Purdue University
Cell-Based Assays: Innovations in Reagents, Technologies &
Screening:
“So what are high content assays anyway?”
J. Paul Robinson
SVM Endowed Professor of Cytomics
Professor of Biomedical Engineering
Purdue University, West Lafayette, IN
Lecture delivered May 5, 2008 in Boston
Please acknowledge any materials used from this presentation
Purdue University
My assay
is bigger than your assay…..
Big
Really Big
Bigger
Biggest
Really, Really, Big
Really, Really, Really, Big
Purdue University
Cell analysis technology state-of-the-art….?
•
•
•
•
•
•
•
Imaging
1930-40s• Cell cytochemistry and staining
1950s • Cell counting
• Cell sorting
1960s
• Cell detection
1970s • Cell separation/classification (MABs)
1980s • Polychromatic (multicolor) cytometry
Imaging
1990s • Automated imaging, cytomics,
2000s metabolomics
• ??...
• 2010s
Purdue University
So what is the difference between:
• High throughput
• High content
Can a high throughput assay also be high
content?
Does it have to be an “image”?
Purdue University
“High content screening (HCS) is an imaging
approach to cell-based assays that has had an
impact in the fields of neurobiology, signaling,
target identification and validation and in vitro
toxicology”
“Approaching High Content Screening and Analysis: Practical
Advice for Users”, S. Keefer and Joseph Zock
in High Content Screening. Edited by Steven Haney
Copyright 2008 John Wiley & Sons, Inc.
Purdue University
The basics of assay design..
Once we used test tubes to perform our
assays.
If we wanted to add more tests,
we just added more tubes……..
Purdue University
Then we moved to 96 well plates…
and standardization….
= 96 tubes
Then we stacked
many 96 well plates
Purdue University
Then we discovered imaging….
Then multicolor
systems…..
and instruments that could image lots of cells
on lots of wells…
Purdue University
and the rot set in.…..
My system
can collect
a million
cells and 50
parameters..
MY system can collect
a gazillion cells and
a billion parameters…
(so there)
Purdue University
Instruments got bigger and $$$$
•
•
•
•
Expectations were high
Costs were high
There were no standards
Many used proprietary data collection or
analysis
• But productivity and Return on Investment did
not match
• Followed by the High Content Recession
Purdue University
and the rot set in….
and incidentally….
Purdue University
So several questions arose…
• Is it better to collect more variables/parameters
on fewer cells..
• Or less variables/parameters and a lot of cells….
• What kinds of analysis do you need and how do
you efficiently achieve an analytical solution?
Note Definition
Variable – something that is actually measured
Parameter – a derived value from a set of variables
Purdue University
and most people answer …
• We want a lot of parameters and a
lot of cells…..
Purdue University
Images taken from: High Content Screening. Edited by Steven Haney
Copyright 2008 John Wiley & Sons, Inc.
Purdue University
Outline
This presentation will discuss the developments in screening tools with
a view to showing how the technologies have matured into highly
advanced approaches driving systems analysis……
•
•
•
•
•
Overview of historical developments in cell analysis
Outline systems available for identifying properties of single cells
Define how populations have been classified traditionally
Identify emerging tools for single cell analysis
Illustrate emerging applications that advance opportunities for using
modeling approaches to data analysis
Purdue University
First Ultraviolet Imaging - A. Kohler 1904
275 nm
280 nm
Salamander maculosa larva epidermal cells - 1300 X
A. Kohler, Mikrophotographische Untersuchungen mit ultraviolettem Licht, Z. Wiss. Mikroskopie 21, 1904
Slide kindly supplied by Elena Holden, Compucyte
Purdue University
UV Measurements of DNA and Cytoplasm - T. Caspersson 1936
Ultraviolet absorption measurements of
a grasshopper metaphase chromosome
Densitometer traces across
a region of the chromosome
Extinction values for chromosome and
cytoplasm plotted against wavelength
Cytoplasmic Chromosomal Background
absorption
absorption
signal
Uber den chemischen Aufbau der Strukturen des Zellkernes, Skand. Arch. Physiol. 73, 1936
Slide kindly supplied by Elena Holden, Compucyte
Progression of Cell Analysis
Purdue University
Wallace H. Coulter’s only
Scientific publication
Cell Analysis – Circa 1956
Purdue University
Kamentsky - Automated Imaging
Dr. Kamentsky
LA Kamentsky & CN Liu, Computer-automated design of multifont print recognition
logic, IBM J. Research & Development 7, 1963
Slide kindly supplied by Elena Holden, Compucyte
Purdue University
Johan Sebastiaan (Bas) Ploem
Epi-illumination
Liver tissue. Nuclei stained with Feulgenpararosaniline for DNA. Epi-illumination with
narrow band green light (546nm) and a
dichroic beam splitter for reflecting green
light. Probably the first example of
microscope excitation with green light (Ploem,
1965). Note large image contrast
Image from wikimedia.org
Leitz PLOEMOPAK illuminator
An epi-illumination cube used in
fluorescence microscopy. Ploem's
vertical illuminator bears his name
and is commonly used today.
Image from micro.magnet.fsu.edu
For his contributions to the practice of microscopy, Ploem has received various honors. He
was elected as a fellow of the Papanicolaou Cancer Research Institute in 1977 and was a
recipient of the C. E. Alken Foundation award in 1982. He is also a member of the Society
of Analytical Cytology, the Dutch Society of Cytology, the International Academy of
Cytology and the Royal Microscopical Society, for which he served as president in 1986. In
1993, he became an Honorary Fellow of the International Society for Analytical Cytology
Purdue University
Progression of Cell Analysis
Equivalent list mode storage of about
200 cells with 1 parameter data
Coulter counter
400 word memory – Fulwyler’s 1965 sorter
Over 50 years of technology development this has led to ………
Purdue University
The progression of cell detection
It’s a cell
It’s a small cell or it’s a big cell
It’s a small cell or it’s a big cell
and it has a DNA content of this
It’s a small cell or it’s a big cell
and it has a certain DNA content
and we can identify this cell as a specific phenotype
Purdue University
It’s a small cell or it’s a big cell
and we can identify this cell as a specific phenotype
within a subset of cells
It’s a small cell or it’s a big cell
and we can identify this cell as a specific phenotype
within a subset of cells – but more colors
It’s a small cell or it’s a big cell
and we can identify each of these phenotypes within a
heterogeneous population simultaneously
It’s a small cell or it’s a big cell
and we can identify each of these phenotypes within a
heterogeneous population simultaneously
We can also evaluation cell function with several
simultaneous parameters. We can label cells with different
intensities of dyes to separate them into groups
Purdue University
Add imaging…
Now we can identify location of target
molecules, We can evaluate the shape,
texture, a variety of complex features and still
measure some different “colors”.
Purdue University
Data processing, analysis & presentation
•
•
•
•
Initial data collection (variables)
Organization of data sets
Reduction of data
Advanced processing
– parameter derivation
– classification algorithms
– presentation of useful data
Purdue University
Image Processing Example
These are all the exact same dataset!!!
Ger van den Engh using different format
Purdue University
Purdue University
What you see is not always what you think it
is…..
Note:
Move not
available
on web
version
Purdue University
E
>2000 scatter patterns from
cultures of 108 Listeria strains
were measured and analyzed
69 – L. monocytogenes
C
16 - L. innocua
12 - L. ivanovii
5 - L. seeligeri
3 - L. welshimeri
3 - L. grayi
B
D
A
Schematic representation of the laser scatterometer used to perform analysis of
bacterial colonies. A – 635-nm diode laser, B – Petri dish containing bacterial
colonies, C – CCD camera, D – Petri-dish holder, and E – detection screen.
29
Purdue University
Every organism has a very specific scatter pattern
L. monocytogenes ATCC19113
L. seeligeri LA 15
L. innocua F4248
L. welshimeri ATCC35897
L. ivanovii ATCC19119
L. grayi LM37
Purdue University Patent Pending
Purdue University
Listeria scatter patterns
L. welshimeri ATCC35897
L. innocua V58
L. ivanovi ATCC19119
L. ivanovi SE98
L. monocytogenes ATCC19113
L. monocytogenes V7
Purdue University Patent Pending
Purdue University
Image analysis using 2D radial Zernike polynomials
Frits Zernike
The Nobel Prize in Physics 1953
The Zernike polynomials are a set of
orthogonal polynomials that arise in the
expansion of a wavefront function for optical
systems with circular pupils. They were
introduced by F. Zernike in 1934: Zernike, F.
"Beugungstheorie des Schneidenverfahrens
und seiner verbesserten Form, der
Phasenkontrastmethode." Physica 1, 689-704,
1934.
Bartek Rajwa, Bulent Bayrakta, Padmapriya P. Banada, Karleigh
Huff, Euiwon Bae, E. Daniel Hirleman, Arun K. Bhunia, J. Paul
Robinson; Phenotypic analysis of bacterial colonies using laser
light scatter and pattern-recognition techniques.Proc. SPIE Vol.
6864, 68640S (Feb. 15, 2008)
Graphical representation of radial Zernike polynomials Zn,m in 2D (image size
128 x 128 pixels), and their magnitudes: A – real part Z10,6; B – imaginary part
Z10,6; C – magnitude Z10,6; D – real part Z13,5; E – imaginary part Z13,5; F –
magnitude Z13,5. The larger the n-|m| difference, the more oscillations are
present in the shape. Features used in this study are the magnitudes of Zernike
polynomials. One may note that the values of the magnitudes do not change
when arbitrary rotations are applied.
Discovering the Unknown: Detection of Emerging Pathogens Using a Label-Free Light-Scattering
System; Author:Bartek Rajwa, M. Murat Dundar, Ferit Akova, Amanda Bettasso, Valery Patsekin, E.
Dan Hirleman, Arun K. Bhunia, J. Paul Robinson; Publication: Cytometry 77A: 1103–1112, 2010
Purdue University Patent Pending
Nonpathogenic
Based on scatter patterns, we can identify everything we
have attempted so far. All of the organisms of interest have
been pathogens – mostly food borne in nature
Purdue University
E. coli K12
EPEC
Escherichia coli
Pattern I
E. coli O157:H7 01
EHEC
Pattern II
E. coli O157:H7 K6
E. coli O142:H6 E851171
E. coli O157:H7SEA 13A53
E. coli E2348169 O127:H6
E. coli O157:H7 EDL933
Pattern III
ETEC
E. coli O157:H7 505B
E. coli O157:H7 G5295
E. coli O25:K19:NM
E. coli O157:H7 K1
E. coli O157:H7 G458
E. coli O78:H11
Purdue University Patent Pending
Color map
visualizing PC
values
Purdue
University
Hierarchical clustering based on Zernike moment invariants
L. grayi
L. seeligeri
L. welshimeri
L. monocytogenes
L. ivanovii
L. seeligeri
Hierarchical clustering of bacterial scatter patterns. Symbols represent six different strains of Listeria belonging
to six species: ■ L. grayi LM37, □ L. seeligeri LA15, r L. welshimeri ATCC35897, ◊ L. monocytogenes ATCC19113,
+ L. innocua F4248, L. ivanovii V199. Numbers represent identified clusters of patterns. Note that identified
clusters coincide with the groups of colonies from different strains.
Purdue University Patent Pending
Purdue University
Technology advances
• High speed sorting
• Advanced Polychromatic analysis
• Hyperspectral (multispectral) Analysis
• HypercyteTM – High content Screening
• Multiparameter systems approach to
pathways and cell signaling
Purdue University
Advanced polychromatic cytometry
Hyperspectral cytometry
Hyperspectral cytometry
14-20 PMTs
40-50 filters
1 multichannel “PMT”
1 “filter”
Purdue University
Amnis cytometer
Images taken from Amnis publicity materials.
Purdue University
Multiplexing
6
36 combinations….but only 6 tubes
and this is just the start
4
3
Sample Number
5
2
1
1
2
3
4
5
6
Purdue University
You don’t have to physically sort cells to apply a systems
approach to cell function….
• Power of a systems approach to cell analysis is shown in work from
Gary Nolan’s Laboratory at Stanford
Mechanistic Insights from the Single Cell:
Inference Engines for Clinically Predictive Indicators
Slide kindly supplied by
Garry P. Nolan, Ph.D.
Stanford University
Dept. of Microbiology & Immunology
Purdue University
Larry Sklar & Bruce Edwards
High content screening using flow cytometry
New Mexico Molecular Libraries Screening Center
HT Flow
Cytometry?
This part is NOT high throughput
(~ 2 samples/min)
This part is high
throughput
50,000 cells/s
14 parameters/cell
Slide from Larry Sklar
Slide from Larry Sklar
HyperCyt
384 wells/10 min
1 ml/sample
Commercially
Available
Slide from Larry Sklar
HyperPlex =
HyperCyt and Luminex
Theoretical potential for 50 plex
in 1536 well format, 10 min
(20M per day per detector)
Selectivity
Color 1, 7 levels
20 bead set,
2 colors, 7 levels each
Color 2, 7 levels
Purdue University
Gary Nolan Lab
High Content Drug Screening using Flow Cytometry
Primary Cell Screening
4 natural products + 4 commercial inhibitors
Titrate 6 concentrations of each compound
Stimulate with IFNg, IL-4, IL-6, IL-7, IL-10, IL-15
Phospho
Flow
Analyze B cells, CD4+ and CD4- T cells,
CD11b-hi neutrophils, CD11b-int macrophages
Measure Stat1, Stat3, Stat5, Stat6 phosphorylation
Slide from Gary Nolan’s Lab
Purdue University
Advanced approaches to modeling based on single cell data - Nolan Lab
•
Question: can you predict a signaling network based on network
connectivity knowledge from single cell analysis?
Purdue University
Result…
• “..we correctly reverse-engineered and rapidly
inferred the basic structure of a classically
understood signaling network that connects a
number of key proteins in human T cell
signaling, a map built by classical biochemistry
and genetic analysis over the past 2 decades.”
Science 308: 527, 2005
Slide from Gary Nolan’s Lab
Purdue University
Summary and Conclusions
• Technologies such as flow cytometry are often assumed to have a
narrow phenotypic or cell cycle application
• New technologies are emerging creating even more detection
opportunities
• High throughput with high content sampling is now a reality
• With multiparameter detection you must have powerful analytic
capabilities
• Systems modeling approaches are clearly the next implementation in
cytometry
• Innovative assay design and software approaches have created a
new paradigm for single cell analysis
Purdue University
Note
• Some slides provided by colleagues with their
data have been deleted from this published
version of this presentation.
Purdue University
Acknowledgements
Staff
Thanks to colleagues who provided slides for this
presentation: Gary Nolan, Larry Sklar, Bruce
Edwards






Jennie Sturgis (Imaging)
Kathy Ragheb (Flow)
Cheryl Holdman (Flow)
Gretchen Lawler (CPC)
Steve Kelley (Network)
Hildred Rochon (C4L)
Senior Scientists: Bartek Rajwa, Yanan Jiang**
Postdocs: Valery Patsekin, Tytus Bernas, Sang Youp
Lee, Lova Rakotomalala
Graduate Students: Wamiq Ahmed, Bulent Bayraktar,
Silas Leavesley, Connie Snyder, Muru
Funding & Support Acknowledgement:
Venkatapathi, Jia Liu
Faculty: Kinam Park, V.J.Davisson, Arun Bhunia, Dan NIH, NSF, USDA, Purdue University
Corporate: Beckman-Coulter, Point-Source, Parker-Hannifin,
Hirleman
Polysciences, Bangs labs, MediaCybernetics, Q-Imaging, Amgen
Kodak Medical Systems, Crystalplex, Becton-Dickinson, eBioscience
Bindley Bioscience Center, Purdue University
PUCL, Purdue University