the majority of IL-2-activated NK cells that
perform common effector functions upon encounter with a target cell also shed CD16,
but that cells that shed CD16 do not necessarily lyse the target or secrete IFN-γ or
CCL4 during the course of the 6 h assay.
3.5
Discussion
Dynamic changes in the expression of cellular surface receptors—and the functional
correlates and consequences of these changes—shape many immunological processes.
Receptor modulation is often induced by stimulatory signals that are received from
direct cell-cell interactions, but it is experimentally challenging to correlate the stimuli
that individual cells receive from discrete periods of contact with the cells’ resulting
changes in receptor expression and functional outcomes. The nanowell-based platform
presented here offers one approach to making these types of correlated measurements
on single cells. In this approach, large numbers (thousands) of individual cells are
dynamically tracked as they participate in cell-cell interactions, and the properties
88
B
P815+Ab
100
*
% of CD16- cells
% of CD16- cells
A
80
60
40
20
0
K562
100
P = .1
80
60
40
20
0
Lysis or
Active
shedding IFN- or CCL4
Lysis or
Active
shedding IFN- or CCL4
Figure 3-9: Not all NK cells that shed CD16 perform common effector functions. The frequency of IL-2-activated NK cells that actively shed CD16 (calculated
as described in the text) in response to engaging an (A) antibody-coated P815 or (B)
K562 target cell was compared with the frequency with which these cells mediated
any of the effector functions measured (cytolysis or the secretion of IFN-γ or CCL4).
Paired data from three donors are shown. *P < .05, paired t-test.
of these interactions—as well as their functional outcomes—are directly linked to the
surface marker phenotype of the cell. Furthermore, the receptor of interest (e.g.,
CD16) is not perturbed by antibody staining until after the functional analysis.
We investigated the downmodulation of CD16 from the surface of NK cells in
response to stimulation from contact with target cells. The patterns of downmodulation that we observed in the nanowell assays were consistent with those observed
after bulk incubation with target cells and analysis by flow cytometry. In both cases,
stimulation with antibody-coated P815 target cells resulted in a more complete loss of
CD16 than did stimulation with K562 target cells, which is in agreement with previous reports 156 and with the observation that signaling through CD16 (triggered here
by antibody-coated P815 cells) provides stronger activation than signaling through
other receptors such as NKG2D 103,104 (triggered here by K562 cells).
After bulk incubation with K562 target cells, CD16 was downregulated in a bimodal fashion (the majority of NK cells either retained high levels of CD16 or completely shed CD16), whereas NKG2D was downregulated in a unimodal fashion (the
89
entire distribution shifted to the left). Despite providing strong activating signals,
stimulation with antibody-coated P815 target cells did not alter the expression of
NKG2D on the surface of NK cells. Differences in the target-induced downregulation
of CD16 and NKG2D may reflect the different mechanisms by which each receptor is
removed from the cell surface. Although both signaling from and downregulation of
CD16 and NKG2D share nodes such as phosphatidylinositol 3-kinase (PI3K), 106,160,161
the loss of CD16 occurs mainly by proteolytic shedding and can be induced by a wide
range of stimuli, 152,153 whereas the loss of NKG2D is thought to involve clathrindependent endocytosis and is induced by ligand binding to NKG2D. 161
Both the bulk and single-cell assays demonstrated that the speed and magnitude
of CD16 shedding from the surface of NK cells is affected by the activation state of
the NK cell. In the bulk assays, CD16 shedding was more rapid from IL-2-activated
NK cells than from resting NK cells. Similarly, shorter periods of contact with target
cells were required to induce CD16 shedding in IL-2-activated NK cells than in resting
NK cells. The observation that activation with IL-2 increases the speed (or lowers
the threshold) of CD16 shedding in response to contact with target cells is consistent
with previous reports that IL-2 alone (if used at higher doses and/or longer treatment
periods than were used here) can induce CD16 shedding. 153,154 The influence of IL-2activation on CD16 shedding and the performance of effector functions is important
to consider in the context of cancer therapies that combine rituximab (a monoclonal
antibody that induces ADCC) with IL-2. 162
By monitoring interactions between individual NK cells and target cells, we demonstrated that prolonged contact with a single antibody-coated P815 target cell induced
the shedding of CD16 in the majority of NK cells. Many of these cells also performed
effector functions such as cytolysis of the target or secretion of IFN-γ or CCL4. These
results are consistent with previous reports showing that target-induced downregulation of CD16 is associated with the production of IFN-γ and degranulation. 153,157
Many other actively shedding NK cells, however, did not perform any of the three
effector functions that we measured (at least during the 6 h assay). Previous studies have also shown that the fraction of NK cells that shed CD16 is higher than
90
the fraction that degranulate or produce IFN-γ. 153,157 The single-cell measurements
presented here extend these previous findings (which used bulk incubations of NK
cells and targets) to demonstrate that while common effector functions arising from
discrete encounters with individual target cells are concentrated within the subset of
NK cells that actively shed CD16, not all actively shedding NK cells perform common
effector functions.
Single-cell measurements (e.g., by flow cytometry) that are performed after bulk
co-incubation of NK cells and target cells enable the phenotype and function of each
individual cell to be resolved, but these types of measurement do not enable the
properties of neighboring cells to be linked in a way that reflects potential intercellular interactions. In contrast, the nanowell-based platform presented here enables
measurements that are made on both a “per well” and a “per cell” basis. Thus, the
expression of CD16 on each NK cell can be compared with that of neighboring NK
cells residing within the same nanowell. By measuring the distribution of CD16 states
in wells containing multiple NK cells (and at least one antibody-coated P815 target
cell), we found that the shedding of CD16 occurred in a more coordinated fashion
than would be expected by chance. In other words, wells containing mixed populations of CD16+ and CD16– NK cells were observed less frequently than expected,
while wells containing all CD16– NK cells were observed more frequently than expected. We did not observe cooperative killing behavior in these cases (similar to
our previous findings 99 ), which suggests that although “bystander” NK cells have an
increased likelihood of shedding their CD16, they do not have significantly increased
cytolytic activity.
Several possible mechanisms could contribute to the coordinated loss of CD16
from the surface of neighboring NK cells. Activated NK cells might secrete proteases
that act not only on the activated cell, but also on other cells in the local vicinity. The two proteases that have so far been associated with the cleavage of CD16
(ADAM17 and MMP25) are membrane proteins 163,164 and thus should not operate by
this mechanism, but it is possible that secreted proteases also contribute to the shedding of CD16. Alternatively, proteases on the surface of an activated NK cell might
91
cleave CD16 from the surface of another cell in trans if the two cells engage in close
physical contact. Shedding in trans has been investigated by mixing NK cells with
T cells that were transduced with CD16 and Fc�RIγ (CD16/γ). 157 In this setting,
shedding occurred predominantly from NK cells when the cell mixture was stimulated
with plate-bound antibodies against NK cell activating receptors but from CD16/γtransduced T cells when the cell mixture was stimulated with anti-CD3. Therefore,
in this case, shedding was confined to the activated cell type (i.e., shedding in cis but
not trans). However, there are two important differences between this experiment
and the ones presented here: first, the stimulations only induced mild downregulation of CD16 (as opposed to the nearly complete downregulation that we observed
when using antibody-coated P815 target cells as the stimuli); second, shedding in
trans mediated by homotypic interactions6 between NK cells (as opposed to NK-T
cell interactions) would not have been detected by the assay. Therefore, it is possible
that shedding in trans could occur in other settings. We frequently observed close
homotypic interactions between NK cells in the same nanowell, suggesting that activated surface proteases on one cell might be able to cleave CD16 from the surface of
another cell.
Coordinated loss of CD16 from the surface of neighboring NK cells could also occur
by a variety of indirect mechanisms. It is possible that activated (shedding) NK cells
secrete cytokines or participate in contact-mediated interactions with other NK cells
that increase the likelihood that these other cells will also shed (e.g., by lowering the
threshold of additional activation that is required for shedding or by bringing other
NK cells near the target(s) 98 ). We did not observe cooperative cytolytic behavior
in wells with multiple NK cells, however, which suggests that NK-to-NK influences
on shedding and on effector functions are at least somewhat disconnected. From
the current set of data we cannot definitively say which—if any—of the possible
mechanisms discussed above account for the coordinated behavior of CD16 shedding,
but future experiments could be performed to address each of these possibilities.
6
Homotypic interactions between NK cells can be mediated by a variety of surface molecules and
are thought to play a role in proliferation and the acquisition of effector functions. 165,166
92
Although we focused here on the shedding of CD16, other surface receptors that
play important roles in the functional properties of NK cells are known to be modulated by proteolytic shedding of their ectodomain. One such receptor is CD62L
(L-selectin), which, like CD16, is cleaved and shed by ADAM17. 153,157,167 Given its
function as a homing receptor targeting secondary lymphoid tissues, it is interesting
to speculate how activation-induced downregulation of CD62L might alter the tissue
distribution of activated NK cells in vivo. CD62L has been identified as a marker
of polyfunctional NK cells that respond strongly to stimulation with both cytokines
and target cells, and it has been suggested that CD56dim CD62L+ NK cells are at
an intermediate state of maturation between the CD56bright and the CD56dim CD62L–
subsets. 123 By enabling the correlation of functional activity, interaction dynamics,
and ectodomain shedding, the platform presented here offers a new approach to investigating the unique functional properties of the CD56dim CD62L+ subset.
The shedding of CD62L is also of interest as a marker of antigen-specific T cell activation through the T cell receptor (TCR). CD62L is highly expressed on naı̈ve T cells
but is shed rapidly upon stimulation through the TCR. 160 In this context, the speed
and extent of CD62L shedding has been shown to reflect the strength and duration of
TCR signaling both in vitro and in vivo. 84 These shedding trends are similar to our
observation that the speed and extent of CD16 shedding from NK cells is correlated
with the strength (i.e., strong stimulation presented by antibody-coated P815 cells or
weaker stimulation presented by K562 target cells) and duration of stimulation with
target cells, as well as by the activation state (resting or IL-2-activated) of the NK
cell. We note that the platform presented here could be extended to monitor CD62L
shedding from naı̈ve T cells after encounters with antigen-presenting cells (APCs).
Using this approach, antigen-specific activation of T cells could be measured in a way
that correlates the rapid shedding of CD62L with subsequent functional responses
such as the secretion of cytokines or proliferation.
In summary, we have developed a nanowell-based platform that enables measurement of the kinetics by which and extent to which CD16 is shed from NK cells following discrete encounters with individual target cells. In addition, these measurements
93
enable correlated analysis of shedding and effector functions and analysis of potential
cooperative shedding behaviors amongst small groups of cells. This platform can be
extended to investigate the dynamic, functional, and cooperative properties of the
shedding (or modulation by other mechanisms) of other surface molecules of interest.
94
Chapter 4
Single-Cell Analysis of the
Dynamic and Functional
Properties of Licensed and
Unlicensed Natural Killer Cells
95
4.1
Abstract
It is thought that natural killer (NK) cells are conferred with enhanced functional
abilities when they receive signals through inhibitory receptors (IRs) that recognize
cognate HLA Class I molecules on host cells. NK cells expressing IRs that recognize
host HLA Class I molecules are considered to be functionally “licensed”, whereas NK
cells that lack these inhibitory signals are considered to be functionally “unlicensed”.
The mechanisms underlying licensing are not well understood, in part because of the
large amount of cell-to-cell heterogeneity among NK cells. Furthermore, licensing is
thought to affect the early stages of interactions between NK cells and target cells
(e.g., conjugation), but it is experimentally challenging to integrate measurements of
these early stages of interaction with downstream functional properties and the IR
phenotypes of individual NK cells. To address these challenges and characterize the
licensing process at the single-cell level, we used nanowell-based assays to measure the
dynamic and functional properties of individual IR+ (licensed) and IR– (unlicensed)
NK cells as they interacted with target cells. Both licensed and unlicensed NK cells
secreted cytokines with a frequency that increased with the longest duration of contact
with a target cell, but unlicensed NK cells were less likely than licensed NK cells to
secrete cytokines, even after prolonged contact with a target. These results suggest
that there are underlying differences in the characteristics of licensed and unlicensed
NK cells that are not completely accounted for by differences in their ability to form
an initial engagement with a target cell.
4.2
Introduction
Natural killer (NK) cells express a diverse range of inhibitory receptors (IRs) that recognize human leukocyte antigen (HLA) molecules on healthy cells. These inhibitory
signals play a critical functional role by preventing NK cells from attacking healthy
cells. Two major HLA-specific IRs on NK cells are NKG2A-CD94 heterodimers (hereafter referred to as NKG2A) and members of the killer immunoglobulin-like receptors
96
(KIR) class of receptors. 168 NKG2A is expressed on CD56bright NK cells early in development and is gradually lost as the NK cells transition to the CD56dim subset. 124
Conversely, KIRs are not highly expressed on CD56bright NK cells but expression is
gradually acquired as the NK cells mature. The complementary patterns of NKG2A
and KIR expression ensure that the majority of NK cells express some form of IR at
each of their developmental stages. 169,170
NKG2A is a C-type lectin receptor that recognizes HLA-E, a non-classical (class
Ib) major histocompatibility (MHC) molecule that presents peptides derived from
the signal sequences of MHC Class I molecules 171,172 and delivers inhibitory signals
through immunoreceptor tyrosine-based inhibitory motifs (ITIMs) on its cytoplasmic domain. 173 The KIR class of receptors contains >12 highly homologous members
that are characterized by two or three (2D or 3D, respectively) extracellular Ig-like
domains and a long or short (L or S, respectively) cytoplasmic tail. 174 The long cytoplasmic tails contain one or more ITIMs, making them inhibitory receptors (although
KIR2DL4 is an exception). The short cytoplasmic tails associate with immunoreceptor tyrosine-based activation motif (ITAM)-containing proteins, making them activating receptors. In general, the inhibitory KIR recognize MHC Class I molecules.
Activating KIR bind weakly to the HLA ligand of their inhibitory counterpart, but
specific high-affinity ligands of the activating KIRs are for the most part unknown.
In the human population, over 20 KIR haplotypes containing different combinations of the KIR genes have been described. 174,175 KIR2DL4, KIR3DL2, and KIR3DL3
are found in almost every haplotype and are known as “framework” genes, while
other KIR genes are present in some individuals but absent in others. In addition
to variations in haplotype across individuals, there is a significant amount of cell-tocell variability in the expression of KIR within any one individual. The subset of
KIR genes that are expressed in each cell appears to be stochastically controlled by
methylation of the KIR promoters. 176,177 Once the promoter methylation has been
established during NK cell development, the pattern of expression remains stable
throughout multiple cell divisions. Analysis of 100 NK cell clones derived from two
individuals showed that each cell expressed RNA for 2 to 9 KIR and CD94:NKG2 re97
ceptors per cell. 178 This variability means that in a person whose haplotype contains
9 KIR genes, one NK cell might express a subset of 3 KIR genes while another NK
cell might express a different subset of 7 KIR genes.
Different inhibitory KIR preferentially bind to different HLA Class I molecules,
and each of these binding interactions has a different affinity and associated degree
of inhibitory signaling. 174 For example, KIR2DL1 binds group 2 HLA-C molecules
(HLA-C2), KIR2DL2 and KIR2DL3 bind group 1 HLA-C molecules (HLA-C1), and
KIR3DL1 binds HLA-B molecules of the Bw4 serotype (HLA-Bw4). Because HLA
and KIR loci are on separate chromosomes (6 and 19, respectively) and segregate
independently, it commonly happens that a subset of NK cells in individuals will
express KIRs that do not recognize any of the individual’s HLA molecules. For
example, a KIR3DL1+ NK cell in an HLA-Bw4/Bw4 or an HLA-Bw4/Bw6 individual
would be able to receive inhibitory signals from the cognate HLA molecules, but the
same NK cell in an HLA-Bw6/Bw6 individual would not.
NK cells that do not express any IRs and those that express IRs that do not
recognize HLA Class I molecules expressed by the individual are in theory at risk for
damaging healthy cells, because these NK cells are unable to receive the inhibitory
signals that the healthy self cells present. It is thought, however, that NK cells guard
against this undesirable autoreactivity and maintain tolerance in an IR-dependent
process known as “licensing”. 179 One model of NK cell licensing posits that during
development, NK cells must express an IR that recognizes self-MHC in order to become licensed to kill and secrete cytokines in response to stimulation with target cells.
NK cells that do not express such an IR are anergic (unlicensed) unless exposed to
an inflammatory environment. 180 The degree of licensing (as measured by functional
activity in response to stimulation with HLA-deficient target cells) that each NK
cell receives depends on which IRs the NK cell expresses and whether the individual
expresses the cognate HLA molecules for those IRs. For example, licensing through
NKG2A (binding to HLA-E, which is expressed by all individuals) is weaker than
licensing through KIR2DL1, KIR2DL3, or KIR3DL1 when the individual expresses
the cognate HLA molecules for these KIR (HLA-C2, HLA-C1, and HLA-Bw4, re98
spectively), and licensing is stronger still in NK cells that express multiple inhibitory
KIR (e.g., KIR2DL1+ KIR3DL1+ double-positive NK cells in an HLA-C2, HLA-Bw4
individual). 181
The effects of licensing can be observed both by comparing IR+ and IR– NK cells
within the same individual, as well as by comparing IR+ NK cells from individuals
with HLA backgrounds that either do or do not contain cognate HLA ligands. Experimental protocols employing pan-IR staining or separation of NK cells (in which
NK cells expressing NKG2A, KIR2DL1/2/3, and/or KIR3DL1 are distinguished from
those that do not express any of these IRs) have demonstrated that IR+ NK cells are
more functionally active against MHC-deficient target cells (K562 or antibody-coated
P815 cells) than IR– NK cells are, as assessed by degranulation (CD107a+ ) and the
production of interferon-γ (IFN-γ) . 182,183 Similar functional differences have been
demonstrated by comparing NK cells that are IR– with NK cells that express a single
KIR (with a matched HLA molecule expressed by the individual). For example, stimulation with K562 target cells induces degranulation in a larger fraction of KIR3DL1+
NK cells than in KIR3DL1– NK cells (gated on CD56dim NKG2A– KIR2D– )1 from
HLA-Bw4 donors. 123 Furthermore, KIR+ NK cells from individuals that express cognate HLA ligands for that KIR are more functionally active against MHC-deficient
target cells than the same subset of KIR+ NK cells from individuals that express
HLA ligands that the KIR does not recognize. For example, KIR2DL1+ NK cells
are more responsive when isolated from HLA-C2 individuals than from HLA-C1/C1
individuals, 181,182 KIR2DL2+ and KIR2DL3+ NK cells are more responsive when
isolated from HLA-C1 individuals than from HLA-C2/C2 individuals, 169,181,182 and
KIR3DL1+ NK cells are more responsive when isolated from HLA-Bw4 individuals
than from HLA-Bw6/Bw6 individuals. 181,184–187 At least in some cases, homozygosity
for a cognate HLA ligand can increase the licensing effect, as it has been shown that
KIR3DL1+ NK cells from HLA-Bw4/Bw4 individuals are more responsive than those
from HLA-Bw4/Bw6 individuals, although both are more responsive than KIR3DL1+
NK cells from HLA-Bw6/Bw6 individuals. 186
1
Here, KIR2D encompasses KIR2DL1/2DS1/2DL2/2DS2/2DL3.
99
The expression of certain KIR-HLA combinations has been shown to influence the
outcome of several diseases, including human immunodeficiency virus (HIV). 188 HIVinfected individuals expressing KIR3DL1 along with HLA-Bw4-80I have a delayed
progression to AIDS, and the effect varies with KIR3DL1 allele. 189 These findings
suggest that licensing interactions can contribute to the control of viral infections.
Although the mechanistic basis of licensing is not yet fully characterized, several
lines of evidence suggest that functional differences between licensed and unlicensed
NK cells arise in part due to differences in their ability to receive activating signals
from target cells. A study using a humanized transgenic mouse model in which NK
cells are licensed through the expression of KIR2DL3 and its interaction with HLA-C1
(specifically, HLA-Cw3) demonstrated that in licensed NK cells, activating receptors
are clustered into nanodomains that are favorable for signaling, whereas in unlicensed
NK cells, activating receptors are confined alongside inhibitory receptors in an actin
meshwork. 190 It has also been shown that unlicensed murine NK cells (isolated from
β2m–/– mice, which are deficient in MHC Class I and thus do not support licensing interactions) and unlicensed (IR– ) human NK cells form fewer stable conjugates
with target cells than their licensed counterparts. 183 This deficiency in forming stable
conjugates contributes to the decreased functional activity of unlicensed cells against
target cells and arises from impaired “inside-out” signaling from activating receptors
to the β2 integrin leukocyte function-associated antigen 1 (LFA-1). Inside-out signaling is responsible for switching LFA-1 from a low-affinity state to a high-affinity
state, and thus controls the ability of an NK cell to adhere to a target cell via interactions between LFA-1 (on the NK cell) and its ligand intercellular adhesion molecule
1 (ICAM-1; on the target cell). Together, these studies illustrate that NK cells that
are able to receive inhibitory signals from cognate HLA molecules are better able to
interact with, interrogate, and receive activating signals from target cells.
Here, we adapt and apply the approaches introduced in Chapter 3 to directly monitor how individual licensed (IR+ ) and unlicensed (IR– ) NK cells interact with target
cells. We correlate the properties of these interactions with the resulting functional
outcomes to demonstrate that licensed and unlicensed NK cells that share certain
100
dynamic properties of interaction (specifically, duration of contact with a target cell)
still differ in their ability to secrete cytokines. We also begin to explore how transcriptional profiling with RNA-Seq can be applied to characterize these differences in
more detail.
4.3
4.3.1
Materials & Methods
Nanowell-based assays
Cells and stimulations
Peripheral blood mononuclear cells (PBMCs), P815 target cells, and K562 target
cells were prepared as described in Section 3.3.1.
Nanowell-based assays to measure the expression of IRs and functional
activity of individual NK cells
Nanowell-based assays were performed as described in Section 3.3.3, but on-chip
staining was performed for a panel of IRs instead of (or, in some experiments, in
addition to) on-chip staining for CD16. The staining solution was prepared in a total
of 200 µL media by adding 1 test volume each of anti-NKG2A-PE, anti-KIR2DL1-PE,
anti-KIR2DL2/L3/S2-PE, and anti-KIR3DL1-PE. We did not stain for KIR3DL2,
as interactions between this receptor and its cognate HLA ligands (HLA-A3 and
HLA-A11) have not been shown to mediate licensing (KIR3DL2 single-positive NK
cells are hyporesponsive in HLA-A3 and HLA-A11 individuals). 169 Antibodies are
described in further detail in the Supplementary Methods (Appendix C.1.1). Arrays
were incubated with the staining solution for 1 h to overnight at 4°C.
For Donor 1 and Donor 2, the complete panel of IRs was used because the KIR
and HLA types of these donors were unknown. For Donor 3, anti-KIR2DL2/L3/S2PE was omitted from the staining panel because this donor was KIR and HLA typed
and known to be homozygous for HLA-C2 (therefore KIR2DL2/L3+ NK cells should
not receive strong inhibitory/licensing signals due to the absence of HLA-C1).
101
Flow cytometry to measure the expression of IRs (for comparison with
imaging cytometry)
PBMCs (0.5 × 106 ) were stained with both (1) the panel of antibodies that was
used to sort the CD56dim NK cells for use in the nanowell-based assays and (2) the
panel of antibodies against IRs that was used in the on-chip imaging cytometry. Antibodies are described in further detail in the Supplementary Methods (Appendix
C.1.1). Staining was performed in media at 4°C for 30 min. Cells were then washed
in media, resuspended in phosphate-buffered saline (PBS) with 5% fetal bovine serum
(FBS), filtered, and acquired by flow cytometry (BD FACSAria III, BD Biosciences).
CD56dim NK cells were gated as CD3– CD14– CD19– CD56dim lymphocytes. Compensation was performed using single-stained beads (Quantum Simply Cellular anti-mouse
IgG beads; Bangs Laboratories).
Data analysis
Data were analyzed as described in Section 3.3.4.
4.3.2
Population-level RNA-Seq
Cells
Cryopreserved PBMCs (Ragon Institute HIV-Negative Cohort) were thawed, washed
with media, and then incubated (37°C, 5% CO2 ) for 1.5 h. P815 target cells were
split the day before the assay. Immediately before the assay, they were coated with
rabbit anti-mouse lymphocyte antibody (1 µL per 100,000 cells; Accurate Chemical
& Scientific Corporation) for 30 min and then washed with media.
Prior to labeling and flow sorting, PBMCs (2 × 106 cells/mL) were either incu-
bated with antibody-coated P815 target cells (0.2 × 106 cells/mL) or with media
alone for 4 h.
Surface marker staining and flow sorting
After the 4 h incubation, cells were stained with antibodies against lineage mark102
ers and IRs (anti-CD3-PECy7, anti-CD14-PECy7, anti-CD19-PECy7, anti-CD56PerCPCy5.5, anti-NKG2A-APC, anti-KIR2DL1-PE, anti-KIR2DL2/L3/S2-PE, and
anti-KIR3DL1-BV421). Antibodies are described in further detail in the Supplementary Methods (Appendix C.1.1).
for 15 min.
Staining was performed in media at 37°C
Viability was assessed using the LIVE/DEAD Fixable Aqua Stain
(Life Technologies).
Cells were washed in media, resuspended in PBS with 5%
FBS, and filtered. KIR3DL1+ and KIR3DL1– cells were sorted from the population of dead– CD3– CD14– CD19– CD56dim NKG2A– KIR2DL1– KIR2DL2/L3/S2– lymphocytes (BD FACSAria III, BD Biosciences). Compensation was performed using
single-stained beads.
Cells were sorted into 1.5 mL DNA LoBind tubes (Eppendorf) containing 102 µL
lysis buffer (100 µL Lysis Buffer RA1, 2 µL tris(2-carboxyethyl)phosphine (TCEP);
Macherey-Nagel NucleoSpin RNA XS RNA isolation kit). After sorting, the tubes
were vortexed, centrifuged (300 g, 30 s), and stored at -80°C.
RNA isolation
The cell lysates were thawed on ice and RNA was isolated using the NucleoSpin
RNA XS RNA isolation kit (Macherey-Nagel) following the manufacturer’s instructions.2 RNA was eluted in 2 × 5 µL RNase-free water and stored at -80°C.
cDNA synthesis and quantification
cDNA was synthesized using the SMARTer Ultra Low RNA Kit for Illumina Sequencing (Clontech) following the manufacturer’s instructions3 (1 µL RNA input, 17
cycles long-distance (LD) PCR using the Advantage 2 PCR Kit). cDNA was eluted
in 12 µL purification buffer and stored at -20°C. cDNA was quantified using the
PicoGreen dsDNA Assay Kit (Life Technologies) and size distributions were verified
for a subset of the samples (Fragment Analyzer; Advanced Analytical).
2
Carrier RNA was not used.
For first-strand cDNA synthesis, 1 µL input RNA was added to a master mix of dilution buffer,
RNase inhibitor, and 3’ SMART CDS Primer II A. This protocol differed slightly from the manufacturer’s protocol, which called for first mixing the dilution buffer and RNase inhibitor, then adding
the input RNA, and then adding the 3’ SMART CDS Primer II A.
3
103
Library preparation and sequencing
Paired-end sequencing libraries were prepared from <1 ng cDNA using the Nextera
XT DNA Sample Preparation Kit (Illumina) with the following minor modifications
to the manufacturer’s protocol: tagmentation was performed for 10 min at 55°C
(instead of 5 min), a 60 s extension was used in the limited-cycle PCR amplification
(instead of 30 s), PCR products were cleaned using 0.9× Agencourt AMPure XP
beads (Beckman Coulter), and PCR products were eluted in 30 µL of resuspension
buffer (instead of 52.5 µL). The bead-based library normalization step was omitted.
Libraries were dual indexed using Index 1 (i7) indices N701 – N710 and Index 2 (i5)
indices S503 – S506.
Libraries were quantified by qPCR and fragment sizes were verified (Fragment
Analyzer; Advanced Analytical) prior to pooling and sequencing (2 × 50 bp paired-
end; HiSeq 2000; Illumina). More than 7 × 106 reads were obtained for each sample.
Data analysis
Sequencing data were pre-processed using the FASTX Toolkit (Version 0.0.13).
Bases with a quality score of less than 28 were trimmed from the end of each sequence (command line usage: fastq quality trimmer -t 28 -l 0). Transcript abundances
were calculated using RSEM (RNA-Seq by expectation maximization; Version 1.2.6;
command line usage: rsem-calculate-expression - -paired-end -p 4 - -output-genomebam - -calc-ci - -phred64-quals) 191 with a reference transcriptome prepared from the
UCSC Known Genes database. 192,193 Differential expression was evaluating using the
R/Bioconductor package DESeq. 194,195
The GFOLD (generalized fold change) algorithm 196 (Version 1.0.9) was used to
rank differentially expressed genes when individual samples were compared. This
ranking was used to compare the expression of KIR3DL1 in the cell populations that
were sorted as KIR3DL1+ and KIR3DL1– from each donor.
104
4.4
4.4.1
Results & Discussion
An integrated platform to correlate the expression of
inhibitory receptors (IRs) with functional activity and
dynamic properties at the single-cell level
We extended the platform introduced in Chapter 2 to enable integrated measurements of functional activity, dynamic properties, and the expression of IRs (NKG2A
and KIRs) for individual NK cells.4 In a typical experiment, NK cells and target cells
are co-loaded onto an array of nanowells, and the cytolytic, dynamic, and secretory
activity is measured by simultaneously performing time-lapse microscopy and microengraving. At the conclusion of the assay, cells are stained on the array for surface
markers of interest (in this case, IRs) and acquired by imaging cytometry (Figure
4-1). This order of operations allows the functional properties of cells to be measured
prior to perturbing them by staining for the receptor of interest.
Several important improvements and modifications were made to the original platform. First, the number of wells (and hence the upper limit on the number of cellular
interaction events of interest) imaged during the time-lapse microscopy portion of the
assay was increased from <9,000 to >35,000 while still maintaining an 8 min imaging interval. This increase was enabled by (1) upgrading the microscope with faster
acquisition capabilities and (2) altering the design of the array and increasing the
field of view5 of the microscope to enable more wells to be acquired per image (121
compared with 49). Second, an image processing script based on ImageJ was developed to increase the accuracy and speed of segmenting images of cells and extracting
their fluorescence intensities in each channel. Third, microengraving was performed
at the same time as the time-lapse microscopy (rather than immediately after, as was
done previously), enabling direct assessment of the secretion that occurred during
4
The approaches described in this chapter were also applied to investigate the shedding of CD16
(Chapter 3).
5
The large field of view format was achieved by using a demagnifying lens, which reduced the
magnification of each image from 10× to 6.3×. NK cells and target cells were still readily identified,
even with this decrease in spatial resolution.
105
Flow cytometry
Imaging cytometry
Pan-­IR PE
Merge
Pan-­IR -­ PE
Pan-­IR -­ PE
eFluor670 (NK cells)
CD56 -­ APC
eFluor670 (NK cells)
Figure 4-1: Expression of inhibitory receptors on NK cells can be quantified
by on-chip staining and imaging cytometry. The distribution of IR expression
on CD56dim NK cells was measured by flow cytometry (left) and by imaging cytometry
(right). Flow measurements were performed by staining an aliquot of PBMCs that
came from the same population that was sorted (without labeling IRs) and used in
the single-cell functional assays and on-chip imaging cytometry shown on the right.
Imaging cytometry was performed at the conclusion of the functional assay by staining
on-chip for IRs. Prior to the functional assay, NK cells were stained with eFluor670
to enable unambiguous identification.
the period of observation. Fourth, a protocol was developed (described in Section
3.3.3) to increase the fraction of cells that were retained in the wells between the
combined time-lapse imaging/microengraving and the analysis of the expression of
IRs by on-chip staining and imaging cytometry. Using this protocol, 79% ± 10% of
single NK cells were retained (Figure C-1), thus enabling the assignment of a surface
marker phenotype to the majority of cells for which dynamic, cytolytic, and secretory
parameters were measured.
Altogether, these improvements made it possible to collect integrated measurements of the dynamic, cytolytic, and secretory properties of thousands of individual
NK cells on each array as they interacted with target cells. Previously (Chapter 2),
these types of measurements were limited to hundreds of cells per array. This orderof-magnitude increase in the number of cells that could be analyzed, combined with
the ability to measure the surface marker phenotypes of cells without perturbing them
prior to the functional assays, enabled the investigation of the dynamic and functional
properties of IR+ and IR– NK cells. IR– cells comprise a minority of the total NK cell
106
population (13% ± 6%); 182 thus, the ability to measure thousands—as opposed to
hundreds—of cells is crucial. After filtering for quality metrics associated with each
measured parameter, on each array we collected complete datasets for an average of
2140 (range 979 to 4815, across 12 experiments) single IR+ CD56dim NK cells interacting with 1–3 target cells, and 225 (range 89 to 583) single IR– CD56dim NK cells.
Thus, the workflow presented here enabled parallel, integrated measurements of large
numbers of single IR+ and IR– NK cells.
4.4.2
NK cells that express IRs are more functionally active
against target cells in nanowells than those that do not
Flow cytometry studies have shown that IR+ NK cells are more likely than IR– NK
cells to produce IFN-γ and degranulate when co-incubated in bulk with K562 target
cells or antibody-coated P815 target cells. 182,183 We therefore investigated whether
functional differences between the two subsets of NK cells could be detected using
our integrated platform, where individual NK cells are confined in nanowells with a
limited number of target cells.
Cytolytic activity was measured by the acquisition of SYTOX (a membraneimpermeable nucleic acid stain) by target cells during the course of the 6 h coincubation, as originally described in Chapter 2. Concurrent secretory activity was
measured by microengraving during the entire period of co-incubation and imaging.6
Using this approach, wells that contained NK cells secreting IFN-γ or CCL4 in response to interactions with target cells could be clearly distinguished from empty
wells on the same array (Figure 4-2).
When the IR phenotype of each NK cell (measured by imaging cytometry after the
functional assays, as in Figure 4-1) was linked to its functional properties, we found
that IR+ NK cells were significantly more likely than IR– NK cells to secrete IFN-γ
or CCL4 upon contact with antibody-coated P815 target cells in nanowells (Figure 46
In this method of microengraving, the glass slide coated with capture antibodies is clamped to
the array in a metal hybridization chamber, as was done in Chapter 2, but the sealed sandwich is
then removed from the chamber and placed on the microscope stage for imaging. Secretion and
imaging data can thus be collected concurrently rather than in a staggered manner.
107
IFNg
CCL4
1
1
Empty
1 NK + 1−3 Targets
0.8
Frequency
Frequency
0.8
0.6
0.4
0.2
0
Empty
1 NK + 1−3 Targets
0.6
0.4
0.2
0
5
10
15
log(MFI)
0
0
5
10
log(MFI)
Figure 4-2: Detection of secreted cytokines by microengraving concurrently
with time-lapse microscopy. Microengraving was performed for 6 h on a sealed
but unclamped array while imaging the array. Histograms of the resulting intensities
of IFN-γ (left) and CCL4 (right) are shown for wells that were empty (blue) or that
contained a single NK cell and 1–3 antibody-coated P815 target cells. The threshold
intensity for positive events (dashed black line) was defined as the median intensity
in empty wells plus two standard deviations. Results from Donor 1 are shown and are
representative of the other donors. MFI; median fluorescence intensity after correction
for local background intensity.
108
15
IFNg
CCL4
1
1
Empty
1 IRpos NK + 1−3 Targets
1 IRneg NK + 1−3 Targets
0.8
Frequency
Frequency
0.8
0.6
0.4
0.2
0
Empty
1 IRpos NK + 1−3 Targets
1 IRneg NK + 1−3 Targets
0.6
0.4
0.2
0
5
10
15
log(MFI)
0
0
5
10
log(MFI)
Figure 4-3: Secretory activity of IR+ and IR– NK cells measured by microengraving. The secretory activity shown in Figure 4-2 was segregated based on
whether the NK cell was IR+ (green) or IR– (red). Results from Donor 1 are shown
and are representative of the other donors.
3). These differences in secretory activity were observed in three independent donors
(Figure 4-4A) and were maintained when the NK cells were activated with IL-2 (100
U/mL, 20 h) prior to the assay (Figure 4-4B).7 IR+ NK cells also tended to lyse
target cells and perform all three effector functions (lysis and secretion of both IFNγ and CCL4) more frequently than IR– NK cells, but these comparisons did not
achieve statistical significance. When K562 cells were used as the targets, we observed
similar trends but with overall lower frequencies of effector function (Figure C-2). We
therefore focused our analysis on the interactions between NK cells and antibodycoated P815 target cells.
We stained for a panel of IRs that included NKG2A, KIR2DL1/2/3, and KIR3DL1.
It is important to note that in the two donors that were not KIR- and HLA-typed
(and thus received the full staining panel), it is possible that a subset of the IR+ NK
cells expressed KIR that did not recognize cognate HLA molecules in the individual. These cells would be unlicensed despite staining positive for IRs. Similarly, in
7
Microarray analysis of NK cells has shown that the expression of KIR genes is downregulated
after 24 h of activation with 100 U/mL IL-2, 197 but other studies have shown that the surface expression of KIRs does not change significantly after short-term activation with IL-2. 198 We observed
similar frequencies of IR+ NK cells in the resting and IL-2-activated conditions.
109
15
Experiment: 20140326 (IL-­2 stim)
Cells with no contact not plotted
Resting
IFN-+ (%)
Lysis+ (%)
15
10
5
0
+
**
40
30
20
10
0
-
+
40
20
0
-
15
60
+
IR
IR
10
5
0
-
P = .08
+
IR
-
IR
IL-­2
IFN-+ (%)
Lysis+ (%)
30
20
10
0
+
-
IR
**
50
40
30
20
10
0
+
*
60
40
20
0
-
IR
+
-
IR
n.s.
40
All+ (%)
n.s.
40
CCL4+ (%)
B
*
80
All+ (%)
n.s.
20
CCL4+ (%)
A
30
20
10
0
+
-
IR
Figure 4-4: IR+ NK cells secrete cytokines more frequently than IR– NK
cells do upon interacting with target cells. The cytolytic and secretory activity
of single NK cells that formed at least one contact with an antibody-coated P815
target cell were measured and segregated based on whether the NK cell was IR+ or
IR– . The performance of all three effector functions (All+ ; far right) was also assessed.
NK cells were either (A) resting or (B) activated with IL-2 (100 U/mL, 20 h) prior
to the assay. Paired data from three donors are shown. The percentage of cells that
were positive for each function were corrected for the background frequencies, which
were calculated from wells with 0 NK cells and 1–3 target cells (for background lysis)
or wells with an NK cell but no targets (for background secretion). Note difference
in scale between graphs. *P < .05, **P < .01, paired t-test.
110
all donors it is possible that a subset of the IR– NK cells expressed other inhibitory
receptors (e.g., the leukocyte Ig-like receptor (LIR)) that were not included in the
staining panel. These cells could in theory have been licensed despite staining negative for IRs, although the inhibitory signals delivered by LIR are less dominant than
those delivered by KIRs or NKG2A. 5
4.4.3
Functional differences between IR+ and IR– NK cells
responding to target cells are maintained across different durations of contact
The increased functional activity of IR+ NK cells compared with IR– NK cells could
have arisen if IR+ NK cells contacted and conjugated with targets more effectively
than IR– NK cells did. Indeed, it has been shown that IR– cells form less stable
conjugates with K562 target cells than IR+ NK cells do, and that this impaired ability
to conjugate with a target cell makes IR– cells less likely to mediate cytolysis. 183
We therefore sought to determine the extent to which differences in contact with
target cells contributed to the differences in secretory activity that we observed. In
these assays, we used the murine P815 cell line as target cells. Importantly, human
LFA-1 binds murine ICAM-1, 199 so any deficiencies in conjugate formation caused by
impaired inside-out signaling to LFA-1 in unlicensed NK cells 183 should be observable.
From the time-lapse microscopy measurements of individual interactions between
NK cells and antibody-coated P815 target cells, we calculated the longest continuous
period of time that each NK cell spent in contact with a target cell (Figure 4-5). Both
IR+ and IR– NK cells exhibited a broad distribution of contact times. Aggregating
data from three donors, the median durations of each NK cell’s longest contact with a
target cell were 144 min (resting IR+ ), 120 min (resting IR– ), 128 min (IL-2-activated
IR+ ), and 104 min (IL-2-activated IR– ). However, although the IR– NK cells tended
to have shorter periods of longest contact, statistical significance was only achieved
in one of the pairwise comparisons after correcting for multiple hypothesis testing
(Figure 4-5).
111
Our finding that the IR+ and IR– NK cells had similar contact time distributions
was somewhat unexpected, because a previous study demonstrated that IR– cells
form less stable conjugates with target cells than IR+ NK cells do. 183 In this study,
however, conjugation was measured by flow cytometry. Measurement of conjugation
by flow cytometry necessarily subjects pairs of potentially interacting cells to stringent
requirements for attachment due to the mechanical perturbation that is involved
in sample preparation. In comparison, measurement of contact in the nanowells
does not involve any mechanical perturbation, and thus does not distinguish between
cells that are weakly or strongly attached. It is possible that unlicensed NK cells
with deficiencies in inside-out signaling to LFA-1 are still able to weakly interact
with target cells but are not able to strongly attach. Because we imaged at low
magnification (6.3×), we could not accurately distinguish between non-productive
contacts and contacts that resulted in stable conjugation, the formation of a synapse,
and reorganization of the actin cytoskeleton. Future studies could be performed
at higher magnification to distinguish between these states of contact. Distinction
between these states of contact could provide one explanation for why not all NK
cells that arrest upon contact with a target are cytolytically active. 97
IR+ and IR– NK cells had a uniformly high likelihood of participating in at least
one contact with a target cell over the course of the assay (94% ± 0.2% for resting IR+
cells, 95% ± 1% for resting IR– cells, 93% ± 3% for IL-2-activated IR+ cells, and 93%
± 2% for IL-2-activated IR– cells), suggesting that they had similar propensities for
initially finding a target. Thus, both IR+ and IR– NK cells were capable of spending
prolonged periods of time in contact with a target cell, but there was considerable
variability in the duration of contact within each population.
We next asked how the variability in contact time was related to the two effector
functions (secretion of IFN-γ or CCL4) that differed between the IR+ and IR– NK
cells (Figures C-3, C-4, C-5).8 The secretion of IFN-γ and CCL4 were both strongly
correlated with the duration of the longest contact that an NK cell made with a target
cell (Figure 4-6). This correlation suggested that in order to make a fair comparison
8
We also assessed the correlation between contact time and cytolysis (Figures C-6 and C-9).
112
300
Resting
200
IL-2
●
●
●
100
●
●
●
●
●
●
●
●
+
−
+
−
●
0
(min)(min)
LongestTime
contact
*
IR
+
−
+
Donor 1
−
+
−
+
−
Donor 2
Donor 3
Figure 4-5: Distribution of contact durations between NK cells and target
cells, segregated by the expression of IRs on NK cells. Interactions between
NK cells (either resting or IL-2-activated, as indicated) and antibody-coated P815
target cells were tracked (8 min/frame) over the course of the 6 h co-incubation in
nanowells. The longest continuous period of contact was scored for single NK cells
in wells with target cells. Only NK cells that had at least one period of contact with
a target cell were included in the analysis. The difference in contact times within
each pair of IR+ and IR– populations were compared by the Mann-Whitney test.
Comparisons that were significant at the α = 0.05 level after applying the BenjaminiHochberg method of controlling the false discovery rate (FDR) are indicated (*).
113
of the target-induced secretory activity of two different cells, it is necessary to also
consider and account for how long each of the cells spends in contact with the target.
We therefore compared the frequency with which resting IR+ and IR– NK cells that
spent the same length of time in contact with a target secreted IFN-γ or CCL4
(Figures 4-7 and C-7). Both populations secreted IFN-γ and CCL4 more frequently
as they spent longer in contact with a target, but despite forming long contacts,
IR– NK cells secreted less frequently9 than IR+ NK cells with matched durations of
contact did. These results suggested that although the expression of IRs (and hence
licensing) affects the ability of NK cells to form stable conjugates with targets, 183
even IR– NK cells that do spend a prolonged period of time in contact with a target
are less functionally active than their IR+ counterparts.
4.4.4
Transcriptional profiling of licensed and unlicensed NK
cells
The results of the single-cell functional studies indicated that there were underlying
differences in the characteristics of IR+ and IR– NK cells that could not be completely
accounted for by differences in their ability to engage target cells. We sought to
uncover these differences by using RNA-Seq to generate transcriptional profiles of
NK cells with distinct repertoires of IRs and their corresponding HLA ligands.
To investigate licensing in a highly defined subset of NK cells, we focused on comparing KIR3DL1+ and KIR3DL1– NK cells from donors that did or did not express
the cognate HLA ligand (HLA-Bw4) for KIR3DL1. To isolate the effects of licensing
through KIR3DL1 as much as possible, we excluded confounding licensing interactions through a combination of donor selection (by excluding donors with certain
KIR-HLA pairings) and flow sorting (by excluding NK cells expressing other IRs).
This strategy allowed us to compare licensed NK cells (KIR3DL1+ cells from HLABw4 donors) and two sets of unlicensed NK cells (KIR3DL1– cells from all donors;
9
The median intensity of positive secretion events was similar across the two subsets of NK cells
(Figure C-8). This similarity suggests that IR+ and IR– NK cells differ in their abilities to commence
secreting rather than in their ability to produce cytokines once they have committed to secretion.
114
Resting
IL-2
1.00
IF Ng+ N K cells (fraction)
IFN-Ȗ+ (fraction)
IF Ng+ N K cells (fraction)
1.00
0.75
0.50
0.25
0.00
Donor
1
0.50
2
3
0.25
0.00
0
100
200
300
0
Longest contact (min)
1.00
100
200
300
100
200
300
Longest contact (min)
1.00
CCL4+ NK cells (fraction)
CCL4+ (fraction)
C C L4+ N K cells (fraction)
0.75
0.75
0.50
0.25
0.00
0.75
0.50
0.25
0.00
0
100
200
300
0
Longest
Longest contact
contact (min)
(min)
Longest contact
contact (min)
(min)
Figure 4-6: NK cells secrete cytokines with a probability that depends on
their longest duration of contact with a target cell. The fraction of NK cells
that secreted IFN-γ (top) or CCL4 (bottom) is plotted as a function of the duration
of their longest continuous period of contact with antibody-coated P815 target cells.
NK cells were either used resting (left) or after activation with IL-2 (right). Trend
lines were estimated by LOESS regression; shading represents the 95% confidence
interval.
115
Donor 1
IFN-Ȗ+ (fraction)
1.00
Donor 2
1.00
1.00
0.75
0.75
0.50
0.50
0.50
0.25
0.25
0.25
IR+
IR-
0.75
0.00
0.00
0
CCL4+ (fraction)
Donor 3
100
200
0.00
0
300
100
200
300
1.00
1.00
1.00
0.75
0.75
0.75
0.50
0.50
0.50
0.25
0.25
0.25
0.00
0.00
0
100
200
300
Longest contact (min)
0
100
200
300
0
100
200
300
0.00
0
100
200
300
Longest contact (min)
Longest contact (min)
Figure 4-7: IR– NK cells are less likely than IR+ cells to secrete cytokines,
even after prolonged contact with a target. The secretory activity of resting NK
cells interacting with antibody-coated P815 target cells is plotted as in Figure 4-6 but
is segregated based on whether the NK cell was IR+ (green) or IR– (orange). Contact
durations were binned due to the low number of events in the IR– population. Trend
lines were estimated by LOESS regression; shading represents the 95% confidence
interval.
116
KIR3DL1+ cells from HLA-Bw6/Bw6 donors).
To select donors with suitable characteristics, we screened a cohort of KIR- and
HLA-typed healthy donors according to the following criteria:
All donors
• Donors who were positive for KIR3DL1 were included, with exceptions as listed
below.
• Donors who were positive for the activating receptor KIR3DS1 were excluded,10
because although the ligand for KIR3DS1 is not definitively known, there is
genetic evidence that KIR3DS1 interacts with HLA-Bw4. 200
• Donors who had NK cells with low surface expression of KIR3DL1 as measured
by flow cytometry were excluded.11 In this cohort, we did not have detailed
KIR typing (at the level of individual alleles) for all donors, and so we used
this surrogate measure to exclude donors who presumably expressed an allele of
KIR3DL1 that results in low surface expression of the receptor (KIR3DL1*005,
*006, *007, *053, *054 ). These low-expressing alleles have been shown to
abrogate licensing effects. 187
• Donors who had NK cells with no surface expression of KIR3DL1 as measured by
flow cytometry were excluded. As described above, we used this surrogate measure to exclude donors that presumably were homozygous for KIR3DL1*004,
an allele that causes KIR3DL1 to be retained intracellularly. 201
• Donors who were HLA-A*03 or HLA-A*11 were excluded, because these HLA
molecules are ligands for KIR3DL2. 202 KIR3DL2 is a framework gene that is
present in almost all KIR haplotypes, and thus we could not exclude donors
that were positive for this KIR. We note that interactions between KIR3DL2
10
These donors were initially included, but sequencing data from these samples were excluded
from the final analysis of licensing interactions.
11
As with KIR3DS1, these donors were initially included, but sequencing data from these samples
were excluded from the final analysis of licensing interactions.
117
and HLA-A*03 or HLA-A*11 are not thought to mediate strong licensing, 169
and so it may not have been necessary to exclude donors with this KIR-HLA
pair.
Donors expressing HLA-Bw4 (cognate HLA ligand for KIR3DL1)
• Donors who were homozygous (HLA-Bw4/Bw4) or heterozygous (HLA-Bw4/Bw6)
for HLA-Bw4 were included.
Donors not expressing HLA-Bw4 (cognate HLA ligand for KIR3DL1)
• Donors who were homozygous (HLA-Bw6/Bw6) for HLA-Bw6 were included.
• Donors who were HLA-A*23, *24, *25, or *32 were excluded, because these
HLA molecules are ligands for KIR3DL1. 202,203
We flow sorted KIR3DL1+ and KIR3DL1– NK cells (gating to exclude confounding
IRs, as described in Section 4.3.2) from three donors who expressed HLA-Bw4 (two
HLA-Bw4/Bw4 and one HLA-Bw4/Bw6) and two who did not (HLA-Bw6/Bw6). In
each population, we typically obtained 103 – 104 cells. After isolating the RNA and
performing RNA-Seq, we compared the expression of KIR3DL1 in the cells that were
sorted as being KIR3DL1+ or KIR3DL1– . In each sample, KIR3DL1 was highly
expressed in the KIR3DL1+ sorted population compared with the KIR3DL1– sorted
population, as expected. In each of these comparisons, KIR3DL1 was in the top 1%
of differentially expressed (over expressed in the KIR3DL1+ population) transcripts
by GFOLD ranking.
Licensing depends on both the KIR status of the NK cell (KIR3DL1+ or KIR3DL1– )
and on the HLA background of the donor (positive or negative for HLA-Bw4, the cognate ligand of KIR3DL1). To identify transcripts whose expression was associated
with this interaction, we approached the problem as a multifactorial experiment and
used DESeq to fit generalized linear models (GLMs) with a design matrix specifying
the KIR status, HLA type, and stimulation condition (with or without antibodycoated P815 target cells) of each sample. Comparison of models with or without the
118
KIR-HLA interaction term returned two differentially expressed transcripts (FDR
<0.1, Benjamini-Hochberg method): IL1R2 and S100A13. IL1R2 was expressed
predominantly in KIR3DL1– NK cells from HLA-Bw4 donors, while S100A13 was
expressed in KIR3DL1+ NK cells from HLA-Bw4 donors and KIR3DL1– NK cells
from HLA-Bw6/Bw6 donors. IL-1R2 is a decoy receptor for IL-1. It has a truncated
cytoplasmic domain (rendering it unable to signal) but competes with IL-1R1 (which
does propagate signals) for ligands. 204 S100A13 is a member of the S100 family of proteins, which are involved in a variety of cellular processes. S100A13 forms a complex
with IL-1α to facilitate the secretion of this cytokine. 205 The biological significance
of the differential expression of proteins related to the IL-1 pathway in NK cells with
different KIR-HLA profiles is unclear, and the results of these transcriptional profiles should be validated at the protein level (e.g., by surface staining for IL-1R2) in
a larger set of donors. Furthermore, it should be noted that because our analysis
included only a small number of donors, it lacked the statistical power necessary to
detect a broader range of differentially expressed transcripts. Similar results were
obtained from pairwise comparisons across KIR3DL1 status and HLA type. These
results are consistent with previous studies indicating that at least some aspects of
licensing are regulated by non-transcriptional mechanisms 190,206
In contrast, when we compared NK cells that were treated with different stimulation conditions (unstimulated or incubated with antibody-coated P815 target cells)
without considering KIR status or HLA type, we recovered a set of 73 differentially
expressed transcripts that were enriched in immune-related gene sets (Gene Set Enrichment Analysis; GSEA), as expected. Overall, the population-level transcriptional
profiling highlighted that a large cohort of donors is needed to properly examine differences that are expected to be subtle (as in the differences between licensed and
unlicensed NK cells) and to have high donor-to-donor variability. In ongoing work,
we are pursuing an alternative approach to investigate potential transcriptional signatures of licensing by applying single-cell RNA-Seq to characterize NK cells with
distinct repertoires of IRs from within the same donor (see Chapter 6).
119
4.5
Conclusions
In this chapter, we described an integrated platform to correlate the expression of IRs
on individual NK cells with their functional activity and dynamic properties during
interactions with target cells. Using this approach, we demonstrated that functional
differences in licensed (IR+ ) and unlicensed (IR– ) NK cells could be detected in the
nanowell-based assay. We found that unlicensed NK cells were less likely than licensed
NK cells to secrete cytokines, even after prolonged contact with antibody-coated P815
target cells. Thus, potential differences in contact behavior are alone not enough to
completely explain the increased ability of licensed NK cells to respond to target cells.
120
Chapter 5
Cellular Barcodes for Efficiently
Profiling Single-Cell Secretory
Responses by Microengraving
This chapter is presented as it appeared in Ref. 207, with minor modifications.
Yamanaka, Y.J.,* Szeto, G.L.,* Gierahn, T.M., Forcier, T.L., Benedict, K.F., Brefo,
M.S.N., Lauffenburger, D.A., Irvine, D.J., Love, J.C. Cellular barcodes for efficiently profiling single-cell secretory responses by microengraving. Analytical Chemistry, 2012, 84(24):10531-10536.
*Equal contribution
121
5.1
Abstract
We present a method that uses fluorescent cellular barcodes to increase the number of
unique samples that can be analyzed simultaneously by microengraving, a nanowell
array-based technique for quantifying the secretory responses of thousands of single
cells in parallel. Using n different fluorescent dyes to generate 2n unique cellular
barcodes, we achieved a 2n -fold reduction in the number of arrays and quantity of
reagents required per sample. The utility of this approach was demonstrated in three
applications of interest in clinical and experimental immunology. Using barcoded
human peripheral blood mononuclear cells and T cells, we constructed dose-response
curves, profiled the secretory behavior of cells treated with mechanistically distinct
stimuli, and tracked the secretory behaviors of different lineages of CD4+ T helper
cells. In addition to increasing the number of samples analyzed by generating secretory profiles of single cells from multiple populations in a time- and reagent-efficient
manner, we expect that cellular barcoding in combination with microengraving will
facilitate unique experimental opportunities for quantitatively analyzing interactions
among heterogeneous cells isolated in small groups (∼2–5 cells).
5.2
Introduction
Immune cells secrete cytokines to coordinate intercellular communication within the
immune network. 17 There is great interest in profiling the secretory activity of immune cells because the cytokines they secrete play a central role in the maintenance
of immune homeostasis, the elimination of infectious pathogens, and the induction
of allergic and autoimmune responses. 41,208 The considerable cell-to-cell variability
present within populations of immune cells underscores the importance of analytical
techniques that enable high-throughput, single-cell secretory measurements.
We previously developed a technique called microengraving that uses dense arrays
of subnanoliter wells (nanowells) to quantify the secretion of multiple cytokines from
thousands of individual cells in parallel. 60–62,209,210 Cells are isolated in an array of
122
nanowells, and a glass slide bearing cytokine-specific antibodies is compressed on the
array to capture the cytokines secreted by the cells in each well. Single-cell secretory
profiles are created by registering the spatial address of each spot on the resulting
microarray of secreted proteins back to the corresponding nanowell, and hence the
cell(s), that produced the cytokines. Microengraving can be repeatedly performed on
the same cells in a nondestructive manner, enabling analytical processes that are not
feasible using destructive or end-point single-cell measurements of cytokine production (e.g., intracellular cytokine staining or ELISPOT). These processes include the
retrieval of viable cytokine-secreting cells 60,61 and longitudinal tracking of single-cell
secretory profiles. 62
To date, microengraving has been performed with a throughput of one sample of
cells per process. In many cases, however, the analysis of multiple samples in parallel
would increase experimental efficiency. One strategy for multiplexing is to use unique
combinations of fluorescent dyes to identify distinct groups of cells. This strategy,
known as fluorescent cellular barcoding, has been used to increase the throughput
of flow cytometry 211,212 and cell-based assays for drug screening, 213 as well as to
track the behavior of specific cells within complex populations. 214–216 In this chapter,
we describe the development and validation of three sets of cellular barcodes that
are compatible with microengraving. Application of these cellular barcodes enables
the simultaneous analysis of multiple samples of cells by microengraving on a single
array of nanowells and thus increases sample throughput, minimizes sample-to-sample
technical variability, and reduces both the number of arrays and the quantity of
reagents used. Moreover, cellular barcoding opens the door to new applications of
microengraving, such as the quantitative analysis of secretory networks governing
cell-cell interactions in multicelled wells.
123
5.3
5.3.1
Materials & Methods
Fabrication of arrays of nanowells
Poly(dimethylsiloxane) (PDMS) (Sylgard 184 Silicone Elastomer Kit; Dow Corning,
Midland, MI) arrays of nanowells comprising 50 µm cubic wells (84,672 wells/array)
were prepared on 75 × 25 mm2 glass slides (Corning, Lowell, MA) following pre-
viously reported protocols 126 with minor adaptations. Details can be found in the
Supplementary Information (Appendix D.1.1).
5.3.2
Cells
Peripheral blood mononuclear cells (PBMCs) were isolated by density centrifugation
using Ficoll-Paque PLUS (GE Healthcare, Piscataway, NJ) from the whole blood
of healthy donors (Research Blood Components, Boston, MA) and were either used
fresh or cryopreserved. Before use, cryopreserved PBMCs were thawed, washed with
complete media (RPMI-1640 (Mediatech, Manassas, VA) supplemented with 10%
heat inactivated fetal bovine serum (FBS; PAA Laboratories, New Bedford, MA), 2
mM l-glutamine, 100 U/mL penicillin, 100 µg/mL streptomycin, and 10 mM HEPES
(all from Mediatech)), and then rested overnight (37°C, 5% CO2 ) in complete media.
T cells were isolated from PBMCs by negative selection (EasySep Human T Cell
Enrichment Kit; STEMCELL Technologies, Vancouver, BC, Canada) and incubated
in complete media until use. Details on T helper (Th) cell polarization can be found
in the Supplementary Information (Appendix D.1.2).
5.3.3
Stimulations
Prior to performing the dye-swap experiment, T cells (106 cells/mL) were stimulated
in a conical tube for 3 h with 25 ng/mL phorbol 12-myristate 13-acetate (PMA) and
1 µM ionomycin (both from Sigma-Aldrich, St. Louis, MO). To evaluate the effect of
dose on the functional responses to PMA/ionomycin, T cells (75,000 cells/well) were
stimulated in a 96-well flat-bottom plate for 5 h with 10 ng/mL PMA and 0, 0.25,
124
0.5, or 1 µg/mL ionomycin. To compare diverse stimulation conditions, PBMCs (106
cells/well) were incubated in a 96-well U-bottom plate in media only or stimulated for
4 h with PMA/ionomycin (10 ng/mL PMA, 1 µM ionomycin), the Toll-like receptor 4
(TLR4) agonist lipopolysaccharide (LPS-EK, 1 µg/mL; InvivoGen, San Diego, CA),
or the TLR7/8 agonist R848 (1 µg/mL; InvivoGen). To test CD4+ Th cell lineages,
Th0-, Th1-, or Th2-polarized cells (105 cells/well) were incubated in a 96-well Ubottom plate in media only or stimulated for 4 h with PMA/ionomycin (10 ng/mL
PMA, 1 µg/mL ionomycin) as indicated.
5.3.4
Cellular barcoding
Three sets of cellular barcodes were applied as described below. Cells were then
washed twice with media and either loaded into a 96-well plate to determine classification accuracy or mixed and loaded onto an array of nanowells for imaging and
microengraving.
1. Antibody-based barcoding. Groups of cells were barcoded by the combinatorial application of anti-CD45-quantum dot (QD) 705 (20 nM; Invitrogen)
and anti-CD45-QD800 (20 nM; Invitrogen) in media for 15 min at room temperature. Four (22 ) unique barcodes were defined on the basis of QD705 and
QD800 staining.
2. Cytosolic barcoding. Groups of cells were barcoded by the combinatorial
application of the membrane-permeable dyes carboxyfluorescein diacetate succinimidyl ester (CFSE, 1 µM; Invitrogen, Grand Island, NY) and CellTracker
Red (CTR, 2.5 µM; Invitrogen). Four (22 ) unique barcodes were defined on
the basis of CFSE and CTR staining. Cells were labeled by directly adding the
appropriate combination of dyes to cell suspensions during the last 30 min of
stimulation.
3. Streptavidin-based barcoding. Cells were suspended in Hanks Balanced
Salt Solution (HBSS) at 106 cells/mL in 15 mL conical tubes that were previously blocked with 0.5% polyvinyl alcohol to prevent the adhesion of cells. The
125
cells were labeled with sulfo-NHS-LC-biotin (0.1 mg/mL; Thermo Scientific,
Waltham, MA) for 30 min at 4°C. After one wash, groups of cells were barcoded
by the combinatorial application of 40 nM of streptavidin-phycoerythrin (PE)Cy7 (BioLegend), streptavidin-ITK-QD705 (Invitrogen), and streptavidin-ITKQD800 (Invitrogen) in HBSS for 15 min at room temperature in a 96-well plate
that was preblocked with 10% FBS. Eight (23 ) unique barcodes were defined on
the basis of PECy7, QD705, and QD800 staining.
5.3.5
Loading of cells onto arrays of nanowells
Cells originating from different groups were combined into a single suspension (5
× 105 cells/mL) after each group had been uniquely barcoded. Then, 300 µL of
cell suspension was deposited onto the array. To minimize the chance of cell-cell
interactions occurring in the mixed cell suspension, cells were deposited on the array
within minutes of being combined together. Cells were allowed to settle by gravity
for 5 min before the array was washed gently with media.
5.3.6
Detection of secreted proteins by microengraving
Immediately after labeling and loading the cells, microengraving was performed for
1 h to detect interferon-γ (IFN-γ), macrophage inflammatory protein-1β (MIP-1β),
interleukin-2 (IL-2), IL-4, IL-6, and tumor necrosis factor-α (TNF) using previously
reported protocols 60,126 with minor adaptations. The specific cytokines analyzed in
each experiment are indicated in the text. Additional details can be found in the
Supplementary Information (Appendix D.1.3).
5.3.7
Staining for viability and surface marker expression
Where indicated, cells were stained on the arrays with anti-CD8-AlexaFluor647 (2
µg/mL; BioLegend) or anti-CD3-PerCP-eFluor710 (2 test volumes; eBioscience, San
Diego, CA). Staining solutions were applied to the arrays for 30 min at room temperature or 4°C and then washed with media. Shortly before imaging, each array
126
was covered with 200 µL of the viability dye calcein violet (2 µM; Invitrogen). Lifter
slips (Electron Microscopy Sciences, Hatfield, PA) were placed on top of the arrays
to prevent drying during imaging.
5.3.8
Imaging cytometry
The arrays were imaged using an automated, inverted epifluorescence microscope
(Axio Observer, 10×/0.3 objective; Carl Zeiss, Jena, Germany; or Eclipse Ti, 10×/0.45
objective; Nikon Instruments, Tokyo, Japan) with an EM-CCD camera (ImagEM;
Hamamatsu Photonics, Hamamatsu, Japan; or iXon3; Andor Technology, South
Windsor, CT). A custom-written MATLAB script (Enumerator) was used to analyze the images. This script returned each cell’s nanowell “address” and its intensity
in each fluorescent channel (Figure D-1).
5.3.9
Data analysis
Custom-written scripts (MATLAB R2010b; MathWorks, Natick, MA) were used to
assign well occupancies and correlate imaging cytometry data with secreted protein
data for each well. Only viable cells (calcein violet+ ) were included in the analysis. Statistical tests were performed in Prism 5 (GraphPad Software, La Jolla, CA).
Additional details can be found in the Supplementary Information (Appendix D.1).
5.4
5.4.1
Results & Discussion
Overview of cellular barcoding applied to multiplex
single-cell secretory measurements
To implement cellular barcoding in conjunction with microengraving, live cells originating from different populations were labeled with unique combinations of fluorescent
dyes, combined, and then loaded onto a single array of nanowells (Figure 5-1). Imaging cytometry was used to determine each cell’s barcode (corresponding to the cell’s
127
population of origin), viability, and expression of surface markers, while microengraving was used to measure the proteins secreted by the cells in each nanowell (Figure
5-2).
5.4.2
Development and validation of cellular barcodes
We developed and validated three sets of cellular barcodes for use in nanowell-based
assays (Figure 5-3). The first set targeted hematopoietic cells (such as immune cells)
expressing CD45 on their surface. Four antibody-based barcodes were created by
labeling CD45+ cells with combinations of two different fluorophore-conjugated antibodies against CD45 (Figure 5-3A). This strategy of barcoding is extendable to other
surface-expressed markers common to all populations of cells of interest but cannot be
applied if cells express surface proteins heterogeneously or if suitable antibodies are
unavailable. We therefore established a second set of cellular barcodes to label cells
with four combinations of two fluorescent, cytosolic dyes (CFSE and CTR) (Figure
5-3B). Unlike antibody-based barcodes, cytosolic barcodes are suitable for general use
with all cell types and are durably retained in cells for hours to days. These features
make cytosolic barcodes useful for experiments that involve mixed cell types or that
require deconvolution of populations after prolonged culture. Finally, to extend the
depth of barcoding, we created streptavidin-based barcodes by biotinylating cells and
then labeling them with eight combinations of three different fluorophore-conjugated
streptavidins (streptavidin-PE-Cy7, streptavidin-QD705, and streptavidin-QD800)
(Figure 5-3C). The high efficiency of cellular labeling with biotin-streptavidin and
the large spectral selection of commercially available streptavidin-fluorophore conjugates make this approach useful when a high depth of barcoding is needed (e.g.,
parallel analysis of disaggregated tissue biopsies).
All three sets of barcodes produced uniform, unambiguous cellular staining (Figure
5-3) and a reproducibly high accuracy of objective classification (Figure 5-4 and Table
D.1; calculations described in Appendix D.1.4). We note that factors such as the
resolution of the microscope, size of the cells, and density of the cells in the nanowells
can influence the accuracy of classification; thus, it is important to measure the
128
Group N
Group 1 Group 2
1
2
...
Apply cellular barcodes
Combine and load cells into nanowells
Imaging cytometry
Microengraved proteins
3
Perform modular operations on array
4
Deconvolve barcodes;; correlate data
i
ĺ Group 1
ii
iii
...
...
...
ĺ Group 2
ĺGroup N
Barcode ĺ Group
Cell
Secreted proteins i -­ iii
Figure 5-1: Schematic for using cellular barcodes to increase the throughput of secretory measurements from single cells. (1) Distinct groups of cells
(e.g., from different treatment conditions) are labeled with unique combinations of
fluorescent dyes (barcodes). (2) The cells are combined and loaded onto the array of
nanowells. (3) Viability and surface marker expression (labeled on-chip), as well as
the barcodes of each cell, are determined by imaging cytometry. Microengraving is
performed to measure the factors secreted by the cells in each well. (4) Barcodes are
deconvolved during image analysis to identify each cell’s group of origin. Data from
imaging cytometry and microengraving are matched on a per-well basis.
129
Imaging cytometry micrograph
viability dye barcode dye 1 barcode dye 2
Microarray of secreted proteins
ȝP
IFN-­Ȗ MIP-­1ȕ IgG background channel Figure 5-2: Matched imaging cytometry and microengraving. Representative
composite micrographs of imaging cytometry (left) and corresponding microarray of
secreted proteins (right) from a 7 × 7 block of nanowells containing barcoded cells. In
this example, calcein violet was used as the viability dye, CFSE was used as barcode
dye 1, and CTR was used as barcode dye 2.
accuracy for each different experimental system in which barcodes are used.
As with any perturbation, the application of barcodes could potentially affect the
biology of the cell. For example, functionalization of the cell surface for labeling
might affect the cell’s ability to respond to autocrine cues. It was therefore necessary to assess whether barcoding perturbed the cellular functions measured over the
time scale of interest here (hours). Accordingly, we performed a dye-swap experiment
to validate that barcoding did not affect the short-term secretory responses of the
cells. Primary human T cells were stimulated with PMA and ionomycin, divided into
aliquots, barcoded, and loaded together onto an array of nanowells. Microengraving
was then performed to quantify the secretion of IFN-γ, MIP-1β, and IL-2 from the
cells in each barcoded group. The percentage of single cells that secreted each cytokine was uniform across all barcoded groups within a given set of barcodes (Figure
5-5). The interbarcode coefficients of variation (CVs) were 3–14% (IFN-γ), 3–14%
(MIP-1β), and 2–12% (IL-2). These CVs are comparable to the interassay variability
130
A
B
C
Cytosolic barcodes
Y
Y
Antibody-­based barcodes
Streptavidin-­based barcodes
cytosolic dye
cell
streptavidin-­
fluorophore
anti-­CD45-­QD
biotin
1 Negative
2 QD705+
3 QD800+
4 QD705+QD800+
1 Negative
2 CFSE+
1 Negative
2 PECy7+
3 QD705+
4 QD800+
4
3
2
3
QD800 (AFU)
800
CTR (AFU)
QD800 (AFU)
4
3 CTR+
4 CFSE+CTR+
2
2
3
QD705 (AFU)
4
2
3
CFSE (AFU)
4
5 PECy7+QD705+
6 PECy7+QD800+
7 QD705+QD800+
8 PECy7+QD705+QD800+
4
3
2
2
PE
Cy 3
7 (
AF 4
U
PECy7 )
2
3
)
5 (AFU
QD70705
4
Figure 5-3: Barcoded T cells. Cells were labeled with (A) antibody-based barcodes, (B) cytosolic barcodes, or (C) streptavidin-based barcodes. The “Negative”
population denotes cells that were not labeled. AFU, arbitrary fluorescence units
(logicle transformation).
of other single-cell assays measuring the production of cytokines, including intracellular cytokine staining 217,218 and ELISPOT. 219 Furthermore, the median fluorescence
intensities (MFIs) for positive secretion events were similar across barcoded groups
(Figure D-2). In each set of barcodes, one group of cells was not labeled until after
microengraving and thus served as an internal, unperturbed control. The secretory
response of this group was not significantly different from the responses of the other
barcoded groups, further indicating that the application of the barcodes did not affect
the short-term secretory biology of the cells. We therefore proceeded to use cellular
barcoding in three applications relevant to experimental and clinical immunology.
5.4.3
Application 1: Efficient construction of dose-response
curves
Many immunological assays rely on stimulating cells with a chemical stimulant or an
antigen and then measuring the resulting secretion of cytokines. In these assays, it
is often beneficial to test a range of doses to determine the optimal concentration
131
A
B
Antibody-­based barcodes
C
Cytosolic barcodes
1 -­ Negative
2 -­ QD705+
3 -­ QD800+
4 -­ QD705+QD800+
Streptavidin-­based barcodes
1 -­ Negative
2 -­ CFSE+
3 -­ CTR+
4 -­ CFSE+CTR+
1 -­ Negative
2 -­ PECy7+
3 -­ QD705+
4 -­ QD800+
5 -­ PECy7+QD705+
6 -­ PECy7+QD800+
7 -­ QD705+QD800+
8 -­ PECy7+QD705+QD800+
Fraction of actual barcode class (%)
2
99.9
3
99.3
4
100.0
1
2
3
1
99.3
2
Actual barcode
1
100.0
Actual barcode
Actual barcode
1
98.6
3
96.9
4
97.8
100
100.0
2
99.9
3
80
100.0
4
60
99.8
5
99.9
6
40
99.9
7
1
4
2
3
4
Called barcode
99.9
1
Called barcode
20
100.0
8
2
3
4
5
6
7
0
8
Called barcode
Figure 5-4: Classification accuracy. Classification accuracy of cells labeled with
(A) antibody-based barcodes, (B) cytosolic barcodes, or (C) streptavidin-based barcodes.
B
1.5
1.0
0.5
0.0
1 2 3 4
1 2 3 4
1 2 3 4
Barcode
IFN-
MIP-1
IL-2
Analyte
C
Cytosolic barcodes
1.5
Normalized secretion
Antibody-­based barcodes
Normalized secretion
Normalized secretion
A
1.0
0.5
0.0
1 2 3 4
1 2 3 4
1 2 3 4
Barcode
IFN-
MIP-1
IL-2
Analyte
Streptavidin-­based barcodes
1.5
1.0
0.5
0.0
12345678 12345678 12345678
IFN-
MIP-1
IL-2
Barcode
Analyte
Figure 5-5: Dye-rotation experiments to validate that the application of
barcoding dyes does not affect the short-term secretory profiles of cells.
Secretory responses were compared among uniformly stimulated T cells that received
different (A) antibody-based barcodes, (B) cytosolic barcodes, or (C) streptavidinbased barcodes. For each analyte, the frequency of secretion observed from cells with
different barcodes was normalized to the unlabeled group of cells (Barcode 1; dark
grey). The mean and range of three replicates are shown. For each analyte and
barcode set, there was no significant difference (P > 0.05) in normalized secretion
among the different barcodes (one-way analysis of variance (ANOVA)).
132
of the stimulating reagent. We used cytosolic barcodes to efficiently measure, on a
single array of nanowells, the secretory response of human T cells activated with 10
ng/mL PMA in combination with four doses of ionomycin (0, 0.25, 0.5, and 1 µg/mL).
After microengraving and deconvolution of barcoded single cells, we observed a dosedependent increase in secretion (Figure 5-6, top panel). By staining for CD8 (a surface
marker that distinguishes subsets of T cells), we identified subset-specific differences
in secretion (Figure 5-6, bottom panel). In response to increasing doses of ionomycin,
the percentage of CD8– T cells secreting IL-2 increased sharply, but the percentage
of CD8+ T cells secreting IL-2 remained at basal levels. In contrast, the percentage
of CD8+ T cells secreting MIP-1β increased strongly in response to increasing doses
of ionomycin, but the percentage of CD8– T cells secreting MIP-1β remained close to
basal levels. These observations are consistent with previous studies. 220,221 Together,
these results demonstrate that cellular barcoding allows dose-response curves to be
efficiently constructed from a single microengraving process.
5.4.4
Application 2: Profiles of secretory responses induced
by diverse stimuli
Clinical studies are currently trying to establish robust methods to monitor human
immune responses. 222,223 One approach is to test how immune cells respond to diverse
stimuli that trigger distinct signaling pathways. 222 To test this scenario, we applied
cytosolic barcodes to PBMCs treated with four different stimulation conditions and
measured their secretory responses by microengraving. We observed secretion profiles consistent with the class of stimuli applied (Figure 5-7, top panel). Unstimulated
cells had low numbers of secreting single cells for all cytokines. Treatment with PMA
and ionomycin stimulated the secretion of IL-2 and TNF, whereas treatment with
TLR agonists LPS or R848 induced the secretion of IL-6 and TNF. By staining for
the T cell-specific surface marker CD3, we identified CD3+ T cells as the dominant
population responding to stimulation with PMA and ionomycin (Figure 5-7, bottom panel). In contrast, the majority of secretory responses to both TLR stimuli
133
Enriched T cells
8
IFN-
MIP-1
IL-2
6
4
1
0.25
0
0.5
2
0
Secreting
single cells (%)
(13605 cells)
Ionomycin (µg/mL) + 10 ng/mL PMA
-­
+
CD8
CD8
Ionomycin (µg/mL)
+ 10 ng/mL PMA
4
0
1
2
0.5
1
0.5
0.25
2
6
0.25
4
8
0
6
0
(4861 cells)
Secreting
single cells (%)
8
0
Secreting
single cells (%)
(8744 cells)
Ionomycin (µg/mL)
+ 10 ng/mL PMA
Figure 5-6: Application of cellular barcodes to construct dose-response
curves. Barcoded T cells were stimulated with 0, 0.25, 0.5, or 1 µg/mL ionomycin
and 10 ng/mL PMA. Microengraving was used to measure single-cell secretory responses, and surface phenotype was distinguished by on-chip labeling with anti-CD8.
The numbers of single cells analyzed are indicated in parentheses.
134
PBMCs
(22885 cells)
Secreting
single cells (%)
20
IL-2
IL-6
TNF
15
10
CD3
15
10
10
5
5
0
0
P/I
R848
LPS
15
P/I
R848
LPS
20
P/I
R848
LPS
(15663 cells)
20
P/I
R848
LPS
+
P/I
R848
LPS
(7222 cells)
Secreting
single cells (%)
P/I
R848
LPS
-­
P/I
R848
LPS
CD3
P/I
R848
LPS
0
P/I
R848
LPS
5
Figure 5-7: Application of cellular barcodes to measure the percentage of
single PBMCs secreting cytokines after treatment with mechanistically
distinct stimuli. Barcoded PBMCs were stimulated with PMA/ionomycin (P/I),
R848, LPS, or left unstimulated (–). Microengraving was used to measure single-cell
secretory responses, and surface phenotype was distinguished by on-chip labeling with
anti-CD3. The numbers of single cells analyzed are indicated in parentheses.
came from CD3– cells. These findings are consistent with existing knowledge on the
secretory responses induced by each of these stimuli 224,225 and show that cellular barcoding can be used to multiplex the analysis of diverse stimulatory conditions for
applications of microengraving in immune monitoring, such as the rapid evaluation
of immunological responses to disease states or vaccination.
135
5.4.5
Application 3: Profiles of lineage-dependent secretory
responses
Different lineages of CD4+ T helper (Th) cells have distinct transcription factors and
secretory profiles 226 but cannot be distinguished by their surface markers. To measure the secretory profiles of multiple lineages on a single array of nanowells, we used
cytosolic barcodes to label PMA/ionomycin-stimulated Th cells that had been cultured under Th0-, Th1-, or Th2-polarizing conditions, which each promote distinct
secretory patterns. Th1 cells can secrete IFN-γ and IL-2, Th2 cells can secrete IL-4
and IL-2, and Th0 cells can secrete both Th1 and Th2 cytokines, 226–229 although in
vitro polarization does not produce 100% conversion. 230 After microengraving and deconvolution of barcoded single cells, we found that the percentages of single cells from
each lineage that secreted IL-2, IFN-γ, or IL-4 (Figure 5-8A), as well as the rates at
which secretion-positive cells from each lineage secreted each cytokine (Figure 5-8B),
were both consistent with the expected lineage-specific secretory patterns. Together,
these results show that cellular barcoding in combination with microengraving enables efficient tracking of the identities of T cells that have distinct secretory profiles
but indistinguishable sets of surface-expressed markers.
5.5
Conclusions
Here, we have demonstrated three sets of cellular barcodes that can be used to increase the throughput of single-cell secretory measurements by 2n -fold, where n is
the number of dyes used to generate the barcodes. Although we focused on single cells, the platform is also well suited for measuring the secretory profiles from
small groups of cells (2–5 cells/nanowell) with precisely defined demographics. We
anticipate that the combination of cellular barcoding and microengraving will enable
the quantitative analysis of cell-cell interactions with a resolution that has not been
possible using traditional experimental systems and will yield novel insights into the
mechanisms governing the behavior of complex cellular systems.
136
Secreting single cells (%)
A
IL-2
IFN-
IL-4
50
40
30
20
10
0
Th0
Th1
Th2
Th1
-
PMA / Ionomycin
B
Th0
Th1
Th2
P/I
n=
390
782
Th1
1000
Th0
77
Th1
Th2
P/I
n=
131
2142
*
100000
10000
456
***
IL-4 (MFI)
10000
1000
***
100000
IFN- (MFI)
IL-2 (MFI)
100000
Th1
10000
1000
Th0
275
Th2
P/I
74
Th1
n=
17
12
Th1
-
74
3
Figure 5-8: Application of cellular barcodes to profile lineage-dependent
secretory responses from CD4+ Th cells. Single-cell secretory responses
from barcoded CD4+ cells biased to Th0, Th1, or Th2 and then stimulated with
PMA/ionomycin (P/I) or left unstimulated (–). (A) Percentage of secreting single
cells from each population of Th cells. (B) Intensities of secretion from single Th
cells. Only positive secretion events are shown. Red lines indicate the mean and the
standard error of the mean. *P < 0.05, ***P < 0.0001, Kruskal-Wallis test followed
by Dunn’s post-test comparing the three groups of Th cells that were stimulated with
PMA/ionomycin. The cells were labeled with cytosolic barcodes in this experiment.
137
138
Chapter 6
Summary & Future Directions
In this thesis, we introduced a set of tools to study cellular interactions at the singlecell level. We applied these tools to measure the cellular dynamics (e.g., motility,
contact time, contact history) of interactions between 100’s – 1000’s of individual
NK cells and target cells. We then correlated these dynamic properties with the
surface-marker phenotype of each NK cell and the functional outcomes (e.g., cytolysis,
secretion of cytokines) that resulted from each interaction. From these integrated
single-cell profiles of dynamics, function, and surface-marker phenotype, we uncovered
new properties by which NK cells perform immune surveillance, attack target cells,
regulate their expression of surface markers, and drive downstream immune responses.
Although here we focused on the interactions between NK cells and target cells, this
platform can be adapted to study a wide range of intercellular interactions of interest.
The nanowell-based platform facilitated the analysis of both individual cells and
multicellular systems. In addition to identifying associations between distinct dynamic, functional, and phenotypic properties in single NK cells, the analysis of large
numbers of nanowells containing different combinations of NK cells and target cells
enabled the identification of multicellular processes that occur either independently
(as we observed with cytolytic activity) or in a manner that is more coordinated than
expected by chance alone (as we observed with the shedding of CD16 from the surface
of NK cells). In contrast, measurements of isolated cells or of bulk mixtures of cells
are not amenable to this type of cooperativity analysis. The strategies for fluorescent
139
cellular barcoding that were initially applied here to increase the throughput of singlecell secretory measurements can be extended to further define the cooperative—or
independent—behaviors of cells in multicellular systems.
The results presented here offer many possibilities for future extensions:
Alternative target cells
In the experiments presented here, we used K562 cells and P815 cells (coated with
antibodies to induce ADCC) as targets for NK cells. Alternative target cell lines, or
even primary cells (e.g., virally infected cells), could be used to provide distinct stimuli
to NK cells. To provide inhibitory signals through specific KIR, MHC Class I deficient
target cell lines (722.221) could be transfected with different HLA Class I allotypes. 198
These cells would be useful for characterizing the inhibition experienced by NK cells
expressing distinct repertoires of IRs. In addition, a variety of different target cells
could be used to further study ADCC, other CD16-mediated effector functions, and
the shedding of CD16. In the context of cancer therapy, B cells (Raji) coated with
rituximab 153 would provide a clinically relevant model of an antibody-coated target
cell, as would HER2+ breast cancer cells coated with trastuzumab. 231 In the context
of HIV, HIV-infected CD4+ T cells coated with serum from an HIV+ patient would
provide a clinically relevant model of an antibody-coated target cell. 156 The use of
these model cell lines would likely influence the functional activity observed in the
assay, as it has been shown that the secretion of cytokines from NK cells mixed with
antibody-coated target cells is correlated with the surface density of antibody on the
target. 12 Furthermore, these target cells would present other relevant ligands (e.g.,
HLA molecules).
Combinations of target cells
In a diseased tissue (e.g., tumor or site of infection), NK cells may encounter target cells coated with antibodies that activate NK cells through CD16 as well as target
cells that express stress-induced ligands that activate NK cells through NKG2D or
other receptors. To simulate these heterogenous mixtures, different types of target
140
cells could be co-loaded into the nanowells. This system would facilitate investigation of how the activation-induced downregulation of these receptors affects historydependent cytolytic responses of NK cells as they serially encounter target cells that
present ligands for either CD16 or NKG2D (or other receptors). At the end-point of
the assay, on-chip fixation, permeabilization, and staining for surface markers as well
as cytolytic molecules such as perforin could be used to infer which NK cells failed to
kill late-encountered targets due to receptor downregulation and which failed to kill
due to depletion of cytolytic granules.
CD16 shedding behavior: Reporters and perturbations
To monitor the activity of the proteases that are responsible for the shedding of
CD16, FRET-based reporter peptides could be spiked into the media and introduced
to the array of nanowells. 232,233 To perturb the activity of the proteases, specific or
broad-spectrum protease inhibitors could be used. 152–157 Together, these approaches
would further define the kinetics of CD16 shedding from NK cells upon contact with
target cells, as well as the functional consequences of CD16 shedding.
Monitor CD16 shedding in live tissues
Dynamic in situ cytometry (DISC) has been used to monitor the shedding of
CD62L from T cells as they encounter and are activated by APCs in living tissues. 84
DISC could be applied in a similar fashion to monitor the shedding of CD16 from NK
cells as they interact with target cells. It should be noted, however, that studies using
murine neutrophils have shown that murine CD16 is not appreciably shed upon cellular activation (although CD62L is), 234 which would complicate the in vivo analysis
of CD16 shedding. This difference in CD16 shedding in humans compared with mice
also raises interesting and clinically important questions about how to interpret data
from mouse models testing the efficacy of antibody therapeutics that induce ADCC.
Single-cell studies of the functional and dynamic properties of NK cells
licensed through specific KIR-HLA interactions
141
The approach from Chapter 4 could be extended to investigate the functional
properties of NK cells that have been licensed through a specific receptor (instead
of a collection of different IRs). For these experiments, donors with specific combinations of KIR and HLA would be selected (e.g., donors positive for KIR3DL1
and either HLA-Bw4 or HLA-Bw6/Bw6). The functional properties of KIR3DL1+
cells from HLA-Bw4 donors would be compared with those from HLA-Bw6/Bw6
donors (and with KIR3DL1– cells from either set of donors). We would expect the
first set of cells to be licensed and more functionally active than the latter two sets.
To obtain the cells for the nanowell assay, cells would be flow sorted by gating on
CD56dim NKG2A– otherKIR– NK cells. After the conclusion of the nanowell assay
(in which cytolysis, secretion, motility, contact, etc. would be measured), on-chip
staining and imaging cytometry would be performed to identify which NK cells were
positive for the specific KIR of interest (e.g., KIR3DL1). Although this approach
would be useful to investigate functional properties of NK cells licensed through specific inhibitory interactions, it would be technically challenging due to the low number
of cells that would be sorted and used on each array of nanowells.
Apply the nanowell assays to investigate the revocable license model
According to the “revocable license” model, licensed NK cells have their license
“revoked” after an activating encounter (by rearrangement of activating receptors
on the cell surface); the license can be reinstated by an inhibitory (licensing) interaction. 206,235 Thus, we would hypothesize that cells that receive an inhibitory signal (via
an anti-NKG2A or anti-KIR antibody) after an initial encounter with a target will
be “relicensed” and more functionally responsive upon a subsequent encounter than
cells that do not receive an inhibitory signal between encounters. To test this hypothesis, we would flow sort CD56dim NK cells, perform the nanowell assay (monitoring
cytolysis, secretion, motility, contact, etc.) while providing inhibitory signals by including antibody-coated beads in the nanowells or by uniformly coating the surface
of the nanowells with the antibodies against the inhibitory receptors, and then stain
on-chip at the end-point of the assay for the inhibitory receptor that was triggered
142
by the antibody (e.g., NKG2A or KIR). We would expect that cells that expressed
the receptor of interest and received an inhibitory signal between interactions would
be more functionally active upon subsequent encounters with target cells than those
that did not receive the inhibitory signal (and hence remained in the “revoked” state).
Investigate the transcriptional profiles of single NK cells with distinct
repertoires of inhibitory receptors
In ongoing work, we are using single-cell RNA-Seq to investigate the transcriptional profiles of individual NK cells with distinct repertoires of inhibitory receptors.
Single-cell RNA-Seq has been used to characterize the population structure of immune
cells and to define subsets of cells based on combinations of expressed transcripts. 68,69
Due to the large amount of heterogeneity present within the NK cell population—
even after gating on NK cells defined by the expression of several surface markers—we
expect that this approach will enable the identification of new signatures and subsets.
In addition, RNA-Seq could be extended to profile individual NK cells or small
groups of NK cells retrieved from the nanowells by micromanipulation at the conclusion of the functional assays. The use of the murine P815 target cells is amenable to
distinguishing transcripts from NK cells and target cells.
143
144
Appendix A
Supplementary Information for
Chapter 2
A.1
A.1.1
Supplementary Methods
Fabrication of arrays of nanowells
Arrays of nanowells containing either 30 µm (248,832 wells/array) or 50 µm (84,672
wells/array) cubic wells were prepared on 75 × 25 mm2 glass slides (Corning) by
injecting a silicone elastomer (poly(dimethylsiloxane) (PDMS; Dow Corning), 10:1
ratio of base:catalyst) into a mold containing a microfabricated silicon master. PDMS
was cured at 80°C for 4 h and then released from the mold. Shortly before depositing
cells onto the array, the arrays were treated with an oxygen plasma (Harrick PDC32G) for 30 s to sterilize the array and render the PDMS hydrophilic. Arrays were
stored in phosphate-buffered saline (PBS) until use, and were washed with complete
media prior to cell loading.
A.1.2
Microengraving to detect secreted proteins
Capture antibodies against MIP-1β (R&D Systems), IFN-γ (Mabtech), and human
immunoglobulin G (hIgG) (ZyMax; Invitrogen) (10 µg/mL each in borate buffer; pH
9) were coated onto poly(L-lysine)-coated glass microscope slides for 1 h at room
145
temperature. Slides were then blocked in 1.5% bovine serum albumin (BSA; EMD
Chemicals) / PBS-TWEEN20 (0.05%; Sigma-Aldrich) (PBST) for 30 min, washed
once in PBS, dipped in water, and spun dry. Prior to microengraving, the cellloaded arrays of nanowells were rinsed with serum-free media containing hIgG (34.5
ng/mL; Athens Research & Technology) to provide a positive background signal in
every well and facilitate the registration of the features during image analysis of the
captured protein microarrays. Excess media was aspirated and the capture antibodycoated glass slides were placed face down on top of the cell-loaded arrays (Figure
2-6). Compression was applied using a microarray hybridization chamber (Agilent).
The clamped arrays were incubated for 1 h (30 µm wells) or 2 h (50 µm wells) to
allow the capture of secreted proteins onto the antibody-coated glass slide. The resulting microarrays of secreted proteins were then separated from the PDMS array,
washed in PBS, blocked with 1.5% BSA-PBST, and hybridized (45 min, room temperature) with the following detection antibodies: biotinylated anti-MIP-1β, anti-IFN-γAlexaFluor555, and anti-hIgG-AlexaFluor700 (1 µg/mL each in 0.1% BSA-PBST, all
from the same manufacturers as the corresponding capture antibodies). Arrays were
washed in PBS and PBST and then hybridized an additional 30 min at room temperature with streptavidin-AlexaFluor647 (1 µg/mL; Invitrogen). After a final series
of washes in PBS, PBST, and water, the protein microarrays were dried and imaged with 5-µm resolution using a commercial microarray scanner (GenePix 4200AL,
Molecular Devices).
The microarray of secreted proteins was analyzed using commercial image processing software (GenePix Pro 6, Molecular Devices). The median fluorescence intensity
(MFI) in each channel was calculated for each spot in the array to determine the
relative intensity of secretion of the cells in the corresponding nanowell. Data were
filtered to exclude spots with saturated pixels or high coefficients of variation (>70–
90). Spots with a high signal-to-noise ratio (>3–5), low relative local background,
and MFI > [global background + 2 standard deviations] were marked as positive
spots.
146
A.2
Supplementary Figures
Live CD3-CD14-CD19lymphocytes
Lymphocytes
250K
NK cells
5
NK cell subsets
5
10
5
10
10
61.7
60.2
100K
50K
11.2
104
103
102
103
85.6
103
102
102
0
0
6.48
0
104
CD56
150K
104
CD56
CD3/14/19
SSC
200K
3.48
0
0
50K
100K
150K
200K
FSC
250K
0
102
103
104
Live/Dead
105
0
103
CD16
104
105
0
50K
100K
150K
SSC
200K
250K
!
Figure A-1: Flow cytometry gating scheme for NK cells. Numbers indicate
percentages within each gate. NK cells were defined as live CD3– CD14– CD19– lymphocytes expressing CD56 and/or CD16 (i.e., CD56+ CD16– (CD56bright NK cells),
CD56+ CD16+ (CD56dim NK cells) or CD56– CD16+ (CD56neg NK cells)).
20110715 K526/221 Recept#491528
Layout-1
!
ULBP2
ULBP3
MICA/B
100
80
80
80
80
60
40
60
40
60
40
20
20
20
0
0
0
0 102
103
104
ULBP-1
105
100
0 102
103
104
ULBP-2
105
100
% of Max
100
% of Max
100
% of Max
100
% of Max
K562
ULBP1
60
40
20
0 102
103
104
ULBP-3
0
105
100
0 102
103
104
MIC A/B
105
100
Figure A-2: Surface expression of NKG2D ligands on K562 target cells.
Black lines represent the K562 cells stained with the indicated antibody; grey shaded
areas show the respective isotype controls.
40
60
40
60
40
20
20
20
0
0
0
0 102
103
104
ULBP-1
105
0 102
103
104
ULBP-2
105
147
80
% of Max
60
80
% of Max
80
% of Max
221
% of Max
80
60
40
20
0 102
103
104
ULBP-3
105
0
0 102
103
104
MIC A/B
105
Wells / array (#)
100000
10000
1000
100
10
1
0.1
NK cells / well (#): 0 1 2 3 0 1 2 3 0 1 2 3 0 1 2 3
K562 cells / well (#):
0
1
2
3
Figure A-3: Distribution of NK cells and K562 target cells in arrayed nanowells. Bar graph of the mean and standard deviation (SD) of the numbers of nanowells
on each array that contained the indicated combinations of NK cells and target cells.
Results were compiled from eight arrays containing 30 µm cubic nanowells. Each
75 × 25 mm2 array contained 248,832 wells. In all experiments, to minimize any
potential edge effects, only wells located in the center of the array were considered
(147,456 wells). Further filtering was performed to exclude wells with dead cells at
the t = 0 h time-point and wells in which the cell occupancy changed between the
images acquired at t = 0 h and t = 4 h.
148
NKG2D
100
Positive NK cells (%)
Positive NK cells (%)
NKp46
80
60
40
20
0
Baseline
Media
IL-2
100
80
60
40
20
0
Baseline
100
80
60
40
20
0
Baseline
Media
IL-2
Perforin
Positive NK cells (%)
Positive NK cells (%)
CD69
Media
IL-2
100
80
60
40
20
0
Baseline
Media
IL-2
Figure A-4: Phenotypes of NK cells after 44 h stimulation with IL-2. NKp46,
NKG2D, CD69, and intracellular perforin were measured by flow cytometry for freshly
isolated NK cells (Baseline) or for NK cells stimulated for 44 h in media or in 50 U/ml
IL-2. Bar graphs show mean and SD for three donors per condition.
149
150
Appendix B
Supplementary Information for
Chapter 3
B.1
B.1.1
Supplementary Methods
Antibodies for flow and imaging cytometry
Cells were stained for flow or imaging cytometry using the antibodies or isotype
controls (ITCs) listed in Table B.1 and described in the text.
Table B.1: Antibodies for flow and imaging cytometry
Surface Marker
CD3
CD14
CD16
CD19
CD56
CD314 (NKG2D)
ITC (mouse IgG1, κ)
ITC (mouse IgG1, κ)
Fluorophore Clone
Supplier
FITC
UCHT1
Biolegend
FITC
HCD14
Biolegend
BV421
3G8
Biolegend
PerCP
3G8
Biolegend
FITC
HIB19
Biolegend
APC
HCD56
Biolegend
PerCP-Cy5.5 B159
BD Biosciences
PE
1D11
Biolegend
PE
MOPC-21 Biolegend
BV421
MOPC-21 Biolegend
151
Use
Flow cytometry
Flow cytometry
Flow cytometry
Imaging cytometry
Flow cytometry
Flow cytometry
Flow cytometry
Flow cytometry
Flow cytometry
Flow cytometry
P815
Frequency (% max)
A
Supplementary Figures
B
K562
Frequency (% max)
B.2
No target
0.5 h
1 h
2 h
4 h
Isotype control
CD16 -­ BV421
CD16 -­ BV421
Figure B-1: Expression of CD16 on resting CD56dim NK cells after stimulation with target cells in bulk. Resting PBMCs (i.e., not activated with IL-2) were
incubated with (A) antibody-coated P815 target cells or (B) K562 target cells for 0,
0.5, 1, 2, or 4 h and then stained, fixed, and acquired by flow cytometry. CD56dim
NK cells were gated as CD3– CD14– CD19– CD56dim lymphocytes.
152
B
P815+Ab
1.00
0.75
0.50
0.25
0.00
●
●
●
● ●● ●
● ●●●
●●
●
●● ● ● ●
●
●
●●●●●
● ● ● ●
● ●●● ●●●●
● ● ●●● ● ● ● ●
●●
●
● ●● ●
●
●
●● ●
●●
●●
●●
●
●
● ●
●●●
●
●
●
●
●
●●
●
●
● ●
●●● ●
●●
●
● ● ●●●●● ● ● ●
●
●●●● ● ● ●
● ●●●● ●● ●
●
● ● ●●
● ●●
●
●●●●●
●●● ●
● ●●● ●
●●
●● ●●● ●●
●●●●●
● ●●
●●●●●
● ●● ●● ● ●●● ●●
0
100
200
+
CD16
CD16+NK
NKcells
cells(fraction)
(fraction)
+
CD16
C D16+ NK
N K cells
cells(fraction)
(fraction)
A
K562
1.00
0.75
0.50
● ●●● ●
● ●●
●●
●●●● ●●
● ●
●●
● ●● ●
● ●●
●●● ● ●
●●
●
● ●● ●
●
●●●● ● ●●● ●●
●● ●● ● ●
●
●● ●●
●● ●●
●●●●
●●
● ●●
●●
●●● ●●
●●
●●●●● ●
● ● ● ● ●●● ●
●●●
● ●● ●
●
●
●
●●
●
●●
● ●
●● ● ● ● ●●
●●●
● ●● ●●
● ●●●
● ●
●
●
●
●●
●
●●● ●
●● ● ●
●
●
●
●● ●
●
●● ● ●
●
●
●● ●
●
●●
●
●● ●
●●●
●
●
IL−2
●
Resting
0.25
0.00
300
0
C umulative contact (min)
100
200
300
Cumulative contact (min)
Figure B-2: NK cells shed CD16 with a probability that depends on their
cumulative duration of contact with a target cell and their activation state.
The fraction of NK cells that retained expression of CD16 is plotted as a function of
their cumulative duration of contact with (A) antibody-coated P815 target cells or
(B) K562 target cells. NK cells were either IL-2-activated (orange, combined data
from three donors) or resting (green, data from one donor, who was also included
in the IL-2-activated data set). Trend lines were estimated by LOESS regression;
shading represents the 95% confidence interval.
Actual fraction
Cytolysis
NK cells
0.8
2
3
0.6
0.4
0.2
0.0
0.0 0.2 0.4 0.6 0.8
Predicted fraction
Figure B-3: NK cells act independently when lysing antibody-coated P815
target cells in nanowells. Wells containing the indicated number of IL-2-activated
NK cells and 1–3 antibody-coated P815 target cells were included in the analysis. The
predicted fractions of wells with at least one dead target were calculated as described
in Ref. 99. Predicted fractions were then compared with the experimentally observed
fractions. Data from three donors are shown. The actual (observed) fractions did not
differ significantly (at the α = 0.05 level) from the predicted fractions according to
Pearson’s chi-squared test (regardless of corrections for multiple comparisons).
153
Experiment: 20140326, 20140328, 20140331 (IL-­2 stim)
B
P815+Ab
100
*
% of CD16- cells
% of CD16- cells
A
80
60
40
20
0
K562
*
100
80
60
40
20
0
Active
shedding
Lysis
Active
shedding
Lysis
Figure B-4: Not all NK cells that shed CD16 lyse target cells. The frequency
of IL-2-activated NK cells that actively shed CD16 (calculated as described in the
text) in response to engaging an (A) antibody-coated P815 or (B) K562 target cell
was compared with the frequency with which these cells exhibited cytolytic activity.
Paired data from three donors are shown. *P < .05, paired t-test.
154
Appendix C
Supplementary Information for
Chapter 4
C.1
C.1.1
Supplementary Methods
Antibodies for flow and imaging cytometry
Cells were stained for flow or imaging cytometry using the antibodies listed in Table
C.1 and described in the text.
C.2
Supplementary Figures
155
Table C.1: Antibodies for flow and imaging cytometry
Surface Marker
CD3
CD14
CD19
CD56
Fluorophore
FITC
PE-Cy7
FITC
PE-Cy7
FITC
PE-Cy7
APC
PerCP-Cy5.5
PE
Clone
UCHT1
UCHT1
HCD14
M5E2
HIB19
HIB19
HCD56
B159
143211
CD158a
(KIR2DL1)
CD158b/j
(KIR2DL2/L3/S2 )
CD158e1
(KIR3DL1)
PE
DX27
BV421
PE
DX9
DX9
CD159a
(NKG2A)
APC
PE
Z199
Z199
156
Supplier
Biolegend
Biolegend
Biolegend
Biolegend
Biolegend
Biolegend
Biolegend
BD Biosciences
R&D Systems
Use
Flow cytometry
Flow cytometry
Flow cytometry
Flow cytometry
Flow cytometry
Flow cytometry
Flow cytometry
Flow cytometry
Flow cytometry
Imaging cytometry
Biolegend
Flow cytometry
Imaging cytometry
Biolegend
Flow cytometry
Biolegend
Flow cytometry
Imaging cytometry
Beckman Coulter Flow cytometry
Beckman Coulter Flow cytometry
Imaging cytometry
Retention (%)
100
80
60
40
20
0
PBMC
NK
NK
(control) (10 min) (>30 min)
Iced and spun
Figure C-1: Chilling the array of nanowells improves the retention of NK
cells. Percentage of single cells retained in nanowells between time-lapse imaging
(or, for PBMC control, mock perturbation by sealing with a glass slide) and analysis
of surface marker staining by imaging cytometry. For the two NK cell data sets, the
array was chilled on ice for the indicated period of time and centrifuged (340 g, 1
min) prior to removing the glass slide that sealed the array during the time-lapse
imaging. Cells were then stained on the array for surface markers and imaged again.
For the PBMC control, all manipulations were performed at room temperature. Bars
indicate median and standard deviation.
157
Experiment: 20140326 (IL-­2 stim)
Cells with no contact not plotted
Resting
n.s.
0.5
+
4
2
0
-
+
4
2
0
-
+
-
IR
IL-­2
n.s.
2.0
1.5
1.0
0.5
0.0
+
n.s.
15
IFN-+ (%)
Lysis+ (%)
6
IR
IR
B
*
8
CCL4+ (%)
1.0
0.0
n.s.
6
IFN-+ (%)
Lysis+ (%)
1.5
-
10
5
0
+
-
IR
IR
n.s.
20
CCL4+ (%)
A
15
10
5
0
+
-
IR
Figure C-2: Functional activity of IR+ and IR– NK cells against K562 target
cells. The cytolytic and secretory activity of single NK cells that formed at least one
contact with a K562 target cell were measured and segregated based on whether the
NK cell was IR+ or IR– . NK cells were either (A) resting or (B) activated with IL2 (100 U/mL, 20 h) prior to the assay. Paired data from three donors are shown.
The percentage of cells that were positive for each function were corrected for the
background frequencies, which were calculated from wells with 0 NK cells and 1–
3 target cells (for background lysis) or wells with an NK cell but no targets (for
background secretion). Note difference in scale between graphs. *P < .05, paired
t-test.
158
log(CCL4)
10
10
10
10
8
8
8
8
6
6
6
6
4
4
4
4
2
2
2
2
log(IFNg)
0
0
200
400
0
0
200
0
400
0
200
400
0
10
10
10
10
8
8
8
8
6
6
6
6
4
4
4
4
2
2
2
2
0
0
200
400
Cumulative
contact time (min)
0
0
200
400
Longest
contact time (min)
0
0
200
400
First time of
contact (min)
0
0
200
400
0
200
400
First time of
longest contact (min)
Figure C-3: Distribution of secretion intensities as a function of contact time
parameters. Each point represents a well containing a single NK cell that interacted
with a target cell with the contact time properties that are indicated at the bottom
of each column. The black line indicates the threshold for positive events. The data
shown here are from Donor 1 (resting NK cells; antibody-coated P815 target cells).
159
CCL4+ (%)
100
100
100
100
50
50
50
50
IFNg+ (%)
0
200
400
0
0
200
0
400
0
200
400
0
100
100
100
100
50
50
50
50
0
Scaled
event frequency
0
0
200
400
0
0
200
0
400
0
200
400
0
1
1
1
1
0.5
0.5
0.5
0.5
0
0
200
400
Cumulative
contact time (min)
0
0
200
400
Longest
contact time (min)
0
0
200
400
First time of
contact (min)
0
0
200
400
0
200
400
0
200
400
First time of
longest contact (min)
Figure C-4: Frequency of secretion as a function of contact time parameters.
The frequencies of single NK cells secreting CCL4 (top row) or IFN-γ (middle row)
are plotted as a function of the contact time properties that are indicated at the
bottom of each column. The overall frequency of events is shown in the bottom row.
The data shown here are from Donor 1 (resting NK cells; antibody-coated P815 target
cells).
160
log(CCL4)
median of positives
log(IFNg)
median of positives
Scaled
event frequency
8
8
8
8
7
7
7
7
6
6
6
6
5
0
200
400
6
5.5
5
0
200
5
400
7
7
6
6
0
200
400
200
400
0
200
400
5.5
5
5
4.5
0
200
400
4
5
0
200
4
400
4.5
0
200
400
4
1
1
1
1
0.5
0.5
0.5
0.5
0
0
6
5
4
5
0
200
400
Cumulative
contact time (min)
0
0
200
400
Longest
contact time (min)
0
0
200
400
First time of
contact (min)
0
0
200
400
First time of
longest contact (min)
Figure C-5: Median intensity of positive secretion events as a function of
contact time parameters. The background-corrected median fluorescence intensity
of positive secretion events from single NK cells secreting CCL4 (top row) or IFN-γ
(middle row) is plotted as a function of the contact time properties that are indicated
at the bottom of each column. The overall frequency of events is shown in the bottom
row. The data shown here are from Donor 1 (resting NK cells; antibody-coated P815
target cells).
161
Lysis (%)
100
100
100
100
80
80
80
80
60
60
60
60
40
40
40
40
20
20
20
20
Scaled
event frequency
0
0
200
400
0
0
200
0
400
0
200
400
0
1
1
1
1
0.8
0.8
0.8
0.8
0.6
0.6
0.6
0.6
0.4
0.4
0.4
0.4
0.2
0.2
0.2
0.2
0
0
200
400
Cumulative
contact time (min)
0
0
200
400
Longest
contact time (min)
0
0
200
400
First time of
contact (min)
0
0
200
400
0
200
400
First time of
longest contact (min)
Figure C-6: Frequency of cytolysis as a function of contact time parameters.
The frequency of single NK cells that lysed a target cell is plotted as a function of
the contact time properties that are indicated at the bottom of each column. The
overall frequency of events is shown in the bottom row. The data shown here are
from Donor 1 (resting NK cells; antibody-coated P815 target cells).
162
CCL4+ (%)
100
100
100
100
50
50
50
50
IFNg+ (%)
0
200
400
0
0
200
0
400
0
200
400
0
100
100
100
100
50
50
50
50
0
Scaled
event frequency
0
0
200
400
0
1
1
0.8
0.8
0.6
0.6
0.4
0.4
0.2
0
200
400
Cumulative
contact time (min)
0.2
0
0
200
0
400
0
200
400
0
1
1
0.5
0.5
200
400
Longest
contact time (min)
0
0
200
400
First time of
contact (min)
0
0
200
400
0
200
400
0
200
400
First time of
longest contact (min)
Figure C-7: Frequency of secretion as a function of contact time parameters,
segregated by the expression of IRs on NK cells. The frequencies of single IR+
(green) and IR– (red) NK cells secreting CCL4 (top row) or IFN-γ (middle row) are
plotted as a function of the contact time properties that are indicated at the bottom
of each column. The overall frequency of events is shown in the bottom row. The
data shown here are from Donor 1 (resting NK cells; antibody-coated P815 target
cells).
163
log(CCL4)
median of positives
7.5
8
7
7
7.5
7
7
6.5
6
5
6.5
6
6
0
200
400
5.5
0
200
5
400
6
0
200
400
5.5
6
6
6
6
5.5
5.5
5.5
5.5
5
5
5
5
4.5
4.5
4.5
4.5
4
0
200
400
4
1
1
0.8
0.8
0.6
0.6
0.4
0.4
0.2
0
200
400
Cumulative
contact time (min)
0.2
0
200
4
400
Scaled
event frequency
log(IFNg)
median of positives
Scaled
event frequency
8
0
200
400
Longest
contact time (min)
0
200
400
4
1
1
0.5
0.5
0
0
200
400
First time of
contact (min)
0
0
200
400
0
200
400
0
200
400
First time of
longest contact (min)
Figure C-8: Median intensity of positive secretion events as a function of
contact time parameters, segregated by the expression of IRs on NK cells.
The background-corrected median fluorescence intensity of positive secretion events
from single IR+ (green) and IR– (red) NK cells secreting CCL4 (top row) or IFN-γ
(middle row) is plotted as a function of the contact time properties that are indicated
at the bottom of each column. The overall frequency of events is shown in the bottom
row. The data shown here are from Donor 1 (resting NK cells; antibody-coated P815
target cells).
164
Lysis (%)
100
100
100
100
80
80
80
80
60
60
60
60
40
40
40
40
20
20
20
20
Scaled
event frequency
0
0
200
400
0
1
1
0.8
0.8
0.6
0.6
0.4
0.2
0
200
0.4
0
200
400
Cumulative
contact time (min)
0.2
0
0
400
0
200
400
0
1
1
0.8
0.8
0.6
0.6
0.4
0.4
0.2
0.2
200
400
Longest
contact time (min)
0
0
200
400
First time
of contact (min)
0
0
200
400
0
200
400
First time of
longest contact (min)
Figure C-9: Frequency of cytolysis as a function of contact time parameters,
segregated by the expression of IRs on NK cells. The frequency of single IR+
(green) and IR– (red) NK cells that lysed a target cell is plotted as a function of the
contact time properties that are indicated at the bottom of each column. The overall
frequency of events is shown in the bottom row. The data shown here are from Donor
1 (resting NK cells; antibody-coated P815 target cells).
165
166
Appendix D
Supplementary Information for
Chapter 6
D.1
D.1.1
Supplementary Methods
Fabrication of arrays of nanowells
Arrays of nanowells comprising 50 µm cubic wells (84,672 wells/array) were prepared
on 75 × 25 mm2 glass slides (Corning, Lowell, MA) following previously reported
protocols 126 with minor adaptations. To fabricate the arrays, the silicone elastomer
poly(dimethylsiloxane) (PDMS) (Sylgard 184 Silicone Elastomer Kit; Dow Corning,
Midland, MI) was mixed at a 10:1 ratio of base:catalyst, degassed under a vacuum at
room temperature for 1 h, and then injected into a mold containing a microfabricated
silicon master. The PDMS was cured at 80°C for 4 h and subsequently released from
the mold to produce a glass slide-backed array of nanowells.
Shortly before use, the arrays of nanowells were treated with oxygen plasma
(Plasma Cleaner PDC-001; Harrick Plasma, Ithaca, NY) for 2 min to sterilize the
array and render the PDMS hydrophilic. Following plasma treatment, the arrays
were stored in phosphate-buffered saline (PBS) and then washed and blocked with
serum-containing media prior to depositing cells onto the array.
167
D.1.2
T helper (Th) cell biasing
Naı̈ve CD4+ T cells were isolated from fresh PBMCs by negative selection (EasySep
Human Naı̈ve CD4+ T Cell Enrichment Kit; STEMCELL Technologies). Purity was
routinely >95%. Cells were plated at 50,000 cells per well in 96-well U-bottom plates
and activated with anti-CD3/28 Dynabeads (Life Technologies, Carlsbad, CA). Th0,
Th1, and Th2 cultures were maintained in Xvivo20 (Lonza, Walkersville, MD) and
activated with a 2:1 bead:cell ratio. Biasing conditions for each subset were as follows:
• Th0: Unsupplemented
• Th1: 20 ng/mL IL-2 (Peprotech, Rock Hill, NJ), 10 ng/mL IL-12 (R&D Systems, Minneapolis, MN), and 10 µg/mL anti-IL-4 (BD Biosciences, Franklin
Lakes, NJ)
• Th2: 20 ng/mL IL-2 (Peprotech), 20 ng/mL IL-4 (R&D Systems), 10 µg/mL
anti-IFN-γ (BD Biosciences), and 10 µg/mL anti-IL-12 (BD Biosciences).
On day 3, cells were counted, split to a concentration of 5 × 106 cells/mL, and
re-fed with fresh media. On day 5, the beads were magnetically removed, and the
cells were re-fed and split. Cells were left resting for 2 days before being restimulated
for another 5 days as described above. After resting for another 2 days, cells were
used for functional assays.
D.1.3
Detection of secreted proteins by microengraving
Microengraving was performed using previously reported protocols 60,126 with minor
adaptations. Poly(L-lysine)-coated glass microscope slides were coated with a capture
antibody against human IgG (ZyMax; Invitrogen) and a set of 3 of the following
capture antibodies, as detailed in the text: anti-IFN-γ (Mabtech, Mariemont, OH),
anti-MIP-1β (R&D Systems), anti-IL-2 (R&D Systems), anti-IL-6 (BD Biosciences),
anti-TNF-α (BioLegend, San Diego, CA), or anti-IL-4 (BioLegend). The capture
antibodies were diluted in borate buffer (pH 9) to a concentration of 10 µg/mL
168
for each antibody immediately prior to being applied to the glass slide. 236 Coating
was performed at room temperature for 1 h or at 4°C overnight. Slides were then
blocked in 1.5% bovine serum albumin (BSA; EMD Chemicals, Gibbstown, NJ) /
PBS-TWEEN20 (.05%; Sigma-Aldrich) (PBST) or in non-fat milk (3% w/v in PBST)
for 30 min, washed once in PBS, dipped in water, and spun or blotted to remove excess
fluid.
Immediately prior to microengraving, the cell-loaded arrays of nanowells were
rinsed with FBS-free media with 0.01% human serum (containing IgG) to provide
a positive background signal in every well. This uniform background signal facilitated the registration of the array during image analysis of the captured protein
microarrays. Capture antibody-coated glass slides were placed face-down on top of
the cell-loaded arrays, and compression was applied using a microarray hybridization
chamber (Agilent, Santa Clara, CA). The clamped arrays were returned to the incubator for 1 h to allow the capture of secreted proteins onto the antibody-coated glass
slide. The resulting protein microarrays of secreted products were then separated from
the PDMS array, washed in PBS, blocked with 1.5% BSA-PBST or 3% milk, and hybridized (45 min, room temperature) with detection antibodies against the analytes
of interest. Solutions of detection antibodies were prepared at 1 µg/mL for each antibody in 0.1% BSA-PBST. The following detection antibodies were used: anti-hIgGAlexaFluor700, anti-IFN-γ-AlexaFluor555, biotinylated anti-MIP-1β (in combination
with streptavidin-AlexaFluor647 (1 µg/mL; Invitrogen) applied during an additional
30 min hybridization step), anti-IL-2-AlexaFluor594, anti-IL-6-AlexaFluor555, antiTNF-α-AlexaFluor488, and anti-IL-4-AlexaFluor647 (all from the same manufacturers as the paired capture antibodies listed above).
The resulting microarrays of secreted proteins were imaged with 5-µm resolution using a commercial microarray scanner (GenePix 4200AL; Molecular Devices,
Sunnyvale, CA). The microarrays were analyzed using commercial image processing software (GenePix Pro 6, Molecular Devices). The median fluorescence intensity
(MFI) in each channel was calculated for each spot on the array to determine the
relative intensity of secretion from the cells in the corresponding nanowell. Data were
169
filtered to exclude spots with saturated pixels or high coefficients of variation (>100).
Spots with a high signal-to-noise ratio (>1), low relative local background, and MFI
> [MFI of local background spots + 2 standard deviations] were marked as positive
spots. Background correction was performed on a per-block (7 × 7 block of nanowells) basis using a custom-written script in MATLAB (R2010b; MathWorks, Natick,
MA).
D.1.4
Calculation of the classification accuracy of cellular
barcoding
Groups of barcoded cells were loaded into separate wells of a 96-well flat-bottom plate;
each well contained a collection of cells with a single, known barcode. Calcein violet
(2 µM) was then added to identify viable cells. The cells in each well were imaged,
and the intensities of each viable cell’s barcoding dyes were determined. Images were
acquired within 30 min of when the images were acquired in the microengraving assays
measuring secretion. The data from all cells were then pooled to produce histograms
of the distribution of fluorescence intensities for each barcoding dye. Barcode classifications for each cell were assigned based on the intensity thresholds determined from
these histograms.
The classification accuracy was calculated as the percentage of cells that received
a given barcode (i.e., total number of viable cells in a given well) that were correctly
classified as having the given barcode. For example, consider a case in which barcode
1 was applied to a group of cells, which were then loaded into a single well of a
96-well plate. If 100 total cells were analyzed from this well, and if the imaging and
analysis procedure classified 98 of these cells as being labeled with barcode 1, then a
classification accuracy of 98% would be assigned to barcode 1.
The analysis described above was performed by imaging cells with uniformly applied barcodes in separate wells of a 96-well plate. In actual experiments, however,
cells with different barcodes are mixed and imaged in nanowells. In these mixed
settings, it is possible that two adjacent cells with different barcodes could be mis170
classified as one double-positive cell. Therefore, to quantify the accuracy of classifying
double-positive cells in typical nanowell experiments, we manually reviewed the images of randomly selected putative double-positive cells and recorded how frequently
their classification as a double-positive cell was incorrect.
171
D.2
Supplementary Figures & Tables
!
Figure D-1: Automated counting of viable cells. Representative image of automated segmentation and counting of cells using Enumerator (written in MATLAB).
The positions of the wells are determined from the transmitted light image (not
shown), and the segmentation of the cells within the wells is determined from the
fluorescence signal of the viability dye (calcein violet; shown in white). Red boxes
mark cells that were identified by the segmentation algorithm. The barcode of each
identified cell is determined from the intensities of the fluorescent cellular barcoding
dyes associated with the cell (not shown).
172
Table D.1: Classification accuracy for double-positive cells
Barcode
set
Antibody
Cytosolic
Streptavidin
Double-positive cells
manually reviewed
(#)
50
50
50
Correctly
classified double-positive
events (#)
49
50
48
173
Accuracy of classifying double-positive
cells (%)
98
100
96
A
B
C
Antibody-­based barcodes
1000
100
1
2
3
10000
1000
100
4
IFN- (MFI)
10000
1
Antibody-based barcode
1
2
3
1000
1
1
2
3
4
3
Antibody-based barcode
4
5
6
7
8
1000
100
4
1
2
3
4
5
6
7
8
Streptavidin-based barcode
100000
IL-2 (MFI)
10000
1000
100
3
10000
Cytosolic barcode
IL-2 (MFI)
IL-2 (MFI)
2
100000
1000
2
100000
Antibody-based barcode
10000
1
Streptavidin-based barcode
10000
100
4
100000
1000
100
4
MIP-1 (MFI)
1000
100
3
100000
10000
100
2
10000
Cytosolic barcode
MIP-1 (MFI)
MIP-1 (MFI)
100000
Streptavidin-­based barcodes
100000
100000
IFN- (MFI)
IFN- (MFI)
100000
Cytosolic barcodes
1
2
3
Cytosolic barcode
4
*
10000
1000
100
1
2
3
4
5
6
7
8
Streptavidin-based barcode
Figure D-2: Intensities of secretion in dye-rotation experiments. Intensities of
secretion from secretion-positive single cells that were exposed to a uniform stimulation (PMA/ionomycin) and labeled with (A) antibody-based barcodes, (B) cytosolic
barcodes, or (C) streptavidin-based barcodes. Boxes indicate the median and the
25th and 75th percentiles, and whiskers indicate the min and max. MFI, median
fluorescence intensity. *P < 0.05, Kruskal-Wallis test followed by Dunn’s post-test.
Note: The borderline-significant (P = 0.039) difference in the intensity of secreted
IL-2 from cells labeled with the streptavidin-based barcodes was only observed in one
of nine replicates of the microengraving process; in all other replicates, there was no
significant difference (P > 0.05) among the streptavidin-based barcodes.
174
Bibliography
[1] Yamanaka, Y. J., Gierahn, T. M. & Love, J. C. The dynamic lives of t cells: new
approaches and themes. Trends Immunol. 34, 59–66 (2013). PMID: 23200626
PMCID: PMC3565007.
[2] Yaln, A., Yamanaka, Y. J. & Love, J. C. Analytical technologies for integrated
single-cell analysis of human immune responses. Methods Mol. Biol. 853, 211–
235 (2012). PMID: 22323150.
[3] Warren, R. L. et al. Exhaustive t-cell repertoire sequencing of human peripheral
blood samples reveals signatures of antigen selection and a directly measured
repertoire size of at least 1 million clonotypes. Genome Res. 21, 790–797 (2011).
PMID: 21349924 PMCID: PMC3083096.
[4] Blattman, J. N. et al. Estimating the precursor frequency of naive antigenspecific CD8 t cells. J. Exp. Med. 195, 657–664 (2002). PMID: 11877489
PMCID: PMC2193761.
[5] Pegram, H. J., Andrews, D. M., Smyth, M. J., Darcy, P. K. & Kershaw, M. H.
Activating and inhibitory receptors of natural killer cells. Immunol. Cell Biol.
89, 216–224 (2011). PMID: 20567250.
[6] Gardiner, C. M. Killer cell immunoglobulin-like receptors on NK cells: the how,
where and why. Int. J. Immunogenet. 35, 1–8 (2008). PMID: 18093180.
[7] Newell, E. W., Sigal, N., Bendall, S. C., Nolan, G. P. & Davis, M. M. Cytometry
by time-of-flight shows combinatorial cytokine expression and virus-specific cell
niches within a continuum of CD8+ t cell phenotypes. Immunity 36, 142–152
(2012). PMID: 22265676 PMCID: PMC3752833.
[8] Horowitz, A. et al. Genetic and environmental determinants of human NK
cell diversity revealed by mass cytometry. Sci Transl Med 5, 208ra145 (2013).
PMID: 24154599 PMCID: PMC3918221.
[9] Varadarajan, N. et al. A high-throughput single-cell analysis of human CD8 t
cell functions reveals discordance for cytokine secretion and cytolysis. J. Clin.
Invest. 121, 4322–4331 (2011). PMID: 21965332 PMCID: PMC3204845.
175
[10] Betts, M. R. et al. Sensitive and viable identification of antigen-specific CD8+
t cells by a flow cytometric assay for degranulation. J. Immunol. Methods 281,
65–78 (2003). PMID: 14580882.
[11] Alter, G., Malenfant, J. M. & Altfeld, M. CD107a as a functional marker for
the identification of natural killer cell activity. J. Immunol. Methods 294, 15–22
(2004). PMID: 15604012.
[12] Fauriat, C., Long, E. O., Ljunggren, H.-G. & Bryceson, Y. T. Regulation of
human NK-cell cytokine and chemokine production by target cell recognition.
Blood 115, 2167–2176 (2010). PMID: 19965656 PMCID: PMC2844017.
[13] Min-Oo, G., Kamimura, Y., Hendricks, D. W., Nabekura, T. &
Lanier, L. L.
Natural killer cells:
walking three paths down
memory lane.
Trends in Immunology 34, 251–258 (2013).
URL
http://www.cell.com/article/S1471490613000318/abstract.
[14] Murphy, K. M. & Stockinger, B. Effector t cell plasticity: flexibility in the
face of changing circumstances. Nat. Immunol. 11, 674–680 (2010). PMID:
20644573 PMCID: PMC3249647.
[15] Dustin, M. L. T-cell activation through immunological synapses and kinapses.
Immunol. Rev. 221, 77–89 (2008). PMID: 18275476.
[16] Dustin, M. L. & Long, E. O. Cytotoxic immunological synapses. Immunol.
Rev. 235, 24–34 (2010). PMID: 20536553 PMCID: PMC2950621.
[17] Frankenstein, Z., Alon, U. & Cohen, I. R. The immune-body cytokine network
defines a social architecture of cell interactions. Biol. Direct 1, 32 (2006). PMID:
17062134 PMCID: PMC1636025.
[18] Boehm, U., Klamp, T., Groot, M. & Howard, J. C. Cellular responses to
interferon-gamma. Annu. Rev. Immunol. 15, 749–795 (1997). PMID: 9143706.
[19] Dragic, T. et al. HIV-1 entry into CD4+ cells is mediated by the chemokine
receptor CC-CKR-5. Nature 381, 667–673 (1996). PMID: 8649512.
[20] Cocchi, F. et al. Identification of RANTES, MIP-1 alpha, and MIP-1 beta as
the major HIV-suppressive factors produced by CD8+ t cells. Science 270,
1811–1815 (1995). PMID: 8525373.
[21] Perfetto, S. P., Chattopadhyay, P. K. & Roederer, M. Seventeen-colour flow
cytometry: unravelling the immune system. Nat. Rev. Immunol. 4, 648–655
(2004). PMID: 15286731.
[22] Krutzik, P. O. & Nolan, G. P. Intracellular phospho-protein staining techniques
for flow cytometry: monitoring single cell signaling events. Cytometry A 55,
61–70 (2003). PMID: 14505311.
176
[23] Sachs, K., Perez, O., Pe’er, D., Lauffenburger, D. A. & Nolan, G. P. Causal
protein-signaling networks derived from multiparameter single-cell data. Science 308, 523–529 (2005). PMID: 15845847.
[24] Lamoreaux, L., Roederer, M. & Koup, R. Intracellular cytokine optimization
and standard operating procedure. Nat Protoc 1, 1507–1516 (2006). PMID:
17406442.
[25] Nagorsen, D. & Marincola, F. M. Analyzing T Cell Responses: How to analyze
cellular immune responses against tumor associated antigens (Springer, 2006).
[26] Rothaeusler, K. & Baumgarth, N. Assessment of cell proliferation by 5bromodeoxyuridine (BrdU) labeling for multicolor flow cytometry. Curr Protoc
Cytom Chapter 7, Unit7.31 (2007). PMID: 18770853.
[27] Yu, Y., Arora, A., Min, W., Roifman, C. M. & Grunebaum, E. EdU incorporation is an alternative non-radioactive assay to [(3)H]thymidine uptake for
in vitro measurement of mice t-cell proliferations. J. Immunol. Methods 350,
29–35 (2009). PMID: 19647746.
[28] Quah, B. J. C., Warren, H. S. & Parish, C. R. Monitoring lymphocyte proliferation in vitro and in vivo with the intracellular fluorescent dye carboxyfluorescein
diacetate succinimidyl ester. Nat Protoc 2, 2049–2056 (2007). PMID: 17853860.
[29] Aubry, J. P. et al. Annexin v used for measuring apoptosis in the early events
of cellular cytotoxicity. Cytometry 37, 197–204 (1999). PMID: 10520200.
[30] Kim, G. G., Donnenberg, V. S., Donnenberg, A. D., Gooding, W. & Whiteside,
T. L. A novel multiparametric flow cytometry-based cytotoxicity assay simultaneously immunophenotypes effector cells: comparisons to a 4 h 51Cr-release
assay. J. Immunol. Methods 325, 51–66 (2007). PMID: 17617419 PMCID:
PMC2040258.
[31] Wlodkowic, D. et al. Chip-based dynamic real-time quantification of druginduced cytotoxicity in human tumor cells. Anal. Chem. 81, 6952–6959 (2009).
PMID: 19572560 PMCID: PMC3977701.
[32] Liu, L. et al. Visualization and quantification of t cell-mediated cytotoxicity using cell-permeable fluorogenic caspase substrates. Nat. Med. 8, 185–189 (2002).
PMID: 11821904.
[33] Packard, B. Z., Telford, W. G., Komoriya, A. & Henkart, P. A. Granzyme b
activity in target cells detects attack by cytotoxic lymphocytes. J. Immunol.
179, 3812–3820 (2007). PMID: 17785818.
[34] Chattopadhyay, P. K. et al. Quantum dot semiconductor nanocrystals for immunophenotyping by polychromatic flow cytometry. Nat. Med. 12, 972–977
(2006). PMID: 16862156.
177
[35] Bandura, D. R. et al. Mass cytometry: technique for real time single cell
multitarget immunoassay based on inductively coupled plasma time-of-flight
mass spectrometry. Anal. Chem. 81, 6813–6822 (2009). PMID: 19601617.
[36] Bendall, S. C. et al. Single-cell mass cytometry of differential immune and drug
responses across a human hematopoietic continuum. Science 332, 687–696
(2011). PMID: 21551058 PMCID: PMC3273988.
[37] Bendall, S. C., Nolan, G. P., Roederer, M. & Chattopadhyay, P. K. A deep
profiler’s guide to cytometry. Trends Immunol. 33, 323–332 (2012). PMID:
22476049 PMCID: PMC3383392.
[38] Qiu, P. et al. Extracting a cellular hierarchy from high-dimensional cytometry
data with SPADE. Nat. Biotechnol. 29, 886–891 (2011). PMID: 21964415
PMCID: PMC3196363.
[39] Angelo, M. et al. Multiplexed ion beam imaging of human breast tumors. Nat.
Med. 20, 436–442 (2014). PMID: 24584119.
[40] Giesen, C. et al. Highly multiplexed imaging of tumor tissues with subcellular resolution by mass cytometry. Nat. Methods 11, 417–422 (2014). PMID:
24584193.
[41] Seder, R. A., Darrah, P. A. & Roederer, M. T-cell quality in memory and
protection: implications for vaccine design. Nat. Rev. Immunol. 8, 247–258
(2008). PMID: 18323851.
[42] Mascher, B., Schlenke, P. & Seyfarth, M. Expression and kinetics of cytokines
determined by intracellular staining using flow cytometry. J. Immunol. Methods
223, 115–121 (1999). PMID: 10037239.
[43] Sandberg, J. K., Fast, N. M. & Nixon, D. F. Functional heterogeneity of
cytokines and cytolytic effector molecules in human CD8+ t lymphocytes. J.
Immunol. 167, 181–187 (2001). PMID: 11418647.
[44] Lagrelius, M., Jones, P., Franck, K. & Gaines, H. Cytokine detection by multiplex technology useful for assessing antigen specific cytokine profiles and kinetics
in whole blood cultured up to seven days. Cytokine 33, 156–165 (2006). PMID:
16529940.
[45] Betts, M. R. et al. HIV nonprogressors preferentially maintain highly functional
HIV-specific CD8+ t cells. Blood 107, 4781–4789 (2006). PMID: 16467198
PMCID: PMC1895811.
[46] Almeida, J. R. et al. Superior control of HIV-1 replication by CD8+ t cells is
reflected by their avidity, polyfunctionality, and clonal turnover. J. Exp. Med.
204, 2473–2485 (2007). PMID: 17893201 PMCID: PMC2118466.
178
[47] Precopio, M. L. et al. Immunization with vaccinia virus induces polyfunctional
and phenotypically distinctive CD8(+) t cell responses. J. Exp. Med. 204,
1405–1416 (2007). PMID: 17535971 PMCID: PMC2118607.
[48] Darrah, P. A. et al. Multifunctional TH1 cells define a correlate of vaccinemediated protection against leishmania major. Nat. Med. 13, 843–850 (2007).
PMID: 17558415.
[49] Hernandez-Fuentes, M. P. & Salama, A. In vitro assays for immune monitoring
in transplantation. Methods Mol. Biol. 333, 269–290 (2006). PMID: 16790856.
[50] Oelke, M. et al. Functional characterization of CD8(+) antigen-specific cytotoxic t lymphocytes after enrichment based on cytokine secretion: comparison
with the MHC-tetramer technology. Scand. J. Immunol. 52, 544–549 (2000).
PMID: 11119258.
[51] Saade, F., Gorski, S. A. & Petrovsky, N.
Pushing the frontiers of t-cell vaccines:
accurate measurement of human t-cell
responses.
Expert Rev Vaccines 11, 1459–1470 (2012).
URL
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3581609/.
PMID:
23252389 PMCID: PMC3581609.
[52] Zuber, B. et al. Detection of human perforin by ELISpot and ELISA: ex vivo
identification of virus-specific cells. J. Immunol. Methods 302, 13–25 (2005).
PMID: 16005014.
[53] Shafer-Weaver, K. et al. The granzyme b ELISPOT assay: an alternative to
the 51Cr-release assay for monitoring cell-mediated cytotoxicity. J Transl Med
1, 14 (2003). PMID: 14697097 PMCID: PMC317386.
[54] Snyder, J. E. et al. Measuring the frequency of mouse and human cytotoxic
t cells by the lysispot assay: independent regulation of cytokine secretion and
short-term killing. Nat. Med. 9, 231–235 (2003). PMID: 12539041.
[55] Ahlborg, N. & Axelsson, B. Dual- and triple-color fluorospot. Methods Mol.
Biol. 792, 77–85 (2012). PMID: 21956502.
[56] Ma, C. et al. A clinical microchip for evaluation of single immune cells reveals
high functional heterogeneity in phenotypically similar t cells. Nat. Med. 17,
738–743 (2011). PMID: 21602800 PMCID: PMC3681612.
[57] Lu, Y. et al. High-throughput secretomic analysis of single cells to assess
functional cellular heterogeneity. Anal. Chem. 85, 2548–2556 (2013). PMID:
23339603 PMCID: PMC3589817.
[58] Ma, C. et al. Multifunctional t-cell analyses to study response and progression
in adoptive cell transfer immunotherapy. Cancer Discov 3, 418–429 (2013).
PMID: 23519018 PMCID: PMC3716460.
179
[59] Love, J. C., Ronan, J. L., Grotenbreg, G. M., van der Veen, A. G. & Ploegh,
H. L. A microengraving method for rapid selection of single cells producing antigen-specific antibodies. Nat. Biotechnol. 24, 703–707 (2006). PMID:
16699501.
[60] Han, Q., Bradshaw, E. M., Nilsson, B., Hafler, D. A. & Love, J. C. Multidimensional analysis of the frequencies and rates of cytokine secretion from single
cells by quantitative microengraving. Lab Chip 10, 1391–1400 (2010). PMID:
20376398 PMCID: PMC3128808.
[61] Varadarajan, N. et al. Rapid, efficient functional characterization and recovery
of HIV-specific human CD8+ t cells using microengraving. Proc. Natl. Acad.
Sci. U.S.A. 109, 3885–3890 (2012). PMID: 22355106 PMCID: PMC3309713.
[62] Han, Q. et al. Polyfunctional responses by human t cells result from sequential
release of cytokines. Proc. Natl. Acad. Sci. U.S.A. 109, 1607–1612 (2012).
PMID: 22160692 PMCID: PMC3277116.
[63] Flatz, L. et al. Single-cell gene-expression profiling reveals qualitatively distinct
CD8 t cells elicited by different gene-based vaccines. Proc. Natl. Acad. Sci.
U.S.A. 108, 5724–5729 (2011). PMID: 21422297 PMCID: PMC3078363.
[64] Tang, F. et al. mRNA-Seq whole-transcriptome analysis of a single cell. Nat.
Methods 6, 377–382 (2009). PMID: 19349980.
[65] Ramskld, D. et al. Full-length mRNA-Seq from single-cell levels of RNA and
individual circulating tumor cells. Nat. Biotechnol. 30, 777–782 (2012). PMID:
22820318 PMCID: PMC3467340.
[66] Picelli, S. et al. Smart-seq2 for sensitive full-length transcriptome profiling in
single cells. Nat. Methods 10, 1096–1098 (2013). PMID: 24056875.
[67] Picelli, S. et al. Full-length RNA-seq from single cells using smart-seq2. Nat
Protoc 9, 171–181 (2014). PMID: 24385147.
[68] Shalek, A. K. et al. Single-cell transcriptomics reveals bimodality in expression
and splicing in immune cells. Nature 498, 236–240 (2013). PMID: 23685454
PMCID: PMC3683364.
[69] Jaitin, D. A. et al. Massively parallel single-cell RNA-seq for marker-free decomposition of tissues into cell types. Science 343, 776–779 (2014). PMID:
24531970.
[70] Lee, J. H. et al. Highly multiplexed subcellular RNA sequencing in situ. Science
343, 1360–1363 (2014). PMID: 24578530.
[71] Mempel, T. R., Scimone, M. L., Mora, J. R. & von Andrian, U. H. In vivo imaging of leukocyte trafficking in blood vessels and tissues. Curr. Opin. Immunol.
16, 406–417 (2004). PMID: 15245733.
180
[72] Bousso, P. T-cell activation by dendritic cells in the lymph node: lessons from
the movies. Nat. Rev. Immunol. 8, 675–684 (2008). PMID: 19172690.
[73] Cahalan, M. D. & Parker, I. Choreography of cell motility and interaction
dynamics imaged by two-photon microscopy in lymphoid organs. Annu. Rev.
Immunol. 26, 585–626 (2008). PMID: 18173372 PMCID: PMC2732400.
[74] Germain, R. N., Robey, E. A. & Cahalan, M. D. A decade of imaging cellular
motility and interaction dynamics in the immune system. Science 336, 1676–
1681 (2012). PMID: 22745423 PMCID: PMC3405774.
[75] Murooka, T. T. & Mempel, T. R. Multiphoton intravital microscopy to study
lymphocyte motility in lymph nodes. Methods Mol. Biol. 757, 247–257 (2012).
PMID: 21909917.
[76] Mempel, T. R., Henrickson, S. E. & Von Andrian, U. H. T-cell priming by
dendritic cells in lymph nodes occurs in three distinct phases. Nature 427,
154–159 (2004). PMID: 14712275.
[77] Miller, M. J., Safrina, O., Parker, I. & Cahalan, M. D. Imaging the single cell
dynamics of CD4+ t cell activation by dendritic cells in lymph nodes. J. Exp.
Med. 200, 847–856 (2004). PMID: 15466619 PMCID: PMC2213293.
[78] Shakhar, G. et al. Stable t cell-dendritic cell interactions precede the development of both tolerance and immunity in vivo. Nat. Immunol. 6, 707–714
(2005). PMID: 15924144 PMCID: PMC1560107.
[79] Celli, S., Lematre, F. & Bousso, P. Real-time manipulation of t cell-dendritic
cell interactions in vivo reveals the importance of prolonged contacts for CD4+
t cell activation. Immunity 27, 625–634 (2007). PMID: 17950004.
[80] Henrickson, S. E. et al. T cell sensing of antigen dose governs interactive behavior with dendritic cells and sets a threshold for t cell activation. Nat. Immunol.
9, 282–291 (2008). PMID: 18204450 PMCID: PMC2698867.
[81] Celli, S., Garcia, Z. & Bousso, P. CD4 t cells integrate signals delivered during
successive DC encounters in vivo. J. Exp. Med. 202, 1271–1278 (2005). PMID:
16275764 PMCID: PMC2213240.
[82] Skokos, D. et al. Peptide-MHC potency governs dynamic interactions between
t cells and dendritic cells in lymph nodes. Nat. Immunol. 8, 835–844 (2007).
PMID: 17632517.
[83] Clark, C. E., Hasan, M. & Bousso, P. A role for the immediate early gene product c-fos in imprinting t cells with short-term memory for signal summation.
PLoS ONE 6, e18916 (2011). PMID: 21552553 PMCID: PMC3084237.
181
[84] Moreau, H. D. et al. Dynamic in situ cytometry uncovers t cell receptor signaling
during immunological synapses and kinapses in vivo. Immunity 37, 351–363
(2012). PMID: 22683126.
[85] Friedman, R. S., Beemiller, P., Sorensen, C. M., Jacobelli, J. & Krummel, M. F.
Real-time analysis of t cell receptors in naive cells in vitro and in vivo reveals
flexibility in synapse and signaling dynamics. J. Exp. Med. 207, 2733–2749
(2010). PMID: 21041455 PMCID: PMC2989766.
[86] Azar, G. A., Lematre, F., Robey, E. A. & Bousso, P. Subcellular dynamics of
t cell immunological synapses and kinapses in lymph nodes. Proc. Natl. Acad.
Sci. U.S.A. 107, 3675–3680 (2010). PMID: 20133676 PMCID: PMC2840513.
[87] Celli, S., Breart, B. & Bousso, P. Intravital two-photon imaging of natural
killer cells and dendritic cells in lymph nodes. Methods Mol. Biol. 415, 119–126
(2008). PMID: 18370151.
[88] Garrod, K. R., Wei, S. H., Parker, I. & Cahalan, M. D. Natural killer cells
actively patrol peripheral lymph nodes forming stable conjugates to eliminate
MHC-mismatched targets. Proc. Natl. Acad. Sci. U.S.A. 104, 12081–12086
(2007). PMID: 17609379 PMCID: PMC1924574.
[89] Deguine, J., Breart, B., Lematre, F., Di Santo, J. P. & Bousso, P. Intravital
imaging reveals distinct dynamics for natural killer and CD8(+) t cells during
tumor regression. Immunity 33, 632–644 (2010). PMID: 20951068.
[90] Foley, M. H. et al. High avidity CD8+ t cells efficiently eliminate motile hivinfected targets and execute a locally focused program of anti-viral function.
PLoS ONE 9, e87873 (2014). PMID: 24551068 PMCID: PMC3923750.
[91] Guldevall, K. et al. Imaging immune surveillance of individual natural killer cells
confined in microwell arrays. PLoS ONE 5, e15453 (2010). PMID: 21103395
PMCID: PMC2980494.
[92] Eriksson, M. et al. Inhibitory receptors alter natural killer cell interactions with
target cells yet allow simultaneous killing of susceptible targets. J. Exp. Med.
190, 1005–1012 (1999). PMID: 10510090 PMCID: PMC2195645.
[93] Bhat, R. & Watzl, C. Serial killing of tumor cells by human natural killer cells–
enhancement by therapeutic antibodies. PLoS ONE 2, e326 (2007). PMID:
17389917 PMCID: PMC1828617.
[94] Culley, F. J. et al. Natural killer cell signal integration balances synapse symmetry and migration. PLoS Biol. 7, e1000159 (2009). PMID: 19636352 PMCID:
PMC2707003.
[95] Vanherberghen, B. et al. Classification of human natural killer cells based
on migration behavior and cytotoxic response. Blood 121, 1326–1334 (2013).
PMID: 23287857.
182
[96] Olofsson, P. E. et al. Distinct migration and contact dynamics of resting and
IL-2-Activated human natural killer cells. Front Immunol 5, 80 (2014). PMID:
24639676 PMCID: PMC3945532.
[97] Khorshidi, M. A. et al. Analysis of transient migration behavior of natural killer
cells imaged in situ and in vitro. Integr Biol (Camb) 3, 770–778 (2011). PMID:
21687858.
[98] Nieto, M. et al. Roles of chemokines and receptor polarization in NK-target
cell interactions. J. Immunol. 161, 3330–3339 (1998). PMID: 9759849.
[99] Yamanaka, Y. J. et al. Single-cell analysis of the dynamics and functional
outcomes of interactions between human natural killer cells and target cells.
Integr Biol (Camb) 4, 1175–1184 (2012). PMID: 22945136.
[100] Cooper, M. A., Fehniger, T. A. & Caligiuri, M. A. The biology of human natural
killer-cell subsets. Trends Immunol. 22, 633–640 (2001). PMID: 11698225.
[101] Vivier, E., Tomasello, E., Baratin, M., Walzer, T. & Ugolini, S. Functions of
natural killer cells. Nat. Immunol. 9, 503–510 (2008). PMID: 18425107.
[102] Lanier, L. L. Up on the tightrope: natural killer cell activation and inhibition.
Nat. Immunol. 9, 495–502 (2008). PMID: 18425106 PMCID: PMC2669298.
[103] Bryceson, Y. T., Ljunggren, H.-G. & Long, E. O. Minimal requirement for
induction of natural cytotoxicity and intersection of activation signals by inhibitory receptors. Blood 114, 2657–2666 (2009). PMID: 19628705 PMCID:
PMC2756125.
[104] Bryceson, Y. T., March, M. E., Ljunggren, H.-G. & Long, E. O. Synergy
among receptors on resting NK cells for the activation of natural cytotoxicity
and cytokine secretion. Blood 107, 159–166 (2006). PMID: 16150947 PMCID:
PMC1895346.
[105] Mesecke, S., Urlaub, D., Busch, H., Eils, R. & Watzl, C. Integration of activating and inhibitory receptor signaling by regulated phosphorylation of vav1
in immune cells. Sci Signal 4, ra36 (2011). PMID: 21632469.
[106] Vivier, E., Nunes, J. A. & Vely, F. Natural killer cell signaling pathways.
Science 306, 1517–1519 (2004). PMID: 15567854.
[107] Fehniger, T. A. et al. CD56bright natural killer cells are present in human
lymph nodes and are activated by t cell-derived IL-2: a potential new link
between adaptive and innate immunity. Blood 101, 3052–3057 (2003). PMID:
12480696.
[108] Ferlazzo, G. et al. The abundant NK cells in human secondary lymphoid tissues
require activation to express killer cell ig-like receptors and become cytolytic.
J. Immunol. 172, 1455–1462 (2004). PMID: 14734722.
183
[109] Nagler, A., Lanier, L. L., Cwirla, S. & Phillips, J. H. Comparative studies
of human FcRIII-positive and negative natural killer cells. J. Immunol. 143,
3183–3191 (1989). PMID: 2530273.
[110] Nagler, A., Lanier, L. L. & Phillips, J. H. Constitutive expression of high affinity
interleukin 2 receptors on human CD16-natural killer cells in vivo. J. Exp. Med.
171, 1527–1533 (1990). PMID: 2139697 PMCID: PMC2187903.
[111] Nguyen, K. B. et al. Coordinated and distinct roles for IFN-alpha beta, IL-12,
and IL-15 regulation of NK cell responses to viral infection. J. Immunol. 169,
4279–4287 (2002). PMID: 12370359.
[112] Cooper, M. A. et al. Human natural killer cells: a unique innate immunoregulatory role for the CD56(bright) subset. Blood 97, 3146–3151 (2001). PMID:
11342442.
[113] Fehniger, T. A. et al. Differential cytokine and chemokine gene expression by
human NK cells following activation with IL-18 or IL-15 in combination with IL12: implications for the innate immune response. J. Immunol. 162, 4511–4520
(1999). PMID: 10201989.
[114] Gosselin, J., TomoIu, A., Gallo, R. C. & Flamand, L. Interleukin-15 as an
activator of natural killer cell-mediated antiviral response. Blood 94, 4210–
4219 (1999). PMID: 10590066.
[115] Marcenaro, E. et al. IL-12 or IL-4 prime human NK cells to mediate functionally
divergent interactions with dendritic cells or tumors. J. Immunol. 174, 3992–
3998 (2005). PMID: 15778356.
[116] Ushio, S. et al. Cloning of the cDNA for human IFN-gamma-inducing factor, expression in escherichia coli, and studies on the biologic activities of the
protein. J. Immunol. 156, 4274–4279 (1996). PMID: 8666798.
[117] Andrews, D. M. et al. Innate immunity defines the capacity of antiviral t cells
to limit persistent infection. J. Exp. Med. 207, 1333–1343 (2010). PMID:
20513749 PMCID: PMC2882831.
[118] Ferlazzo, G. et al. Human dendritic cells activate resting natural killer (NK)
cells and are recognized via the NKp30 receptor by activated NK cells. J. Exp.
Med. 195, 343–351 (2002). PMID: 11828009 PMCID: PMC2193591.
[119] Piccioli, D., Sbrana, S., Melandri, E. & Valiante, N. M. Contact-dependent
stimulation and inhibition of dendritic cells by natural killer cells. J. Exp. Med.
195, 335–341 (2002). PMID: 11828008 PMCID: PMC2193592.
[120] Lang, P. A. et al. Natural killer cell activation enhances immune pathology and
promotes chronic infection by limiting CD8+ t-cell immunity. Proc. Natl. Acad.
Sci. U.S.A. 109, 1210–1215 (2012). PMID: 22167808 PMCID: PMC3268324.
184
[121] Waggoner, S. N., Cornberg, M., Selin, L. K. & Welsh, R. M. Natural killer
cells act as rheostats modulating antiviral t cells. Nature 481, 394–398 (2012).
PMID: 22101430 PMCID: PMC3539796.
[122] Jacobs, R. et al. CD56bright cells differ in their KIR repertoire and cytotoxic
features from CD56dim NK cells. Eur. J. Immunol. 31, 3121–3127 (2001).
PMID: 11592089.
[123] Juelke, K. et al. CD62L expression identifies a unique subset of polyfunctional
CD56dim NK cells. Blood 116, 1299–1307 (2010). PMID: 20505160.
[124] Yu, J. et al. CD94 surface density identifies a functional intermediary between
the CD56bright and CD56dim human NK-cell subsets. Blood 115, 274–281
(2010). PMID: 19897577 PMCID: PMC2808153.
[125] Wulfing, C., Purtic, B., Klem, J. & Schatzle, J. D. Stepwise cytoskeletal polarization as a series of checkpoints in innate but not adaptive cytolytic killing.
Proc. Natl. Acad. Sci. U.S.A. 100, 7767–7772 (2003). PMID: 12802007 PMCID:
PMC164662.
[126] Ogunniyi, A. O., Story, C. M., Papa, E., Guillen, E. & Love, J. C. Screening
individual hybridomas by microengraving to discover monoclonal antibodies.
Nat Protoc 4, 767–782 (2009). PMID: 19528952.
[127] Tomescu, C., Chehimi, J., Maino, V. C. & Montaner, L. J. Retention of viability,
cytotoxicity, and response to IL-2, IL-15, or IFN-alpha by human NK cells after
CD107a degranulation. J. Leukoc. Biol. 85, 871–876 (2009). PMID: 19237639
PMCID: PMC2669410.
[128] Taub, D. D., Sayers, T. J., Carter, C. R. & Ortaldo, J. R. Alpha and beta
chemokines induce NK cell migration and enhance NK-mediated cytolysis. J.
Immunol. 155, 3877–3888 (1995). PMID: 7561094.
[129] Schroder, K., Hertzog, P. J., Ravasi, T. & Hume, D. A. Interferon-gamma: an
overview of signals, mechanisms and functions. J. Leukoc. Biol. 75, 163–189
(2004). PMID: 14525967.
[130] Orange, J. S. Human natural killer cell deficiencies and susceptibility to infection. Microbes Infect. 4, 1545–1558 (2002). PMID: 12505527.
[131] Spencer, S. L., Gaudet, S., Albeck, J. G., Burke, J. M. & Sorger, P. K. Nongenetic origins of cell-to-cell variability in TRAIL-induced apoptosis. Nature
459, 428–432 (2009). PMID: 19363473 PMCID: PMC2858974.
[132] Trapani, J. A. & Smyth, M. J. Functional significance of the perforin/granzyme
cell death pathway. Nat. Rev. Immunol. 2, 735–747 (2002). PMID: 12360212.
185
[133] Lehmann, C., Zeis, M. & Uharek, L. Activation of natural killer cells with
interleukin 2 (IL-2) and IL-12 increases perforin binding and subsequent lysis
of tumour cells. Br. J. Haematol. 114, 660–665 (2001). PMID: 11552995.
[134] Landowski, T. H. et al. Myeloma cells selected for resistance to CD95-mediated
apoptosis are not cross-resistant to cytotoxic drugs: evidence for independent
mechanisms of caspase activation. Blood 94, 265–274 (1999). PMID: 10381522.
[135] Montel, A. H., Bochan, M. R., Hobbs, J. A., Lynch, D. H. & Brahmi, Z. Fas
involvement in cytotoxicity mediated by human NK cells. Cell. Immunol. 166,
236–246 (1995). PMID: 7497525.
[136] Munker, R. et al. Expression of CD95(FAS) by gene transfer does not sensitize
k562 to fas-killing. Hematol Cell Ther 39, 75–78 (1997). PMID: 9168303.
[137] Hietakangas, V. et al. Erythroid differentiation sensitizes k562 leukemia cells
to TRAIL-induced apoptosis by downregulation of c-FLIP. Mol. Cell. Biol. 23,
1278–1291 (2003). PMID: 12556488 PMCID: PMC141136.
[138] Chauveau, A., Aucher, A., Eissmann, P., Vivier, E. & Davis, D. M. Membrane
nanotubes facilitate long-distance interactions between natural killer cells and
target cells. Proc. Natl. Acad. Sci. U.S.A. 107, 5545–5550 (2010). PMID:
20212116 PMCID: PMC2851811.
[139] Reefman, E. et al. Cytokine secretion is distinct from secretion of cytotoxic
granules in NK cells. J. Immunol. 184, 4852–4862 (2010). PMID: 20368273.
[140] Kurago, Z. B., Lutz, C. T., Smith, K. D. & Colonna, M. NK cell natural
cytotoxicity and IFN-gamma production are not always coordinately regulated:
engagement of DX9 KIR+ NK cells by HLA-B7 variants and target cells. J.
Immunol. 160, 1573–1580 (1998). PMID: 9469412.
[141] Hesslein, D. G. T., Takaki, R., Hermiston, M. L., Weiss, A. & Lanier, L. L.
Dysregulation of signaling pathways in CD45-deficient NK cells leads to differentially regulated cytotoxicity and cytokine production. Proc. Natl. Acad. Sci.
U.S.A. 103, 7012–7017 (2006). PMID: 16627620 PMCID: PMC1459010.
[142] Billadeau, D. D., Upshaw, J. L., Schoon, R. A., Dick, C. J. & Leibson, P. J.
NKG2D-DAP10 triggers human NK cell-mediated killing via a syk-independent
regulatory pathway. Nat. Immunol. 4, 557–564 (2003). PMID: 12740575.
[143] Egen, J. G. et al. Intravital imaging reveals limited antigen presentation and
t cell effector function in mycobacterial granulomas. Immunity 34, 807–819
(2011). PMID: 21596592 PMCID: PMC3164316.
[144] Davis, D. M. Mechanisms and functions for the duration of intercellular contacts
made by lymphocytes. Nat. Rev. Immunol. 9, 543–555 (2009). PMID: 19609264.
186
[145] Sips, M. et al. Altered distribution of mucosal NK cells during HIV infection.
Mucosal Immunol 5, 30–40 (2012). PMID: 21993602 PMCID: PMC3740353.
[146] Bonorino, P. et al. Fine characterization of intrahepatic NK cells expressing
natural killer receptors in chronic hepatitis b and c. J. Hepatol. 51, 458–467
(2009). PMID: 19596474.
[147] Nimmerjahn, F. & Ravetch, J. V. Fcgamma receptors as regulators of immune
responses. Nat. Rev. Immunol. 8, 34–47 (2008). PMID: 18064051.
[148] Lanier, L. L., Le, A. M., Civin, C. I., Loken, M. R. & Phillips, J. H. The
relationship of CD16 (leu-11) and leu-19 (NKH-1) antigen expression on human
peripheral blood NK cells and cytotoxic t lymphocytes. J. Immunol. 136, 4480–
4486 (1986). PMID: 3086432.
[149] Ravetch, J. V. & Perussia, B. Alternative membrane forms of fc gamma
RIII(CD16) on human natural killer cells and neutrophils. cell type-specific
expression of two genes that differ in single nucleotide substitutions. J. Exp.
Med. 170, 481–497 (1989). PMID: 2526846 PMCID: PMC2189395.
[150] O’Shea, J. J., Weissman, A. M., Kennedy, I. C. & Ortaldo, J. R. Engagement of
the natural killer cell IgG fc receptor results in tyrosine phosphorylation of the
zeta chain. Proc. Natl. Acad. Sci. U.S.A. 88, 350–354 (1991). PMID: 1703295
PMCID: PMC50808.
[151] Lanier, L. L. Natural killer cell receptor signaling. Curr. Opin. Immunol. 15,
308–314 (2003). PMID: 12787756.
[152] Harrison, D., Phillips, J. H. & Lanier, L. L. Involvement of a metalloprotease in
spontaneous and phorbol ester-induced release of natural killer cell-associated fc
gamma RIII (CD16-II). J. Immunol. 147, 3459–3465 (1991). PMID: 1834741.
[153] Romee, R. et al. NK cell CD16 surface expression and function is regulated by
a disintegrin and metalloprotease-17 (ADAM17). Blood 121, 3599–3608 (2013).
PMID: 23487023 PMCID: PMC3643761.
[154] Peruzzi, G. et al. Membrane-type 6 matrix metalloproteinase regulates the
activation-induced downmodulation of CD16 in human primary NK cells. J.
Immunol. 191, 1883–1894 (2013). PMID: 23851692 PMCID: PMC3745217.
[155] Grzywacz, B., Kataria, N. & Verneris, M. R.
CD56dimCD16+
NK cells downregulate CD16 following target cell induced activation
of matrix metalloproteinases.
Leukemia 21, 356–359 (2007).
URL
http://www.nature.com/leu/journal/v21/n2/full/2404499a.html.
[156] Liu, Q. et al. Matrix metalloprotease inhibitors restore impaired NK cellmediated antibody-dependent cellular cytotoxicity in human immunodeficiency
virus type 1 infection. J. Virol. 83, 8705–8712 (2009). PMID: 19553339 PMCID:
PMC2738177.
187
[157] Lajoie, L. et al. ADAM17-mediated shedding of FcRIIIA on human NK cells:
identification of the cleavage site and relationship with activation. J. Immunol.
192, 741–751 (2014). PMID: 24337742.
[158] Cecchetti, S., Spadaro, F., Lugini, L., Podo, F. & Ramoni, C. Functional role of
phosphatidylcholine-specific phospholipase c in regulating CD16 membrane expression in natural killer cells. Eur. J. Immunol. 37, 2912–2922 (2007). PMID:
17899539.
[159] Gierahn, T. M., Loginov, D. & Love, J. C. Crossword: a fully automated
algorithm for the segmentation and quality control of protein microarray images.
J. Proteome Res. 13, 362–371 (2014). PMID: 24417579 PMCID: PMC3979532.
[160] Sinclair, L. V. et al. Phosphatidylinositol-3-OH kinase and nutrient-sensing
mTOR pathways control t lymphocyte trafficking. Nat. Immunol. 9, 513–521
(2008). PMID: 18391955 PMCID: PMC2857321.
[161] Ogasawara, K. et al. Impairment of NK cell function by NKG2D modulation
in NOD mice. Immunity 18, 41–51 (2003). PMID: 12530974.
[162] Khan, K. D. et al. A phase 2 study of rituximab in combination with recombinant interleukin-2 for rituximab-refractory indolent non-hodgkin’s lymphoma.
Clin. Cancer Res. 12, 7046–7053 (2006). PMID: 17145827.
[163] Seals, D. F. & Courtneidge, S. A. The ADAMs family of metalloproteases:
multidomain proteins with multiple functions. Genes Dev. 17, 7–30 (2003).
PMID: 12514095.
[164] Velasco, G. et al. Human MT6-matrix metalloproteinase: identification, progelatinase a activation, and expression in brain tumors. Cancer Res. 60, 877–882
(2000). PMID: 10706098.
[165] Assarsson, E. et al. NK cells stimulate proliferation of t and NK cells through
2B4/CD48 interactions. J. Immunol. 173, 174–180 (2004). PMID: 15210772.
[166] Lee, K.-M. et al. Requirement of homotypic NK-cell interactions through
2B4(CD244)/CD48 in the generation of NK effector functions. Blood 107,
3181–3188 (2006). PMID: 15905190 PMCID: PMC1895752.
[167] Condon, T. P. et al. ADAM17 but not ADAM10 mediates tumor necrosis factoralpha and l-selectin shedding from leukocyte membranes. Antisense Nucleic
Acid Drug Dev. 11, 107–116 (2001). PMID: 11334139.
[168] Elliott, J. M. & Yokoyama, W. M. Unifying concepts of MHC-dependent natural
killer cell education. Trends Immunol. 32, 364–372 (2011). PMID: 21752715
PMCID: PMC3151350.
188
[169] Fauriat, C. et al. Estimation of the size of the alloreactive NK cell repertoire:
studies in individuals homozygous for the group a KIR haplotype. J. Immunol.
181, 6010–6019 (2008). PMID: 18941190.
[170] Freud, A. G. & Caligiuri, M. A. Human natural killer cell development. Immunol. Rev. 214, 56–72 (2006). PMID: 17100876.
[171] Borrego, F., Masilamani, M., Marusina, A. I., Tang, X. & Coligan, J. E. The
CD94/NKG2 family of receptors: from molecules and cells to clinical relevance.
Immunol. Res. 35, 263–278 (2006). PMID: 17172651.
[172] Lee, N. et al. HLA-E is a major ligand for the natural killer inhibitory receptor
CD94/NKG2A. Proc. Natl. Acad. Sci. U.S.A. 95, 5199–5204 (1998). PMID:
9560253 PMCID: PMC20238.
[173] Kabat, J., Borrego, F., Brooks, A. & Coligan, J. E. Role that each NKG2A
immunoreceptor tyrosine-based inhibitory motif plays in mediating the human
CD94/NKG2A inhibitory signal. J. Immunol. 169, 1948–1958 (2002). PMID:
12165520.
[174] Bashirova, A. A., Martin, M. P., McVicar, D. W. & Carrington, M. The killer
immunoglobulin-like receptor gene cluster: tuning the genome for defense. Annu
Rev Genomics Hum Genet 7, 277–300 (2006). PMID: 16824023.
[175] Uhrberg, M. et al. Human diversity in killer cell inhibitory receptor genes.
Immunity 7, 753–763 (1997). PMID: 9430221.
[176] Chan, H.-W. et al. DNA methylation maintains allele-specific KIR gene expression in human natural killer cells. J. Exp. Med. 197, 245–255 (2003). PMID:
12538663 PMCID: PMC2193817.
[177] Chan, H.-W., Miller, J. S., Moore, M. B. & Lutz, C. T. Epigenetic control of
highly homologous killer ig-like receptor gene alleles. J. Immunol. 175, 5966–
5974 (2005). PMID: 16237090.
[178] Valiante, N. M. et al. Functionally and structurally distinct NK cell receptor
repertoires in the peripheral blood of two human donors. Immunity 7, 739–751
(1997). PMID: 9430220.
[179] Yokoyama, W. M. & Kim, S. How do natural killer cells find self to achieve
tolerance? Immunity 24, 249–257 (2006). PMID: 16546094.
[180] Kim, S. et al. Licensing of natural killer cells by host major histocompatibility
complex class i molecules. Nature 436, 709–713 (2005). PMID: 16079848.
[181] Yu, J. et al. Hierarchy of the human natural killer cell response is determined by
class and quantity of inhibitory receptors for self-HLA-B and HLA-C ligands.
J. Immunol. 179, 5977–5989 (2007). PMID: 17947671.
189
[182] Anfossi, N. et al. Human NK cell education by inhibitory receptors for MHC
class i. Immunity 25, 331–342 (2006). PMID: 16901727.
[183] Thomas, L. M., Peterson, M. E. & Long, E. O. Cutting edge: NK cell licensing
modulates adhesion to target cells. J. Immunol. 191, 3981–3985 (2013). PMID:
24038086 PMCID: PMC3795972.
[184] Parsons, M. S., Zipperlen, K., Gallant, M. & Grant, M.
Killer cell
immunoglobulin-like receptor 3DL1 licenses CD16-mediated effector functions
of natural killer cells. J. Leukoc. Biol. 88, 905–912 (2010). PMID: 20664023.
[185] Boulet, S. et al. HIV protective KIR3DL1 and HLA-B genotypes influence NK
cell function following stimulation with HLA-devoid cells. J. Immunol. 184,
2057–2064 (2010). PMID: 20061407.
[186] Kim, S. et al. HLA alleles determine differences in human natural killer cell
responsiveness and potency. Proc. Natl. Acad. Sci. U.S.A. 105, 3053–3058
(2008). PMID: 18287063 PMCID: PMC2268583.
[187] Kamya, P. et al. Receptor-ligand requirements for increased NK cell polyfunctional potential in slow progressors infected with HIV-1 coexpressing
KIR3DL1*h/*y and HLA-B*57. J. Virol. 85, 5949–5960 (2011). PMID:
21471235 PMCID: PMC3126301.
[188] Carrington, M., Martin, M. P. & van Bergen, J. KIR-HLA intercourse in
HIV disease. Trends Microbiol. 16, 620–627 (2008). PMID: 18976921 PMCID:
PMC3463869.
[189] Martin, M. P. et al. Innate partnership of HLA-B and KIR3DL1 subtypes
against HIV-1. Nat. Genet. 39, 733–740 (2007). PMID: 17496894.
[190] Guia, S. et al. Confinement of activating receptors at the plasma membrane
controls natural killer cell tolerance. Sci Signal 4, ra21 (2011). PMID: 21467299.
[191] Li, B. & Dewey, C. N. RSEM: accurate transcript quantification from RNA-Seq
data with or without a reference genome. BMC Bioinformatics 12, 323 (2011).
PMID: 21816040 PMCID: PMC3163565.
[192] Fujita, P. A. et al. The UCSC genome browser database: update 2011. Nucleic
Acids Res. 39, D876–882 (2011). PMID: 20959295 PMCID: PMC3242726.
[193] Hsu, F. et al. The UCSC known genes. Bioinformatics 22, 1036–1046 (2006).
PMID: 16500937.
[194] Anders, S. & Huber, W. Differential expression analysis for sequence count
data. Genome Biol. 11, R106 (2010). PMID: 20979621 PMCID: PMC3218662.
[195] Anders, S. et al. Count-based differential expression analysis of RNA sequencing data using r and bioconductor. Nat Protoc 8, 1765–1786 (2013). PMID:
23975260.
190
[196] Feng, J. et al. GFOLD: a generalized fold change for ranking differentially
expressed genes from RNA-seq data. Bioinformatics 28, 2782–2788 (2012).
PMID: 22923299.
[197] Dybkaer, K. et al. Genome wide transcriptional analysis of resting and IL2
activated human natural killer cells: gene expression signatures indicative of
novel molecular signaling pathways. BMC Genomics 8, 230 (2007). PMID:
17623099 PMCID: PMC1959522.
[198] Draghi, M. et al. Single-cell analysis of the human NK cell response to missing
self and its inhibition by HLA class i. Blood 105, 2028–2035 (2005). PMID:
15528315.
[199] Johnston, S. C., Dustin, M. L., Hibbs, M. L. & Springer, T. A. On the species
specificity of the interaction of LFA-1 with intercellular adhesion molecules. J.
Immunol. 145, 1181–1187 (1990). PMID: 2199576.
[200] Krner, C. & Altfeld, M. Role of KIR3DS1 in human diseases. Front Immunol
3, 326 (2012). PMID: 23125843 PMCID: PMC3485674.
[201] Pando, M. J., Gardiner, C. M., Gleimer, M., McQueen, K. L. & Parham, P.
The protein made from a common allele of KIR3DL1 (3DL1*004) is poorly
expressed at cell surfaces due to substitution at positions 86 in ig domain 0 and
182 in ig domain 1. J. Immunol. 171, 6640–6649 (2003). PMID: 14662867.
[202] Capittini, C. et al. Possible KIR-driven genetic pressure on the genesis and
maintenance of specific HLA-A,B haplotypes as functional genetic blocks. Genes
Immun. 13, 452–457 (2012). PMID: 22573115.
[203] Foley, B. A. et al. The reactivity of bw4+ HLA-B and HLA-A alleles with
KIR3DL1: implications for patient and donor suitability for haploidentical stem
cell transplantations. Blood 112, 435–443 (2008). PMID: 18385451.
[204] Peters, V. A., Joesting, J. J. & Freund, G. G. IL-1 receptor 2 (IL-1R2) and
its role in immune regulation. Brain Behav. Immun. 32, 1–8 (2013). PMID:
23195532 PMCID: PMC3610842.
[205] Mohan, S. K. & Yu, C. The IL1alpha-S100A13 heterotetrameric complex structure: a component in the non-classical pathway for interleukin 1alpha secretion. J. Biol. Chem. 286, 14608–14617 (2011). PMID: 21270123 PMCID:
PMC3077658.
[206] Long, E. O., Kim, H. S., Liu, D., Peterson, M. E. & Rajagopalan, S. Controlling natural killer cell responses: integration of signals for activation and
inhibition. Annu. Rev. Immunol. 31, 227–258 (2013). PMID: 23516982 PMCID: PMC3868343.
191
[207] Yamanaka, Y. J. et al. Cellular barcodes for efficiently profiling single-cell
secretory responses by microengraving. Anal. Chem. 84, 10531–10536 (2012).
PMID: 23205933 PMCID: PMC3691955.
[208] O’Shea, J. J., Ma, A. & Lipsky, P. Cytokines and autoimmunity. Nat. Rev.
Immunol. 2, 37–45 (2002). PMID: 11905836.
[209] Choi, J., Love, K. R., Gong, Y., Gierahn, T. M. & Love, J. C. Immunohybridization chain reaction for enhancing detection of individual cytokinesecreting human peripheral mononuclear cells. Anal. Chem. 83, 6890–6895
(2011). PMID: 21812465 PMCID: PMC3175753.
[210] Song, Q. et al. On-chip activation and subsequent detection of individual
antigen-specific t cells. Anal. Chem. 82, 473–477 (2010). PMID: 20000848
PMCID: PMC2828941.
[211] Krutzik, P. O., Clutter, M. R., Trejo, A. & Nolan, G. P.
Fluorescent cell barcoding for multiplex flow cytometry.
In Current
Protocols in Cytometry (John Wiley & Sons, Inc., 2001).
URL
http://onlinelibrary.wiley.com/doi/10.1002/0471142956.cy0631s55/abstract.
[212] Krutzik, P. O. & Nolan, G. P. Fluorescent cell barcoding in flow cytometry
allows high-throughput drug screening and signaling profiling. Nat. Methods 3,
361–368 (2006). PMID: 16628206.
[213] Mattheakis, L. C. et al. Optical coding of mammalian cells using semiconductor
quantum dots. Anal. Biochem. 327, 200–208 (2004). PMID: 15051536.
[214] Kulesa, P. M. et al. Multispectral fingerprinting for improved in vivo cell dynamics analysis. BMC Dev. Biol. 10, 101 (2010). PMID: 20868502 PMCID:
PMC2954993.
[215] Gan, W. B., Grutzendler, J., Wong, W. T., Wong, R. O. & Lichtman, J. W.
Multicolor ”DiOlistic” labeling of the nervous system using lipophilic dye combinations. Neuron 27, 219–225 (2000). PMID: 10985343.
[216] Jaiswal, J. K., Mattoussi, H., Mauro, J. M. & Simon, S. M. Long-term multiple
color imaging of live cells using quantum dot bioconjugates. Nat. Biotechnol.
21, 47–51 (2003). PMID: 12459736.
[217] Nomura, L. E., Walker, J. M. & Maecker, H. T. Optimization of whole blood
antigen-specific cytokine assays for CD4(+) t cells. Cytometry 40, 60–68 (2000).
PMID: 10754518.
[218] Horton, H. et al. Optimization and validation of an 8-color intracellular cytokine staining (ICS) assay to quantify antigen-specific t cells induced by vaccination. J. Immunol. Methods 323, 39–54 (2007). PMID: 17451739 PMCID:
PMC2683732.
192
[219] Boulet, S. et al. A dual color ELISPOT method for the simultaneous detection
of IL-2 and IFN-gamma HIV-specific immune responses. J. Immunol. Methods
320, 18–29 (2007). PMID: 17222422 PMCID: PMC3627477.
[220] Podtschaske, M. et al. Digital NFATc2 activation per cell transforms graded t
cell receptor activation into an all-or-none IL-2 expression. PLoS ONE 2, e935
(2007). PMID: 17895976 PMCID: PMC1978524.
[221] Grob, M. et al. Altered intracellular expression of the chemokines MIP-1alpha,
MIP-1beta and IL-8 by peripheral blood CD4+ and CD8+ t cells in mild allergic
asthma. Allergy 58, 239–245 (2003). PMID: 12653799.
[222] Longo, D. M. et al. Single-cell network profiling of peripheral blood mononuclear
cells from healthy donors reveals age- and race-associated differences in immune
signaling pathway activation. J. Immunol. 188, 1717–1725 (2012). PMID:
22246624 PMCID: PMC3517183.
[223] Querec, T. D. et al. Systems biology approach predicts immunogenicity of the
yellow fever vaccine in humans. Nat. Immunol. 10, 116–125 (2009). PMID:
19029902.
[224] Wagner, T. L. et al. Modulation of TH1 and TH2 cytokine production with the
immune response modifiers, r-848 and imiquimod. Cell. Immunol. 191, 10–19
(1999). PMID: 9918682.
[225] Muzio, M. et al. Differential expression and regulation of toll-like receptors
(TLR) in human leukocytes: selective expression of TLR3 in dendritic cells. J.
Immunol. 164, 5998–6004 (2000). PMID: 10820283.
[226] Zhu, J., Yamane, H. & Paul, W. E. Differentiation of effector CD4 t cell populations. Annu. Rev. Immunol. 28, 445–489 (2010). PMID: 20192806 PMCID:
PMC3502616.
[227] Gajewski, T. F., Joyce, J. & Fitch, F. W. Antiproliferative effect of IFN-gamma
in immune regulation. III. differential selection of TH1 and TH2 murine helper
t lymphocyte clones using recombinant IL-2 and recombinant IFN-gamma. J.
Immunol. 143, 15–22 (1989). PMID: 2525146.
[228] Gajewski, T. F., Lancki, D. W., Stack, R. & Fitch, F. W. ”Anergy” of TH0
helper t lymphocytes induces downregulation of TH1 characteristics and a transition to a TH2-like phenotype. J. Exp. Med. 179, 481–491 (1994). PMID:
8294862 PMCID: PMC2191376.
[229] Luchansky, M. S. & Bailey, R. C. Rapid, multiparameter profiling of cellular
secretion using silicon photonic microring resonator arrays. J. Am. Chem. Soc.
133, 20500–20506 (2011). PMID: 22040005 PMCID: PMC3241882.
193
[230] Murphy, E. et al. Reversibility of t helper 1 and 2 populations is lost after longterm stimulation. J. Exp. Med. 183, 901–913 (1996). PMID: 8642294 PMCID:
PMC2192360.
[231] Kohrt, H. E. et al. Ibrutinib antagonizes rituximab-dependent NK cellmediated cytotoxicity. Blood 123, 1957–1960 (2014). PMID: 24652965 PMCID:
PMC3962169.
[232] Miller, M. A. et al. Proteolytic activity matrix analysis (PrAMA) for simultaneous determination of multiple protease activities. Integr Biol (Camb) 3,
422–438 (2011). PMID: 21180771 PMCID: PMC3173501.
[233] Yao, X. et al. Functional analysis of single cells identifies a rare subset of
circulating tumor cells with malignant traits. Integr Biol (Camb) 6, 388–398
(2014). PMID: 24522233.
[234] Wang, Y. et al. ADAM17 cleaves CD16b (FcRIIIb) in human neutrophils.
Biochim. Biophys. Acta 1833, 680–685 (2013). PMID: 23228566 PMCID:
PMC3556181.
[235] Huse, M., Catherine Milanoski, S. & Abeyweera, T. P. Building tolerance
by dismantling synapses: inhibitory receptor signaling in natural killer cells.
Immunol. Rev. 251, 143–153 (2013). PMID: 23278746.
[236] Ronan, J. L., Story, C. M., Papa, E. & Love, J. C. Optimization of the surfaces
used to capture antibodies from single hybridomas reduces the time required for
microengraving. J. Immunol. Methods 340, 164–169 (2009). PMID: 19028499.
194