The Dynamics of Surface Detachment and Quorum Sensing
in Spatially Controlled Biofilm Colonies of Pseudomonas aeruginosa
by
Hongchul Jang
Bachelor of Engineering in Chemical and Biological Engineering
Korea University, South Korea, 2001
Master of Science in Chemical Engineering
Yale University, 2004
MASSAcHU
0F T EC
INS
NOGY
Master of Science in Mechanical Engineering
Massachusetts Institute of Technology, 2010
LIBRARIES
Master of Science in Civil and Environmental Engineering
Massachusetts Institute of Technology, 2010
Submitted to the Department of Civil and Environmental Engineering
in Partial Fulfillment of the Requirements for the Degree of
Doctor of Philosophy in the field of Civil and Environmental Engineering
at the
Massachusetts Institute of Technology
February 2014
C 2014 Massachusetts Institute of Technology
All rights reserved
Signature of Author: ......................
Departmen
Certified by
................
Associ te Profe
Civil and Environmental Engineering
January 15, 2014
.......................
Roman Stocker
of Civil and Environmental Engineering
Thesis Supervisor
A ccepted by: ...........................
CHe di M. Nepf
Chair, Departmental Committee for Graduate Students
E
The Dynamics of Surface Detachment and Quorum Sensing
in Spatially Controlled Biofilm Colonies of Pseudomonas aeruginosa
by
Hongchul Jang
Submitted to the Department of Civil and Environmental Engineering on January 15, 2014
in partial fulfillment of the requirements for the
Degree of Doctor of Philosophy in the field of Civil and Environmental Engineering
ABSTRACT
Biofilms represent a highly successful life strategy of bacteria in a very broad range of
environments and often have negative implications for industrial and clinical applications, as
their removal from surfaces and the prevention of biofouling in the first place represent
formidable and to date unmet challenges. At the same time, biofilms modulate important natural
processes, including nutrient cycling in rivers and streams and the clogging of porous materials.
Biofilm development is a dynamic process, dependent on a host of cellular and environmental
parameters that include, among others, hydrodynamic environment and the communication
among cells (QS).
Here we used microfluidics and micro-contact printing, paired with video-microscopy and
image analysis, to study several aspects relating to the temporal dynamics of bacterial biofilms.
Beyond reporting on specific findings from these experiments, we demonstrate how these
innovative technologies can aid in obtaining a new layer understanding on biofilm processes and
biofilm removal, thanks to unprecedented control over the biofilm's microenvironment.
In a first set of experiments (chapter 2), we cultured Pseudomonasaeruginosabiofilms in a
microfluidic channel for different times, after which we used the passage of an air plug as a
mechanical insult to drive detachment. We found that the adhesion properties of an early-stage
biofilm have a strong correlation with the time of growth and that biofilm detachment can occur
in a spatially heterogeneous manner characterized by a regular pattern of 'holes' in the original
biofilm. The resulting spatial distribution of bacteria correlates with the distribution of the extra
polymeric substance (EPS) matrix before the insult, indicating that the locations of the holes
correspond to where there was the least EPS. These results demonstrate that the detachment
mechanism is a competition between the shear force exerted by the external flow and the local
adhesion force of a given patch of biofilm, in turn governed by the local amount of EPS. This
mechanism, and the observed heterogeneity in the detachment, imply that even at rather high
shear rates, where the bulk of the biofilm is removed, local strongholds survive detachment, and
may represent seeds for the colony to reform.
In a second set of experiments (chapter 3), we examined the effects of patch size and
hydrodynamic environment on QS induction on spatially defined patches of Pseudomonas
aeruginosabiofilm. We found that the smaller biofilm patches start QS earlier than those in the
larger patches. However at later times the proportion of auto-induced bacteria (normalized by the
surface area covered by the cells) is higher in the larger patches. We expanded on these findings
by investigating the contribution of flow to QS induction on the patches. The effect of ambient
fluid flow was to accelerate the induction of quorum sensing compared to static conditions at
moderate flow rates, due to the increase in the convective supply of nutrients and to quench
quorum sensing at high flow rates, due to the autoinducer signal being washed out by flow.
These findings establish microfluidics as a new tool in the study of biofilms, which enables both
accurate control over microenvironmental conditions and direct observation of the dynamics of
biofilm formation and disruption.
Chapter 4 deepens the exploration of the role of micro-spatial heterogeneity on microbial
processes by presenting a numerical model of how heterogeneity and microbial behavior
(chemotaxis) affect trophic in a microbial food web. Results show that the intensity of the trophic
transfer strong depends on the motile behavior of the different trophic levels: trophic transfer is
enhanced when directional motility towards resource patches outweighs the random component
of motility inherent in all microbial locomotion.
Finally, in the Appendices, we demonstrate how the methods developed in this thesis can help
in the assessment of the antifouling capabilities of new-generation surfaces, designed to prevent
fouling and in the assessment of the cell adhesion on surfaces, fabricated with an arrangement of
spatially localized hydrophobic patterns.
In summary, this thesis demonstrates that use of new micro-technology and associated
mathematical modeling enables new insights into the colonial life form of microorganisms and
may provide impetus for new approaches to prevent biofouling on surfaces or remove biofilms
from surfaces.
Thesis Supervisor: Roman Stocker
Title: Associate Professor of Civil and Environmental Engineering
Acknowledgements
This work would not have been possible without help and support of Professor Roman Stocker,
whom I consider one of the best ever advisors. I would like to extend my sincere thanks and
profound sense of gratitude to him for his intellectual guidance and constructive criticism. He is
a fantastic mentor and a constant source of inspiration. It has been my great honor to work under
his guidance.
Professor Roger Kamm, I always remember his generosity and kindness. I would like to thank
him for his helpful suggestions and guidance.
Professor Janelle Thompson is appreciated for being in my committee and for her advice and
guidance.
I have benefited greatly from my association with all past and present members of the Stocker
lab, in particular Dr. Roberto Rusconi, Dr. Steve Smriga, Dr. Jeffrey Guasto, Kwangmin Son,
Marcos, Tanvir Ahmed, Dr. Mack Durham, Michael Barry, and Dr. Justin Seymour.
I also would like to express many thanks to my friends and colleagues for making my life at MIT
special and enjoyable.
Finally, I owe my deepest gratitude to my parents and family who have provided me with great
opportunities, everlasting support, and big encouragement.
Most importantly, I would like to give my special thanks and love to my wife Sujin Kim, my
daughter, Ashley Chelyn Jang and my son, Aiden Joonsuh Jang for all the love, support and
encouragement. They have been a source of inspiration and this work is a tribute to them.
HongchulJang
January2014
4
Table of Contents
Chapter 1
Introduction and Background........................................17
1.1
Microbial Biofilms........................................................................17
1.2
Biofilm Detachment.........................................................................21
1.3
The Role of Cell Signaling in Biofilm Dynamics...................................23
1.4
Microfabrication..........................................................................27
1.5
pContact Printing for Biofilm Research................................................29
1.6
Motivation and Thesis Scope...........................................................35
Chapter 2
Disruption of biofilm patches by an air plug reveals microscale
heterogeneity ............................................................................
2.1
Introduction.............................................................................
2.2
Results and Discussion..................................................................51
2.3
Conclusion.................................................................................66
2.4
Materials and Methods.....................................................................67
Chapter 3
48
49
The Effect of Patch Size and Fluid Flow on Quorum Sensing in
Biofilms...................................................................................
3.1
Introduction...................................................................................74
3.2
Results and Discussion...................................................................77
5
74
3.3
Conclusion....................................................................................91
3.4
Materials and Methods..................................................................92
Chapter 4
Enhanced Trophic Transfer Rates in a Chemotactic Food
Chain.......................................................................................
4.1
Introduction...............................................................................
4.2
Methods.......................................................................................101
4.3
Results........................................................................................106
4.4
Discussion and Conclusions..............................................................123
98
APPENDIX A
Synergistic Effects of Chlorine and Antifouling Surfaces........................130
APPENDIX B
Spatial Control of Cell Attachment by M ucin Coatings..........................139
6
98
List of Figures
Figure 1-1 Conceptualization of biofilm structure, development and dynamic behaviors. Image
credit: P. Dirckx, Center for Biofilm Engineering, Montana State University, USA ................
18
Figure 1-2 LasI/LasR-RhlI/RhlR quorum sensing system in Pseudomonas aeruginosa.The Las!
protein produces the homonserine lactone signaling molecule N-(3-oxododecanoyl)-homoserine
lactone, autoinducer 1 (All, blue circles), and the RhlI protein synthesizes N-(butryl)homoserine lactone, autoinducer 2 (A12, red triangles). Activated LasR (with All) activates
transcription of rhlR, which means the las quorum sensing system controls RhlR at the
transcriptional level. All also controls the activity of RhlR at the post-translational level by
binding to RhlR, which decreases A12 binding to RhlR, therefore inhibiting expression of rhlA
and other genes. The Pseudomonas quinolone signal (PQS) is an additional regulatory link
between the Las and Rhl quorum sensing circuit. ..................................................................
24
Figure 1-3 (Left) Schematic of the microcontact printing (ptCP) technique used for depositing
hydrophobic patches onto glass substrates. (a) Wet inking is achieved by placing a patterned
PDMS stamp onto the OTS ink pad. (b) After inking, the stamp is left in a chemical hood for 2
min to let the solvent (hexane) evaporate. (c) By bringing the inked stamp in conformal contact
with the glass surface, the hydrophobic long-chain (C-18) silane is transferred to the glass
surface. Due to the patterned structure of the stamp, only the areas with protrusions contact the
glass surface, and the silane is area-selectively transferred according to the pattern of the stamp.
(d) The microfluidic channel is positioned on top of the patterned substrate. (Right) Fluorescent
7
images of FITC-labeled bovin serum albumin (BSA) absorbed onto the hydrophobic patches,
demonstrating the specificity of the pCP technique. The top row shows squares of 20 pm and 50
pm size. The bottom row shows lines of 100 pm thickness and the MIT logo........................
32
Figure 1-4 Phase-contrast images of P aeruginosa PA01 selectively attached on hydrophobic
patches of different shapes and sizes. The top row shows squares and the bottom row shows
vertical lanes of biofilm. Note the strong difference in the density of attached cells between the
patches and the regions of the surface outside the patches, demonstrating the specificity achieved
by this patterning process..........................................................................................................
34
Figure 2-1 An air plug creates a characteristic pattern of holes in a biofilm. (a) Schematic of the
microfluidic setup, showing the geometry of the microchannel and the experimental method.
Pseudomonasaeruginosabacteria preferentially attached to the hydrophobic patches. After that a
certain growth time (4 h, 8 h, or 12 h), a controlled air plug was injected in the channel at the
mean flow speed of 250 pm/s. (b) Residual biofilm after the passage of the air plug, showing the
semi-regular hole pattern for different patch sizes (from left to right: 400, 300, 200 and 100 pm).
The biofilm growth time was 8 h. Scale bar = 100 pm. ...........................................................
51
Figure 2-2 Schematic of the microcontact printing (ptCP) technique used for depositing .....
52
Figure 2-3 The effect of the air plug strongly depends on biofilm age. (a-f) Biofilms grown for
different times (4 h, 8 h, 12 h), shown before (a-c) and after (d-f) the passage of an air plug. The
air plug traveled from left to right. (g) Surface coverage (fraction of the surface of a patch
covered by bacteria) before and after the passage of an air plug, for biofilms of different age
(grayscale bars). The red curve shows the normalized change in surface coverage. (h) Free
8
surface distance between cells (grayscale bars) and fractal dimension of the cell distribution (red
symbols) measured before and after the passage of an air plug, for biofilms of different age..... 54
Figure 2-4 Paeruginosaadsorption on the OTS step-wise gradient. The contact-printing time of
OTS is gradually increased from inner out. The color plot (top figure) was generated using the
value of fluorescence intensity over the distance (bottom figure). The higher intensity indicates
the higher adsorption density of Paeruginosa.........................................................................
55
Figure 2-5 Porosity and equivalent hole radius on the size of patches. The top four panels
correspond to the size of patches. .............................................................................................
56
Figure 2-6 AFM characterization of a ruptured P aeruginosa biofilm surface. (a) The residual
biofilm is shown in yellow, the region from which bacteria have detached is in brown. (b) The
height profile of the surfaces along the white line in panel (a). The red and green arrowheads
correspond to the same symbols in panel (a). Note the different height of the surface in the hole
(left green arrow) and on the glass surface (right green arrow), showing that OTS coating in the
hole has not detached ....................................................................................................................
57
Figure 2-7 (a) Secondary electron micrographs with spot EDX analysis of P aeruginosabiofilm
patches. . EDX is an analytical technique used for the chemical characterization of a sample.
Briefly, the interaction between X-ray excitation and the sample surface is used to determine the
surface composition. Carbon-containing substances are dominant in the detached regions of cell
patches (a and b, top plot), whereas silicon-containing substances are dominant in the uncoated
glass regions (a and b, bottom plot). Since the biofilm holes do not show abundant Si, it infers
that the regions are still covered by OTS or by cells ...............................................................
9
58
Figure 2-8 Biofilm holes and hole formation dynamics. (a) Close-up view (60X objective) of the
bacterial distribution at the edge of a4 00x400 ptm 2 biofilm patch after the passage of an air plug.
The dotted line denotes the edge of patch. (b) Zoomed-in view of the dynamics of hole formation
at 1 s intervals. The boundaries of three holes are identified with red, blue and yellow at each of
the four time points. (c) Time course of the hole area, for the three holes shown in (b). Black
symbols denote the average of the three measurements. .........................................................
60
Figure 2-9 The hole pattern correlates with the EPS distribution. (a-b) Distribution of
fluorescently-labelled P aeruginosa after 8 h of biofilm growth, immediately before (a) and
immediately after (b) the passage of an air plug. (c-d) Distribution of EPS, fluorescently labeled
using lectins (Methods), at the same times and position as (a-b). (e) Overlay of the EPS
distribution before the air plug (red) and the bacteria distribution after the air plug (green). Data
are from a different experiment than (a-d). (f) Normalized cross-correlation between the EPS
distribution before the air plug and the cell distribution before (second bar form left) and after
(fourth bar from left) the air plug. Error bars correspond to the standard error over three replicate
experiments. The first and third bars represent results from a bootstrapping analysis (BS), where
the EPS distribution before the air plug is correlated with random permutations of the cell
distribution before (first bar) and after (third bar) the air plug. Note the lack of correlation in
th ese controls. ...............................................................................................................................
65
Figure 3-1 LasI/LasR-RhlI/RhlR quorum sensing system in Pseudomonas aeruginosa.The LasI
protein produces the homoserine lactone signaling molecule N-(3-oxododecanoyl)-homoserine
lactone, autoinducerl (All, blue circles), and the RhlI protein synthesizes N-(butryl)-homoserine
lactone, autoinducer2 (A12, red triangles). Activated LasR (with All) activates transcription of
10
rhiR, which means the las QS system controls RhiR at the transcriptional level. All also controls
the activity of RhlR at the post-translational level by binding to RhlR, which decreases A12
binding to RhlR, therefore inhibiting expression of rhlA and other genes. Presumably, this action
ensures that the LasI/LasR circuit is established prior to the establishment of the RhlI/RhlR
circuit. The Pseudomonas quinolone signal (PQS) is an additional regulatory link between the
L as and R hl Q S circuits. ...............................................................................................................
75
Figure 3-2 Comparison of fluorescent intensity under static (a, b) and flowing (c, d) conditions
for three P aeruginosastrains: PAOl mini Tn5-rhlA::GFP (QS marker; yelloq symbols), PAOl
wild-type (WT; greeen symbols), and PAOl pSMC21 (constitutively expressing GFP; blue
symbols). Under static conditions, fluorescent intensities increase over time for all three strains,
with the greatest increase expressed by the PAO1 pSMC21 strain. This is shown both using
quantitative fluorescence data (a) and visually in an assay plate (b). In contrast, under flowing
condition, the fluorescence of PAO1 pSMC21 was much higher and was expressed more rapidly
than the other two strains, and PAOl WT displayed negligible fluorescence. This is shown via
quantitative imaging (c) and visually in microscopic snapshots taken after 18, 42, and 45 h of
biofilm grow th in the m icrochannel.........................................................................................
79
Figure 3-3 Quorum sensing signal (GFP fluorescent intensity) for square biofilm patches of
different sizes, for a flow rate of 1 pl/min corresponding to a Peclet number of 8.4. The top panel
shows the time series of the fluorescent intensity for each patch size, as the mean and standard
deviation over 6 patches. The lower set of four panels shows the fluorescent intensity at four time
points. In each of the lower panels, the upper row corresponds to the raw fluorescent data, and
the lower row presents the same data in pseudo-coloring, to clarify differences. In the pseudo-
11
coloring, purple to purple-blue tones correspond to weaker QS, whereas yellow to red tones
correspond to stronger QS. The left column in each of the lower set of four panels corresponds to
a patch size of 50 ptm, the middle one to a patch size of 150 pm, and the right one to a patch size
o f 4 0 0 p m ......................................................................................................................................
81
Figure 3-4 Time series of QS intensity for different flow rates. The intensity of QS was
measured as the GFP fluorescence intensity, reported here as a percentage of QS induction. Flow
rates corresponded to average flow velocities of (a) 83 pim/s, (b) 250 pim/s, and (c) 1,000 pim/s.
In each graph, the Peclet number is given, which expresses the relative importance of transport
by advection and by diffusion (see text). For each flow rate, two different patch sizes were
investigated (150 jim, 400 jim). Data correspond to means and standard deviations over six
biofilm patches, for each case. Panel a corresponds to the top panel in Figure 3-3. Panel d
corresponds to the situation in which flow was stopped after the data in panel c (Pe = 102.0)
were collected, and shows that QS resumed upon stopping the flow. .......................................
86
Figure 3-5 Simultaneous confocal measurements of biofilm biomass and QS intensity. (a) Eight
time-lapse images showing both biomass (left, shown in grayscale) and the corresponding
fluorescent intensity (right, shown in green) captured by CRM and FCLSM, respectively. The
patch size is 150 jim
x
150 jim and the flow rate was 1 pl/min, corresponding to a Peclet number
of 8.4. The voxel size is (0.29 pm)x
x
(0.29 pm)y
x
(1.31 jim)z = 0.11 jim3. (b) The temporal
evolution of the biofilm's biomass (red) and fluorescent intensity (green) recorded at 3 h
intervals (top), and the fluorescent intensity normalized by the biomass (bottom)................... 90
Figure 4-1 Effects of dimensionality and domain size. The diffusion of a resource patch
12
generates different profiles of concentration (A) in 1 D, 2D and 3D. Radial profiles are shown at t
= 0.005. The initial half-width of the Gaussian patch was c-o = 0.02. In the remainder of this work,
3D simulations are discussed. Inset: the resource patch reaches the boundary of the simulation
domain (r = 1) after t - 0.02, which was adopted as the temporal horizon for all simulations.. 107
Figure 4-2 Three possible microbial food chains: uniform, passive and active. (a) Shown is the
radial concentration of resources (A, dashed lines) and consumers (B, solid lines) at time t = 0.02
for three cases, uniform (green), passive (blue), and active (red). In a uniform food chain, the
resource is uniform over space, hence consumers remain uniformly distributed (with a value
close to their initial value of 3/(4r)). In a passive food chain (KB
=B=
0), a resource patch
triggers a consumer patch, but the latter is small because it is caused purely by growth and the
duration of the process is short. In an active food chain, consumers respond to a resource patch
using chemotactic motility (here: KB
cases,
7B =
1, ,B
=
100), markedly enhancing accumulation. For all
1. Note that the green dashed line is indistinguishable from the blue dashed line. (b)
These different responses can significantly affect resource uptake, as shown by the spatially
integrated, instantaneous transfer ( f ABdV). Uptake is considerably larger and faster when the
consum ers respond actively to the resource patch......................................................................
109
Figure 4-3 Spatiotemporal dynamics of a two-level food chain. Color maps represent the
distribution of consumers (B) and resource uptake (AB) in space (r) and time (t), for
(a,b) a
passive food chain and (c,d) an active food chain (for definitions of 'passive' and 'active', see
Fig. 2). For comparison, in a uniform food chain (not shown) consumers (B) are uniformly
distributed with B ~ 3/(47r) = 0.239 (growth is weak, hence the value is nearly constant).
Different consumer responses generate not only different consumer distributions (a,c), but also
13
different uptake landscapes (b,d). Parameters are the same as in Fig. 4-2. ................................. 111
Figure 4-4 Effect of chemotactic sensitivity on trophic transfer. The increase in consumer
biomass, ABT, is shown for a two-level food chain (A, B), for six cases: uniform (no resource
patch), passive (no consumer motility), purely diffusive (motility, but no chemotaxis;
%B=
and three cases of chemotactic
=
(XB
=100,
1000 or 10000) food chains. For all cases KB
= 10, except for the passive case for which KB
=
0,
qB
0)
10, qB
10. Numbers on the right indicate the
increase in consumer biomass at t = 0.02. Notice that the trophic transfer is larger in a passive
food chain than in an active one with XB =100 (but less than the ;r B= 1000 case)...................113
Figure 4-5 Effect of random motility on trophic transfer. The increase in consumer biomass, ABT,
is shown for three two-level (A, B) food chains: passive (no consumer motility) and chemotactic
(XB
100,
B = 10), the latter with
KB
= 1 or 10. For KB = 10, the figure further shows the effect
of the saturation gradient 8: the upper and lower boundaries of the shaded region correspond to 6
=
10 to 10 4..................................................................................................................................114
Figure 4-6 Trophic transfer for different chemotactic food chains. Shown is the increase in
consumer biomass, ABT, at time t = 0.02 for a two-level food chain (A, B), two values of the
biomass transfer rate, (a) qB = 1 and (b) qB = 10, and a matrix of motility values, KB and %B.The
case
%B=
KB
=
0 corresponds to a passive food chain, while cases with %B= 0 correspond to
'diffusive' consum ers (m otile but non-chemotactic). ..................................................................
115
Figure 4-7 Increase in the trophic transfer rate due to chemotactic motility, for a two-level (A, B)
food chain with KB = 10 and qB
=
10. Rc represents the ratio of times at which 50% of the
resource (A) has been taken up by the consumers (B), for a passive (i.e., non-motile) food chain
14
relative to an active (i.e., chemotactic) one. When Rc = 1 (dotted line) the transfer rate is equal to
that of a passive food chain. Biomass can be transferred over ten times faster due to chemotactic
m o tility.........................................................................................................................................
117
Figure 4-8 The degree of patchiness in the consumer distribution depends on the consumer's
response. Shown are (a) the patchiness index,
PBB,
and (b) the resource-consumer correlation
index, PAB, for a passive (i.e., non motile) and an active (XB = 1000, KB = 1) two-level food chain,
for two values of the biomass transfer coefficient, qB
=
1 and
qB =
10. Note that the patchiness in
the consumer distribution, PBB, can be larger for the passive case when biomass transfer is rapid
(qB = 10; local growth), but the consumer-resource correlation
PAB
(and thus trophic transfer) is
always larger for the active case (except in the earliest stages)...................................................118
Figure 4-9 Increase in consumer patchiness due to chemotactic motility. Shown is the patchiness
index,
at time t = 0.02 for a two-level food chain (A, B), two values of the biomass transfer
PBB,
rate, (a)
qB
= 1 and (b)
77B =
10, and a matrix of motility values, KB and %B. The case
VB =
KB
=
0
corresponds to a passive food chain, while cases with %B= 0 correspond to 'diffusive' consumers
(m otile but non-chem otactic).......................................................................................................119
Figure 4-10 Effect of chemotaxis on trophic transfer in a three-level (A, B, C) food chain. The
increase in the biomass of the top consumer, AC 1 , is shown for three consumer responses:
uniform (no resource patch), passive (no motility of the consumers, B and C), and chemotactic
(ZB
KB
=
Xc =
1000). For all cases KB
KC
0, qB
=
=
KC
=
10, qB = 7c = 10, except for the passive case for which
C = 10. Numbers on the right indicate the increase in the biomass of the top
consumer (C) at t = 0.02. Biomass transfer is most rapid in the active food chain and least rapid
15
in the uniform food chain............................................................................................................
120
Figure 4-11 Effect on trophic transfer of a differential chemotactic response by different trophic
levels, for a three-level food chain. Curves (panels a, e, i, m) show the time course of biomass
change for all three trophic levels (A, B, C). Numbers on the right of each panel show the
increase in the biomass of the top consumer at t = 0.02. Color maps
represent the
spatiotemporal distribution of the lower consumer, B (panels b, f, j, n), the top consumer, C
(panels c, g, k, o), and the co-localization between B and C, BC (panels d, h, 1, p). The four rows
correspond to four consumer responses. Row 1: XB
Row 3:
%B
100, Xc = 1000. Row 4:
%B--
XC = 100. Row 2:
= 1000. For all cases, KB =
%B=
KC =
B
1000,
c
C
=
100.
10... 122
Figure 4-12 Trophic dynamics in a four-level chemotactic food web. The time course of the
change in biomass of all four trophic levels (A, B, C, D) is shown for KB
10, KC = 0.1,
C= 100, 1C = 10, KD
=
10,
%D =
100, qD=
=
10,
%B=
1000,
B=
10. The case of a fully passive food
chain (all consumers B, C and D are non-motile) is shown for comparison, limited to the biomass
change of the top predator (D). While population dynamics become more complex in the
presence of multiple trophic levels, an active response can still result in a significant
enhancem ent in trophic transfer rates.........................................................................................125
16
Chapter 1
Introduction and Background
1.1 Microbial Biofilms
Microorganisms are ubiquitous in the environment. Despite their tiny size, their combined
biomass is enormous. For example, a milliliter of seawater contains a million bacteria, making
for a total of 1029 bacteria in the world's oceans 1, alone. Concentrations of bacteria are even
higher in other environments, including the subsurface and the human body, and it is now well
accepted that this huge bacterial biomass performs a plethora of crucial functions in nearly every
environment.
Many microbes are associated with solid surfaces. These interfacial microbial communities are
termed 'biofilms' and are structurally and dynamically complex biological systems (Fig. 1.1).
The structures consist primarily of a matrix of extracellular polysaccharide substance (EPS) that
the bacterial cells produce and within which they are embedded
2. The
EPS matrix provides
several functional purposes for the biofilm, including protecting bacteria from environmental
threats, providing mechanical stability, and degrading macromolecules to be used by the cells3
17
Figure 1-1 Conceptualization of biofilm structure, development and dynamic behaviors. Image
credit: P. Dirckx, Center for Biofilm Engineering, Montana State University, USA
Biofilm formation is a sequential process, consisting of a series of steps that occur when cells
encounter favorable nutrient and surface conditions. The steps are:
(i)
transport of microbes to a surface by swimming, sinking or flow
(ii)
initial reversible attachment, affected by hydrodynamic conditions, where bacteria
move on the surface by means of flagellar swimming or pili-mediated twitching
motility 4, seeking a niche to settle
(iii)
6,7
irreversible attachment, accompanied by profound physiological changes, including
18
rapid down-regulation of flagellar synthesis
3,8-10,
reduced growth rates and up-
regulation of alginate synthesis to produce EPS, which promotes cell adhesion and
formation of 3D cell clusters, protecting cells from desiccation,, antibiotics,
bacteriophages and predators
(iv)
7,11
biofilm maturation, when cell clusters grow to thicknesses of order 100 pm and
complex architectures emerge, with water channels transporting nutrients across the
biofilm
(v)
motility-mediated dispersion
2,5,12
or shear induced erosion 13 of cells from the biofilm,
possibly to colonize new habitats and initiate secondary infections
14
Fully developed biofilms have complex architectures and have been compared to 'cities of
microbes' ". Mushroom-shaped microcolonies, tens of microns thick 16,17, are encased in an EPS
matrix and criss-crossed by interstitial water channels 2,9. Their spatial heterogeneity harbors a
range of microenvironments, providing niches for multiple phenotypes 18, and affects community
structure and productivity 19. This environment is dynamic: structures appear and disappear as
new cells attach, others migrate among colonies 20, cooperate for biofilm stability or compete for
nutrients 2, and ultimately disperse 9.
Biofilms are a major component of world-wide bacterial biomass
19
2'2
and they often have
deleterious effects in natural, industrial and medical settings. In natural settings, they can impact
environments on a global level, including producing and consuming atmospheric gases;
mobilizing toxic elements such as mercury, arsenic and selenium causing stream or soil
contamination
2;
and producing toxic algal blooms and creating oxygen depletion zones in lakes,
rivers and coastal environments
25
. In many industrial processes, the formation of biofilms on
nearly every surface is responsible for huge economics losses (in the billions of dollars yearly in
the US alone
26)
resulting from biofouling and biocorrosion
clogging and damage, product contamination and energy losses
27,
which leads to equipment
28
In medicine, biofilms cause
.
deleterious effects in many medical devices including catheters and prosthetic heart valves 2
There are, however, applications that take advantage of biofilm growth. The ability of the
biofilm polymer matrix to entrap minerals and nutrients is exploited in purification of drinking
water
29,
detoxification of oil spills, removal of heavy metals and biodegradation of hazardous
xenobiotics in contaminated waters and soils
biosensors
3.
30,
as well as environmental monitoring using
Many biofilms play an important role in the ecology of the earth and the
sustainability of life. In soil, for instance, almost every chemical transformation involves active
contributions from microorganisms, many of which are surface-attached. In particular, microbes
play an active role in soil fertility by mediating the cycle of nutrients, like carbon and nitrogen,
20
which are required for plant growth
32,33.
An understanding of the mechanisms underlying
biofilm formation, development and detachment is a crucial step towards elucidating the role of
biofilms in ecosystems and developing new strategies for biofilm control.
1.2 Biofilm Detachment
Of these phenomena, one of the least understood is detachment from a surface, even though
understanding detachment mechanisms has great significance for controlling biofilms.
Detachment refers to the release of microbial cells, and their associated matrix polymer (EPS),
from a solid surface into the bulk fluid. The detachment process is an important component of
the biofilm life cycle and plays a fundamental role in dissemination and contamination and
ultimately the long-term survival in either natural 34 or man-made settings of industrial and
medical importance
. Some of the factors that have been suggested to be important in biofilm
detachment include matrix-degrading enzymes
levels and microbial growth status
40,
38,
microbially generated gas bubbles 39, nutrient
fluid shear stress
41,42
and quorum-sensing signals 43.
Passing air-liquid interfaces (air bubbles or air plugs) have been demonstrated to yield extremely
efficient detachment forces, as in the oral cavity during eating, speaking, drinking and
swallowing 44, on the eye and on contact lenses during blinking 45, and on rocks and ship hulls in
21
marine environments 46. Since biofilm age increases biofilm adhesive strength due to EPS
production, which in turn increases the biofilm's resistance to cleaning treatments
protects biofilm cells against antimicrobial agents and desiccation
49,
47,48
and
biofilms become gradually
more difficult to remove over time. This is a huge problem also because, even if a small part of
the biofilm survives the cleaning treatment, it can represent a seed from which the biofilm
rapidly grows back. Thus, treatment should ideally be applied during the early stage of biofilm
development before the biofilm is strongly sheltered by its EPS.
In Chapter 2 of this thesis, I present experiments to understand how the age of the biofilm
affects removal by shear, applied through passage of an air plug. These experiments led to the
discovery of a new 'detachment mode', in which biofilm removal is highly heterogeneous and
characterized by a semi-regular pattern of holes, from which cells are removed, separated by
regions in which the biofilm remains unscathed. High-resolution observations in space and time,
enabled by the microfluidic model system, are used to rationalize the detachment dynamics,
leading to the conclusion that the local amount of EPS sets the ability of any given area of the
biofilm to withstand removal by mechanical forces.
22
1.3 The Role of Cell Signaling in Biofilm Dynamics
Molecular 'quorum sensing' (QS) systems are an important mechanism by which bacteria
form and maintain complex biofilms through the ability to communicate and to alter behavior in
response to the presence of other bacteria 502. Bacterial cells produce signaling molecules that
are secreted into the environment, from which they can disperse by diffusion or advection. The
producer cells respond to their own signals, which are therefore called autoinducers
-
Although the signaling mechanisms differ, QS systems are utilized by both gram positive and
gram negative bacterial species s. In general to sense their population density gram negative
bacteria use small chemical molecules called acylated homoserine lactones (AHLs)
56
and gram
positive bacteria use oligo-peptides 57, both collectively referred to as autoinducers (AIs).
Typically, an autoinducer molecule induces the transcription of a set of genes that includes the
gene encoding the autoinducer-producing enzyme, which results in a positive feedback loop 58.
The autoinducer can trigger a response by the cell if the autoinducer concentration - regulated by
the rates of production, decay and mass transfer integrated over time - reaches a threshold at the
cell's location (Fig. 1-2).
23
*
p
ALAHL
3-0 i2HS
A HL diffuse
out
AHllIus
I
p
('4NL(,2
diffuse in
i*t
01
6,
-----
Time
Figure 1-2 LasILasR-RhlIIRhlR quorum sensing system in Pseudomonas aeruginosa.The LasI
protein produces the homonserine lactone signaling molecule N-(3-oxododecanoyl)-homoserine
lactone, autoinducer 1 (All, blue circles), and the RhlI protein synthesizes N-(butryl)homoserine lactone, autoinducer 2 (A12, red triangles). Activated LasR (with All) activates
transcription of rhIR, which means the las quorum sensing system controls RhlR at the
transcriptional level. All also controls the activity of RhlR at the post-translational level by
binding to RhIR, which decreases A12 binding to RhIR, therefore inhibiting expression of rhlA
and other genes. The Pseudomonas quinolone signal (PQS) is an additional regulatory link
between the Las and Rhl quorum sensing circuit.
The existence of a threshold cell concentration for the production of bioluminescence in the
marine symbiotic bacteria Vibrio fischeri 5, the formation of fruiting bodies in myxobacteria 59
and the development of competence for genetic transformation in Streptococci 60, are among the
lines of evidence for cell-to-cell signaling systems in bacterial communities. At present, QS is
defined as a cell-density-dependent bacterial intercellular signaling mechanism that enables
24
bacteria to coordinate the expression of certain genes to coordinate group behaviors.
Several physical, biological, and chemical factors have the potential to influence QS in biofilm
systems 61. For example, genetically modified mutants of Pseudomonas aeruginosa unable to
perform quorum sensing do not form the complex mushroom-shaped structures typical of wildtype biofilms
50,
which suggests that the production of EPS in Pseudomonas aeruginosa is
influenced by QS. Other studies have shown that QS regulates components of EPS which
contribute to the sticky level, thus impacting the structural components of biofilm matrix. 62,63
The hydrodynamic environment can also influence QS. Janakiraman et al. 64 studied biofilm
growth and QS in uniform microfluidic chambers and developed a one-dimensional
mathematical model. They found that the flow rate greatly affects QS: at high flow rates, the
biofilm thickness is smaller due to detachment, and the transport rate of AHL out of the biofilm
by fluid flow can be so high that the AHL concentration does not reach the threshold required for
induction. Purevdorj and Costerton
65
emphasize the importance of transport processes, in
particular highlighting that the washout of signal molecules from the bulk liquid surrounding
bacterial microcolonies is expected to increase the concentration gradient between the biofilm
and the bulk liquid, in turn driving a larger diffusive flux of signal molecules out of the biofilm.
Their work indicates that diffusion through the biofilm under different hydrodynamic conditions
25
is strongly correlated with the AHL signal concentration in microcolonies. Muller et al. 66
mathematically studied QS induction on biofilm patches. They found that induction can occur for
a lower number of cells in a patchy biofilm than in a homogeneous biofilm, because the cell
density (and consequently, AHL concentration) when cells are compacted in small clusters.
Building on this evidence of the key roles of the hydrodynamic environment and patchiness
for QS in biofilms, I expand on these findings by investigating the contribution of flow to QS
induction on spatially defined patches of biofilm. These experiments, presented in chapter 3, take
advantage of the technique developed as part of this thesis for controlling and monitoring
biofilms in microfluidic devices, including their spatial arrangement on a solid surface. Although
QS plays an important role in biofilm formation and maturation
67,
the methods currently
available for creating and studying biofilms do not provide reproducible, spatially controllable
biofilms for comparative analysis to study the chemical, physical, and environmental factors that
influence biofilm development in a statistically relevant experimental format. The major
advantage of the technique that I present and apply in this thesis is that it allows placement
hydrophobic chemical groups with a well-defined structure on a solid surface, enabling control
over the amount of adsorbed cells via the density of these hydrophobic groups.
26
1.4 Microfabrication
Methods used to study biofilms should ideally enable tight control of physical conditions.
Traditional approaches based on 96- or 384-well plates are only suitable for studying biofilm
development in static conditions. However, the environment that microbes inhabit is anything
but static. Microorganisms are exposed to a range of flow environments, from creeping flow in
soil, to variable flow conditions in rivers and streams, to the strong shear experienced in stirred
bioreactors and industrial pipelines. The presence of flow not only affects how microorganisms
are transported and dispersed, but also their ability to interact with their local habitat, including
attaching and detaching from surfaces, casting chemical signals for chemical communication,
and resisting penetration by antibiotics.
Microfabrication is a flexible and powerful approach to create and control microenvironments
in which cell behavior can be studied. The ability to culture cells in microfluidic systems offers
numerous advantages over traditional benchtop-scale technologies. Bioreactors used for
conventional cell culture typically operate in a batch or semi-batch mode that requires frequent
media changes and the use of mechanical stirring to enhance mass transfer. These processes
introduce mechanical stresses and steep concentration gradients that deviate significantly from
those encountered in natural microbial habitats. Consequently, the cellular microenvironment is
27
either severely disrupted or completely unmaintainable, altering the observed growth and
behavior in culture.
Since flow and nutrients transport is precisely controllable in a microfluidic system, cells are
exposed to an environment that, for many aspects, more closely mimics their natural one.
Furthermore, the optical access enabled by the transparency of microfluidic devices means that
single cells or collections of cells can be directly observed microscopically as they respond to
stimuli under conditions that are more consistent with actual physiological settings.
Microsystems also offer the capability of performing studies in a highly parallel manner by
incorporating a few to several dozens of cell culture chambers within a single chip.
In this work, I used microfabrication to produce experimental systems that allowed me (i)
spatial control over biofilm formation; (ii) accurate control over fluid flow rates and nutrient
delivery rates; (iii) control over environmental insults, such a air plugs; (iv) the direct monitoring
of biofilm dynamics through automated imaging and image analysis, producing unprecedented
amounts of data on the spatial and temporal dynamics of biofilms; and (v) simultaneous, parallel
experiments to include on-chip controls as well as on-chip variation of relevant parameters. This
work, this, contributes to establish microfabrication and microfluidics as an ideal tool for biofilm
research.
28
1.5 pContact Printing for Biofilm Research
Most bacteria attach to surfaces in spatially heterogeneous or 'patchy' patterns 2, because
distributions of hydrodynamic shear on surfaces and important chemical constituents are
heterogeneous in natural systems 68, and because of natural variability in the time of colonization
of different areas on a surface. Therefore, it would be desirable to study properties (e.g., quorum
sensing, detachment) of biofilms as a function of their spatial structure, and thus to be able to
control and manipulate this spatial structure.
Many approaches to bacterial patterning have been reported, including dip-pen lithography 69
inkjet printing
70,
photolithography
71,
and spotting 72,73. Although spotting, photolithography,
and inkjet printing can create micropatterns of cells and dip-pen lithography is suitable for
fabricating nanopatterns, these methods depend heavily on specialized microfabrication facilities
and, as a result, have been used very little to address fundamental questions about biofilm
dynamics. Also, these robot-controlled printing methods are serial and require long processing
that is undesirable to prepare single-use, disposable devices for flexible experimentation.
Microcontact printing (pCP) provides an alternative method, which affords great flexibility to
readily alter the size and shape of features that control cell adhesion, density and geometry.
Pioneering work in pCP was performed by Whitesides' group (Harvard University) for the
29
patterned
transfer
of thiols
onto
poly(dimethylsiloxane) (PDMS) stamp
gold
74-76.
surfaces
by means
of a
microstructured
The process of pCP is schematically depicted in Fig.
1-3 (left-hand side). PDMS is the material most frequently used to make stamps, as a slab of
polymer that bears a microscale relief pattern on one side, since it can be molded using a master.
PDMS is flexible enough to make conformal contact for pCP even with rough surfaces but still
has enough mechanical stiffness to reproduce patterns in the micrometer range. The PDMS
stamp is "inked" and put in contact with the substrate surface. Ideally, the ink is transferred from
stamp to substrate only in the area of contact. The successful application of tCP of chemicals
from a microstamp onto a surface is dependent upon the time required to progress from drying
the stamp to printing onto substrates, and upon the property of the immobilizing surfaces
77
Insufficient pressure might result in poor printing and excessive pressure will induce sagging of
the stamp, causing chemicals to print outside of the features.
Microcontact printing to produce spatially controlled biofilm patches
As part of the work for this thesis, I created bacterial biofilm patches within microfluidic
channels by patterning hydrophobic substrate patches within the channels onto which biofilmforming bacteria preferentially attached. I successfully created arbitrarily-shaped hydrophobic
patches on a glass surface encased within a PDMS microfluidic chamber. Fig. 1-3 (left-hand side)
30
shows the schematic of the process. This surface treatment has been obtained through microcontact printing (piCP) of organosiloxane (R-Si-Xn, where X is a hydrolysable group such as
halogen, alkoxy, acyloxy, or amine) self-assembled monolayers (SAMs) onto glass substrates, by
using a PDMS stamp. Specifically, the organosiloxane I found to be optimal for this procedure is
octadecyltrichlorosilane (OTS), which is an amphiphilic molecule consisting of a long-chain alky
group (C18H37-) and a polar head group (SiCl 3 -), because it can be easily covalently attached to
the microscopic glass slide and it has excellent stability and resistance to high shear flow and
ambient light. The OTS-modified surfaces were found to be appropriate for the task since they
have a -CH
3
group that makes them hydrophobic and hydrophobic surfaces typically favor
bacterial attachment
78,79.
Two layers of testing have demonstrated the viability of this method to produce biofilm
patches. First, I have tested the performance of the microstamping process by assessing the
presence and integrity of OTS patches using bovin serum albumin (BSA) labeled with
fluorescein
isothiocyanate
(FITC-BSA),
as shown
in Fig. 1-3
(right-hand side). The
epifluorescence images of FITC-BSA adsorption indicate that the non-specific protein adsorption
can be attributed to the hydrophobic interactions between the OTS-modified regions and the
hydrophobic portions of BSA molecules, which are commonly accepted as the dominant driving
31
force for adsorption between the substrate and proteins. 80
a)
PDMS microstamp
OTS
resin
b)
PDMS microstamp
OTS resin
Glass substrate
C)
PDMS chip with
d)
microchannels
Inlet
Outlet
hydrophobic
hydrophilic
Microchannel
Figure 1-3 (Left) Schematic of the microcontact printing (pCP) technique used for depositing
hydrophobic patches onto glass substrates. (a) Wet inking is achieved by placing a patterned
PDMS stamp onto the OTS ink pad. (b) After inking, the stamp is left in a chemical hood for 2
min to let the solvent (hexane) evaporate. (c) By bringing the inked stamp in conformal contact
with the glass surface, the hydrophobic long-chain (C-18) silane is transferred to the glass
surface. Due to the patterned structure of the stamp, only the areas with protrusions contact the
glass surface, and the silane is area-selectively transferred according to the pattern of the stamp.
(d) The microfluidic channel is positioned on top of the patterned substrate. (Right) Fluorescent
images of FITC-labeled bovin serum albumin (BSA) absorbed onto the hydrophobic patches,
demonstrating the specificity of the pCP technique. The top row shows squares of 20 ptm and 50
pm size. The bottom row shows lines of 100 pm thickness and the MIT logo.
Second, I have tested the ability of this method to create well-defined biofilm patches by
flowing a suspension of Pseudomonas aeruginosathrough the microchannel and determining the
spatial distribution of adherent cells. Results show that this technique allows the creation of
32
patches of biofilm of different shape and size, with high specificity (Fig. 1-4), in that there is
strong preference for cells to attach within, but not outside, the patches. Furthermore, the degree
of cell adhesion can be controlled by varying the surface wettability, through control over contact
printing time of OTS.
Cells preferentially adhere to the hydrophobic patches, via hydrophobic interactions between
the bacterial cell wall and the SAM, and the degree of cell adhesion can be controlled by varying
the surface wettability. Since wettability and surface chemistry are supposed to be the key
parameters that control the adhesion of bacteria, I investigated various alkylsilane self-assembled
monolayer inks to find the best candidate for stamping hydrophobic patterns onto glass
substrates. I have achieved the best selectivity in the adhesion of bacteria on patches by using
octadecyltrichlorosilane (OTS), and found that thin layers of this compound possess high
chemical and mechanical stability under shear stress, even when the latter is strong. Thus, using
the technique described here, it is possible to create biofilm patches with defined size and shape,
without the chemical substrate being washed away by flow. Furthermore, this approach
seamlessly allows replication of patterns in arbitrary arrangements and in large numbers over a
surface (Fig. 1-4), without any additional effort beyond the initial computer-based design of the
stamp.
33
40X
60X
Figure 1-4 Phase-contrast images of P aeruginosa PAOl selectively attached on hydrophobic
patches of different shapes and sizes. The top row shows squares and the bottom row shows
vertical lanes of biofilm. Note the strong difference in the density of attached cells between the
patches and the regions of the surface outside the patches, demonstrating the specificity achieved
by this patterning process.
Since this approach creates reproducible biofilm patches, it overcomes a limitation of
traditional methods in which biofilms are typically grown from cells that have adsorbed to
surfaces through a process that is largely stochastic and nearly impossible to control or reproduce.
Thus, the patterning of bacteria on surfaces described here provides a unique opportunity to
study biofilm formation in controlled, spatially heterogeneous environments. This technique has
allowed me to create spatially controlled, functional surfaces that capture and localize cells
34
within selected regions, while limiting nonspecific adsorption of cells to adjacent, 'background'
regions.
1.6 Motivation and Thesis Scope
The motivation of this research is to contribute to a better understanding of bacteria attached
to surfaces, in view of their importance in a broad range of settings and in particular of their
heightened antiobiotic resistance compared to planktonic cells. In nature, microbial biofilms are
characterized by strong heterogeneity and are continuously subject to fluid flow and the resulting
environmenetal insults (e.g., detaching shear forces). This heterogeneity ranges from the
physiological differences between each of the myriad of cells in the biofilm, to the individual
colonies of the same species or of different species that make up the biofilm, to the variability
imposed by the topography of the supporting surface. Nevertheless, biofilm studies often assume
that environmental conditions are homogeneous, as this small-scale spatial variability is difficult
to model and control in a laboratory setting. Thus, the effect of microscale heterogeneity on the
development and stability of biofilms is largely unknown. Fluid flow is similarly ubiquitous in
biofilm environments, because completely quiescent environments are exceedingly rare, and
even relatively slow fluid motion at the macroscale can generate strong velocity gradients (shear)
35
at the microscale. The resulting hydrodynamic shear, coupled with the inherent variability within
the biofilm (e.g., as will be seen, the local strength of adhesion), can determine the detachment
dynamics of the biofilm from the surface.
In this thesis, I investigate the role of heterogeneity and of fluid flow on bacterial biofilms.
To address these questions, I devised a new microfluidic technique (sections 1.5 and 2.4) to
precisely control spatial biofilm structure and hydrodynamic conditions. With this system in
hand, I have investigated the effect of patch size and fluid flow on biofilm dynamics in
Pseudomonas aeruginosaPAO1 - an opportunistic pathogen in humans and a model organism
for biofilm studies - including the susceptibility to disruption by shear (Chapter 2), and the
ability to perform quorum sensing (Chapter 3).
In Chapter 2, I demonstrate a novel mode of detachment of a biofilm from a surface. After the
passage of an air plug, the break-up of the residual thin liquid film scrapes and rearranges
bacteria on the surface, such that a 'Swiss cheese' pattern of holes is left in the residual biofilm. I
demonstrate that this pattern results from the competition between hydrodynamic forces from
shear, tending to dislodge cells, and adhesion forces originations from the local amount of EPS
produced by the cells. This new detachment mode needs to be considered when shear-based
removal strategies
are designed, as biofilm removal might be confused with biofilm
36
rearrangement, and persisting biofilm strongholds could rapidly seed biofilm regrowth, rendering
treatment vain.
In Chapter 3, I examine the effects of biofilm colony size and ambient fluid flow on the QS
induction in a developing Pseudomonas aeruginosa biofilm. I hypothesize that as the fluid flow
rate over a biofilm increases, the amount of biofilm biomass required for the induction of QS
within the population also increases. I found that quorum sensing sets in earlier in smaller
biofilm patches, yet its intensity at long times is greater in larger patches. The effect of ambient
fluid flow was to accelerate the induction of quorum sensing compared to static conditions at
moderate flow rates, due to the increase in the convective supply of nutrients and to quench
quorum sensing at high flow rates, due to the autoinducer signal being washed out by flow.
These findings establish microfluidics as a new tool in the study of biofilms, which enables both
accurate control over microenvironmental conditions and direct observation of the dynamics of
biofilm.
Heterogeneity is a broad feature of microbial populations, and can have multiple consequences
on microbial processes beyond those specific to biofilms. As a first attempt to characterize the
trophic consequences of heterogeneity, in Chapter 4 I present a spatially explicit mathematical
model to better understand how heterogeneity affects nutrient transfer across different trophic
37
levels. I find that chemotaxis accelerates nutrient transfer when chemotactic sensitivity exceeds a
threshold, as otherwise random motility acts to slow nutrient transfer. I find that nutrient transfer
is enhanced by chemotaxis by up to 10-fold, and that chemotaxis can increase patchiness within
the microbial community. While in its current implementation the model is only conceptually
linked to heterogeneity in biofilms, as the model focuses specifically on planktonic communities,
this chapter broadly addresses the role of heterogeneity on the interactions between different
populations of microbes, focusing on predator-prey interactions. It is hoped that, in the future,
models of this kind could provide guidance also for the analysis of heterogeneity on biofilm
dynamics.
38
Bibliography
1. Whitman, W.B., Coleman, D.C. & Wiebe, W.J. Prokaryotes: the unseen majority. Proceedings
of the National Academy of Sciences 95, 6578-6583 (1998).
2. Costerton, J.W., Lewandowski, Z., Caldwell, D.E., Korber, D.R. & Lappin-Scott, H.M.
Microbial biofilms. Annual Reviews in Microbiology 49, 711-745 (1995).
3. Flemming, H.C., Neu, T.R. & Wozniak, D.J. The EPS matrix: the "house of biofilm cells".
Journal of Bacteriology 189, 7945-7947 (2007).
4. Harshey, R.M. Bacterial motility on a surface: many ways to a common goal. Annual Reviews
in Microbiology 57, 249-273 (2003).
5. Klausen, M., et al. Biofilm formation by Pseudomonas aeruginosa wild type, flagella and type
IV pili mutants. Molecular microbiology 48, 1511-1524 (2003).
6. Agladze, K., Wang, X. & Romeo, T. Spatial periodicity of Escherichia coli K-12 biofilm
microstructure initiates during a reversible, polar attachment phase of development and requires
the polysaccharide adhesin PGA. Journal of Bacteriology 187, 8237-8246 (2005).
7. O'Toole, G., Kaplan, H.B. & Kolter, R. Biofilm formation as microbial development. Annual
Reviews in Microbiology 54, 49-79 (2000).
8. Parsek, M.R. & Singh, P.K. Bacterial biofilms: an emerging link to disease pathogenesis.
Annual Reviews in Microbiology 57, 677-701 (2003).
9. Donlan, R.M. Biofilms: microbial life on surfaces. Emerging infectious diseases 8, 881-890
(2002).
10.
Davies, D. & Geesey, G. Regulation of the alginate biosynthesis gene algC in
Pseudomonas aeruginosa during biofilm development in continuous culture. Applied and
39
Environmental Microbiology 61, 860-867 (1995).
11.
Duddridge, J.E., Kent, C. & Laws, J. Effect of surface shear stress on the attachment of
Pseudomonas fluorescens to stainless steel under defined flow conditions. Biotechnology and
Bioengineering 24, 153-164 (1982).
12.
Sauer, K., Camper, A.K., Ehrlich, G.D., Costerton, J.W. & Davies, D.G. Pseudomonas
aeruginosa displays multiple phenotypes during development as a biofilm. Journal of
Bacteriology 184, 1140-1154 (2002).
13.
Purevdorj-Gage, B., Costerton, W. & Stoodley, P. Phenotypic differentiation and seeding
dispersal in non-mucoid and mucoid Pseudomonas aeruginosa biofilms. Microbiology 151,
1569-1576 (2005).
14.
Ymele-Leki, P. & Ross, J.M. Erosion from Staphylococcus aureus biofilms grown under
physiologically relevant fluid shear forces yields bacterial cells with reduced avidity to collagen.
Applied and environmental microbiology 73, 1834-1841 (2007).
15.
Watnick, P. & Kolter, R. Biofilm, city of microbes. Journal of Bacteriology 182, 2675-2679
(2000).
16.
Korber, D.R., Lawrence, J.R., Sutton, B. & Caldwell, D.E. Effect of laminar flow velocity
on the kinetics of surface recolonization by Mot+ and Mot- Pseudomonas fluorescens. Microbial
ecology 18, 1-19 (1989).
17.
Stewart, P., Peyton, B., Drury, W. & Murga, R. Quantitative observations of heterogeneities
in Pseudomonas aeruginosa biofilms. Applied and environmental microbiology 59, 327-329
(1993).
18.
Fux, C., Costerton, J., Stewart, P. & Stoodley, P. Survival strategies of infectious biofilms.
Trends in Microbiology 13, 34-40 (2005).
40
19.
Hansen, S.K., Rainey, P.B., Haagensen, J.A. & Molin, S. Evolution of species interactions
in a biofilm community. Nature 445, 533-536 (2007).
20.
Tolker-Nielsen, T., et al. Development and dynamics of Pseudomonassp. biofilms. Journal
of Bacteriology 182, 6482-6489 (2000).
21.
Xavier, J.B. & Foster, K.R. Cooperation and conflict in microbial biofilms. Proceedings of
the National Academy of Sciences 104, 876-881 (2007).
22.
Costerton, J., Stewart, P. & Greenberg, E. Bacterial biofilms: a common cause of persistent
infections. Science 284, 1318 (1999).
23.
Costerton, J. The biofilm primer, (Springer Verlag, 2007).
24.
Kohusovi, K., Havel, L., Vlasik, P. & Tonika, J. A long-term survey of heavy metals and
specific organic compounds in biofilms, sediments, and surface water in a heavily affected river
in the Czech Republic. Environmental monitoring and assessment 174, 555-572 (2011).
25.
Lawrence, J., et al. Microscale and molecular assessment of impacts of nickel, nutrients,
and oxygen level on structure and function of river biofilm communities. Applied and
environmental microbiology 70, 4326-4339 (2004).
26.
Flemming, H.-C. Microbial biofouling: unsolved problems, insufficient approaches, and
possible solutions. in Biofilm highlights 81-109 (Springer, 2011).
27.
Flemming, H.-C. Biofilms and environmental protection. Water Science & Technology 27,
1-10 (1993).
28.
Palmer, J., Flint, S. & Brooks, J. Bacterial cell attachment, the beginning of a biofilm.
Journal of Industrial Microbiology and Biotechnology 34, 577-588 (2007).
29.
Simpson, D.R. Biofilm processes in biologically active carbon water purification. Water
research 42, 2839-2848 (2008).
41
30.
Singh, R., Paul, D. & Jain, R.K. Biofilms: implications in bioremediation. Trends in
Microbiology 14, 389-397 (2006).
31.
Fernandes, R., et al. Biological nanofactories facilitate spatially selective capture and
manipulation of quorum sensing bacteria in a bioMEMS device. Lab Chip 10, 1128-1134 (2010).
32.
Mader, P., et al. Soil fertility and biodiversity in organic farming. Science 296, 1694 (2002).
33.
Wardle, D.A., et al. Ecological linkages between aboveground and belowground biota.
Science 304, 1629-1633 (2004).
34.
Nickel, J.C., Costerton, J.W., McLean, R.J. & Olson, M. Bacterial biofilms: influence on
the pathogenesis, diagnosis and treatment of urinary tract infections. Journal of Antimicrobial
Chemotherapy 33, 31-41 (1994).
35.
Zottola, E.A. & Sasahara, K.C. Microbial biofilms in the food processing industry-
Should they be a concern? International Journal of Food Microbiology 23, 125-148 (1994).
36.
Walker, J., et al. Control ofLegionella pneumophila in a hospital water system by chlorine
dioxide. Journal of Industrial Microbiology 15, 384-390 (1995).
37.
Piriou, P., Dukan, S., Levi, Y. & Jamge, P. Prevention of bacterial growth in drinking water
distribution systems. Water science and technology 35, 283-287 (1997).
38.
Allison, D.G., Ruiz, B., SanJose, C., Jaspe, A. & Gilbert, P. Extracellular products as
mediators of the formation and detachment of Pseudomonas fluorescens biofilms. Fems
Microbiology Letters 167, 179-184 (1998).
39.
Ohashi, A. & Harada, H. Characterization of detachment mode of biofilm developed in an
attached-growth reactor. Water science and technology 30, 35-45 (1994).
40.
Stewart, P.S. A model of biofilm detachment. Biotechnology and Bioengineering 41, 111-
117 (1993).
42
41.
Peyton, B.M. Effects of shear stress and substrate loading rate on Pseudomonas aeruginosa
biofilm thickness and density. Water Research 30, 29-36 (1996).
42.
Picioreanu, C., van Loosdrecht, M.C.M. & Heijnen, J.J. Two-dimensional model of biofilm
detachment caused by internal stress from liquid flow. Biotechnology & Bioengineering 72, 205218 (2001).
43.
Hentzer, M., et al. Inhibition of quorum sensing in Pseudomonas aeruginosa biofilm
bacteria by a halogenated furanone compound. Microbiology 148, 87-102 (2002).
44.
Busscher, H., Cowan, M. & Van der Mei, H. On the relative importance of specific and
non-specific approaches to oral microbial adhesion. FEMS microbiology letters 88, 199-209
(1992).
45.
Landa, A.S., van der Mei, H.C., van Rij, G. & Busscher, H.J. Efficacy of ophthalmic
solutions to detach adhering Pseudomonas aeruginosa from contact lenses. Cornea 17, 293
(1998).
46.
Cooksey, K. & Wigglesworth-Cooksey, B. Adhesion of bacteria and diatoms to surfaces in
the sea: a review. Aquatic Microbial Ecology 9, 87-96 (1995).
47.
Okabe, S., Kuroda, H. & Watanabe, Y. Significance of biofilm structure on transport of
inert particulates into biofilms. Water science and technology 38, 163-170 (1998).
48.
Chen, M., Zhang, Z. & Bott, T. Effects of operating conditions on the adhesive strength
of< i> Pseudomonas fluorescens</i> biofilms in tubes. Colloids and Surfaces B: Biointerfaces
43, 61-71 (2005).
49.
., H.-C.F., et al. & Flemming, H.-C.,. Cohesiveness in biofilm matrix polymers
Community structure and co-operation in biofilms, (Cambridge University Press, 2000).
50.
Davies, D.G., et al. The involvement of cell-to-cell signals in the development of a
43
bacterial biofilm. Science 280, 295 (1998).
51.
Miller, M.B. & Bassler, B.L. Quorum sensing in bacteria. Annual Reviews in
Microbiology 55, 165-199 (2001).
52.
Irie, Y. & Parsek, M. Quorum sensing and microbial biofilms. Bacterial Biofilms, 67-84
(2008).
53.
Nealson, K.H., Platt, T. & Hastings, J. Cellular control of the synthesis and activity of the
bacterial luminescent system. Journal of Bacteriology 104, 313 (1970).
54.
Eberhard, A. Inhibition and activation of bacterial luciferase synthesis. Journal of
Bacteriology 109, 1101 (1972).
55.
Nealson, K.H. & Hastings, J. Bacterial bioluminescence: its control and ecological
significance. Microbiology and Molecular Biology Reviews 43, 496 (1979).
56.
Fuqua, C., Parsek, M.R. & Greenberg, E.P. Regulation of gene expression by cell-to-cell
communication: Acyl-homoserine lactone quorum sensing. Annu. Rev. Genet. 35, 439-468
(2001).
57.
Kuipers, O.P., de Ruyter, P.G.G.A., Kleerebezem, M. & de Vos, W.M. Quorum sensing-
controlled gene expression in lactic acid bacteria. Journal of Biotechnology 64, 15-21 (1998).
58.
Whitehead, N.A., Barnard, A.M.L., Slater, H., Simpson, N.J.L. & Salmond, G.P.C.
Quorum-sensing in Gram-negative bacteria. FEMS Microbiology Reviews 25, 365-404 (2001).
59.
Dworkin, M. & Kaiser, D. Cell interactions in myxobacterial growth and development.
Science 230, 18 (1985).
60.
Dawson, M.H. & Sia, R.H.P. In vitro transformation of pneumococcal types: I. a technique
for inducing transformation of pneumococcal types in vitro. The Journal of experimental
medicine 54, 681 (1931).
44
61.
Horswill, A.R., Stoodley, P., Stewart, P.S. & Parsek, M.R. The effect of the chemical,
biological, and physical environment on quorum sensing in structured microbial communities.
Analytical and bioanalytical chemistry 387, 371-380 (2007).
62.
Morohoshi, T., et al. Inhibition of quorum sensing in Serratia marcescens AS-l
by
synthetic analogs of N-acylhomoserine lactone. Applied and environmental microbiology 73,
6339-6344 (2007).
63.
Marketon, M.M., Glenn, S.A., Eberhard, A. & Gonzilez, J.E. Quorum sensing controls
exopolysaccharide production in Sinorhizobium meliloti. Journal of Bacteriology 185, 325-331
(2003).
64.
Janakiraman, V., Englert, D., Jayaraman, A. & Baskaran, H. Modeling growth and quorum
sensing in biofilms grown in microfluidic chambers. Annals of biomedical engineering 37, 1206-
1216 (2009).
65.
Purevdorj, B., Costerton, J. & Stoodley, P. Influence of hydrodynamics and cell signaling
on the structure and behavior of Pseudomonas aeruginosa biofilms. Applied and environmental
microbiology 68, 4457 (2002).
66.
MUller,
J., Kuttler, C.,
Hense, B.A.,
Rothballer, M.
& Hartmann,
A. Cell-cell
communication by quorum sensing and dimension-reduction. Journal of mathematical biology
53, 672-702 (2006).
67.
Hardie, K.R. & Heurlier, K. Establishing bacterial communities by'word of mouth': LuxS
and autoinducer 2 in biofilm development. Nature Reviews Microbiology 6, 635-643 (2008).
68.
Singer, G., Besemer, K., Schmitt-Kopplin, P., H6dl, I. & Battin, T.J. Physical heterogeneity
increases biofilm resource use and its molecular diversity in stream mesocosms. Plos One 5,
e9988 (2010).
45
69.
Ginger, D., Zhang, H. & Mirkin, C. The evolution of dip-pen nanolithography.
Angewandte Chemie-International Edition 43, 30-45 (2004).
70.
Doraiswamy, A., Dunaway, T., Wilker, J. & Narayan, R. Inkjet printing of bioadhesives.
Journal of biomedical materials research. Part B, Applied biomaterials 89, 28 (2008).
71.
Shin, D., et al. Protein patterning by maskless photolithography on hydrophilic polymer-
grafted surface. Biosensors and Bioelectronics 19, 485-494 (2003).
72.
Olle, E., et al. Comparison of antibody array substrates and the use of glycerol to
normalize spot morphology. Experimental and molecular pathology 79, 206-209 (2005).
73.
Zhu, H. & Snyder, M. Protein arrays and microarrays. Current Opinion in Chemical
Biology 5, 40-45 (2001).
74.
Kumar, A. & Whitesides, G.M. Features of gold having micrometer to centimeter
dimensions can be formed through a combination of stamping with an elastomeric stamp and an
alkanethiol "ink"followed by chemical etching. Applied Physics Letters 63, 2002-2004 (1993).
75.
Kumar, A., Biebuyck, H.A. & Whitesides, G.M. Patterning self-assembled monolayers:
applications in materials science. Langmuir 10, 1498-1511 (1994).
76.
Kumar, A. & Whitesides, G.M. Patterned condensation figures as optical diffraction
gratings. Science 263, 60 (1994).
77.
Tan, J., Tien, J. & Chen, C. Microcontact printing of proteins on mixed self-assembled
monolayers. Langmuir 18, 519-523 (2002).
78.
Alavi, M.R., Shukla, H.D., Whitaker, B., Arnold, J. & Shahamat, M. Attachment and
biofilm formation of Mycobacterium marinum on a hydrophobic surface at the air interface.
World Journal of Microbiology and Biotechnology 23, 93-101 (2007).
79.
Miller, R., Gr6ger, G., Hiller, K.A., Schmalz, G. & Ruhl, S. Fluorescence-based bacterial
46
overlay method for simultaneous in situ quantification of surface-attached bacteria. Applied and
environmental microbiology 73, 2653-2660 (2007).
80.
Ostuni, E., Grzybowski, B.A., Mrksich, M., Roberts, C.S. & Whitesides, G.M. Adsorption
of proteins to hydrophobic sites on mixed self-assembled monolayers. Langmuir 19, 1861-1872
(2003).
47
Chapter 2
Disruption of biofilm patches by an air plug reveals
microscale heterogeneity
Bacteria often adhere to surfaces, where they form communities known as biofilms that
display high resistance to both chemical and mechanical insults. Despite the importance of
biofouling in health and ecology, the mechanisms governing biofilm development remain poorly
understood, particularly for what concems biofilm disruption by mechanical forces. Here we
present experiments with surface-treated microfluidic devices that reveal a new mode of biofilm
disruption by fluid flow. We show that, after the passage of an air plug, the break-up of the
residual thin liquid layer scrapes and rearranges bacteria on the surface, resulting in a
characteristic, semi-regular pattern of holes in the biofilm. We demonstrate that this pattern
depends on the local biofilm age and correlates with the spatial distribution of the extracellular
polymeric matrix, revealing that biofilm detachment can be highly heterogeneous and is
governed by the microscale variability in adhesion strength. Because few survivors suffice to
regrow a biofilm, these results point at the importance of considering microscale heterogeneity
when designing and assessing the effectiveness of biofilm removal strategies by mechanical
forces.
48
2.1 Introduction
Biofilms are surface-associated microbial communities encased in a self-secreted matrix of
extracellular polymeric substances (EPS). Biofilms account for the largest fraction of bacterial
biomass on the planet, and often have deleterious effects in natural, industrial and medical
settings 1. In the environment, biofilms can mobilize heavy metals such as mercury and arsenic
causing stream or soil contamination
2
and create oxygen depletion zones in lakes, rivers and
coastal habitats 3. In industrial processes, the formation of biofilms is responsible for huge
economic losses (in the billions of dollars yearly in the US alone) resulting from biofouling and
biocorrosion 4, which leads to equipment clogging and damage, and product contamination 5. In
medicine, biofilms represent the major source of infections associated with catheters and
implanted devices 6. Despite the importance of finding effective methods for biofilm removal in
these and other applications, our understanding of biofilm development and in particular of the
mechanisms responsible for biofilm detachment remains far from complete.
Detachment refers to the release of bacterial cells from the surface-associated biofilm into the
bulk fluid. Several factors can contribute to detachment, including matrix-degrading enzymes 7
nutrient levels and microbial growth status 8, fluid shear stress 9,10 and quorum-sensing signals " .
Mechanical forces associated with fluid flow have long been used to detach biofilms from
surfaces
8,12,
with a prominent example being potential, effective approach to remove biofilms
49
from surfaces involves the passage of bubbles or air plugs over the surface
13.
Bubbles pluck
bacteria off a surface when the three-phase line (separating the liquid, the air and the solid
surface) contacts the bacteria
14.
By producing high shear stresses, traveling air-liquid interfaces
generate large detachment forces in a broad range of environments, including in the oral cavity
during eating, speaking, drinking and swallowing
blinking
16
15,
on the eye and on contact lenses during
, and on rocks and ship hulls in aquatic systems 17I . However, previous research on
biofilm detachment by air plugs has focused on endpoint measurements to quantify the net
amount of biofilm removed, whereas the mode of biofilm disruption has remained unexplored.
Here we show that, for early-stage biofilms (i.e., when bacterial colonies are essentially
organized as monolayers), insult by mechanical forces results in a new phenomenon, whereby
the passage of an air plug opens regular holes in the biofilm but fails to completely remove it. We
rationalize this finding in terms of the competition between dislodging shear forces and the
spatially varying adhesion strength resulting from intrinsic heterogeneity in the age and matrix
production of different areas within a biofilm.
50
2.2 Results and Discussion
We studied the formation and disruption of controlled PseudomonasaeruginosaPAO1 biofilm
patches on the glass bottom of a microfluidic channel (Fig. 2-1 a).
(a)
inlet
controlled air plug f
m
channe/
20 PM/s
H
outlet
hydrophobi
f
4Wb
*
I.
W,~4
41
Si&L
Figure 2-1 An air plug creates a characteristic pattern of holes in a biofilm. (a) Schematic of the
microfluidic setup, showing the geometry of the microchannel and the experimental method.
Pseudomonasaeruginosabacteria preferentially attached to the hydrophobic patches. After that a
certain growth time (4 h, 8 h, or 12 h), a controlled air plug was injected in the channel at the
mean flow speed of 250 pm/s. (b) Residual biofilm after the passage of the air plug, showing the
semi-regular hole pattern for different patch sizes (from left to right: 400, 300, 200 and 100 gm).
The biofilm growth time was 8 h. Scale bar = 100 pm.
51
Biofilm patches formed by the preferential adhesion of P aeruginosacells to hydrophobic
square patches, previously created on the glass by microcontact printing (P CP;
18-20;
Fig. 2-2;
Methods).
(a)
(b)
(c)
Glass substrate
Placing PDMS microstamp inked with OTS
Stamping
(d)
OTS pattern transfer
(e)
Covered by microfluidic channel
()
Filled by bacterial solution
Figure 2-2 Schematic of the microcontact printing (pCP) technique used for depositing
hydrophobic patches onto glass substrates.
A dilute bacterial suspension (optical density OD600 ~ 0.2) was injected in the channel and
incubated under quiescent conditions for 1 h, allowing cells to attach to the channel's surfaces.
52
Then, the bacterial suspension was replaced by a minimal culture medium (M63), which was
flown continuously at 3 pl/min (average flow velocity = 250 pm/s) to supply adhering cells with
nutrients. Over the course of a few hours, P aeruginosacells progressively covered the surface
of the hydrophobic patches. While some bacteria attached to the surface outside of the patches,
most bacteria adhered onto the patches (Fig. 2-3 a,b,c, ), where cell adhesion is greatly favored
by the substrate's strong hydrophobicity
21,22.
The concentration of adhering cells could be
controlled by varying the concentration of the chemical (OTS; see Methods) used in printing the
patches (Fig. 2-4).
To determine the effect of a mechanical insult on an early-stage biofilm, biofilm patches were
exposed to the controlled passage of an air plug after 8 h of growth. The air plug was created by
rapidly switching injection from media to air, creating a 2.5 mm long air plug traveling at 250
[tm/s (Methods). The air plug traveled over each patch in less than 10 s, causing a dramatic and
highly characteristic disruption of the original biofilm patch: the resulting biofilms were in the
shape of a semi-regular pattern of holes, from which bacteria had been entirely removed,
separated by 'bacterial levees', consisting of a concentrated monolayer of cells (Fig. 2-1 b).
53
II
Uf- Oz
.
1.
.*f
-0
AM
(d)
(e
(I)
7
k
AW
two-
As
0.
t..
(g)
60
Before air bubble
After air bubble
50
a:
30
20
.
8
ue4
20
10
10
8
a
6
16
4
14
.
10
2
0
L.
0
0
4
20
212
Z
0
22
After air bubble
18
C
40
30
U)
Before air bubble
60
50
40
(h)
10 g
70
08
4
12
Time (hour)
8
12
Time (hour)
Figure 2-3 The effect of the air plug strongly depends on biofilm age. (a-f) Biofilms grown for
different times (4 h, 8 h, 12 h), shown before (a-c) and after (d-f) the passage of an air plug. The
air plug traveled from left to right. (g) Surface coverage (fraction of the surface of a patch
covered by bacteria) before and after the passage of an air plug, for biofilms of different age
(grayscale bars). The red curve shows the normalized change in surface coverage. (h) Free
surface distance between cells (grayscale bars) and fractal dimension of the cell distribution (red
symbols) measured before and after the passage of an air plug, for biofilms of different age.
54
480
4601
~4401
420
400
380
IL
360
340
0
200
400
600
800
1000
Distance (in pixel, 0.4prIpixeI)
Figure 2-4 Paeruginosaadsorption on the OTS step-wise gradient. The contact-printing time of
OTS is gradually increased from inner out. The color plot (top figure) was generated using the
value of fluorescence intensity over the distance (bottom figure). The higher intensity indicates
the higher adsorption density of Paeruginosa.
Image analysis showed that holes, which varied in shape but displayed no obvious asymmetry
associated with the flow direction, had an equivalent radius of 6.5 ~ 8 pim and the porosity of the
end-state biofilm was ~68%. This pattern was highly consistent among different patches within
the same microchannel and among replicate experiments (Fig. 2-5). Experiments with patches of
different size (100, 200, 300 and 400 pim squares) showed that the pattern and its porosity were
independent of patch size within the range tested (Fig. 2-5).
55
100
20
80
I
C
T
60
IT
15u,
10
0
0
40
5
20
0
100x 100
200x 200
300 x300
400x400
0
Patch size (sm2 )
Figure 2-5 Porosity and equivalent hole radius on the size of patches. The top four panels
correspond to the size of patches.
To confirm that the holes in the biofilm resulted from the detachment of bacteria, rather than
the detachment of the underlying OTS layer, we scanned the ruptured biofilms using atomic
force microscopy (AFM). Results showed that the height of the surface inside the holes was
greater than the uncoated glass surface outside the patches by an amount corresponding to the
original thickness of the OTS layer (Fig. 2-6), indicating that cell detachment - not OTS
detachment - was responsible for the hole patterns. In addition, chemical analysis of the surface
using scanning electron microscopy coupled with energy dispersive X-ray spectroscopy
(SEM/EDX) showed that the OTS coating is stable even under high shear flows (Fig. 2-7).
56
(a)
(b)
Figure 2-6 AFM characterization of a ruptured P aeruginosabiofilm surface. (a) The residual
biofilm is shown in yellow, the region from which bacteria have detached is in brown. (b) The
height profile of the surfaces along the white line in panel (a). The red and green arrowheads
correspond to the same symbols in panel (a). Note the different height of the surface in the hole
(left green arrow) and on the glass surface (right green arrow), showing that OTS coating in the
hole has not detached.
57
(b)
(a)
(a)
C peak
+
1'
~
Si peak
aa3' IQiEJJbh5W&5
nV5 RII
LO
-XAVA IIBHIII~~
1U1
s
Figure 2-7 (a) Secondary electron micrographs with spot EDX analysis of P aeruginosabiofilm
patches. . EDX is an analytical technique used for the chemical characterization of a sample.
Briefly, the interaction between X-ray excitation and the sample surface is used to determine the
surface composition. Carbon-containing substances are dominant in the detached regions of cell
patches (a and b, top plot), whereas silicon-containing substances are dominant in the uncoated
glass regions (a and b, bottom plot). Since the biofilm holes do not show abundant Si, it infers
that the regions are still covered by OTS or by cells.
The adhesion properties of an early-stage biofilm are strongly dependent on the biofilm
growth time. This is clearly revealed by comparing experiments in which the air plug was
injected after 4, 8 and 12 h of biofilm growth, respectively (Fig. 2-3). For 4 h old biofilms, the air
plug reduced the surface coverage within patches by nearly two thirds, from 32% to 13%. For 8
h old biofilms, the reduction in surface coverage caused by the air plug was considerably smaller
58
(from 45% to 39%). For 12 h old biofilms, the reduction was essentially negligible (from 49% to
47%) and the air plug produced no visible change of the biofilm structure.
This trend is supported by an analysis of the mean distance between individual cells in the
biofilm before and after the air plug passed. Imaging at 40X allowed us to resolve individual
cells in the biofilm monolayer. The mean distance between cells was computed as
N
L=
4
N
after having measured the area Ai of each hole, for all N holes in a patch, via image analysis
(Methods). This analysis revealed that the mean distance between cells changed considerably
upon passage of the air plug for the younger biofilms: for 4 h and 8 h old biofilms, L increased
from 4.1 pm to 7.9 pm and from 3.1 pm to 5.8 pm, respectively. In contrast, L remained nearly
unchanged for the 12 h biofilm (from 2.1 pm to 2.3 pm). Taken together, these measurements
indicate that an air plug largely removes 4 h old biofilms, primarily rearranges cells in 8 h old
biofilms without detaching them, and has little effect on 12 h old biofilms. The rearrangement of
the 8 h old biofilm is evident from the imaging (Fig. 2-3e) and is further supported by a
considerable change in the fractal dimension of the cell distribution, from 1.92 to 1.68.
59
44.-
11(
350
Qbo
300
260
1
d'
2
01
E
.~200
S160
100
(b)c
fte
ortm
oit.()Tm
(c)rs
ofthho
__are,_fothethrehlsshwni_(b)
50
0
1
2
3
4
Time (9n)
Figure 2-8 Bioflm holes and hole formation dynamics. (a) Close-up view (60X objective) of
the bacterial distribution at the edge of a4 00x400 gim 2 bioflm patch after the passage of an air
plug. The doffed line denotes the edge of patch. (b) Zoomed-in view of the dynamics of hole
formnation at I s intervals. The boundaries of three holes are identified with red, blue and yellow
at each of the four time points. (c) Time course of the hole area, for the three holes shown in (b).
Black symbols denote the average of the three measurements.
60
Imaging of the hole formation dynamics at high temporal resolution (50 frames/s) revealed
that the hole pattern resulted from the rupturing of the residual thin liquid film between the
channel surface and the air plug at discrete locations. The ensuing movement of the contact line
scraped bacteria outwards from the holes, to form 'bacterial levees' between adjacent holes (Fig.
2-8).
An air plug traveling over a solid surface remains separated from it by a thin layer of liquid,
whose thickness depends on the capillary number, Ca = pU/a, which measures the relative
importance of viscous forces and capillary forces. Here, p is the dynamic viscosity of the liquid,
U is the air plug velocity, and c is the interfacial tension between liquid and air. In our
experiments, the air plug traveled at U = 250 pm/s in water (p = 10-3 Pa s;
Y= 72x10-3 N/m),
resulting in Ca = 3.5x 10- . For Ca << 1, the liquid film thickness, h, follows Bretherton's law 2,,
h/H z Ca/ 3 23 where H = 50 pm is the microchannel's height 24 This results in an estimated
liquid film thickness of <0.1 pm. Thus, the air plug would create, on a flat surface, a liquid film
that is thinner than the thickness of the bacteria (- 1 pm). This suggests that the uneven carpet of
surface-adhering bacteria, over which the liquid film must travel, represents a collection of
comparatively tall obstacles for the advancing liquid film, providing a mechanism for its rupture.
61
Calculations show that the rupture of the thin film is consistent with an evaporation-driven
instability. An evaporation rate of 5x 100 Kg/m-2 s-1, determined assuming room temperature and
50% relative humidity (RH)
25,
indicates that the 0.1 pm thick water film evaporates in -2 s. A
sensitivity analysis on the relative humidity, RH, which is unfortunately unknown in our
experiments, reveals that our conclusions are robust for even large variations in RH. For RH
values ranging from 10% to 90%, the evaporation rate ranges from 10- 5 to 10-4 Kg/m-2 s-,which
gives an evaporation time of Is to 4s for a 0.1 pm thick water film
26.
These times are in all cases
smaller than the ~10 s passage time of the air bubble, indicating that, irrespective of the precise
value of the RH in the air bubble, the proposed mechanism allows the evaporation of the entire
film over the time of the bubble's passage. Thus, over the 10 s that the air plug resides over a
given point on the surface (the plug's length divided by its speed), the thin film can completely
evaporate. Its evaporation timescale is consistent with the dynamics of hole opening (Fig. 2-8b)
and indicates that evaporation can rapidly thin the liquid layer, leading to the deformation of its
free surface in the voids between bacteria and to its ultimate rupture. The resulting three-phase
contact line moves radially outward form the point of rupture to minimize surface energy. This
process is akin to "confined dewetting lithography", where the thinning and rupture of thin liquid
film is used to arrange surface-residing colloidal particles into defined patterns
62
26
.
Because
bacteria are thicker than the liquid film, they protrude above the air-liquid interface of the
dewetting hole's receding contact line: the resulting capillary force can dislodge individual cells,
which are transported outwards by the receding contact line until the latter becomes pinned when
the accumulated bacteria form a levee between the opening hole and an adjacent hole (Fig. 2-8b).
The key role of evaporation in this process is supported by the observation that the hole's area
grows linearly with time (Fig 2-8c). This quasi-linear dependence of the hole are on time is
broadly consistent with a role of evaporation in driving the rupture
27.
This mechanism works at
intermediate adhesion strengths (8 h old biofilm), whereas for younger biofilms cells are swept
away and for older biofilms the capillary force is insufficient to scrape cells across the surface.
While microorganisms maintain their individual functioning during biofilm development, EPS
formation initiates irreversible attachment
28,29
. The EPS matrix provides several functions for
the biofilm, such as protecting bacteria from environmental threats, providing mechanical
stability, and degrading macromolecules to be used by the cells 28. Some reports indicate that
EPS content is not associated with biofilm adhesion strength
I8.
This apparent discrepancy is
likely due to the complex dependency of detachment on temperature, nutrient availability,
hydrodynamics, and presence of chemical toxins
30.
Alternation in extracellular components
during biofilm development can have a potential effect on biofilm detachment
63
.
A correlation analysis shows that the spatio-temporal distribution of EPS is the primary
determinant of the local strength of adhesion, strongly suggesting that it is therefore responsible
for the observed heterogeneous patterns. To quantify the distribution of EPS, we injected
fluorescently labeled lectins (wheat germ agglutinin (WGA) conjugated with tetramethyl
rhodamine isothicyanate (TRITC);
32)
in the microchannel at different times of the biofilm's
growth (2 h, 4 h, 6 h and 8 h). Interestingly, although the distribution of cells in each patch is
rather uniform (Fig. 2-9a), the distribution of EPS before passage of the air plug is heterogeneous
(Fig. 2-9c), likely reflecting intra-population variability in EPS production and different surface
colonization times, in line with recent discoveries about the role of EPS in guiding the formation
of micro-colonies
.
Furthermore, there is a significant correlation between the pre-air-plug
EPS distribution and the post-air-plug cell distribution, and this correlation is 5-fold larger than
the correlation between the pre-air-plug EPS distribution and the pre-air-plug cell distribution
(Fig. 2-9e-f). Furthermore, we found that the passage of the air plug had a negligible effect on
the distribution of EPS on the surface (Fig. 2-9d), which further confirms the role of the EPS in
protecting biofilm-forming bacteria from mechanical forces.
64
cells
(a)
before
(b)
EPS
after
(e)
(d)
before
(c)
after
(f)
0.5
C:
0.4
0
0.3
0
0
0
0.2
T
0.1
0
BS
EPS bef.
Cells bef.
BS
EPS bef.
Cells aft.
Figure 2-9 The hole pattern correlates with the EPS distribution. (a-b) Distribution of
fluorescently-labelled P aeruginosa after 8 h of biofilm growth, immediately before (a) and
immediately after (b) the passage of an air plug. (c-d) Distribution of EPS, fluorescently labeled
using lectins (Methods), at the same times and position as (a-b). (e) Overlay of the EPS
distribution before the air plug (red) and the bacteria distribution after the air plug (green). Data
are from a different experiment than (a-d). (f) Normalized cross-correlation between the EPS
distribution before the air plug and the cell distribution before (second bar form left) and after
(fourth bar from left) the air plug. Error bars correspond to the standard error over three replicate
experiments. The first and third bars represent results from a bootstrapping analysis (BS), where
the EPS distribution before the air plug is correlated with random permutations of the cell
distribution before (first bar) and after (third bar) the air plug. Note the lack of correlation in
these controls.
65
These results indicate that holes in the biofilm structure formed where there was the least
amount of EPS before the air plug arrived, and that cells from regions of low EPS concentration
were scraped into the regions of highest EPS concentration, where they - together with the EPS
- formed levees that prevented further hole expansion.
2.3 Conclusion
These findings may have significant impact and potential applications when shear-based
removal strategies are designed, as biofilm removal might be confused with biofilm
rearrangement. In the past, biofilm has been assumed to be homogeneous and biofilm models
have been developed based on that assumption. However, the results presented in this paper
show that the heterogeneous distribution of extracellular matrix play a crucial role in the
detachment mode of biofilm under a mechanical insult or hydrodynamic shear. A recent study
showed that P aeruginosasecretes Psl during migration
. The resulting Psl trails influence the
motions of succeeding P aeruginosacells on the surface. Characteristic length in separated holes
in structures showed ~ 10 ptm at 5h, which is similar to our finding, 6.5~8pm hole radius after
the bubble passage on 8h biofilm growth. Thus, the report supports that where more matrix is
initially deposited, that is where the biofilm is strongest at the microscale, and those will be the
strongholds that will resist mechanical insult, even acting as collection points for other bacteria.
66
This study demonstrates that high fluid shear can be effective in stimulating bacterial
detachment. The addition of air bubble to the flow allows the detachment of tenaciously adhering
bacteria not detached by flow alone. The principles outlined are expected to have general validity
for bacterial detachment from surfaces by fluid flow and passing air bubble.
2.4 Materials and Methods
Materials. Octadecyltrichlorosilane (OTS)
[CH 3(CH 2 ) 17 SiCl 3] (Aldrich, 97%), hexane
(Aldrich, anhydrous, 99%), and fluorescein isothiocyanate-bovine serum albumin (FITC-BSA)
and tetramethyl rhodamine isothiocyanate-wheat
germ agglutinin (TRITC-WGA) were
purchased from Sigma-Aldrich (St. Louis, MO, USA). Polydimethylsiloxane (PDMS, Sylgard
184) was purchased from Dow Coming.
Bacterial strain and growth condition. The wild-type strain of Pseudomonas aeruginosa
PAOI was used for this study (courtesy of George O'Toole, Dartmouth University). For the
preparation of the cell culture, cells from freezer stocks were inoculated in LB medium (10 g/L
NaCl, 5g/L yeast extract, 10 g/L tryptone) at 30'C under shaking (150 rpm). Cells were
resuspended in fresh LB medium and incubated at 37 0 C under shaking (180 rpm) up to OD6 00 =
0.2, corresponding to early exponential phase. An aliquot of this cell suspension was injected for
I h in the microfluidic channel. Thereafter, continuous injection of M63 minimal medium (100
67
mM of KH 2 PO 4 ,; 15mM of (NH 4 ) 2 SO 4 ; 1.8 uM of FeSO 4 ; 1.0mM of MgSO 4 ) and 0.5% glucose
at a constant volumetric flow rate of 3 pl/min began, and was maintained for 4, 8 or 12 hours, to
produce biofilms of different stage of maturation. This flow rate corresponds to an average flow
velocity of 250 tm/s and a shear rate at the bottom glass wall of-30 s- .
Air plug generation. To introduce controlled air plugs in the microchannel, we used a threeway valve: one inlet for the bacteria solution (used for initial injection), one inlet for the bacteriafree M63 media (used for 4 h, 8 h or 12 h), and one inlet for atmospheric air. To control the
velocity of the air plug, the valve and a syringe pump were connected to a microchannel inlet and
outlet. By opening and closing the valve port, air plugs were introduced into the microchannel.
PDMS stamp fabrication for surface patterning. PDMS stamps were fabricated by curing
the prepolymer on silicon masters patterned with SU-8 photoresist (SU-8 2050, MicroChem, MA,
USA) using conventional photolithography. The masters used for patterning had recessing
features, which resulted in PDMS replicas with protruding features. To assist in removal of cured
PDMS from the masters, the SU-8 masters were silanized overnight by exposure to the vapor of
1,1,2,2,-tetrahydrooctyl-1-trichlorosilane, CF 3 (CF 2) 6(CH 2) 2SiCl 3 . To cure the PDMS prepolymer,
a mixture of 10:1 silicon elastomer and the curing agent was poured on the master and held at
65'C for 2 h. The PDMS replica was then peeled from the silicon master.
68
Generation of patterned hydrophobic coatings. Hydrophobic patterns of OTS on the glass
substrate were made by using microcontact printing (pCP), as shown in Fig. 2-2. The PDMS
stamp was inked with a 2 mM hexane solution of OTS and dried in air for 5 min, then placed in
contact with the glass substrate at room temperature for 30 s. The stamp was carefully peeled off
and the substrate was rinsed with 2-propanol (IPA) and DI, then dried. Because the
trichlorosilane reagents are sensitive to the water content and temperature of the printing ambient,
the relative humidity and the temperature of the room were kept constant at -50.-55% and ~2224 0 C, respectively. Additional experiments showed that the concentration of OTS alters the
adsorption density of P aeruginosa (Fig. 2-3). The contact-printing time of OTS is gradually
increased from inner out to generate the OTS step-wise gradient.
Microscopy. Conventional epifluorescence microscopy imaging was performed using an
inverted microscope (Nikon TE-2000E) fitted with GFP and RFP filter sets. Images were
acquired with 40x and 60x objectives and an Andor iXon CCD camera (50 frame / s) cooled to 65'C. Image analysis was performed using built-in plugins of the ImageJ software
(http://rsbweb.nih.gov/ij/).
69
Bibliography
1. Costerton, J.W., Lewandowski, Z., Caldwell, D.E., Korber, D.R. & Lappin-Scott, H.M.
Microbial biofilms. Annual Reviews in Microbiology 49, 711-745 (1995).
2. Kohusovi, K., Havel, L., Vlasik, P. & Tonika, J. A long-term survey of heavy metals and
specific organic compounds in biofilms, sediments, and surface water in a heavily affected river
in the Czech Republic. Environmental monitoring and assessment 174, 555-572 (2011).
3. Lawrence, J., et al. Microscale and molecular assessment of impacts of nickel, nutrients, and
oxygen level on structure and function of river biofilm communities. Applied and environmental
microbiology 70, 4326-4339 (2004).
4. Flemming, H.-C. Biofilms and environmental protection. Water Science & Technology 27, 110(1993).
5. Brooks, J.D. & Flint, S.H. Biofilms in the food industry: problems and potential solutions.
International Journal of Food Science & Technology 43, 2163-2176 (2008).
6. Costerton, J., Stewart, P. & Greenberg, E. Bacterial biofilms: a common cause of persistent
infections. Science 284, 1318 (1999).
7. Allison, D.G., Ruiz, B., SanJose, C., Jaspe, A. & Gilbert, P. Extracellular products as
mediators of the formation and detachment of Pseudomonas fluorescens biofilms. Fems
Microbiology Letters 167, 179-184 (1998).
8. Stewart, P.S. A model of biofilm detachment. Biotechnology and Bioengineering 41, 111-117
(1993).
9. Peyton, B.M. Effects of shear stress and substrate loading rate on Pseudomonas aeruginosa
biofilm thickness and density. Water Research 30, 29-36 (1996).
70
10.
Picioreanu, C., van Loosdrecht, M.C.M. & Heijnen, J.J. Two-dimensional model of biofilm
detachment caused by internal stress from liquid flow. Biotechnology & Bioengineering 72, 205218 (2001).
11.
Hentzer, M., et al. Inhibition of quorum sensing in Pseudomonas aeruginosa biofilm
bacteria by a halogenated furanone compound. Microbiology 148, 87-102 (2002).
12.
Picioreanu, C., van Loosdrecht, M.C. & Heijnen, J.J. Two-dimensional model of biofilm
detachment caused by internal stress from liquid flow. Biotechnology & Bioengineering 72, 205218 (2001).
13.
Pitt, W.G., McBride, M.O., Barton, A.J. & Sagers, R.D. Air-water interface displaces
adsorbed bacteria. Biomaterials 14, 605-608 (1993).
14.
G6mez-Suirez, C., Busscher, H.J. & van der Mei, H.C. Analysis of bacterial detachment
from substratum surfaces by the passage of air-liquid interfaces. Applied and environmental
microbiology 67, 2531-2537 (2001).
15.
Busscher, H.J. & van der Mei, H.C. How do bacteria know they are on a surface and
regulate their response to an adhering state? PLoS Pathogens 8, e1002440 (2012).
16.
Bruinsma, G., Van der Mei, H. & Busscher, H. Bacterial adhesion to surface hydrophilic
and hydrophobic contact lenses. Biomaterials 22, 3217-3224 (2001).
17.
Cooksey, K. & Wigglesworth-Cooksey, B. Adhesion of bacteria and diatoms to surfaces in
the sea: a review. Aquatic Microbial Ecology 9, 87-96 (1995).
18.
Kumar, A., Biebuyck, H.A. & Whitesides, G.M. Patterning self-assembled monolayers:
applications in materials science. Langmuir 10, 1498-1511 (1994).
19.
Kumar, A. & Whitesides, G.M. Features of gold having micrometer to centimeter
dimensions can be formed through a combination of stamping with an elastomeric stamp and an
71
alkanethiol "ink"followed by chemical etching. Applied Physics Letters 63, 2002-2004 (1993).
20.
Kumar, A. & Whitesides, G.M. Patterned condensation figures as optical diffraction
gratings. Science 263, 60 (1994).
21.
Alavi, M.R., Shukla, H.D., Whitaker, B., Arnold, J. & Shahamat, M. Attachment and
biofilm formation of Mycobacterium marinum on a hydrophobic surface at the air interface.
World Journal of Microbiology and Biotechnology 23, 93-101 (2007).
22.
Ostuni, E., Grzybowski, B.A., Mrksich, M., Roberts, C.S. & Whitesides, G.M. Adsorption
of proteins to hydrophobic sites on mixed self-assembled monolayers. Langmuir 19, 1861-1872
(2003).
23.
Angeli, P. & Gavriilidis, A. Hydrodynamics of Taylor flow in small channels: a review.
Proceedings of the Institution of Mechanical Engineers, Part C: Journal of Mechanical
Engineering Science 222, 737-751 (2008).
24.
Han, Y. & Shikazono, N. Measurement of liquid film thickness in micro square channel.
International Journal of Multiphase Flow 35, 896-903 (2009).
25.
Sartori, E. A critical review on equations employed for the calculation of the evaporation
rate from free water surfaces. Solar energy 68, 77-89 (2000).
26.
Celio, H., Barton, E. & Stevenson, K.J. Patterned assembly of colloidal particles by
confined dewetting lithography. Langmuir 22, 11426-11435 (2006).
27.
Churaev, N.V. Liquid and vapor flows in porous bodies: surface phenomena, (Taylor &
Francis, 2000).
28.
Flemming, H.-C., Neu, T.R. & Wozniak, D.J. The EPS matrix: the "house of biofilm cells".
Journal of Bacteriology 189, 7945-7947 (2007).
29.
Flemming, H.-C. & Wingender, J. The biofilm matrix. Nature Reviews Microbiology 8,
72
623-633 (2010).
30.
Xavier, J.B., Picioreanu, C., Rani, S.A., van Loosdrecht, M.C. & Stewart, P.S. Biofilm-
control strategies based on enzymic disruption of the extracellular polymeric substance matrix-a
modelling study. Microbiology 151, 3817-3832 (2005).
31.
Liu, Y. & Tay, J.-H. Detachment forces and their influence on the structure and metabolic
behaviour of biofilms. World Journal of Microbiology and Biotechnology 17, 111-117 (2001).
32.
Strathmann, M., Wingender, J. & Flemming, H.-C. Application of fluorescently labelled
lectins for the visualization and biochemical characterization of polysaccharides in biofilms of
Pseudomonasaeruginosa.Journal of Microbiological Methods 50, 237-248 (2002).
33.
Zhao, K., et al. Psl trails guide exploration and microcolony formation in Pseudomonas
aeruginosa biofilms. Nature 497, 388-391 (2013).
73
The Effect of Patch Size and Fluid Flow on Quorum
Chapter 3
Sensing in Biofilms
Biofilm forming bacteria are industrially and medically relevant organisms that are
exceptionally resistant to a wide variety antimicrobial treatments. This resistance is due in part to
a biofilm forming bacteria's ability to sense and communicate with neighboring bacterial. As a
result of this intercellular communication, i.e. Quorum Sensing (QS), bacteria are able to
cooperate as a complex community. This communication system is used to modulate important
facets of biofilm behavior and thus is an attractive target for biofilm control and potential
antimicrobial agents. Understanding the significance of hydrodynamics in QS is crucial if we
consider the habitats in which biofilms tend to form, where fluid flow is often prevalent.
3.1 Introduction
Quorum sensing (QS) is an important mechanism by which bacteria form and maintain complex
biofilms through the ability to communicate and to alter behavior in response to the presence of
other bacteria
[-3.
Bacterial cells produce signaling molecules that are secreted into the
environment, where they disperse by diffusion or advection. In general, to sense their population
density, gram negative bacteria use small chemical molecules called acylated homoserine
74
lactones (AHLs) 4 and gram positive bacteria use oligo-peptides 5, both collectively referred to as
autoinducers (ALs). Typically, an autoinducer molecule induces the transcription of a set of genes
that includes the gene encoding the autoinducer-producing enzyme, which results in a positive
feedback loop
6.
The autoinducer can trigger a response by the cell if its concentration - regulated
by the rates of production, decay and mass transfer integrated over time - reaches a threshold at
the cell's location (Fig. 3-1).
AH~~AH
Lifs
dinseot---
AHL diffuse ou
.--
--
p
*p-
Ti me
Figure 3-1 LasI/LasR-RhlI/RhlR quorum sensing system in Pseudomonas aeruginosa. The LasI
protein produces the homoserine lactone signaling molecule N-(3-oxododecanoyl)-homoserine
lactone, autoinduceri (All, blue circles), and the RhlI protein synthesizes N-(butryl)-homoserine
lactone, autoinducer2 (AI2, red triangles). Activated LasR (with All) activates transcription of
rhlR, which means the las QS system controls RhlR at the transcriptional level. A~l also controls
the activity of RhlR at the post-translational level by binding to RhlR, which decreases AI2
binding to RhlR, therefore inhibiting expression of rhlA and other genes. Presumably, this action
ensures that the LasI/LasR circuit is established prior to the establishment of the RhlI/RhlR
circuit. The Pseudomonas quinolone signal (PQS) is an additional regulatory link between the
Las and Rhl QS circuits.
75
The existence of a threshold cell concentration for the production of bioluminescence in the
marine symbiotic bacteria Vibrio fischeri 7, the formation of fruiting bodies in myxobacteria
8
and the development of competence for genetic transformation in Streptococci 9, are among the
lines of evidence for cell-to-cell signaling systems in bacterial communities. At present, QS is
defined as a cell-density-dependent bacterial intercellular signaling mechanism that enables
bacteria to coordinate the expression of certain genes to coordinate group behaviors.
Several physical, biological, and chemical factors have the potential to influence QS in biofilm
systems lo. Among these, the hydrodynamic environment can have a significant influence on QS.
Janakiraman et al. " studied biofilm growth and QS in microfluidic chambers and with a onedimensional mathematical model, at low Reynolds numbers (Re<<l). They found that the flow
rate greatly affects QS: at high flow rates, the transport rate of AHL out of the biofilm can be so
high that the AHL concentration does not reach the threshold required for induction. Purevdorj
and Costerton
12
emphasize that the washout of signal molecules from the bulk fluid surrounding
bacterial microcolonies is expected to increase the concentration gradient between the biofilm
and the bulk liquid, in turn driving a larger diffusive flux of signal molecules out of the biofilm.
Muller et al.
13
mathematically studied QS induction in biofilm patches. They found that
induction can occur for a lower number of cells in a patchy biofilm than in a homogeneous
76
biofilm, because the cell density (and consequently, the AHL concentration) is higher in small
cell clusters.
QS is gaining increasing interest as a promising treatment target, e.g., for pathogenic bacteria
14,
which raises the interest in understanding environmental controls on QS. QS may depend on
the spatial properties of the environment. In particular, it has been recognized that not only a
high population density but also a confined space favor QS
15,16.
This feature cannot be captured
by most studies of QS, which focus on batch liquid cultures. Here we explore the effect of the
size of the biofilm on QS.
Although QS plays an important role in biofilm formation and maturation
17,
the combined
effect of hydrodynamics and spatial heterogeneity are still poorly understood, mostly due to
experimental difficulties. We present here microfluidic experiments to investigate the effects of
biofilm size and fluid flow on QS induction in Pseudomonas aeruginosabiofilms.
3.2 Results and Discussion
Comparison of GFP intensity between static and flowing conditions
We used three P aeurginosaPAO1 strains, which are wild-type (WT), PAO1 carrying pSMC21
(constitutively expressing GFP), PAOI carrying pMH520 (only expressing GFP when QS system
is activated. The model organism P aeruginosa PAO1 mini Tn5-rhlA::GFP (carrying pMH520)
77
excretes a variety of fluorescent compounds, including pyocyanin, pyoverdine, and pyorubin.
Since the experiments presented here rely on the detection of GFP expression as a measure of QS,
we needed to differentiate green fluorescence signals from the different sources in order to
reliably use GFP expression as a measure of QS. Under non-flowing culture conditions, the wildtype PAO1 strain (green triangles in Fig. 3-2a) shows considerable green fluorescence, with an
intensity that is similar to that of a QS marker construct (orange circles), but weaker than the
fluorescence displayed by PAO 1 pSMC21 (blue squares), which constitutively expresses GFP.
Conversely, when these strains are grown in a microchannel under flow conditions (mean
speed = 250 pm/s), fluorescence intensities for both the wild-type PAO1 strain and the QS marker
construct are considerably weaker than the PAOI pSMC21 strain (Fig. 3-2c). These observations
demonstrate that under the flow conditions used in the patch-biofilm experiments, pyocyanin and
other green fluorescent compounds are washed away by flow and contribute negligibly to overall
fluorescence intensity. Comparison of the static and flowing conditions allows us to conclude
that the fluorescence measured under fluid flow in the microchannels can be attributed to and
used as a metric of QS.
78
(a)
(c)
Static
10000
OS marker
0
---
8000
(mini-Tn"-.IA GFP)
Flowing
1400
Confined microwells
Microfluidic channels
PA1 (woGFP frmRtbe)
U-PAOI (w/GFP, from Fosler)
1: 200
6000
1' 000
I
4000
800
2000
600
0
0
(b) .
10
20
Time Mhrs)
30
0
40
10
20
30
40
Time (hrs)
(d)
Figure 3-2 Comparison of fluorescent intensity under static (a, b) and flowing (c, d) conditions
for three P aeruginosastrains: PAO1 mini Tn5-rhlA::GFP (QS marker; yelloq symbols), PAOl
wild-type (WT; greeen symbols), and PAOl pSMC21 (constitutively expressing GFP; blue
symbols). Under static conditions, fluorescent intensities increase over time for all three strains,
with the greatest increase expressed by the PAO1 pSMC21 strain. This is shown both using
quantitative fluorescence data (a) and visually in an assay plate (b). In contrast, under flowing
condition, the fluorescence of PAO1 pSMC21 was much higher and was expressed more rapidly
than the other two strains, and PAO1 WT displayed negligible fluorescence. This is shown via
quantitative imaging (c) and visually in microscopic snapshots taken after 18, 42, and 45 h of
biofilm growth in the microchannel.
79
The effect of patch size on QS
Fluorescence measurements were used as a metric for QS in biofilm patches of different size,
monitored in microfluidic channels over time. Results are compared for 3 patch sizes in Figure 33, including square patches of 50, 150 and 400 pm side length. We found that bacteria in the
smaller biofilm patches start QS earlier than those in the larger patches. However, at later times
the QS signal (normalized by patch size) is higher in the larger patches.
Specifically, we observed that the smaller patches (50 and 150 Pim) are the first to show the
induction of QS, which first occurs at t = 500 min for the 50 pm patches and at t = 600 min for
the 150 pm patches. At t = 700 min, the QS signal from the 150 pim patches overtakes the QS
signal from the 50 pm patches, and remains larger until the end of the experiment at t = 2000
min. Meanwhile, the QS signal from the 400 pim patches has a considerably slower start,
beginning only at t = 1000 min. Thereafter, however, the 400 Pim patches exhibit the fastest
growth in the QS signal. At t = 1400 min, the QS signal from the 400 patches overtakes the QS
signal from the 50 pim patches. At t = 1600 min, the QS signal from the 400 patches overtakes
also the QS signal from the 150 pim patches. Thereafter, the QS from the 400 Pim patches
remains the strongest among all three patch sizes, until the end of the experiment at t = 2000 min.
80
100
-
4002 Am2
1502
80 1-
S5()2
U)
0
C
0
U
C
0
U
U)
0
0
ir
401
20
0
I.....
600
800
1000
1200
1400
1600
1800
2000
Time (min)
Figure 3-3 Quorum sensing signal (GFP fluorescent intensity) for square biofilm patches of
different sizes, for a flow rate of 1 pl/min corresponding to a Peclet number of 8.4. The top panel
shows the time series of the fluorescent intensity for each patch size, as the mean and standard
deviation over 6 patches. The lower set of four panels shows the fluorescent intensity at four time
points. In each of the lower panels, the upper row corresponds to the raw fluorescent data, and
the lower row presents the same data in pseudo-coloring, to clarify differences. In the pseudocoloring, purple to purple-blue tones correspond to weaker QS, whereas yellow to red tones
correspond to stronger QS. The left column in each of the lower set of four panels corresponds to
a patch size of 50 pm, the middle one to a patch size of 150 pm, and the right one to a patch size
of 400 pm.
81
These observations reveal a clear and consistent pattern, never reported before, which applies
to the comparison among any two of the three patch sizes tested. The smaller patch always begin
to exhibit QS earlier, but is then overtaken in terms of the strength of the QS signal by the larger
patch, and after this cross-over it is the larger patch that continues to exhibit the strongest QS
signal for the entire duration of the observations.
That larger patches exhibit stronger QS at long times corresponds to one's intuitive
understanding of QS, which is favored by higher cell numbers in the same region. In virtue of
their larger surface area, the larger patches provide greater real estate for cell growth, hence for a
greater concentration of QS molecules. What is not intuitive and, to the best of our knowledge
has not been reported previously, is the dominance of smaller patches in terms of QS at earlier
times. This transient dominance of smaller patches represents a new feature of QS and highlights
the importance of taking explicitly into account the size of a biofilm when investigating QS. In
the paragraphs that follow, we present a conceptual model that, while not entirely verifiable with
the data at hand, provides a reasonable framework with predictions that are in line with the
unique cross-over we have observed.
The initial cell densities are highly comparable among all patch sizes. The vertical
confinement, represented by the depth of the microchannel (50 pim) is also the same for all patch
82
sizes. The main difference between the different patch sizes, then, is the higher ratio between
perimeter and surface area of the smaller patch sizes. This varies from 200 Pm / 502
[tm 2
= 0.08
pm-1 for the 50 pm patches, to 0.027 pm-1 for the 150 ptm patches, to 0.01 pm-' for the 400 pim
patches, an 8-fold variation overall. Because the vertical dimension is equal or smaller than the
horizontal extent of the patches, this suggests that cells in the smaller patches are on average
more readily supplied by nutrients in the surface-parallel (x,y) direction, and that cells in the
largest patches are on average (and particularly at the center of the patch) limited by nutrient
supply. This argument suggests that biofilms in the smaller patches should grow faster than
biofilms in the larger patches. While this conflicts with early-time measurements of biofilm
surface coverage, these early times are more likely to be dominated by vertical nutrient supply,
thus not showing the predicted nutrient limitation. Unfortunately, we lack data on biofilm
biomass at later times to corroborate this assumption.
A closely related same argument allows us to speculate on the reason for the stronger QS
signal in the larger patches at later times. The argument is again based on the differential
magnitude of lateral (i.e., surface-parallel) transport, only in this case the transport of the
autoinducer molecules. Because of the higher perimeter to surface area ratio, we speculate that
the autoinducer molecules produced by the small patches more readily escape from the region
83
directly above the cells, whereas they remain more 'trapped' in the case of the larger patches.
Together, these two arguments - based on a single mechanism applied to the two solutes that
affect QS, i.e., nutrients and autoinducer molecules - indicate that smaller patches are less
limited by nutrient supply at early times, thus growing faster and expressing QS earlier, whereas
larger patches grow more slowly, but the autoinducers that they express remain more strongly
localized, thus leading to stronger QS at later times. The differential perimeter to surface area
ratio, then, can explain the crossover between the QS signals for different patch sizes observed in
the experiments. We envisage that a mathematical model of QS and mass transport would
confirm this prediction, but this model is beyond the scope of this thesis.
The effect of fluid flow on QS
Here we investigate the effect of fluid flow on QS. Relative to static conditions, flow has two
primary roles: (1)
it contributes to biomass growth by increasing the convective supply of
nutrients, which enables increased cell growth and thus accelerates the induction of QS; and (2)
it contributes to AI mass transfer. Convective transport of A! molecules can affect the onset of
QS through the following: (1) Al molecules that are produced in a biofilm colony and diffuse
from the colony into the aqueous phase are transported downstream (i.e., locally removed),
84
which delays local up-regulation of QS; (2) downstream colonies receive Al molecules from
those upstream, which enhances up-regulation of QS in downstream colonies.
The effect of fluid flow on the transport of Al is best quantified in comparison to the transport
of Al by diffusion alone, i.e., the no-flow case. This comparison is captured by a dimensionless
parameter called the Peclet number, defined as Pe=
D
, where U is a characteristic fluid
velocity (here taken to be the mean speed of the flow), L is a characteristic length scale (here
taken to be the depth of the microchannel), and D is the molecular diffusivity of the Al. For Al in
water,
DAI
z 4.9x10 10 m2/s 10. Here we performed experiments over a range of values of the
Peclet number, Pe = 8.4, 25.5, 102.0, capturing a broad range of scenarios, from one in which
transport by fluid flow is predicted to be of comparable importance to transport by diffusion, to
two cases (Pe = 25.5, 102.0) in which flow is predicted to trump diffusion.
To study the effects of flow on QS, time series of fluorescent intensity were recorded within
biofilm patches for different flow velocities, corresponding to different Peclet numbers (Fig.3-4).
This was done for two different patch sizes. Panel a in Figure 3-4 corresponds to the case
analyzed above (Fig. 3-3, top panel), and the discussion here centers around differences
compared to the previous case.
85
(a)
(c)
Pe= 8.4
100
+
rzz~]
400
80
80
C
60
C
0
a
Pe= 102.0
100
60
C
40
40
0L
20
20
LL
A
-
600
800
(b)
1000 1200 1400
1600
0
1800 2000
0
500
(d)
Pe= 25.5
100
1500
1000
200C
Pe= 0
100
ii- 40-
0-
80
-.
80
150
S
60
60
C
0
40
40
LL
U.
20
0
20
0
500
ni
1000
1500
2000
0
Thie (min)
100
200
300
400
500
Tim (mIn)
Figure 3-4 Time series of QS intensity for different flow rates. The intensity of QS was
measured as the GFP fluorescence intensity, reported here as a percentage of QS induction. Flow
rates corresponded to average flow velocities of (a) 83 pm/s, (b) 250 pm/s, and (c) 1,000 pm/s.
In each graph, the Peclet number is given, which expresses the relative importance of transport
by advection and by diffusion (see text). For each flow rate, two different patch sizes were
investigated (150 pm, 400 pm). Data correspond to means and standard deviations over six
biofilm patches, for each case. Panel a corresponds to the top panel in Figure 3-3. Panel d
corresponds to the situation in which flow was stopped after the data in panel c (Pe = 102.0)
were collected, and shows that QS resumed upon stopping the flow.
First, for a decrease in flow rate (Pe=8.4, Fig. 4a), the 150 pm patches exhibit a later onset of
QS and the 400 pm patches also exhibit a later onset of QS compared to the case of Pe=25.5. The
86
two patch sizes, in this case, display onset of QS at around t = 600 min, t = 1000 min. Still, as
before, the bigger patches 'catch up', and the QS signal from the 400 pm patches overtakes that
from the 150 pm patches at t = 1600 min, a cross-over time that is not substantially different
from the Pe=25.5 case (Fig. 3-4b).
Second, for an increase in flow rate (Pe=102.0, Fig. 3-4c), none of the patches begins QS, and
the fluorescent signal decays over time (no increase in fluorescent intensity). This likely
indicates that the advective transport by fluid flow washed out the Al molecules, preventing them
from accumulating to a sufficient degree to initiate QS. Stopping the flow, in this latter
experiment, led to the induction of QS, confirming that a reversible process, such as the washout
of the Al molecules, was responsible for the lack of QS at Pe = 102.0.
Measurement of the QS intensity per unit biomass by confocal reflection microscopy
The results presented here demonstrate the importance of hydrodynamics in modulating biofilm
structure and QS expression. An important aspect that warrants further attention in this respect is
the quantification of biomass in the biofilms, in particular to determine whether the observed QS
results from increased biomass (e.g., due to higher nutrient supply) or from increased QS
expression per unit biomass (e.g., due to higher accumulation of autoinducer molecules).
87
The biofilms' biomass can be estimated at early times from the area coverage using phase or
bright-field microscopy. However, these methods are not suitable past the time at which the
biofilm grows beyond a monolayer. For this reason, confocal microscopy techniques were used
to determine the structure and biomass of the biofilms in three dimensions over time.
The technique presented here combines confocal reflection microscopy (CRM) and
fluorescent confocal laser scanning microscopy (FCLSM). CRM is a variation of reflection
microscopy that employs a confocal laser scanning microscope, permitting the visualization of
three-dimensional samples without fixation and fluorescent labeling 1 . This is a technique that
has not been widely employed in biofilm research in recent times, but is very powerful for
obtaining a quantitative measure of biomass in three dimensions, via a non-intrusive approach
(no fixation, no labeling). For CRM, biofilms were illuminated with a 488 or 514 nm argon laser
and the reflected light was collected through a 470-500 or 505-530 nm band pass filter,
respectively, to avoid the influence of autoflorescence. Time-lapse observations of reflectance
(CRM) and fluorescence (FCLSM) were performed simultaneously to examine the relation
between biofilm biomass (CRM) and QS expression (FCLSM). Conveniently, illumination with
a single laser generates both reflectance and fluorescence, which are quantified separately using
band-pass filters. Measurements were acquired at 3 h intervals.
88
The biomass was quantified in terms of the biofilm's bio-volume. For this purpose, CRM
images were analyzed using the COMSTAT software, which runs in MATLAB ". Thresholding
an image stack (which makes up the three-dimension CRM measurement of the biofilm at a
given point in time) results in a three-dimensional matrix with a value of one in positions where
the pixel value in the original image exceeds a threshold (presence of biofilm), and zero
otherwise (absence of biofilm). The cumulative number of biomass pixels ('ones') in all images
of a stack is multiplied by the voxel size (pixel size times vertical distance between images) and
divided by the cross-sectional area of the image stack. The resulting value is the biofilm's
volume divided by the surface area (pim 3/ pm 2), which can be interpreted as an equivalent
average biofilm thickness. Quorum sensing intensity was measured in terms of the intensity of
GFP expression, using FCLSM as customary in biofilm research.
Figure 3-5a shows how the temporal dynamics of the biofilm biomass, measured in 3D using
CRM, is correlated with the temporal dynamics of QS, measured in 3D using FCLSM. The QS
intensity is negligible for the first 18 h, even though the biomass more than doubles during that
time, going from an equivalent biofilm thickness of 1.0 to over 2.0 pim (Fig. 3-5b, top panel).
89
(b)
(a)
hrs
phase
1SO
3_
fluorescent
30
024
6130
30
62
20
6
0
0
20
lime
30
40
(hfs)
Figure 3-5 Simultaneous confocal measurements of biofim biomass and QS intensity. (a) Eight
time-lapse images showing both biomass (left, shown in grayscale) and the corresponding
fluorescent intensity (right, shown in green) captured by CRM and FCLSM, respectively. The
patch size is 150 pim
x
150 pim and the flow rate was 1pjl/min, corresponding to a Peclet number
of 8.4. The voxel size is (0.29 pim)x
x
(0.29 pim)y
x
(1.31 pim)z
=
0.11 pim3. (b) The temporal
evolution of the biofilm's biomass (red) and fluorescent intensity (green) recorded at 3 h
intervals (top), and the fluorescent intensity normalized by the biomass (bottom).
The
QS intensity per unit biomass (Fig. 3-5b, bottom panel), therefore, is very low for the first
18 h of biofllm growth. The picture changes dramatically thereafter. While the biomass still
increases, it does so at a markedly slower rate, as the biofilm increases in mean equivalent
thickness by another micrometer over 27 h, reaching a mean equivalent thickness of ~1 pim at the
end of the observations (t
expression takes off at
=
45 h). Meanwhile, the fluorescent intensity that measures
QS
~18 h, and thereafter increases at a rapid, constant rate until the end of
90
the observations. From 18 to 45 h, the fluorescent intensity goes from ~5% to ~120%, a greater
than 20-fold increase. Together, these observations result in a strong increase in QS expression
per unit biomass (Fig. 3-5b, lower panel). This indicates that (i) the increase in fluorescent
intensity is not due simply to an increase in the biofilm's biomass, but to an actual increase in the
per-unit-biomass (i.e., per cell) QS signal; and (ii) there is a rather clear point in time (18 h) at
which this increase in QS activity sets in. In summary, these results confirm that the fluorescent
signal used as a metric for QS in the first sections of this chapter is indeed to be attributed to QS
induction (presumably, after Al concentration reach the threshold to activate the QS reporter
gene), and not to a simple increase of population density. Thus, fluorescence intensity increases
in microchannel experiments are very likely due to upregulation of QS among P aeruginosaand
not simply to an accumulation of biomass.
3.3 Conclusion
These findings establish microfluidics as a new tool for the study of biofilms. This new tool
enables both accurate control over microenvironmental conditions and direct observation of the
dynamics of biofilm formation. Specifically, we have here reported observations that shed new
light on the role of physical parameters of biofilms on the biological process of quorum sensing.
91
First, we have reported the counterintuitive discovery that QS in smaller biofilms can be
triggered earlier than in larger biofilms, and have proposed a rationalization of this discovery
based on the limitations in the transport of both nutrient molecules and autoinducer molecules.
Second, we have demonstrated that intermediate flow rates are optimal for the onset of QS, and
have proposed a rationalization of this finding that is again based on mass transport: intermediate
flow rates alleviate nutrient transport limitations associated with molecular diffusion alone, but
without strongly washing out the autoinducer molecules as higher flow rates do. These two set of
experiments, then, rest on a common theme: that the interplay of nutrient transport and
autoinducer transport - both controlled by the same physical parameters (biofilm size, fluid flow,
etc) - is a crucial determinant of fundamental biofilm processes such as quorum sensing. We
speculate that a broad range of other phenomena within biofilms will be affected by the
processes elucidated here, and that microfluidics and direct, real-time imaging will provide a
highly valuable approach for the investigation of these dependencies.
3.4 Materials and Methods
Materials. Octadecyltrichlorosilane
(OTS)
[CH 3(CH 2)17SiCl 3] (Aldrich,
97%), hexane
(Aldrich, anhydrous, 99%), fluorescein isothiocyanate-bovine serum albumin (FITC-BSA) and
tetramethyl rhodamine isothiocyanate-wheat germ agglutinin (TRITC-WGA) were purchased
92
from Sigma-Aldrich (St. Louis, MO, USA). Poly(dimethylsiloxane) (PDMS, Sylgard 184) was
ordered from Dow Coming.
Strains, Growth Conditions, and Microfluidic Experiments. For the QS experiments, we
used P aeruginosaPAO1 mini Tn5- rhlA::GFP strain (courtesy of George O'Toole, Dartmouth
University) harboring the QS reporter, which expresses GFP when it is actively undergoing QS.
Using epifluorescent microscopy, we monitored QS by measuring the fluorescent intensity over
time in biofilm patches of different size, fabricated in microfluidic channels with 50 ptm height
and 4 mm width. In a typical experiment, we first inject for 1 h in the microfluidic channel a
low-density bacterial solution (OD 600 =0.2-0.3), which is grown in 5 ml tryptone broth (TB) in a
rotary shaker incubator at 180 rpm at 37 0C for 4 h, after being inoculated with 50 Pl of an
overnight culture (OD 600 ~ 1.0) in TB broth in a rotary shaker incubator at 150 rpm at 30 "C.
After 1 h, the injection is switched from the bacterial solution (in Trypton broth) to pure culture
medium (minimal M63 medium) at a constant flow rate of 3 pl/min, which corresponds to a
mean flow velocity of 250 pm/s and wall shear rate of -30 s-1. This flow condition is maintained
for up to 48 h, during which time the biofilm grows without the further external addition of cells.
The same protocol and media are simultaneously applied to separate (up to 5) microchannels, on
the same chip (i.e., glass slide), each having square hydrophobic patches of differing size (50,
93
150 and 400 pim, respectively). This parallel approach permits a direct comparison among
patches of different size, since the bacterial suspension, its age, and the environmental conditions
(e.g., temperature) are identical among all channels. Initial cell attachment and surface coverage
are monitored using phase-contrast microscopy, using computer-controlled routines to move the
microscope stage and automatically monitor all microchannels over time. The QS signal is
monitored using epifluorescent illumination and a GFP filter, with a 20 ms exposure. The
fluorescence intensity is quantified in terms of GFP signal strength.
PDMS stamp fabrication for surface patterning. PDMS stamps were fabricated by curing
the prepolymer on silicon masters patterned with SU-8 photoresist (SU-8 2050, MicroChem, MA,
USA) using conventional photolithography. The masters used for patterning had recessing
features, which resulted in PDMS replicas with the opposite sense. To assist in removal of cured
PDMS from the masters, the SU-8 masters are silanized by exposure to the vapor of 1,1 ,2,2,tetrahydrooctyl-1-trichlorosilane,
CF 3(CF 2) 6(CH 2) 2SiCl 3
overnight.
To cure
the
PDMS
prepolymer, a mixture of 10:1 silicon elastomer and the curing agent was poured on the master
and placed at 65'C for 2 h. The PDMS replica was then peeled from the silicon master.
94
Bibliography
1. Davies, D.G., et al. The involvement of cell-to-cell signals in the development of a bacterial
biofilm. Science 280, 295 (1998).
2. Miller, M.B. & Bassler, B.L. Quorum sensing in bacteria. Annual Reviews in Microbiology
55, 165-199 (2001).
3. Irie, Y. & Parsek, M. Quorum sensing and microbial biofilms. Bacterial Biofilms, 67-84
(2008).
4. Fuqua, C., Parsek, M.R. & Greenberg, E.P. Regulation of gene expression by cell-to-cell
communication: Acyl-homoserine lactone quorum sensing. Annu. Rev. Genet. 35, 439-468
(2001).
5. Kuipers, O.P., de Ruyter, P.G.G.A., Kleerebezem, M. & de Vos, W.M. Quorum sensingcontrolled gene expression in lactic acid bacteria. Journal of Biotechnology 64, 15-21 (1998).
6. Whitehead, N.A., Barnard, A.M.L., Slater, H., Simpson, N.J.L. & Salmond, G.P.C. Quorumsensing in Gram-negative bacteria. FEMS Microbiology Reviews 25, 365-404 (2001).
7. Nealson, K.H. & Hastings, J. Bacterial bioluminescence: its control and ecological
significance. Microbiology and Molecular Biology Reviews 43, 496 (1979).
8. Dworkin, M. & Kaiser, D. Cell interactions in myxobacterial growth and development.
Science 230, 18 (1985).
9. Dawson, M.H. & Sia, R.H.P. In vitro transformation of pneumococcal types: I. a technique for
inducing transformation of pneumococcal types in vitro. The Journal of experimental medicine
54, 681 (1931).
10.
Horswill, A.R., Stoodley, P., Stewart, P.S. & Parsek, M.R. The effect of the chemical,
95
biological, and physical environment on quorum sensing in structured microbial communities.
Analytical and bioanalytical chemistry 387, 371-380 (2007).
11.
Janakiraman, V., Englert, D., Jayaraman, A. & Baskaran, H. Modeling growth and quorum
sensing in biofilms grown in microfluidic chambers. Annals of biomedical engineering 37, 12061216 (2009).
12.
Purevdorj, B., Costerton, J. & Stoodley, P. Influence of hydrodynamics and cell signaling
on the structure and behavior of Pseudomonas aeruginosa biofilms. Applied and environmental
microbiology 68, 4457 (2002).
13.
M~ller, J., Kuttler, C., Hense, B.A., Rothballer, M. & Hartmann, A. Cell-cell
communication by quorum sensing and dimension-reduction. Journal of mathematical biology
53, 672-702 (2006).
14.
Schuster, M., Sexton, D.J., Diggle, S.P. & Greenberg, E.P. Acyl-Homoserine Lactone
Quorum Sensing: From Evolution to Application. Annual review of microbiology 67(2013).
15.
Boedicker, J.Q., Vincent, M.E. & Ismagilov, R.F. Microfluidic Confinement of Single Cells
of Bacteria in Small Volumes Initiates High-Density Behavior of Quorum Sensing and Growth
and Reveals Its Variability. Angewandte Chemie International Edition 48, 5908-5911 (2009).
16.
Carnes, E.C., et al. Confinement-induced quorum sensing of individual Staphylococcus
aureus bacteria. Nature chemical biology 6, 41-45 (2009).
17.
Hardie, K.R. & Heurlier, K. Establishing bacterial communities by'word of mouth': LuxS
and autoinducer 2 in biofilm development. Nature Reviews Microbiology 6, 635-643 (2008).
18.
Yawata, Y., Uchiyama, H. & Nomura, N. Visualizing the effects of biofilm structures on
the influx of fluorescent material using combined confocal reflection and fluorescent microscopy.
Microbes and environments, 912220146 (2009).
96
19.
Heydorn, A., et al. Quantification of biofilm structures by the novel computer program
COMSTAT. Microbiology 146, 2395-2407 (2000).
97
Chapter 4
Enhanced Trophic Transfer Rates in a Chemotactic
Food Chain
4.1
Introduction
In most natural habitats, predator-prey interactions occur in a spatially heterogeneous
environment. One of the major consequences of this heterogeneity is patch formation of
resources and organisms, which results in increased persistence of the prey population, compared
to homogeneous conditions, and increased consumption by higher-level organisms 1,2. Benefits
conferred by patchiness include a higher probability of mating, but patchiness can be harmful by
favoring predator foraging success..In the marine zooplankton Nyctiphanes australis, swarming
is seasonal and linked to breeding 3,4, suggesting that organisms aggregate in order to find mates.
Ritz has speculated that the hydrodynamics in aquatic crustacean schools might optimize the
capture rate of suspended prey 5. Modelling studies also suggest that collective motion can
improve ability to forage in a noisy resource field 6,7
There are several negative effects associated with patch formation. An obvious cost is that
available resources are shared with other group members. When in aggregation, krill are more
vulnerable to large predators that have evolved efficient strategies to detect and exploit patches
8,9. In the St-Laurent estuary, for example, mammals including rorqual whales are attracted by
aggregations of krill 10.
Whales are observed in regions where krill densities are generally high
and patchy 10,13. Dense aggregations also favor the spread of infectious diseases, which can lead
to mass mortality 14. On the other hand, organisms able to sense the approach of predators can
avoid patches to decrease predation risk. In the Southern Ocean, schools of Antarctic krill are
98
observed to disperse rapidly when approached by underwater vehicles
15,16,
which is thought to
be a strategy for escaping predator attacks by confusing the predator.
Granbaum proposed a non-dimensional indicator, the Frost number Fr as a predictor of the
performance of an organism in exploiting patches in its resources 17. The Frost number expresses
the ratio between the timescale of patch dissipation and the timescale characterizing the
predator's quest for and exploitation of the patch. When Fr 1, prey patches are highly available
to foragers. Conversely, Fr 1 indicates that prey patches dissipate too rapidly for foragers to
exploit them
'a
Although spatial heterogeneity has significant implications for ecological processes,
population models in the ocean are often spatially homogeneous
. Marine ecologists, however,
increasingly recognize that a consideration of heterogeneous spatiotemporal dynamics is
indispensible because there is growing evidence that nutrient patchiness is pervasive at all scales
in the ocean
20.
Recent work has shown that patchiness extends down to the microscale (pm-mm),
where motile marine microbes are believed to forage preferentially on point nutrient sources 2.
Motility in the microbial world is characterized by a random, diffusive component, used to
explore the environment, e.g., for nutrients, and a directional, chemotactic component used to
exploit nutrient patches upon detection. Chemotaxis is the ability of consumers to direct their
movement toward regions of high resource concentration. The performance of a chemotactic
organism depends on the timescale of chemotaxis relative to dispersion of the resource. This
concept is akin to the Frost number introduced by Grfinbaum. Since planktonic microbes in the
microbial loop play an important role in energy flow and biogeochemical cycles
21,25-28,
investigating the foraging abilities of the microbes is a key issue to quantify their effects on
trophic transfer and biogeochemical rates in the environment. Here we present a modeling study
99
that suggests that patchy interactions at the microscale can have major effects on macroscopic
processes.
Blackburn & Fenchel (1999) showed that planktonic microbes can find nutrient patches by
chemical cues over distances of a few millimeters to a centimeter 29. The nutrient patches are
often ephemeral, limited by the erosive effects of diffusion. The capacity of planktonic microbes
to actively locate and exploit the nutrient patches will be important in environment within
Fr 1. Seymour at al. (2008) observed strong responses to microbial patches by three sequential
levels of marine microorganisms in the microbial food web: a phytoplankton (Dunaliella
tertiolecta), a heterotrophic bacterium (Pseudoalteromonashaloplanktis), and a phagotrophic
protist (Neobodo designis)30. Due to technical complexities associated with measuring the
foraging behaviour of multiple microbes simultaneously, they quantified the chemotactic
response of each microorganism separately, resulting in a clear and intense patch of each
microorganism, corresponding to the initial position of the nutrient patch. Such patches lasted in
the order of 20 minutes. These scales will guide our choices in the simulations, where we will
explore times from 10 to 103 seconds, domains from L = 1 to 10 mm, and diffusion coefficients
appropriate for low-molecular-weight organic substrates, DA = 10-9 m2 s-. The experiments by
Seymour et al. (2008) directly inspired the simulations presented here, which provide a model for
a food web where each trophic level responds by chemotaxis to the population of its prey.
Contrary to Seymour et al. (2008), the modeling framework allows us to directly explore the
simultaneous presence of all three trophic levels, and - by comparison with a non-chemotactic
food web -the consequences on resource transfer through the food web.
100
4.2 Methods
Mathematical Model
We modelled a chemotactic food chain as a set of coupled partial differential equations, one
for each trophic level. The lowest level represents dissolved nutrients, of concentration A (but
could also represent a concentration of non-motile prey organisms). The subsequent levels have
concentrations of organisms denoted by B, C, D. We are assuming that all concentrations have
been normalized to the same units, e.g., expressed in terms of carbon concentration (molc/m 3).
The length of the food chain is varied in the simulations, from 2 (A,B) to 4 (A,B,C,D) trophic
levels. Equations are given in vector form, for application to ID, 2D and 3D domains. This work
focuses on 3D domains, except where indicated, to simulate aquatic environments. Because we
are primarily interested in the comparison among different scenarios (active, passive, uniform),
rather than the prediction of absolute quantities, we adopt a number of simplifications intended
to make the model simpler and reduce the number of parameters.
To illustrate the model we focus on the 2-level food chain, for which the equations governing
diffusion, chemotactic motility, and predation read:
aA = P V2 A --A AB
(1 a)
at
at
=IBVB
(Ib)
-V (VBB)±YBfiAAB
Bold symbols denote vector quantities. The first term on the right hand side models diffusion.
For the dissolved nutrient (A), this is molecular diffusion and diffusivity is &A. For the organisms
(B), diffusion arises from the 'random motility' component of swimming and is characterized by
a diffusivity PB. The latter can be estimated by modelling the movement pattern as a random
walk with swimming speed UB and characteristic reorientation time T, yielding iB =
101
rUB2 /3.
The
second term on the right hand side of equation (lb) represents chemotaxis. VB is the
chemotactic velocity of the organisms towards the resource (A) and is here modeled as
VB
=UB tanh
gA
VA .,VAI
where UB is again the organisms' swimming speed. Strictly speaking,
IVAI
UB should be the maximum chemotactic speed, which is a fraction of the swimming speed. Their
theoretical ratio is of order 1 (e.g., 2/3 for bacteria, see Ahmed & Stocker, 2008), hence we will
here take UB to be equal to the swimming speed for simplicity. The hyperbolic tangent accounts
for an increasing response with increasing magnitude of the resource gradient and saturation
beyond a threshold gradient g (Rivero et al., 1989). The magnitude of VB is 76% of UB when
VAI = g. This chemotaxis model is an extension of the classic one by Keller and Segel
31,
with
provision for the saturation of the chemotactic velocity. The term VA / IVAI sets the direction of
the chemotactic velocity to be along the resource gradient.
The last term in equations (la,b) models resource transfer between trophic levels. The shape of
this term depends on the transfer mode, for example osmotrophic uptake versus predation. For
simplicity, we adopt a transfer function that is linearly proportional to encounter rate between
organisms and their resources (here, AB) and describe the transfer process generally as
'predation'. The magnitude of predation is measured by / and a fraction
YB
(assimilation
efficiency) of the prey biomass is turned into predator biomass. Because ;B is typically of order
1 (0.1-0.6; e.g., Fenchel 1982, Kirchman et al. 1992, Conan et al. 1999, Cole & del Giorgio
2000), we use
YB
= 1 for simplicity.
102
Higher trophic levels (C and D) were modeled in the same manner as B, including random
motility, chemotaxis and predation. Equations were then made dimensionless by rescaling
concentrations A, B, C and D with their initial values A0 , BO, Co and Do; lengths with the domain
length scale L; time with the nutrient diffusion time scale L2 /
a;
and thus velocities by plAIL. The
full equations for a four-level food chain, in dimensionless form, are then:
o =V 2 A- 7AB
at
(2a)
aB= KBV 2 B -%BV-
I VjB
tanh L
at
aC
a=t
aD=
cV2C-%CV- tanhI
|_B
2D -
V
tanh
I
VC
at
VB
B7AAB-hBBC
C +wC7BBC-qCCD
IVB
6B
D
+
|VA|
A
(5c
(2b)
(2c)
)
VCD + WDCCD
(2d)
|V C|
Where the parameters are
KB
C
,IKC
_
PA
PA
UCL
UBL 9
'%C
%B
PA
CB
B0
A0
BO
PA
2
0/ B 2
B
7C
C
9CL
CO
B0
CO
5C
DOCL2
(3c)
PA
PA
tB
(
ZD
PA
B7IAL 2
9BL
UDL
'
PA
5
D(3a)
D
PA
_
SD
(d
Do
C0
D
Do
103
These dimensionless parameters govern the population dynamics and trophic transfer. K iS the
ratio of the organisms' diffusivity by random motility relative to the nutrient diffusivity; X is a
chemotactic Peclet number, describing the ratio of time scales for organism chemotaxis relative
to nutrient diffusion (akin to the Frost number
23),
77
is a dimensionless biomass transfer
coefficient, 5 is a dimensionless saturation gradient and co measures the ratio of initial organism
concentrations.
Numerical solution, initial conditions and parameter values
Equations (2a-d) were solved using a finite element methods software (Comsol Multiphysics,
Natick, MA). The spatiotemporal evolution of A, B, C and D was computed subject to Neumann
boundary conditions (no-flux) at the edge of the domain. We computed solutions for different
domain dimensionalities (lD, 2D and 3D), but focus primarily on the 3D case here. Polar and
spherical symmetry were assumed in the 2D and 3D domains, respectively, such that the solution
only depends on the distance r from the origin. The computational domain is therefore the radial
axis. The volume of the domain is 2, nr 2 and 4nr 3/3 in ID, 2D and 3D, respectively.
The radial axis was discretized into 9676 elements, with higher density of elements at smaller
values of r, where sharper changes were expected. Convergence was tested by ensuring that
differences in the results were negligible upon refining the mesh by a factor of four. The
simulations time step was automatically set by the numerical solver, ensuring accuracy and
stability. Data were stored at time intervals dt = 10-4 (from t = 0 to 0.005) and dt = 10-3 (from t =
0.005 to 0.02).
104
The initial condition for the nutrients is a Gaussian patch of half-width cy and total initial mass
of 1:
A(r,t =0)=
1
e
(7)
2
where n is the dimensionality of the system (1, 2 or 3). We chose a = 0.02 for all simulations.
This corresponds for example to a 200 tm nutrient patch in an L = 10 mm domain. For
comparison, we also run cases with a uniform distribution of A, again with total initial mass of 1.
The initial distribution of B, C and D was uniform with total biomass of 1, i.e., Bo = Co = Do=
1/2, 1/7t and 3/(47c) in ID, 2D and 3D, respectively.
We explored a wide range of values for
Ki,
X, and 3,, estimated from typical values of
biophysical parameters and movement characteristics of marine microorganisms and their
environment. Scales of interest (L) range from 1 to 10 mm. Nutrient diffusivity pA ranges from
10-9 m2 s-I for low molecular weight substrates (e.g., sugars and amino acids) to 1 0-11
m
2
s
for
high molecular weight ones (e.g., polysaccharides and carbohydrates) (Benner et al, 1997, 2003).
Swimming speeds range from 10 to 1000 ptm s-1, and random motility coefficients from 10-10 to
10-8 m2 s (Visser & Kiorboe, 2006). The strength of predation, 8, approximately ranges from
10-6 to 10-3 s-1 . The threshold gradient, g, ranges from 10-2 to 102 mm-1 (Rivero et al, 1989). This
resulted in estimates for the range of K, (10- 1 to 103), y, (101 to 106),
T1,
(10-4 to 103), and ei5 (10-
to 104). We further assumed that each trophic level initially had the same biomass ((o, = 1).
To quantify the degree of patchiness of each trophic level, i, and the correlation between two
connected trophic levels (which determines uptake), we computed the zero-lag correlation and
the cross-correlation metrics at any given time, t:
105
1 , t) =
} (t) =
E Ai
_
i
i=E[i], Ai=i-i, j=E[j],
E A
ij
Aj=j-j
where E[] is a spatial average over the domain, i and j are the mean values of biomass in
trophic levels i andj, and Ai and Aj are the spatial deviations from those averages. We call Pij
the patchiness index and Pij the resource-consumer correlation index.
4.3 Results
We investigated chemotaxis-mediated trophic dynamics in three-dimensional space. This
represents an appropriate model, for example, for aquatic ecosystems away from surfaces, where
diffusion is typically not constrained by boundaries. For other situations, including spreading of
microorganisms on surfaces or gamete encounters at air-water interfaces, a two-dimensional
approach would be more appropriate. The concentration profiles generated by diffusion depend
on the dimensionality of the problem. This is shown in Fig. 4-1, where the radial concentration
profiles of the dissolved nutrient A are plotted at time t = 0.005 in 1D, 2D and 3D scenarios.
Higher dimensionality results in stronger gradients. On the other hand, gradients in higher
dimensions fill a smaller volume (the fraction of the domain volume within distance r from the
origin is r, r2 and r3 in ID, 2D and 3D, respectively) and thus are available to a smaller fraction
of a (uniformly distributed) population of organisms. As chemotaxis hinges on the response to
gradients, a quantification of the effect of chemotaxis on trophic transfer requires first a choice of
system dimensionality. While the essential process described here - the enhancement of trophic
transfer rates by chemotaxis - occurs independently of the dimensionality of the system, an
106
estimate of its quantitative impact is dimension-dependent. Therefore, with aquatic ecosystems in
mind, where microbial interactions occurs mostly in the unbounded water column, we focus here
on three-dimensional systems, and use the appropriate mathematical expressions for the vector
operators in Eq. (1).
60
50
t=0.005
ID
-'-
I-0-2DI
*
+
3D
0.0025r1
0.0020
40
0.0015
A
A
30
0.0010
0.0005
20
0.0000 -
0.000
0.005
0.010
0.015
0.020
0.025
0.030
10
0
0. 0
0
0.2
0.4
43
~
0.6
0~
0.8
1.0
r
Figure 4-1 Effects of dimensionality and domain size. The diffusion of a resource patch
generates different profiles of concentration (A) in 1 D, 2D and 3D. Radial profiles are shown at t
= 0.005. The initial half-width of the Gaussian patch was a-o = 0.02. In the remainder of this work,
3D simulations are discussed. Inset: the resource patch reaches the boundary of the simulation
domain (r = 1) after t ~ 0.02, which was adopted as the temporal horizon for all simulations.
107
Because patch utilizations via chemotaxis occurs over short time scales, typically several
minutes for marine microorganisms
32,33,
we focus on simulations over short temporal horizons.
The latter was chosen based on a consideration of the boundary conditions. The inset in Fig. 4-1
shows that the diffusing patch begins reaching the boundary of the domain (r = 1) at t ~ 0.02.
Thus, we chose to run all simulations up to t = 0.02. For example, when the nutrient is made of
small molecules (DA = 10-9 m 2 s-') diffusing in an L = 10 mm domain, this corresponds to a
dimensional time of 0.02L 2/DA= 33 min. For large molecules (DA
=
10-11 m 2 -1), the simulation
horizon is 5.5 hours.
We begin by studying a two-level food chain (A,B), to gain insight into the processes involved
in the response to a resource patch. The key idea is exemplified in Fig. 4-2, where three
fundamental scenarios are considered: a uniform distribution of resource A ('uniform'; green
line); a patch of resource A with a non-motile consumer B ('passive'; blue line); and a patch of
resource A with a motile consumer B ('active'; red line).
Consumers are initially distributed uniformly (B = 3/(47r)) in all cases. The passive case
corresponds to
KB = YB =
(for all cases, 11
B =
0 (no motility), whereas for the active case we chose K B = 1, x
B =
100
1). In a uniform environment, nutrient availability and consumption occur
uniformly, hence motility is irrelevant: there are no cues that make the consumer distribution
depart from uniform. Consumers therefore grow and deplete the resource uniformly. Because the
temporal horizon is short, the growth is small (i.e., B remains ~ 3/(47[); Fig. 4-2a, green solid
line).
108
(a)
(b)
100
3.5
3.0
80
active
2.5
60
A&B
2.0
fAB dV
1.5
40
1.0
20
0
1.0
0.5
0.0
0.5
0.5
passive
0.0
uniform
1.0
0.000
0.005
0.010
0.015
0.020
t
r
Figure 4-2 Three possible microbial food chains: uniform, passive and active. (a) Shown is the
radial concentration of resources (A, dashed lines) and consumers (B, solid lines) at time t = 0.02
for three cases, uniform (green), passive (blue), and active (red). In a uniform food chain, the
resource is uniform over space, hence consumers remain uniformly distributed (with a value
close to their initial value of 3/(4n)). In a passive food chain (KB =
= 0), a resource patch
triggers a consumer patch, but the latter is small because it is caused purely by growth and the
duration of the process is short. In an active food chain, consumers respond to a resource patch
using chemotactic motility (here: KB = I, ;B = 100), markedly enhancing accumulation. For all
cases, r/B= 1. Note that the green dashed line is indistinguishable from the blue dashed line. (b)
These different responses can significantly affect resource uptake, as shown by the spatially
integrated, instantaneous transfer (.ABdV). Uptake is considerably larger and faster when the
consumers respond actively to the resource patch.
The passive and active scenarios share a heterogeneous resource distribution in the form of a
patch, but only the active scenario incorporates a chemotactic response of the consumers. A
resource patch can trigger an accumulation (hence, a patch) of consumers by one of two
mechanisms: local growth (passive case) or recruitment of surrounding organisms by motility
(active case). When the time scale of patch diffusion and organism response is short compared to
the organisms' doubling time, as is the case for most microscale solute patches in the ocean,
active accumulation by motility (Fig. 4-2a, red solid line) is considerably stronger than local
109
growth (Fig. 4-2a, blue solid line). By moving into the center of the patch, active consumers are
exposed to higher resource concentrations than passive ones, enhancing uptake. This results in a
faster depletion of the resource (compare the red and blue dashed lines in Fig. 4-2a). Although
the time over which an active response is superior to a passive one is finite, because resource
gradients are eroded by diffusion and ultimately vanish, we will see that typical response time
scales nonetheless afford a marked uptake advantage to active responders.
It is essential to distinguish between rates and amounts. In the absence of additional factors
(e.g., changes in the resource's bioavailability or the existence of a lower threshold for uptake),
the entire resource will ultimately be consumed both in the passive and in the active case. Hence,
the primary effect of an active response will be on the rate at which the resource is consumed,
which ultimately determines the rate at which resources are transferred further up the food chain
and can potentially affect the amounts transferred (see Discussion). Equivalently, one can
determine the amount of resources transferred instantaneously or on average over a given finite
time. The instantaneous measure is obtained by computing the instantaneous integrated uptake
over the domain, .ABdV, since resource consumption is proportional to the product AB (Eqs. 1,2).
The average measure is then simply the amount of resource transferred over a finite time, here t
= 0.02. Figure 4-lb shows that an active response can significantly enhance instantaneous and
average uptake: for the case in Fig. 4-lb, both measures of trophic transfer are about an order of
magnitude larger for the active food chain than both the passive and uniform food chains. This is
a direct result of the strong motility-mediated accumulation of the active consumers within the
resource patch (Fig. 4-l a).
110
k 1 AB
(b)
(a)
5
0.02
o B
0.02
2
3
2
t 0.01
t 0.01
0.005
0
0.006
0
S
(c)
0.05
0.1
r
0.15
00
.20
IoeB
0.02
0.1
r
0.15
0.2
0A
(d)
50.02
4
2
0.016
0.015
3
t0.01
t 0.01
0.005
0
0.005
2
0
00
0.06
0.1
0.15
0.2
0o
.1
r
0.05
0.1
0.15
0.2
.1
r
Figure 4-3 Spatiotemporal dynamics of a two-level food chain. Color maps represent the
distribution of consumers (B) and resource uptake (AB) in space (r) and time (t), for (a,b) a
passive food chain and (c,d) an active food chain (for definitions of 'passive' and 'active', see
Fig. 2). For comparison, in a uniform food chain (not shown) consumers (B) are uniformly
distributed with B ~ 3/(47r) = 0.239 (growth is weak, hence the value is nearly constant).
Different consumer responses generate not only different consumer distributions (a,c), but also
different uptake landscapes (b,d). Parameters are the same as in Fig. 4-2.
The comparison between active and passive scenarios is presented in more detail in Fig.4-3,
which shows spatiotemporal maps of consumer distribution (panels a,c) and trophic transfer
(panels b,d). Accumulation is considerably stronger for active consumers (panel c) compared to
passive ones (panel a), and the spatial distribution of active consumers follows the diffusion of
111
the resource patch (diffusion-like spreading in panel c). Consequently, trophic transfer, which is
mainly limited to early times and organisms initially in the patch for the passive case (panel b), is
not only very strong at early times in the center of the patch for the active case, but lasts longer
in the center of the patch and extends to surrounding regions (panel d), as motile organisms
follow gradients of the diffusing resource by chemotaxis.
The critical parameter determining the strength with which a consumer is attracted to a
resource is the chemotactic sensitivity. In Fig. 4-4 we examine the effect of increasing the
chemotactic sensitivity, XB, on trophic transfer for a two-level food chain, by computing the time
course of the increase in consumer biomass, ABT.
Trophic transfer increases with XB, as stronger chemotaxis affords earlier accumulation in the
resource patch. This increase is dramatic between XB
=
100 and 1000, when ABRT at t = 0.02
jumps from 42% at XB = 100 to 95% at XB = 000. A further increase in XB saturates transfer, and
further investment by the organisms in improved sensitivity is thus unnecessary. The comparison
with the passive and uniform food chains is insightful. ABT is larger in all three active food
chains than in the uniform food chain (for which ABT = 5% at t = 0.02): motility and patchiness
thus markedly enhance transfer rates. Unsurprisingly, chemotaxis (XB
> 0) is also beneficial
compared to random motility alone (XB = 0; 'diffusive' food chain in Fig. 4-4). On the other hand,
Fig. 4-4 surprisingly shows that chemotactic organisms can perform worst than non-motile ones.
For example, AB 1 is lower for the food chain with ZB = 100 (42% at t = 0.02) than for the passive
food chain (57% at t = 0.02). As expected, organisms with stronger chemotaxis (XB
1000)
outperform (ABT = 95% at t = 0.02) passive ones. This indicates that chemotactic motility is only
beneficial when the chemotactic sensitivity exceeds a minimum threshold.
112
100
97%
95%
80
-
B
- -- diffusive
- passive
uniform
-o---
00%60
7
4
I- R
, 42%-4
.10
42%
40
20
-
05
0.000
0.010
0.005
0.015
0.020
t
Figure 4-4 Effect of chemotactic sensitivity on trophic transfer. The increase in consumer
biomass,
ABT,
is shown for a two-level food chain (A, B), for six cases: uniform (no resource
patch), passive (no consumer motility), purely diffusive (motility, but no chemotaxis;
,
and three cases of chemotactic (X B = 100, 1000 or 10000) food chains. For all cases KB
=
=
10, except for the passive case for which KB
=
0, qB
B =
0)
10, qB
10. Numbers on the right indicate the
increase in consumer biomass at t = 0.02. Notice that the trophic transfer is larger in a passive
food chain than in an active one with X B = 100 (but less than the X B = 1000 case).
To shed light on this apparently counterintuitive result, we examined the effect of random
motility,
KB.
Fig. 4-5 shows that the relative performance of chemotactic and non-motile (i.e.,
passive) consumers depends on the magnitude of the random swimming component. The passive
strategy is superior only when random motility is large (KB = 10 in Fig. 4-5), while chemotaxis is
superior when random motility is small
(KB =
1 in Fig. 4-5). Furthermore, trophic transfer of non-
motile cells depends on the biomass transfer rate, lB: whereas for lB
113
=
10 the passive strategy is
superior to one with moderate chemotaxis (XB
'lB =
=
100), as seen in Fig. 4-4, the opposite occurs for
1 (not shown). In summary, chemotaxis confers a competitive advantage to motile species
over non-motile ones only if (i) the directional component of movement is suitably larger than
the random component, and (ii) accumulation, rather than local growth, is the dominant
mechanism for trophic transfer. Motile but weakly chemotactic cells
(XB =
100 in Fig. 4-4),
therefore, accumulate poorly in the resource patch and tend to diffuse rapidly out of it,
diminishing trophic transfer by local growth.
100
80 -B
-o
B=10
passive
--60
..
1
0.000
0.005
.
---
--
-
-
0.010
0.015
0.020
t
Figure 4-5 Effect of random motility on trophic transfer. The increase in consumer biomass, ABT,
is shown for three two-level (A, B) food chains: passive (no consumer motility) and chemotactic
(XB = 100, qB = 10), the latter with KB = I or 10. For KB = 10, the figure further shows the effect
of the saturation gradient 8: the upper and lower boundaries of the shaded region correspond to 8
= 10- to 104.
114
Although primarily determined by the chemotactic sensitivity, the chemotactic ability of our
model organisms is also affected by the threshold resource gradient, g. The lower g, the higher
the resource transfer (Fig. 4-5, shaded region), because cells respond with the full chemotactic
velocity even to weaker gradients. In reality, there is a trade-off between the ability to respond to
weaker gradients and the robustness of weak signals to noise
, preventing organisms from
responding to exceedingly weak gradients. This was not pursued here, and we chose instead to
focus on the case 6 = 102 for all simulations.
To quantify the effect of random and directed (i.e. chemotactic) motility on transfer, we carried
out simulations for a grid of values of KB and XB. A summary of the resource transfer ABT at t =
0.02 is shown in Fig. 4-6, both for IlB = 1 (Fig. 4-6a) and rIB
(a)
=
10 (Fig. 4-6b).
(b)
35
100
30
60
20
80
415<
40
10
Figure 4-6 Trophic transfer for different chemotactic food chains. Shown is the increase in
consumer biomass, AB1 , at time t = 0.02 for a two-level food chain (A, B), two values of the
biomass transfer rate, (a) q'B = 1 and (b) qr/ = 10, and a matrix of motility values, KB and %B. The
case ZB = KB = 0 corresponds to a passive food chain, while cases with %B = 0 correspond to
'diffusive' consumers (motile but non-chemotactic).
115
The opposing effects of chemotaxis and random motility are evident: an increase in
enhances trophic transfer, whereas an increase in
KB
ZB
reduces it. Both parameters ultimately
depend on swimming speed, showing that an increase in swimming speed alone does not
necessarily increase resource uptake. Instead, it is the relative importance of directed versus
random motility that primarily affect transfer rates. The results for
1BB =
10 confirm that this
conclusion is largely independent of the biomass transfer rate and reiterate the existence of a
chemotactic sensitivity threshold when
lB
is large.
Quantitatively, we find that strongly chemotactic consumers can increase their biomass by up
to 30% within the temporal horizon t = 0.02 for small biomass transfer rates
(lB
=
1) or,
equivalently (given our assumption of perfect consumption, yB = 1, Eq. 1), that they consume 30%
of the nurtients within t = 0.02. For imperfect consumption, numbers can be easily rescaled by YB,
so that a
YB =
0.5 consumption efficiency corresponds here to a maximal biomass increase of 15%
from a patch. For large biomass transfer rates
(1B =
10), even moderately chemotactic consumers
can consume the entire patch within t = 0.02, hence obtain a marked biomass boost from single
patches.
To what extent, then, does chemotaxis affect the rates of trophic transfer? To quantify this, we
computed the ratio RC between the time to transfer 50% of the resources to the consumers in the
passive case, versus the same time computed for the active case. When RC > 1, the active food
chain transfers resources faster than the passive one. Fig. 4-7 shows RC for increasing values of
XB,
for KB
=
10 and flB
=
0. For weak chemotaxis
(XB
100) transfer is slower in the active case,
consistent with our findings above (Fig. 4-4). In contrast, moderate to strong chemotaxis
(XB =
1000-10000) can enhance transfer rates by over an order of magnitude, relative to a passive food
chain.
116
12
10
8
RC 6
4
2
1
...
---------------------------..
>-----
---
------
01
0
100
1000
10000
XB
Figure 4-7 Increase in the trophic transfer rate due to chemotactic motility, for a two-level (A, B)
food chain with KB = 10 and qB = 10. Rc represents the ratio of times at which 50% of the
resource (A) has been taken up by the consumers (B), for a passive (i.e., non-motile) food chain
relative to an active (i.e., chemotactic) one. When Rc = 1 (dotted line) the transfer rate is equal to
that of a passive food chain. Biomass can be transferred over ten times faster due to chemotactic
motility.
As shown in Fig. 4-3, the mode of response to a resource patch affects not only the rate of
resource transfer, but also the spatial distribution of consumers. This, in turn, affects resource
transfer at the next level up in the food chain. We used the correlation metrics PBB and PAB (see
Methods) to quantify this heterogeneity: PBB measures the heterogeneity of consumers, while
PAB measures the co-localization of resources and consumers. Fig. 4-8 presents the temporal
dynamics of the correlation metrics. Fig. 4-8a shows that, at low biomass transfer rates
moderate chemotaxis
(XB =
(rIB
=
1),
1000) can markedly enhance consumer heterogeneity compared to
the case of a food chain where heterogeneity arises only via local growth (passive case in Fig. 4117
8a), by a factor of nearly 200 at t = 0.02. Only when biomass transfer rates are high (flB =10) can
local growth result in stronger patchiness than chemotaxis. The covariance PAB, on the other
hand, is larger for a chemotactic population (XB
=
1000) compared to a non-motile one for both
1 and 10 (Fig. 4-8b).
lB =
(b)
(a)
50
30
400
.- *..
30
-- 4
-o-e=1,
X,=1,000, K,=1
1 0
118=10, /,= , 00 K,=1
25
q,=1.
20
passive
l=10, passive
15
a
20
10
10
0.000
0.005
0.010
0.015
0.000
0.020
0.005
t
0.010
0.015
0.020
t
Figure 4-8 The degree of patchiness in the consumer distribution depends on the consumer's
response. Shown are (a) the patchiness index, PBB, and (b) the resource-consumer correlation
index, PAB, for a passive (i.e., non motile) and an active (lB = 1000, KB = 1) two-level food chain,
for two values of the biomass transfer coefficient, IB 1 and qB= 10. Note that the patchiness in
the consumer distribution, PBB, can be larger for the passive case when biomass transfer is rapid
(qrB = 10; local growth), but the consumer-resource correlation PAB (and thus trophic transfer) is
always larger for the active case (except in the earliest stages).
This superior co-localization with the resources is what allows chemotactic consumers to
obtain the advantage in resource transfer, shown in Fig. 4-6. A grid of simulations (Fig. 4-10)
shows similar trends for consumer patchiness PBB at t
=
0.02 than resource transfer ABT (Fig. 4-
6). As expected, directed motility enhances consumer patchiness, random motility reduces it.
However, in contrast to resource transfer (Fig. 4-6), consumer patchiness is reduced at higher
values of lB (Fig. 4-9b), because of higher local growth effects.
118
(a)
(b)
18
5F
16
7B
4
12
31
10
P
'B =10
14
P.
8
2
6
4
2
01000
0
lift
10)
10
1000
100
10
100
10
41
0
...........
...
1
1C.
Figure 4-9 Increase in consumer patchiness due to chemotactic motility. Shown is the patchiness
index, PBB, at time t = 0.02 for a two-level food chain (A, B), two values of the biomass transfer
rate, (a) qB = 1 and (b) '1B = 10, and a matrix of motility values, KB and %B.The case %B= KB = 0
corresponds to a passive food chain, while cases with %B= 0 correspond to 'diffusive' consumers
(motile but non-chemotactic).
How are these dynamics affected by an increase in the length of the food chain? In Fig. 4-10
we examine a three-level food chain, by showing the increase in biomass of the top consumer,
ACT, for the active, passive and uniform scenarios. ACT is expressed in % of the initial biomass
of C, hence it can reach up to 200% (when all resources, A and B, have been transferred to C).
Consistent with the two-level food chain, a heterogeneous resource distribution results in
considerably faster transfer compared to a uniform one. A strong chemotactic response by both
consumers
(XB =
XC
=
1000) at intermediate values of random motility
(KB =KC =
10)
results in a
three-fold larger resource transfer to the top consumer compared to an entirely passive food
chain
(XB =
XC = KB = KC = 0), even when transfer rates are high (1B
119
=
C=
10). Interestingly, at
early times (up to t = 0.004) the chemotactic response is inferior to the passive response, a result
of the fact that predation by C limits uptake of A by B in the central, nutrient-rich region, slowing
overall trophic transfer. At larger times (0.004 < t < 0.02), however, this is reversed and the
chemotactic food chain channels a significantly larger fraction of the resources to the top
consumer within t = 0.02. This conclusion is even stronger than that for the two-level food chain
(Fig. 4-4), for which the advantage of a chemotactic response with XB
=
1000 was less than twice
that of the passive food chain in terms of nutrient transfer to the top consumer.
180
ISO
XNB,C=
-
140
-
167%
000
passive
uniform
120
0
100
80
60 40
20
.-
50%
g
O__
0.000
0.006
0.010
0.016
0.020
t
Figure 4-10 Effect of chemotaxis on trophic transfer in a three-level (A, B, C) food chain. The
increase in the biomass of the top consumer, AC 1 , is shown for three consumer responses:
uniform (no resource patch), passive (no motility of the consumers, B and C), and chemotactic
XC 1000). For all cases KB = KC = 10, qB= qc = 10, except for the passive case for which
(ZB
KB
=
KC
0, 7B=
= 10. Numbers on the right indicate the increase in the biomass of the top
consumer (C) at t = 0.02. Biomass transfer is most rapid in the active food chain and least rapid
in the uniform food chain.
120
Organisms occupying different levels of the food chain likely have different levels of motility
and chemotaxis. In Fig. 4-10 we assumed the same motility parameters for both consumers (B
and C), but when this assumption is relaxed a rich set of possible outcomes emerges. Fig. 4-11
shows the temporal dynamics of the change in biomass for all three trophic levels (AA 1 , ABT,
ACT), along with the spatiotemporal distribution of the two top consumers (B,C) and their colocalization (the product BC). This was repeated for four combinations of the chemotactic
response of the two consumers: (i) weak chemotactic response by both
stronger response of the lower consumer
consumer
KC
(XB =
= liB =TC =
(XB
(XB
=
XC
=
100);
0i)
1000; c= 100); (iii) stronger response of the top
00; Xc = 1000); strong response by both
(XB
XC
1000). For all cases,
KB
10.
The main conclusion that emerges is that the trophic transfer is large (167%; Fig. 4-11, panel
m) only when both consumers exhibit strong chemotaxis (panels m-p). When B responds weakly
(panels a-d and i-l), the resource A is depleted slowly (panels a, i) and B grows slowly and
weakly (panels b,
j). When
the ensuing response of the top predator is also weak (panels a-d),
there is significant delay in the transfer of biomass to C (panel c) and the overall trophic transfer
is very small (17%). A strong response of the top predator to a weakly responding lower predator
(panels i-1) creates a strong accumulation of top predators (panel k), but localized to a narrow
region in the center of the patch. Ultimately, the overall transfer is still small (31%). On the other
hand, when the lower consumer responds strongly but the top one does not (panels e-h), B
thrives (panels e,f) but C cannot take advantage of it (panel g) and the overall transfer is again
small (22%). Comparison of panels e-h with i-l reveals that an increase in the chemotaxis of the
top predator has a stronger effect on resource transfer compared to an increase in the chemotaxis
of the lower predator.
121
(a)
(d)
(c)
(b)
0.02
2
2
4
17%
1.
t
3
1
1
2
0.6
0.6
0.01
10
-20m
0.01
t
1
0
00
- -
40
O'M
0005
0010
OV1G
0020
0
0
0
40
lg B
(h)
4
2
4.6
.0
1
t
-20
lp)'
0.012
122%
-0.1
2
r
2
0.016
20
0.2
B
()
0.02
log,
0.1
0
r
r
(a)
0
1
1
0.5
0.5
0
00
0.008
-60
0.004
.6
000 000
0,010
0 015
0
1
U)
(1)
.61
0
020
.
40,
(k)
0.02
0.1
-
02
(I)
2
2
1.
1.5
4
13%
30
0.0161
3
10
2
0.0121
0.6
0.6
0.006
-20
0
0
40
-0.5
-0.6
0000
000
0o010
0015
0020
0
3.2
I1-11
0
t
0.1
0
0.2
0.1
0.2
0
(n)
(in)
I'm
0021
4
2
.167%
150
106
0.01
3
.5
too
2
0.01
0.6
0.J
1
0
0.
-60
4.
-1.6
000
0,005
0010
Obis
0020
.1
5
-2
0
t
0.1
0.2
0
0.1
r
0.2
0
0.1
r
0.2
Figure 4-11 Effect on trophic transfer of a differential chemotactic response by different trophic
levels, for a three-level food chain. Curves (panels a, e, i, m) show the time course of biomass
change for all three trophic levels (A, B, C). Numbers on the right of each panel show the
increase in the biomass of the top consumer at t = 0.02. Color maps represent the
spatiotemporal distribution of the lower consumer, B (panels b, f, j, n), the top consumer, C
(panels c, g, k, o), and the co-localization between B and C, BC (panels d, h, 1, p). The four rows
correspond to four consumer responses. Row 1: B = Zc = 100. Row 2: ZB = 1000, XC = 100.
Row 3: XB = 100, XC = 1000. Row 4: %B =Xc = 1000. For all cases, KB = KC = qB = C = 10122
On the other hand, when both predators exhibit a strong chemotactic response (panels m-p)
resource transfer is greatly magnified. The resource A is rapidly depleted by B (panel m), which
exhibits fast growth (panel m), only to be curbed by C. Consumption by the top predator remains
focused near the center after a mild broadening, resulting in a vase-shaped spatiotemporal profile
(panels n-p). In this regime, the top predator gains rapid access to nearly the entire resource
biomass and further takes advantage of the chemotaxis of B towards A.
From the point of view of the top predator, strong chemotaxis is clearly beneficial in all cases.
Interestingly, from the point of view of the lower predator (B), a strong chemotactic response is
advantageous when the predator is weakly chemotactic (compare panels e-h to panels a-d),
whereas in the presence of a strongly chemotactic top predator a weak chemotactic response is
preferable, to avoid 'concentrating' the resources for the top predator (compare panels i-l to
panels m-p). This represents an additional trade-off for chemotaxis: not only does the directional
component of motility have to be sufficient to overcome the diffusive tendency imparted by
random motility, but the strength of the chemotactic response must be commensurate with the
chenotactic abilities of one's predator.
4.4 Discussion and Conclusions
In this study, we have developed a spatially explicit mathematical model of a food chain with
multiple trophic levels with the purpose of quantifying the extent to which the active response of
organisms to resource patches increases resource transfer and propagates patchiness in the ocean.
We found that chemotaxis can enhance the rate at which resources are transferred to the top
consumer by up to tenfold. On the other hand, chemotaxis does not automatically confer motile
species a competitive advantage over non-motile ones. A competitive advantage only occurs
123
when the directional component of movement (chemotaxis) outweighs the randomness
intrinsically associated with organism movement, particularly at small scales. Furthermore,
optimal strength of chemotaxis depends on the chemotaxis of one's predator - creating a
complex optimization problem, in a trophic cascade of chemotaxis.
When background resource availability is low, a consumer waiting for prey (passive case) is
less likely to find resources than one actively foraging. On the other hand, active foraging incurs
an energetic cost, since a faster response implies a great power expenditure for motility. This
power increases with the square of the swimming speed and further affects the choice of the
optimum movement behavior of an organism. Hence, one might expect motile behavior to reflect
a trade-off between benefits (higher nutrient gain) and costs (power expenditure). We are
currently working on this.
Predation dynamics become more complex with an increase in the food chain length. Already
with 3 levels (Fig. 4-12) we found a range of outcomes when the relative magnitude of
chemotaxis was varied among the intermediate and top predator.
124
150
-
A
-- B
-a-
C
---
D
100 - -+-
D, passive
50
**44*48%1
0<
-50
-100
0.000
0.005
0.010
0.015
0.020
t
Figure 4-12 Trophic dynamics in a four-level chemotactic food web. The time course of the
change in biomass of all four trophic levels (A, B, C, D) is shown for KB = 10, XB = 1000, B =
10, icc = 0.1, Xc = 100, zc = 10, KD = 10, XD = 100, qrD = 10. The case of a fully passive food
chain (all consumers B, C and D are non-motile) is shown for comparison, limited to the biomass
change of the top predator (D). While population dynamics become more complex in the
presence of multiple trophic levels, an active response can still result in a significant
enhancement in trophic transfer rates.
125
Bibliography
1. Levin, S.A. Patchiness in marine and terrestrial systems - from individuals to populations. 99103 (Royal Soc London, 1994).
2. Arditi, R., Tyutyunov, Y., Morgulis, A., Govorukhin, V. & Senina, I. Directed movement of
predators and the emergence of density-dependence in predator-prey models. Theoretical
Population Biology 59, 207-221 (2001).
3. Blackburn, M. Observations on the distribution of Nyctiphanes australis Sars (Crustacea,
Euphausiidae) in Australian waters. Rep. Div. Fish. Oceanogr. CSIRO Aust 119, 1-10 (1980).
4. O'brien, D. Surface schooling behaviour of the coastal krill Nyctiphanes australis(Crustacea:
Euphausiacea) off Tasmania, Australia. Marine ecology progress series. Oldendorf 42, 219-233
(1988).
5. Ritz, D. Is social aggregation in aquatic crustaceans a strategy to conserve energy? Canadian
Journal of Fisheries and Aquatic Sciences 57, 59-67 (2000).
6. Granbaum, D. Schooling as a strategy for taxis in a noisy environment. Evolutionary Ecology
12, 503-522 (1998).
7. Wood, A., Ackland, G. & Lenton, T. Mutation of albedo and growth response produces
oscillations in a spatial Daisyworld. Journal of theoretical biology 242, 188-198 (2006).
8. Nicol, S. & O'Dor, R. Predatory behaviour of squid (Illex illecebrosus) feeding on surface
swarms of euphausiids. Can. J. Zool 63, 15-17 (1985).
9. Ritz, D. Social aggregation in pelagic invertebrates. Advances in marine biology 30, 156-156
(1993).
10.
Sourisseau, M., Simard, Y. & Saucier, F. Krill aggregation in the St. Lawrence system, and
126
supply of krill to the whale feeding grounds in the estuary from the gulf. Marine EcologyProgress Series-314, 257 (2006).
11.
De Lafontaine, Y., Demers, S. & Runge, J. Pelagic food web interactions and productivity
in the Gulf of St. Lawrence: a perspective. The Gulf of St. Lawrence: small ocean or big estuary,
99-123 (1991).
12.
Kingsley, M. & Reeves, R. Aerial surveys of cetaceans in the Gulf of St. Lawrence in 1995
and 1996. Canadian Journal of Zoology 76, 1529-1550 (1998).
13.
Simard, Y. & Lavoie, D. The rich krill aggregation of the Saguenay-St. Lawrence Marine
Park: hydroacoustic and geostatistical biomass estimates, structure, variability, and significance
for whales. Canadian Journal of Fisheries and Aquatic Sciences 56, 1182-1197 (1999).
14.
Gomez-Gutierrez, J., Peterson, W., De Robertis, A. & Brodeur, R. Mass mortality of krill
caused by parasitoid ciliates. Vol. 301 339-339 (American Association for the Advancement of
Science, 2003).
15.
Hamner, W., Hamner, P., Strand, S. & Gilmer, R. Behavior of Antarctic krill, Euphausia
superba: chemoreception, feeding, schooling, and molting. Science 220, 433-435 (1983).
16.
O'Brien, D. Description of escape responses of krill (Crustacea: Euphausiacea), with
particular reference to swarming behavior and the size and proximity of the predator. Journal of
Crustacean Biology, 449-457 (1987).
17.
Grinbaum, D. Predicting availability to consumers of spatially and temporally variable
resources. Hydrobiologia 480, 175-191 (2002).
18.
Menden-Deuer, S. & Gr?baum, D. Individual foraging behaviors and population
distributions of a planktonic predator aggregating to phytoplankton thin layers. Limnology and
Oceanography 51, 109-116 (2006).
127
19.
Hutchinson, G. The paradox of the plankton. The American Naturalist 95, 137 (1961).
20.
Haury, L., McGowan, J., Wiebe, P. & Steele, J. Patterns and processes in the time-space
scales of plankton distributions. Collected Reprints. Woods Hole. Dec. 1979 (1978).
21.
Azam, F. Microbial control of oceanic carbon flux: The plot thickens. Science 280, 694-
696 (1998).
22.
Fenchel, T. Microbial behavior in a heterogeneous world. Science 296, 1068-1071 (2002).
23.
Azam, F. & Malfatti, F. Microbial structuring of marine ecosystems. Nature Reviews
Microbiology 5, 782-791 (2007).
24.
Seymour, J. & Stocker, R. Resource patch formation and exploitation throughout the
marine microbial food web. The American Naturalist 173(2008).
25.
Azam, F., et al. The ecological role of water-column microbes in the sea. Mar. Ecol.-Prog.
Ser. 10, 257-263 (1983).
26.
Blackburn, N., Azam, F. & Hagstrom, A. Spatially explicit simulations of a microbial food
web. Limnology and Oceanography 42, 613-622 (1997).
27.
Blackburn, N. & Fenchel, T. Influence of bacteria, diffusion and sheer on micro-scale
nutrient patches, and implications for bacterial chemotaxis. Mar. Ecol.-Prog. Ser. 189, 1-7 (1999).
28.
Azam, F. & Long, R.A. Oceanography - Sea snow microcosms. Nature 414, 495-+ (2001).
29.
Blackburn, N., Fenchel, T. & Mitchell, J. Microscale nutrient patches in planktonic habitats
shown by chemotactic bacteria. Science 282, 2254 (1998).
30.
Seymour, J. & Stocker, R. Resource patch formation and exploitation throughout the
marine microbial food web. The American Naturalist 173, 15-29 (2008).
31.
Keller, E. & Segel, L. Model for chemotaxis. Journal of Theoretical Biology 30, 225-234
(1971).
128
32.
Stocker, R., Seymour, J.R., Samadani, A., Hunt, D.E. & Polz, M.F. Rapid chemotactic
response enables marine bacteria to exploit ephemeral microscale nutrient patches. Proceedings
of the National Academy of Sciences 105, 4209-4214 (2008).
33.
Blackburn, N., Fenchel, T. & Mitchell, J. Microscale nutrient patches in planktonic habitats
shown by chemotactic bacteria. Science 282, 2254-2256 (1998).
34.
Dusenbery, D.B. Sensory ecology: how organisms acquire and respond to information,
(WH Freeman New York, 1992).
129
APPENDIX A
Synergistic Effects of Chlorine and Antifouling Surfaces
The experiments on surface biofouling described in this paper, which resulted from a
collaboration with Prof. Karen Gleason's group in Chemical Engineering at MIT and was
published in Advanced Materials in 2013, were conducted by me, taking advantage of the
approaches and expertise developed in this thesis. Specifically, I contributed Figures 3 and 4 (and
associated text and Methods sections) in this paper. This publication is included here with
permission of all the authors.
130
4
$4A
www.MaterialsViews.com
4
i
ERV
im-5WWW.2dvmat-de
Synergistic Prevention of Biofouling in Seawater
Desalination by Zwitterionic Surfaces and Low-Level
Chlorination
Rong Yang, Hongchul jang, Roman Stocker,* and Karen K. Gleason*
Water scarcity affects one in three people in the world.!' With
nearly 98% of the wvorld's available water supply being seawater or brackish water, desalination has become an important
means to address the scarcity of freshwater resources. Thin
film composite (TFC) reverse osmosis (RO) membranes enable
the removal of salt ions from seawater at room temperature by
applying pressure to the seawater feed. TFC-RO has quickly
become the dominating desalination method since its commercialization in the 1980s and is now used in nearly all RO desalination plants. TFC-RO is considered to have the greatest
water permeability with high salt rejection rate. 2 The bottleneck for TFC-RO to produce freshwater via seawater desalination at a comparable price to natural freshwater is severe
membrane fouling, which impairs water permeation and salt
rejection and thus reduces freshwater yield. Currently, marine
biota and in particular bacteria are removed from the feed by
pretreatment, the most energy-intensive (responsible for >36%
of total plant energy consumption) and chemical-intensive step
in a desalination plant and one that poses environmental risks
to marine organisms when treated water is discharged back
into the ocean. Fouling-resistant RO membranes would bring
major improvements in energy usage. process reliability and
lower the environmental impact of seawater desalination.
Zwitterions are a type of molecular structures with ultra-low
fouling properties, demonstrated in applications ranging from
bio-assays to artificial tissues1'I originating frot the extreme
hydrophilicity induced by electrostatic interaction with watt-n
nolecules, '" which makes the replacement of surface-bound
water molecules by foulants euthalpically unfavorable. However. the zwitterionic coatings fabricated so far are not sufficient in long-term antifouling applications due to the limited
stability in real-world environments. 1
The major challenge in the surface modification of TEC-RO
membranes is to implement antifouling chemistries without
R.Yang, Prof- K.
K. Gleason
Department of Chemical Engineering
Massachusetts Institute of Technology
77 Massachusetts Avenue, Cambridge
Massachusetts. 02139, USA
E-mail: kkg@mit.edu
Dr H. Jang, Prof. R. Stocker
Ralph M. Parsons Laboratory
Department of Civil and Environmental Engineering
Massachusetts Institute of Technology
77 Massachusetts Avenue, Cambridge
Massachusetts. 02139. USA
E-mail: romans@mit.edu
DOI: 10.1002/adma.201 304386
Adv Mair, 2013
DOI 10.1002,adma.201304386
ii
compromising salt rejection andI high water flux The imiting
step for the transport of water and salt across nembranes is
the extremely thin (- 100 to 200 tn) polyamidce selective layer
(Figure la). Pin-holes or defects in the polyamide layer are
routes for non-selective salt transport and thus quench the salt
rejection performance of the membranes. Surface modification methods involving solvents or exposUre to high temperatures (Table 1) can generate or enlarge the undesirable pin-hole
defects." Surface modification layers produce an additional
resistance to water permeability. We have shown previously that
the coatings on RO membranes should be 30 nm or thinner
and thicknesses >100 nm are undesirable because they cause
t
>40% reduction in the water flux.1
Recently, we showed that anti-biofouling coatings of various
compositions can be grafted and directly deposited on commercial TFC-RO membranes via an all-dry process. called initiated
chemical vapor deposition (iCVD).)
The low-temperature.
solvent-free processing leaves the delicate polyamide intact and
thus maintains the high salt rejection. Water flux is maintained
by utilizing ultrathin (30 im) iCVD layers. However. these
acrylate-based films do not resist the degradation by chlorine,
the most prevalent disinfection reagent in water treatment.
We report here a novel pyridine-based zwitterionic surface chemistry that displays significantly improved resistance
against a variety of molecular foulants and improved tolerance
to chlorine exposure as compared to acrylate-based analogues.
The chlori n-resistant surface provides a new perspective for
achieving long-term antifouling. The pyridine-based iwittrionic surfaces demonstrate a synergy with dritiking-water-level
chlorination (5 ppm), resulting in exceedingly high antifouling
performance Synergistic effects have often been observed in
the interactions between pairs of molecules such as pairs of
drugs or toxins, or pairs of surface properties. such as surface
energy and roriglhness. However, to our knowledge, synergistic
effects have not been specifically identified between a fnlictional surface and a solution species. The chlorine-resistant
antifouling surfaces are derived from ultrathin iCVD poly(4vinylpyridine) (P4VP) and its copolymers. i "I The vapor dep-
osition allows the synthesis of insoluble cross-linked coatings
as thin films directly on a Surface in a single step. Enhanced
durability results from cross-linking co-msonomers and in situ
grafting. The in sitri reaction with 1.3-propanesultone (PS)
vapors produces pyridine-based sulfobetaine zwitterioiiic functional groups, having a balanced surface charge. The iCVD synthesis is carried out at loss surface temperature (20 'C) to produce robustly adhered, smooth, ultrathin layers (30 rm) directly
on even delicate substrates. such as TFC-RO membranes
without damaging them. Accelerated testing against marine
t2013 Wil F' VCH Verlag GmbH & Co KGaA Weinheim
131
wileyonlinelibrary.com
1
AAXd
wwwadviatde
ww.atral~es.com
iCVD
20
100 -()
(d)
350 Quaternized pyridine,. Pyridine
-401.5 V
*
-399.5 eV
15 *(f)
-80
300
10
250
?40
5
200
.- .
150
406
-
........
404
402 400 398
Binding energy (9V)
0
396
Bare
membranes
0
Coated
membranes
Bare
membranes
Coated
membranes
Figure 1. Antifouhng zwitterionic coatings applied onto commercial RO membranes via iCVD. ab) Cross-sectional SEM image of (a) bare and (b) iCVD
coated RO membrane Panel (a) shows the porous supportive polysulfone layer (colored in orange) beneath the nonporous, 200-nm-thick, selective
polyamide layer of the RO membrane. In (b), the smooth top layer is the iCVD zwitterionic coating, which is grafted to the selective layer. c) AFM scan
of coated membrane and (inset) bare membrane. Both surfaces are exceptionally smooth, with -1 nm RMS roughness. d) N(ts) XPS high resolution
scan of the iCVD P4VP as-deposited (blue) and derivatized by PS (red), demonstrating full conversion of pyridine to zwitterion. e) Salt rejection of bare
and coated membranes The comparable values of salt rejection indicate that the coating leaves the thin selective layer of the delicate RO membranes
intact f) Water flux through bare and coated membranes Membranes coated with 30-nm functionalized copolymer 1maintain 86% of the original
water flus. Error bars (ef) represent the standard deviations obtained with 3 parallel tests.
bacteria in multichannel inicrofluidic devices shows 100-fold
generation of RO membranes'E
i7
and of antifouling strategies
reduction in biofouling on the coated surface compared to bare
glass. The unique resistance of the pyridine-based films against
in desalination plants, and find additional utility on the hulls of
ships and for submerged marine structures. ix
Ultrathin (30 to 300 rnm) iCVD coatings are successfully
grafted and deposited directly onto commercial TFC-RO menbranes (Figure l), followed by the vapor phase derivatization
(Supporting Information, Figure Sl). The all-dry-processed
coating conforis to the geometry of the underlying substrate
(Figure lb.c), because surface tension and de-wetting are
avoided. The root-tnean-sqtiare (RMS) roughness of bare and
coated RO membranes is 1,1±0.3 nm (Figure ic, inset) and
degradation by chlorine allows a new synergistic approach to
antifouling, which substantially enhances longer-term fouling
resistance compared to either surface modification or chlorination alone, and has the potential to reduce or eliminate
pretreatment of seawater, the most energy- and chenical-intensive step in desalination plants, and thus to reduce the cost of
freshwater production and its collateral toxicity to marine biota.
This approach can facilitate the rational design
of the
next
Table 1. Comparison of the important characteristics of surface modification techniques for zwitterionic antifouling chemistries
Methods
SAMs
Atom -tiansfer
radical-polymenzationir
Bulk solution
polymerizatton
Layer-by-layer
,CVD
All dry processing
X
X
X
X
V
Substraie-mdependene
x
X
V
x
V
10,
10
Synthesis speed
Inm mn
2
10
Small post-treatment roughness
V
V
Conformal (oating
V
V
ultra-thin coating
V
V
X
High surface concentration of zwittenonic groups
V
V
x
wileyonlinelibraryxom
k 2013 WiLEY-VCH Veriag GmbH & Co KGaA, Weinheim
10
V
V
V
V
V
V
V
Ad, Mxtr 2013,
DOI
132
1
x
X
10 1002/adma 201304386
ADVAN ED
MATER ALS
www.advmat-de
www.MaterialsViews.com
0.8±0.1 ni (Figure tc), respectively. This exceptional smoothness is critical to the fouling resistance of the membrane sur12
face, 1 because larger surface areas aid more binding sites
are available for foulants to attach on a rougher surface. Its
addition. nano- atid micro-scale roughness cats entrap proteins
atid bacteria. respectively. arsd provide a "shield" to attached
foulaits from shear forces.1iI The beiigni reaction conditions
allow retention of the zwitterionic groups. as evidenced by the
N(1s) high-resolution scan by X-ray photoelectron spectroscopy
(XPS, Figure Id). The binding ettergy of the pyridine tsitroges
species in the as-deposited iCVD layer is -399.5 eVtI with
a small tail around 402 eV that is attributed to the inevitable
post-treatinteit adsorption of atisiuphieric CO,1 1Tle binding
enterg) ofquaternized pyridine nitrogen is -401.5 eV'' atid the
sytsmetric peak profile indicates complete quatertization by
[te derivatization with PS.
The salt rj ction of the surfce-iodified RO membranes
is inialtered. confirtmig the betnigit nature of the solvent-free
process (Figure le). This substrate-itdepeniceit method allows
simultaneous depositioin ots multiple substrates. This feature is
used to simultaneously deposit on RO meibranes arid on a silicon wafer in order to achieve precise control of coating thickness, which is critical because thi coatings are essential
to
12
maintaining tile high water flux through RO membranes. The
coating thickness ot RO membranes is compared to that otn a
silicon wafeiY' which is monitored via in sits interferotsetry.
With the 30-nm coating thickness achieved by this method. the
water flux is reduced only by -14% compared to untreated RO
rmembranes (Figure 10. This high water flux is achieved with
an amount of dse cross-linker (4%; copolymer 1; Supporting
Information, Figure S). divinylbenzene (DVB). sufficiently
high to ensure the stability of the coatinsg arid sifficiestly low
to effect a minimal reduction in water flux, As expected. water
flux is reduced (by 72%; Supporting Information, Figure S2)
for higher DVB content (17%: copolymer 2) aid also (by 84%)
for the homsopolymser PDVB, owing to its high cross-linking
denssity. It is worth noting that copolymer 1 despite the higher
cross-littkinsg density. has sinilat water Oux as hioiopolyier
P4VP (Supporting Iifornatioti Figures Sl atsd S2). This is
likely a result of surface chain reorganiation of copolymer
I upoi contacting water. Therefore. copolymer t is chosen as
providing the optimssal trade-off betweei coating stability aitd
water flux. Taken together, these rstllts
Sh1w that lit proposed
approach cats overcome the major challenge in the field of surface modification for desalination. by itmplemienting atitifouling
chemistry without cotttprosising water flux and salt rejectiots
of the resulting membranes.
To reveal the cheministry of the autifoulitig coatings, lite rtetition of fisctional groups arsd compositions of as-dtposited and
liitctionalized iCVD polymers are analyzed using Fourier traitsform infrared (FTIR). Excellent agreement is observed betveen
the spectra of iCVD- ansd solution-polyierized PDVB atsd
M4
"" indicating that the tson-vinyl organic functionalities
in the monomers are retained in the iCVD films. Successful
polymerization of DVB (Figure 2b) is evidenced by the reduction of the 903 cms
peak in the PDVB spectrun (Figure 2a.
black). which results from the out-of-plane CH, deformation
its vinyl groups. The existence of this peak its the PDVB spectrum is due to the presence of unreacted petsdant vinyl
Adv
DOI
MOer 2013
10.10021adra.201304386
-
2013 W11 FY VcH
Verlag
bonds.l -"} For the iCVD polymers PDVB (black), copolymer
2 (wine). copolymer I (magenta) and P4VP (blue), there
is a decreasing trend in the area under the 710 cm-' peak, a
neasure of the number of In-substituted aromatic rings in the
DVB repeat units (Figure 2a). This is utilized to calculate compositions of the iCVD copolymers"I which are confirmed by
XPS survey scans. The composition of the copolymers can be
tuned simply by varying the flow rate ratios of 4VP and DVB
monomers (Supporting Information, Figure SI). In the spectra
of P4VP and copolymers, the strong peak at 1600 cm ' is attributed to the C-C and C-N stretching vibrations in the pyridine
ring (Figure 2c).,"' whose intensity increases with more
P4VP repeat units (Supporting Information, Figure Sl). FTIR
spectra collected after the PS derivatization (Figures 2a and
SI) confirms the formation of the pyridine-based sulfobetaine
(Figure 2d) via ring-opening of PS, as evident by the appearanice of a peak at 1036 cm in the spectra of functionalized
P4VP (red) and copolyici I (orange) (Figure 2a). This peak
21
is attributed to the symmetric stretching of the SO, group 1
Therefore. pyridiie-based zwitterionic structures designed to
resist oxidative damages are successfully synthesized using the
solvent-free scheme.
To evaluate the chlorine resistance of the iCVD films. we subject the functionalized homopolymer P4VP, copolymer I and
copolymer 2 to treatment with a 1000 ppm solution of sodium
hypochlorite and we acquire FTIR spectra after different treatment durations. From the spectra. we measure the areas Under
the 1600 cm ' peak (Figure 2a,e) to quantify te functional
retention of the zwitterionic structure; the strong peak intensity
renders the quantification more accurate. The excellent chlorie resistance of copolymer 1 (4% DVB) is evident from the
negligible changes in its spectrum after 2 (green) and 24 (grey)
hours of chlorine treatment (Figure 2a). In contrast.
homopolymer P4VP is rendered soluble by a 10-hour exposure, as shown by the absence of functional peaks in the FTIR
spectrum (Supporting Information.Figure S3). Importantly, the
addition of 4% DVB cross-linker produces a major increase in
the resistance to chlorine. whereas additiotis beyond 4% result
in minor additional resistance (Figure 2c): after 10000 ppim 1i
exposure to chlorine, -94% and -99% pyridine functionalities
remain in functionalized copolymers 1 (4% DVB) and 2 (17%
DV-), respectively. Functionalized copolymer I is thus most
desirable because it resists chlorine very effectiveiy while lea iiig
the water flux nearly intact (Supporting InfortiationFigure S2).
These observations are corroborated by dynamic contact
atgle mteasuren nts on functionalized P4VP and copolymer
1 before atid after chlorine treatment. which yields a comprehe-nsivc evaluatii of thc effects of chlorine on the coatings.
because the dynamic contact angles of coated surfaces are
affected by coating chemistry. surface roughness, swelling,
12
and surface chain reorganization
For the functionalized
P4VP. before chlorine treatment we measure advancing atid
receding contact angles of 31" and 20 whereas after 2000 ppms
h chlorine exposure these values become 48' aid 18 , respectively (Supporting Inforsation,Figure S4). These considerable
changes in dynamic contact angles reflect the poor chlorine
resistance of the fiutctionalized P4VP films. In contrast, the
advancing arid receding contact angles of the functionalized
copolymer I are 51< and 24, respectively. In spite of tle higher
GmbH
133
& Co KCaA. Weinheirn
wileyoninelibrary.com
3
AS
wwwadvmat.de
www.MaterialsViews.com
(a)
(.)
-
000 ppm chlorine 2 h
UnCtionalfred COPoymer I
d
Funttetonalized
u
PAVP
-d--Functonab-z" copolymer
2
20
1E
P
P4VP
IL
copotyme, I
0
0
5000
10000 15000 20000 25000
Chlorine exposure (ppm h)
Copolymeor 2
6f
PDVB
3000
2500
(b) DVB
2000
Wavenumber
(c) 4VP
1500
(cm')
(d) ZwIttaeion
1000
i
i
1-
030
as
e±4
=
ON
A,
01U4ti;!I
40.
f
C
4~4~
201
5
10
is
Drop vol-m (p)
20
25
Figure 2. Chlorine-resistant zwitterionic chemistry. a).FTIR spectra of homopolymers and copolymers as-deposited, after PS functionalization, and
after chlorine exposure. Copolymer 1 and 2 contain 4% and 17% DVB repeat units, respectively. The spectra of functionalized P4VP and copolymer
1 display a peak corresponding to the zwitterionic moiety (1036 cm 1), Copolymer 1 shows unchanged spectra after 1000 ppm chlorine treatment for
2 hours and 24 hours, demonstrating excellent chlorine resistance Spectra are offset vertically for clarity. b-d) Molecular structure of the cross-linker
DV8, 4VP and the zwitterionic moiety obtained after functionalhzation e) The polymers' chlorine resistance, quantified as the area under the 1600 cm 1
peak (corresponding to the pyridine ring). Functionalized homopolymer P4VP does not resist the oxidative damage of chlorine, whereas functionalized copolymer 1, containing merely 4% cross-linker repeat units, resists chlorine considerably better. Increasing cross-linker repeat units beyond 4%
improves chlorine resistance only slightly. f.) Advancing (0) and receding (A) contact angles of the functionalized copolymer 1 before and after chlorine
treatment. The drop volume is the volume of the water droplet used to measure the contact angle. Contact angles are unchanged by chlorine treatment,
confirming the chlorine reststance observed via FTIR (a).
cross-linking density, copolymer I has similar receding contact
with PS is diffusion-limited and the zwitterionic moieties are
only present in the top few nanometers. 'I The consistent
fouling resistance under low (PBS buffer) and high (2 M NaC
added to PBS buffer, corresponding to -117,000 ppm NaCl) salt
concentrations implies that the functionalized copolymer 1sur-
angle as P4VP. This is a sign of surface chain reorganization1
and corroborates the comparable water flux obtained with functionalized P4VP and copolymer I films. The dynamic contact
angles remain unchanged after as much as 24000 ppti h chlorine treatment (Figure 20, confirming the excellent chlorine
resistance of the funrctionalized copolymer I films: the coating
chemistry. surface roughness, swelling, and surface chain
reorganization all remain essentially unaltered even upon pro-
longed exposure
face is charge-neutral (Figure S6).
The good fouling resistance against dissolved chemicals
leads us to test the surfaces against fouling by marine bac-
teria. We use both natural seawater samples and a culture
of Vibrio cylitrophicus. a species broadly representative of
to chlorine.
We demonstrate the anti-biofouling properties of the new
surface chemistry both with dissolved foulants and with marine
bacteria. Quantification of the surface adsorption of I tig mlbovine serum albumin (BSA) in phosphate-buffered saline
(PBS) is conducted via quartz crystal nicrobalance with dissipa-
bacteria prevalent in coastal waters, from where seawater
for desalination typically originates. The dynamics of bacterial attachment are studied in a microfluidic flow system and
imaged with an inverted microscope equipped with a CCD
camiera
Images are extracted from full movies (Supporting
I nformation, Movies) and quantified by image analysis, Microchannels of 600 x 100 pm rectangular cross-section are fabricated out of polydimethylsiloxane (PDMS) using standard
soft lithography techniques"T and mounted on a microscope
glass slide that has been coated with a -300-nm-thick film of
functionalized copolymer 1. Fresh seawater is harvested and
used on the same day as the feed solution for the microfluidic
fouling tests, without any pretreat ment, through continuous
injection at a rate of 2 nil trin' (corresponding to a mean flow
tion monitoring (QCM-D). BSA is a widely used test protein for
antifcouling studies. '"" Analogous tests are carried out with a
representative polysaccharide. I tig mil sodium alginate, the
major component of extracellular materials that lead to membrane biofouling. 2 QCM-D tests reveal no adsorption of either
foulant over 200 minutes on the frmctionalized copolymer
1 surface
(Supporting Information. Figure S5). The thickness
of the coating does not have an impact on the fouling resistance
(Supporting Infortrmation,Figure S6), because the derivatization
4
wileyonlinelibrary.cor
C 2013 Wit EYVcH Vedag
GmbH
PcCo
KGaA.
Weinheim
DO
134
10
Ad Mater 2013.
102/1adma 20130438b
A
BarI
m
5
S
www advmat.de
www.MaterialsViews.com
chlorine
1
5 ppm chlorine
h
12 h
farC
naterI
0 min
85
min
120 min
Figure 3. Enhanced fouling resistance by zwitterionic surfaces and low-level chlorination. a-h) Attachment of concentrated suspensions of the marine
bacteria v cyclhtrophicus to glass surfaces with (ae) no
(bf) the zwitterionic coating (functionalized copolymer 1), (c,g) chlorination (5 ppm),
and (dh) the zwitterionic coating plus chlorination, after 5 hours (a-d) and 12 hours (e-h). The zwitterionic coating shows no signs of fouling after
5 hours under accelerated biocifouling tests conditions (b). whereas after the same arnount of time the bare surface has significant surface coverage by
bacteria (a) After 12 hours, neither the coating alone (f) nor chlorination alone (g) is effective at resisting biofouling, whereas the combined treatment
exhibits dramatically increased fouling resistance and maintains a clean surface (h). Relative fouling indices, F, ib,f) the fraction of surface coverage
for the coated surface compared to the bare glass control - and F (cg) - the fraction of surface coverage in the presence oc chlorination, compared
to that in the absence of chlorination for a bare glass surface are used to quantify the effects of coating and chlorination, respectively- The synergistic
fouling prevention is quantified by the synergistic odex, S (dh). where S - 1 indicates synergy between the coating and chlorination See also Sup
porting Information, Movies 3 and 4 Images in (a-I) are captured with the sane magnification and the scale bar represents 50 pm iq, Comparison of
the attachment and proliferation of a V cyclitrophicus bacterium on (Io) a bare surface, (Im,p) a bare surface with chlorination, and (linq) a coated
surface with chlorination The 5 ppm chlorine addition did not prevent bacterial proliferation on the surface (m,n,p) The -witterionic chemistry is
sritical for the synergistic fouling resistance, as bacteria are readily removed from the zwitterion-coated surface by even laminar flow (Reynolds number
0 1) (q) See also Supporting information, Movie 5 The scale bar represents 5 pm
treatment:
velocity of -(560
prl s 1). Fabricatiot of multiple (2-4) microChannels oct the same chip allows parallel sinmultatteous experiments and thus a dire t comiparisoti of different treatments atid
the minimization of confouiding factors. Because experiments
lasting vip to 100 hours reveal tio dliseerttble surface attachtient fStipporitig Information. Figure 57: Sttpporting Irrforciation Mov ies I attd 2), irrcsp-ctiv of surface conditions, we
rutn accelerated fouling experiments with concentrated cultures
Of V Cylsirplicus, grorwtt isverntight in artificial seawater arid
0.2: 2 x 10 cells
concentrated to an optiral dnsity (OD,,,
md ') corresponding to i-arty expoietial phas-. This bacterial
corceritration is 200 tises that of tvpical seawater
Ma., 20153
001 Ii 1002/vrv -i201s0at8
Ad,
201. A
EiYVCH
lit fle accelerated tests the iCVD zwitterione
oatings
show muich greater resistance to bacterial attachment than
bare glass (Figure 3a,b.f. Supporting Information. Figrtre 58:
Supporirtig Information. Movie i).
Foulitig is quatitfi-ed by
time-lapse imaging of the surfac-, followed by image arsals sis
to determine the number of attached cells arid the percrit surface cosrage tiy bacteria- As tht variables are time-depc1nt,
thf- behavior at 5 hours atid 12 hours will be csussed, but
general conclusions apply also to the data at other tites.
Despite the inTtrinnsis fouling resistance of glass surfaces,
th- number of attached cells oil bare glass increass steadily
over ttie, atid exponentially after 50 mnintites. After 5 hours,
GmbH&
V'cclgi
135
caA weIhi-
wileyonlinelibraryeom
S
www advmat.de
www.MateriaIsViewscom
"ID
2
the cell count over a 0.16-mn area of the bare glass surface
reaches -7500, (Figure 3a), whereas it renains close to zero
otn the coated surface (Figure 3b). Defining a relative fouling
index, F1 , as the fraction of surface coverage for the coated
surface compared to the bare glass control, we find that F,
decreases drastically over time for functionalized copolymer
I and drops to -0.01 after 5 hours (Supporting Information.
Figure S8). This result detionstrates the exceptional fouling
resistance of iCVD zwitterionic coatings, in particular in view
of the fact that smooth, bare glass is already a rather good antifouling surface.
The surfaces' antifouling ef-ects are further boosted by lowlevel chlorination,
(a)2
0
60
0
-
200
reslting in a new synergistic approach
400
0.6 (b)
No
chW
(Figure 3cdgh; Supporting Information, Movie 4), a concentration cormparable to the residual chlorine level in the USA
0.4
E0
national drinking water standards..i To quantify the effect of
chlorination we define a second fouling index, F2 , cotiputed as
the fraction of surface coverage in the presence of chlorination,
compared to that in the absence of chlorination, for the case of
a bare glass surface. Although chlorination overall reduces sur-
70
E
.
30
0
200
400
600
Figure 4. Synergistic prevention of bacterial fouling by the combination
treatment, a) Surface coverage by V cyclitrophicus bacteria under different
conditions The synergistic treatment - integrating iCVD zwitterionic
coating with low-level (5 ppm) chlorination - shows exceptional longterm antifouling activity even under accelerated biofouling conditions
(i.e., dense bacterial suspensions), when each method in isolation begins
to fail. Inset: Time series of the synergistic index (5), quantifying synergistic effect of the two antifouling strategies. Values of S < 1 indicate a
positive synergy between the two treatments. The monotonic decrease
of S, with no signs of saturation over 15 hours, demonstrates the importance of synergistic effect on long-term fouling resistance. b) Viability
of V cyclitrophicus upon addition of chlorine at different concentrations.
1 ppm chlorine does not significantly impact bacterial growth, whereas
5 ppm chlorine reduces the optical density by 42%, but does not kill bacteria. Killing by chlorine is thus not the dominant factor in the success of
the synergistic treatment. c) Mean swimming speed of V cyclitrophicus,
obtained by tracking of individual cells. Addition of up to 5 ppm chlorine
does not significantly change the bacteria's swimming speed, suggesting
that prevention of attachment is not due to a reduction of encounter rates
with surfaces.
'
1
to -0.1 after 900 minutes. No signs of saturation in the decrease
are observed, demoonstrating the long-term nature of the synergy. Values of S over the first 5 hours are not reported because
the surface chemsistry alone reduces fouling to non-detectable
Vedag
5 ppm chlorine
Time (min)
%Surfaceoinvrage
wiLEY-vcH
choeln.
I ppm chiori
..
The temporal dynanics of S (Figure 4a, inset) reveal values
S < after -400 minutes, and a subsequent steady decrease
e 201
No
p 40
%Surfacecoverage,
wileyonlinelibriry.com
800
so
%eSurfceionv-rag-
6
600
(C)
6o
a
%Surfacesonvevrage
of
400
blctwia)
Time (min)
(Figure 4a, inset). Synergistic indices have been used among
others to describe the effects of multi-strategy anti-tunor treatients. wheie S <
indicates a synergistic effect in killing
F
200
0
osent in isolation (Figure 3d,h. After 12-hour exposure to the
V cyclitrophicits suspension, the surface coverage is 35.5 t 1.7%
on bare glass in the presence of 5 ppn chlorine (F2 - 058),
14.1 ± 3.4% on the coated surface without chlorine (F1014),
and only 1.5 ± 0.4% on the coated surface in the presence of
9 ppns chlorine. The percent surface coverage in the synergistic
treatient is 0.02 of that of a bare glass surface without chlorine,
four-fold staller than the prediction (F1 x F2) obtained if the
effest was simply miultiplicatis .
To quantify the synergistic effect of the two antifouling
strategies, we curpute an antifouling synergistic index, S
F
Control (no
0.1
terionic coating in preventing bacterial attachment. However.
the synergistic effect of the zwitterionic coating and chlorination dramnatically increases fouling resistance over eac treat-
S
- ppm chlorine
00 .2
glass in the presence of
5 ppm chlorine enserge after 5 hours (F 2 - 0.45: Figure 3c) and
after 12 hours fouling is severe (F2 - 0 58: Figure 3g). Therefore.
chlorination at a level of 5 pprn is less effective than the zwit-
treatnent.
I ppMf cilone
o 0.3
C.
of fouling on bare
1
SIMI
600
0.4
good resistance to chlorine (Figure 2aef). We run addi-
tumior cells by the different strategies in the
Here we define S as
c
ppm Cl
0
tional, accelerated oticrofluidic tests where the suspension of
V.cyclitrophicus is amtended with 5 ppn of sodiun hypochlorite
face fouling, signs
p
00
against fouling tmade possible by functionalized copolymer
1's
Thm (minI
0.
CmbH & co
GCaA, Weimril
Adv Mate 2013.
DOI 10 1002/adma.201304386
136
ADVANCED
MATER ALS
www.advmat.de
www.MaterialsViews.com
levels (i.e., F,
0) during this time and thus the quantification
of S is not mneaningful.
In the attempt to reveal the mnechanism underpinning the
synergistic effect, the cell-surface interaction is investigated by
observing a single bacteriumn for its proliferation and motility
on the surface for the different treatments (Figure 3i-i Supporting Information. Movie 5). After 85 minutes, replication
has occurred under all conditions (Figure 31-n), at a tnildly
lower rate in the presence of 5 ppm chlorine (Figure 3mi.n), suggesting that the low dose of chlorine has only small effects on
cell growth. This hypothesis is supported by direct viability tests
(Figure 4b), showing that the growth of V qclitrophicis (measured as the optical density of cell cultures) is negligibly aflected
by addition off ppm chlorine and exhibits a 42% reduction with
5
ppm chlorine addition. Furthermore, tracking of individual
cells shows that motility is not significantly affected by I ppm
or 5 ppm chlorination (Figure 4c). Although growth in batch
culture itiglit difTer from growth on a microchannel surface.
taken together these results (Figures 41) and 31-n) denionstrate
that the observed antifouling and synergistic effect of chlorine
are not based on killing of the bacteria. Instead, the primary difference among the three single-cell cases (Figure 3i-q) resides
in the dependence of cell removal froni the surface on the starface chemistry (Figure 3o-q): whereas bacteria remain largely
attached to the bare glass surface, they are easily removed from
the coated surface by anmbient fluid flow. independent of the
presence of chlorine. In particular. bacterial renoval rom the
iCVD /witterionic coating occurs rcadily eventinder the low
laminar flow conditions within the microchannel (Reynolds
number 0.1).
We have demonstrated the ability of
ultrathin.
chlorine-
resistant iCVD zwitterionic copolymners to act as antifouling
coatings arid, based on livir resistance to chlorine, we have
proposed a novel, multi-strategy approach to antifouling, which
hinges on the synergy between surface chemistry and chlorination. The zwitterionic coating prevents the attachment of'
V cyclitrophicuas almost 100 times more effectively than glass
after 5 hours (Figures 3ab and 4a: Supporting Information.
Figure S8), while chlorination. with concentrations as low
as the regulated chlorine residue in drinking water, is able to
enhance the long-term fouling resistance of the zwitterionic
coating by 9.4-fold after 12 hours (Figures 3d,h and 4a). with no
signs of satUration.
A key advantage of the zw itterionic coatings reported here
is the substrate-independence of the vapor application process.
which makes these coatings easily applicable to a broad range
of surfaces. In particular, these coatings may be applied on the
latest salt-reject ing layers,
x" iwhih resist exposure to dhlorine, providing a path towards solving the desalination indtustry's bottleneck of the susceptibility of TFC-RO metnibranies to
oxidative danage by chlorine. The surface treatnent is benign.
easily scalable " and conpatible with the infrastructure in
ni-ibrain- industry.'t whit Ii gives rise to a stabli, non-toxic and
inexpensive ultrathin coating. The good fouling resistancc and
chlorine resistance of this coating can help clininate the most
energy- and cheinical-intensive step (pretreatient of seawater)
in a RO desalination plant~l and reduce the environmcental
imipacts of brin- discharge. This approach therefoin proinises
to lower the price of freshwater in water-scarce countries, \%here
Adv
DOI
Mater 2013
10,1002/adrna.201304386
2013
t
WIt EY. VCH
desalination may serve as the only viable means to provide the
water supply necessary to sustain agriculture, support personal
consuiption. and promote economic development.
Experimental Section
Film Deposition and Derivatization: All
films were deposited
chamber. The glass slides were treated with trichlorovinylsilane (Aldrich,
97%), as described previously.ii During iCVD depositions, 4VP
(Aldrich, 95%) and DVB (Aldrich. 80%) monomers were heated up to
50 C and 65 C in glass
respectively and delivered into the reactor
jars,
using mass flow controllers (1150 MFC, MKS Instruments). Argon
patch flow was metered into the reactor through a mass flow controller
(1479 MFC, MKS Instruments) and the flow rate was varied to keep the
residence time constant. Systematic variation of the flow rate ratios of
the two monomers was performed to yield high-zwitterionic-percentage,
yet chlorine-resistant films of poly(4-vinylpyridine-co-divinylbenzene)
(PVD). Films were deposited at a filament temperature of 250 C and a
stage temperature of 20 C Total pressure in the vacuum chamber was
maintained at 0.8 Torr for all depositions.
In situ interferometry with a 633 nm HeNe laser source (JDS
Uniphase) was used to monitor the film growth and deposit desired
thicknesses on Si substrates. A more accurate film thickness on the Si
wafer substrates was measured post-deposition using a .A. Woollam
M-2000 spectroscopic ellipsometry at three different incidence angles
(65 . 70 . 75 ) using 190 wavelengths from 315 to 718 nm. The data
were fit using a Cauchy-Urbach model. After deposition, the PVDcoated substrates were derivatized as reported previously"
FTIR.
XPS and contact angle measurements were performed as described
previously "I
Permeation and Salt Rejection Tests: Tests of the coated/bare
membranes were performed using a commercial dead-end membrane
filtration unit (Sterlitech Corp., HP4750) with a nitrogen cylinder to
supply feed pressure, which was kept at 700 psi for all tests. The flow
rates of the permeate were determined using a 100 ml metered flask.
For the salt rejection tests, 35000 ppm sodium chloride dissolved in
deionized water was used as feed solution. A conductivity meter (CDH152. Omega Engineering Inc.) was used to measure the conductivities of
the feed and permeate to calculate the salt rejection.
Chlorine Resistance Tests: Samples subject to chlorine resistance
tests were soaked in deionized water for 2 hours, to remove the surface
absorbed PS molecules and loosely attached oligomers of 4VP. Samples
were dried with nitrogen gas and soaked in aqueous solution of sodium
hypochlorite with the concentration of 1000 ppm
various treatment
durations FTIR spectra and dynamic contact angle measurements were
taken before and after treating with chlorine solutions.
for
Bacterial Adhesion Tests: V cyclitrophicus was used as the model
microorganism. Bacteria cells from freezer stocks were inoculated
and grown overnight in artificial seawater at 30 C to an optical
density (ODy,) of 1 while agitated on a shaker (150 rpm). Cells were
suspended in fresh artificial seawater and incubated at 37 C on a
shaker (180 rpm) until the optical density reached 0.2. The bacterial
solution was then injected into the microfluidic channels at a constant
flow rate of 2 pf min , which corresponds to an average flow velocity
of 560 pm s , During the combination treatment, chlorine was directly
added to the vessel containing the media with bacteria to a final
concentration of 5 ppm. Note that in this case the images (Figure 3)
acquired at a certain time (t hours) captured bacteria that have been
exposed to chlorine for i h.
Verlag GrmbH & Co KGaA. Weinheim
137
iCVD
in a custom-built vacuum reactor (Sharon Vacuum), as previously
described.,' "I All the chemicals were used as purchased without further
purification. Siticon (Si) wafers (Wafer World, test grade) were coated
with P4VP or the copolymer of 4VP and DVB without pre-treatment
Prior to deposition, commercial RO membranes (Koch Membrane
System. TFC-HR) were cleaned with filtered nitrogen, and then treated
with oxygen plasma for 1 minute and then placed in the reactor
wileyonlinelibrary.com
7
ADViNE
AMR ALS
www.advmat.de
www.MaterialsViews.com
Supporting Information
Supporting Information is available from the Wiley Online Library or
from the author
Acknowledgements
The authors thank the support from DOE Office of ARPA.E under
award AR0000294 and the King Fahd University of Petroleum and
Minerals in Dhahran. Saudi Arabia, for funding the research reported
in this paper through the Center for Clean Water and Clean Energy at
MIT and KFUPM (to K.K.G.) as well as support through NSF grants
OCE-6917641 -CAREER and CBET-6923975, and NIH grant 6917755
(to R.S.) Dr. Hongchul Jang has been partially supported by a Samsung
Fellowship. We thank Jonathan Shu from the Cornell Center for Materials
Research (CCMR) for help with XPS measurements
Received: August 31. 2013
Revised: October 16, 2013
Published online:
[1] WHO Book 10 facts about water scarcity. .http://www.who.int/featuresjfactfiles/water/en/ -,
[2] M. Elimelech, W A. Phillip, Science 2011, 333. 712 717
[3] S. Jiang, Z Cao, Adv. Mater. 2010, 22. 920-932.
Zhang, Z. Cao, T. Bai. L. Carr. [ R Ella-Menye. C Irvin.
|4] L
B D. Ratner, S. Jiang, Not Biotechincl. 2013. 37, 553-556.
S.
Chen,
J. Zheng, L. Li, S. Jiang, J. Am. Chem. Soc. 2005. 727,
[5]
14473-14478.
[6] H Kitano, T. Mort. Y. Takeuchi. S Tada, M Gemmeilde.
Y. Yokoyama. M. Tanaka, Macromol Biosci 2005 5. 314 321
[7] R Quintana, M Gosa. D. Ja czewski, E Kutnyanszky. G. J. Vancso,
Longmiuir 2013
[8]
[9]
[10]
[11]
[12]
[13]
8
J. S. Louie, I. Pinnau, M. Reinhard, J Membr Sc 2011, 367.
249-255.
R. Yang, I Xu, G Ozaydin-Ince, S. Y. Wong, K K. Gleason, Chem.
Mater. 2011, 23, 1263-1272
R Yang, K K. Gleason. Langmuir 2012, 28, 12266 12274.
G. Ozaydin-Ince, A Matin, Z. Khan. S. M. J Zaidi. K K Gleason,
Thin Solid Films 2013, 539, 181 187.
H B Park, B. D. Freeman, Z.-B Zhang, M. Sankir. J E McGrath,
Angew. Chem Int Ed 2008, 47. 6019-6024.
C D Petruczok, R Yang, K K Gleason Macromolecles 2013, 46.
1832-1840.
wileyonlinelibrary.com
[141 M E Alf, A. Asatekin, M C Barr, S H Baxamusa. H Chelawat,
G Ozaydin-ince, C. D Petruczok, R. Sreenivasan, W. E. Tenhaeff,
N. j. Trujillo. S. Vaddiraju, J Xu, K K. Gleason, Adv. Mater. 2010. 22,
1993-2027.
[15] D. Cohen-Tanugi, J. C. Grossman, Nano Lett. 2012, 12, 3602-3608.
[16] M. Majumder, N. Chopra, R. Andrews. B. J.Hinds, Nature 2005, 438, 44.
[17] J K. Holt. H. G. Park, Y Wang, M. Stadermann, A. B. Artyukhin,
C. P. Grigoropoulos, A. Noy, 0. Bakajin, Science 2006. 372.
1034-1037.
[18] F Natalio, R. Andre, A. F. Hartog, B. Stoll, K P. Jochum, R. Wever,
W Tremel, Nat Nanotechnol. 2012, 7, 530-535
[19] D Rana. T. Matsuura, Chem. Rev. 2010, 770, 2448-2471
[20] G Beamson. D Briggs, in High Resolution XPS ofOrganic Polymers
the Scienta ESCA300 Data Base, John Wiley & Sons, New York 1992.
[21] C G Spanos, J P S Badyal, A J.Goodwin, P. J. Merlin, Polymer
2005, 46, 8908-8912.
[22] A. S. Goldmann, A. Walther, L. Nebhani, R. [oso. D Ernst, K. Loos.
C. Barnei-Kowollik. L Barner, A H. E. Muller. Macrorrolecules 2009.
42. 3707-3714.
]23] L. Wang, Z. Wang, X. Zhang, ). Shen, L Chi. H. Fuchs. Macrornol.
Rapid Commun. 1997. 78 509 514.
[24] D Lin-Vien. N. Colthup, B W Fateley, G J Grasselli, in
The Handbook of Infrared and Raman Characteristic Frequencies
ofOrganic Molecules. Academic Press, San Diego 1991
[25] R. 1. Good,J. Adhes. Sci. Technol. 1992, 6, 1269-1302.
[26 C J. Weinman, N. Gunari, S Krishnan, R Dong, M. Y Pai,
K. E Sohn, G C Walker, E J.Kramer, D. A Fischer, C K Ober, Soft
Matter 2010, 6. 3237 3243
[27] 5. H Baxamusa. K K Gleason. Ad. Funct. Mater 2009, 79,
3489-3496.
[28] A. Asatekin, A Mennitti, S Kang, M Elimelech, E. Morgenroth,
A. M. Mayes.J Membr. Sci. 2006, 285. 81-89
[29] R. Stocker, J R Seymour, A Samadani, D. E. Hunt, M F Polz,
Proc Nat/. Acad. Sci USA 2008. 705, 4209 4214
[30] EPA Drinking Water Contaminants <http-//water epa.gov/drink/
contaminants/index.cfm>
[31] Z. Zhang, I C Lee, L Lin, V Olivas, V Au, T LaFramboise,
M. Abdel-Rahman, X. Wang, A D. Levine, J K. Rho,
Y. J.Choi, C..M. Choi, S.-W. Kim, S. J. lang, Y. S Park. W. S. Kim,
D H Lee, J -S Lee, V. A Miller, M. Arcila, M. Ladanyi, P Moonsamy,
C. Sawyers T [. Boggon. P. C Ma. C Costa. M. Taron, R Rosell.
B. Halmos, T G Bivona. Nat. Genet 2012. 44, 852 860.
]32] C T. Keith, A A. Borisy, B. R Stockwell. Not Rev. Drug Discov 2005,
4, 71-78
[33] M. Gupta. K K Gleason. Then Solid Films 2006. 515, 15 79-1584
[34] N J. Trujillo, S H Baxamusa, K K Gleason. Chrm Mater 2009, 27.
742-750.
2013 WILEY.VCH Vdig rmiDH & Co
KC~A
Wemheii
1i
138
Ad, Mor 2013.
tO 1002jdmad 20130486
APPENDIX B
Spatial Control of Cell Attachment by Mucin Coatings
The heterogeneously coated surfaces described in this paper, which resulted from a
collaboration with Prof. Katharina Ribbeck's group in Biological Engineering at MIT and was
published in Biomicromolecules in 2013, were prepared by me, taking advantage of the
approaches and expertise developed in this thesis. Specifically, I fabricated microchannels with
an arrangement of spatially localized hydrophobic surfaces, which formed the basis for the
experiments on confined attachment presented in this paper. This publication is included here
with permission of all the authors.
139
pubs
acs orq/Biornac
Cell Patterning with Mucin Biopolymers
T. Crouzier, H. Jang,t J. Ahn,t R. Stocker, and K. Ribbeck*
Massachusetts Institute of Technology, 77 Massachusetts Avenue, Cambridge, Massachusetts 02139, United States
0
Supporting Information
Flow of mucin solution
ABSTRACT: The precise spatial control of cell adhesion to
surfaces is an endeavor that has enabled discoveries in cell
biology and new possibilities in tissue engineering. The
generation of cell-repellent surfaces currently requires
advanced chemistry techniques and could be simplified. Here
we show that mucins, glycoproteins of high structural and
chemical complexity, spontaneously adsorb on hydrophobic
substrates to form coatings that prevent the surface adhesion
of mammalian epithelial cells, fibroblasts, and myoblasts. These
mucin coatings can be patterned with micrometer precision using a microfluidic device, and are stable enough to support
myoblast differentiation over seven days. Moreover, our data indicate that the cell-repellent effect is dependent on mucinassociated glycans because their removal results in a loss of effective cell-repulsion. Last, we show that a critical surface density of
mucins, which is required to achieve cell-repulsion, is efficiently obtained on hydrophobic surfaces, but not on hydrophilic glass
surfaces. However, this limitation can be overcome by coating glass with hydrophobic tluorosilane. We conclude that mucin
biopolymers are attractive candidates to control cell adhesion on surfaces.
N
INTRODUCTION
such as gold or silicon and often requires the use of organic
3
solvents, acids, and other hazardous chemicals.' Hence, it is
desirable to develop more efficient and biocompatible strategies
to control the adhesion of mammalian cells to surfaces.
Mucin biopolymers constitute the main gel-forming
4
components of the mucus barrier' and are interesting
candidates for regulating mammalian cell adhesion. Secreted
mucins are densely glycosylated, thread-like molecules with
random coil confgurations,1
which can reach molecular
weights of up to tens of MDa. The diversity of chemical groups
provided by the protein backbone and glycan side chains enable
mucins to adsorb to a number of different substrates, including
polystyrene, gold, and silica. On polystyrene, which is used for
flasks, multiwell plates, and tubes, mucins adsorb through their
exposed hydrophobic protein core portions while protruding
their hydrophilic and glycosylated domains into the aqueous
5
solvent.' This configuration results in highly hydrophidic' 8'2
and hydrated monolayers, with around 95% of the adsorbed
mass being water.' Of note is that mucin coatings are able to
20
decrease the adhesion of certain bacteria,"'
neutrophils, and
222
fibroblasts, - but their use in mammalian cells patterning has
not been explored. Here, we investigate the potential of mucinbased coatings to pattern the attachment of mammalian cells.
Adhesion of mammalian cells to synthetic materials is the first,
necessary step in many cellular processes including migration,
differentiation, and activation of the foreign body reaction, a
process that leads to the rejection of implanted materials.[
Controlling the adhesion of mammalian cells to solid surfaces is
thus of relevance for biomaterials science, as it can improve
overall biocompatibility of implants and facilitate tissue
formation in tissue engineering applications. For example,
spatial control of premuscle myoblast adhesion can accelerate
the formation of contractile muscle tissue and facilitate the
2
study of cell alignment and differentiation." Tuning cell
adhesion to surfaces is also becoming increasingly important for
cell biologists. Controlling cell shape and positioning on
substrates has facilitated the study of tissue formation and
organization, cell-cell interactions" and has enabled wellcontrolled coculture experiments.s
One important aspect of cell patterning is the coating of
substrates with cell-repellent molecules in spatially confined
ways. Effective cell repulsion can be achieved with a number of
different synthetic molecules including polyethylene glycol
(PEG)
and zwitterionic molecules."
Additionally,
surface-
immobilized biopolymers such as hyaluronic acid, alginic
acids or carboxymethyl-dextran, can also be repulsive toward
cells. However, these molecules typically do not spontaneously
UV- and photolithogaphy-based patterning techniques such
as micro stamping-- and local etching2'
or activation9 of
cell-repellent polymer coatings are routinely used by cell
biologist and tissue engineers. An even more accessible and low
cost alternative to UV and photolithography is Micro Molding
adsorb on surfaces with sufficient density to exert effective cell-
repulsion. Common strategies to overcome this limitation are
the addition of surface-anchoring groups to the cytophobic
2
molecules" '
or the chemical activation of the surface to
retain the cytophobic molecules to the surface for extended
periods of time. These efforts can be time-consuming and
costly. Surface grafting in particular is limited to few substrates
ACS Publications
C2013 Amercan Ochemial Socety
Received: March 29, 2013
Revised:
June 6, 2013
l'ublished: July 9, 2013
3010
140
Io
: ,,q/IO1121 /Ln'
ooW4
sIBionrmx5,sk 2013. 1. 3010 3016
Biomacromolecules
in Capillaries,3which is based on placing a mold (often
polydimethylsiloxane (PDMS)) with geometrically defined
open channels on a surface. Cell-repellent molecules are
distributed through the channels, adsorb to the surface, and
leave behind patterned cell-repellent coating once the mold is
removed. Here we seek to combine this relatively simple
patterning technique with the use of mucin biopolymers to
explore novel protocols for cell patterning.
Our work shows that mucins can readily adsorb to
hydrophobic substrates and prevent cells from adhering to,
and spreading, on these substrates. Using a simple microfluidic
device to pattern mucin coatings, precise control of cell
mucin was flushed from the device with washing buffer (20 mM
HEPES, pH 7.4) before the microfluidic device was peeled off the
substrate and the coating immersed in washing buffer until used. For
the quantification of protein adsorption and cell adhesion, the muc
coatings were generated in untreated polystyrene 96 well plates
(Becton Dickinson, 351172). Prior to seeding the cells on the coatings,
the surfaces were treated with 10pg/mL fibronectin in PBS for 30
minutes to promote cell adhesion and differentiation.
Call Culture. HeLa cells, the C2C12 myoblasts and the NIH-3T3
fibroblasts (ATTC CCL92) were maintained at subconfluency in T25
flasks with Dulbecco's modified Eagle medium (DMEM) media
supplemented with 10% fetal bovine serum (Invitrogen) and 1%
antibiotics (25 U/mL penicIdin, 25 pg/mL streptomycin (Invitrogen).
The cells were detached using trypsin-EDTA (Invitro
The
').
detached cells were plated at a density of 45000 cells/cm on mudn
coatings or plastic controls. After 24 h, the cells were washed twice
with PBS before being observed. Viability of the cells was assessed by
Live/Dead (Invitrogen) assay by exposing cells to 2 pM calcein AM
and 4 pM ethidium homodimer-l in serum free DMEM for 45 min,
before observation under the microscope.
To monitor C2C12 differentiation, the cells were grown for 2 days
in growth medium (DMEM + 10% FBS), followed by 5 days in
differentiation medium (DMEM + 1 % Hore Serum), which was
exchanged every two days. Thereafter, the cells were washed with PBS
and fixed for 10 minutes in 3% formaldehyde (Sigma-Aldrich). First,
actin was labeled with phaloidin-rhodamie (Invitrogen). Then
troponin T was labelled by immune flnorescence using a primary antiTroponin T antibody (Goat Anti Mouse antibody, Santa-Cruz sc365575) and a fluorescein-labeled secondary antibody (Donkey anti
positioning and attachment can be achieved. Moreover, the
mucin pattern can be maintained over 14 days in serum-
containing media and is stable enough to spatially confine
myoblasts during their differentiation into myoblast over a 7day culture. Last, the repulsive effect toward cells is likely
mediated by mucin-associated glycans because their removal
eliminates the cell-repellent effect. Although cell-repellent
mucin coatings could not be generated on hydrophilic glass
surfaces, modifing the glass surface with a hydrophobic
fluorosilane layer before generating the mucin coating
recovered the effect. This makes the technology interesting
for a wide variety of materials and applications.
* MATERIALS AND METHODS
Materials and Reagents. Commerdal bovine submaxillary mucin
(BSM Sigma-Aldrids; lot 039K7003) and pig gastric mudn (PGM,
Sigma-Aldrich, lot 116K7040) were dialyzed against water (Spectra/
Por dialysis membrane; 100 3kDa molecular weight cutoff, Spectrum
Labs) to remove impurities, 2 before lyopliffetion for storage. Skim
milk (Becton Dickinson), fibronectin (Sigma-Aldrich), bovine serim
albumin (BSA, Sigma-Aldrich) and lysozyme (Sigma-Aldrich) were
33
used a received. Pig gastric mucins were purified in lab as reported,
omitting the cesium chloride density gradient centrifugation.
BSM fibronectin, BSA, and lysozyme were labeled with Aleza488 or
Rhodamine Red Ibuorophore using the carboxylic add succinimidyl
ester amine-reactive derivatives (Invitrogen). Briefly, the proteins were
dissolved in carbonate-bicarbonate buffer (0.2 M pH 8.5) and
incubated with the dye at a protein-to-dye mass ratio of 1:10 for 1
hour at room temperature. The excess dye was removed either by
centrifuge filtration (Pall, likDa MWCO) for fibronectin and BSM or
in desalting columns (PD-10, GE Healthcare Lifesciences) for BSA
and lysozyme.
PDMS Mold Fabrication for Surface Patterning. PDMS
Mouse, Santa-Crua, sc-2099). The nudei were stained with DAPI
(Santa-Cruz), which was present in the mounting solution.
Cell adhesion was measured using the CyQpant DNA quantification
kit (Invitrogen). In brief, cells were seeded in 96 well plates. After 24
hours, non-adherent cells were removed by washing the wells twice
with 200 pL of PBS. The wells were air dried and frozen at -80 *C
overnight Then the plates were thawed, and 100 pL of Cy~pant
reagent was added to the wells before the fluorescence intensity
(Exc480/Em520 nm) was measured using a fluorescence plate reader
(Spectramax M3, Molecular Device).
Deglycosylation of Mucins. BSM was deglycosylated by
treatment with trifluoromethanesulfonic acd (TFMS), followed by
oxidation and beta-elimination of the residual sugars as previously
described.' Briefly, S mg of BSM was mixed with 375 pL of an icecold solution of TFMS (Acros Organics) containing 10% (v/v) anisole
(Sigma-Aldrich) and stirred on ice. After 2 h the solution was
neutralized by the addition of a solution containing 3 parts pyridine
(Alfa Aesar), 1 part methanol, and 1 part water. The solution was
dialyzed against water for 2 days using a 20 kDa MWCO membrane
(Thermo Scientific). NaC and acetic aid were added to the solution
to a final concentration of 0.33 M and 0.1M, respectively, and the pH
was adjusted to 4.5 with NaOH. The oxidation step was initiated by
adding NaIO 4 to a final concentration of 100 mM, and incubation at 4
*C for 5 h in the dark. The unreacted periodate was neutralized by
adding a 1/2 volume of "neutralizing solution" (400 mM Na 2S2O3 , 100
mM Nal, 100 mM NaHCO 3 ). Beta elimination of the remaining
glycans was started by adjusting the pH to 10.5 using 1 M NaOH.
After a 1-h incubation at 4 *C, the solution was first dialyzed overnight
at 4 *C against a 5 mM NaHCO,3 buffer (pH 10.5) then against
ultrapure water for 2 days. The resulting apo-mucins were
concentrated and dissolved in the appropriate buffer. The
deglycosylation was verified with a Periodate-Schiff say.
Characterization of Apo-Mucin Coatings by QCM-D. Qpartz
crystal microbalance with dissipation monitoring was used (QCM-D,
E4 system, QSense, Sweden) to characterize the apo-mucin coatings.
The crystals used were purchased coated with polystyrene (QSX30S,
Qsense, Sweden). All QCM-D experiments were carried out in noflow conditions. The crystal vibration was followed at its fundamental
frequency (about S MHz) and the six overtones (15, 25, 35, 45,5 and
65 MHz). Changes in the resonance frequencies, and in dissipation of
the vibration once the excitation was stopped, were followed at the
(Sylgard 184, Dow coming, USA) molds were fabricated by curing;
the prepolymer on silicon masters patterned with SU-8 photoresist
(SU-8 2050, MicroChem, MA, USA) using conventional photolithography. The masters used for patterning had recessing features,
which resulted in PDMS replica with the opposite sense. To assist in
removal of cured PDMS from the masters, the SU-8 masters are
silanized by exposure to the vapor of 1,1,2,2,-tetrahydrooctyl-1trichlorosilane, CF3(CF2) 6(CH) 2SiCl, overnight To cure the PDMS
prepolymer, a mixture of 10:1 silicon elastomer and the curing agent
was poured on the master and placed at 65 'C for 2 I. The PDMS
replica was then peeled from the silicon masters.
Generating Micropatterned Mucin Coatings. The PDMS
molds used for patterning were composed of a single miaochannel
(4 mm wide, 6 an long, 50Mm high), posts of selected design and size,
an outlet, and an inlet The device was placed on an untreated
polystyrene surface (Thermo Scientific, 267061), and gentle pressure
was applied to form a reversible seal to the surface. The channel was
filled with a 20 mM HEPES buffer solution (pH 7.4) through suction
at the outlet, filling in the solution from a reservoir placed at the inlet.
Once the channel was filled, the buffer was exchanged for a lmg/mL
solution of mucin (BSM in 20 mM HEPES; pH 7.4), which was left in
the channel for 30 minutes to enable mucin adsorption. Unbound
3011
141
dxhAg/O.021/bma4s447gJomfasmohdads
2013,14 tO-
OS
Biomacromolecules
seven frequencies. Shifts in frequency are related to changes in
adsorbed mass (due to adsorbed polymer, water and ions), whereas
changes in dissipation are influenced by the mechanical properties of
the adsorbed layer. The mucin coatings are highly hydrated, hence, the
dissipation values are significant, and the Sauerbrey relation, which
relates frequency to the adsorbed mass through a linear relationship
does not apply. The Voigt based model was used to accurately
estimate the hydrated thickness. The density of the adsorbed layer
was fixed at 1050 kg m ', which lies between that of pure water (1000
kg in ') and pure protein (1350 kg i '). This approximation was
validated for a related system' and was reported to have minimal
3
impact on the final modeled thickness.'N' "
Fluorosilane Coatinp. Glass microscope slides were coated with
fluorosilane as reported) In brief, microscope slides (VWR Scientific)
were sonicated in a 4 vol % Micro-90 cleaning solutions (International
Products) for 15 min, which was followed by two subsequent steps of
sonication in Mili-Q water. After being dried with compressed air,
glass slides were placed and sealed in a Teflon canister with one drop
of I H,I H,2H,2H-perfluoro-decyl-trichloro-silane (Sigma-Aldrich).
The Tellon canisters were in a sealed glovebox under dry nitrogen
atmosphere during the fluoroalkyl silane treatment, and then placed in
a heater overnight at I10 C.
Microscopy. For imaging, the coatings were submerged in 1B1S
(pH 7.4), using an Observer Z. inverted fluorescent microscope
(Zeiss, Oberkochen, Germany) and lOX 0.3 NA objective (Zeiss,
Germany). Image analysis was performed with built-in plugins of the
inageJ software.
Quantification of Mucin Adsorption. To measure the quantity
of adsorbed BSM on polystyrene, glass and fluorosilane coated glass
surfaces, a 50 ol. drop of fluorescently labeled muchis (Img/mL in
PBS) was placed onto each surface and the inucin and was allowed to
adsorb for 30 minutes. The slides were rinsed and kept in NIS for
observation. Each surface was prepared in duplicate, and 10 images
were taken per condition. The fluorescence intensity was averaged
over all images taken per surface.
0
-
160
140
1 2
0
=
BSA
Skim milk
PGM
S Sigma
-
B
LabPGM
Sigma BSM
c 100
80
5
_
i
n
60
40
20
ea
S0
HeLa
C2C12
3T3
Figure 1. Mucin coatings are cytophobic. (A) HeLa epithelial cells,
3T3 fibroblast cells and C2C12 myoblasts were seeded on a
polystyrene surface, or on coatings generated from IISM (Signia) or
PGM (in-lab punfied) and labeled with a live (green)/dead (red)
stain. Scale bars: 200 pom. (l) Quantification of cell adhesion on mucin
coatings generated with commercial BSM and IPGM (Sigma ISM/
IGM), in lab purified PGM (Lab PGM), Skim milk, and [ISA. The
highlighted region between 94 1201t is the standard deviation of tine
reference adhesion to polystyrene.
RESULTS AND DISCUSSION
Mucin Coatings Prevent Adhesion of Several Cell
procedure may result in an altered biochemistry of the PIGM,
Types. To test whether mucins are suitable substrates to
regulate cell adhesion in vitro, we generated mucin coatings by
placing a 0.1% mucin solution in contact with a polystyrene
surface. We used commercial bovine submaxillary mucin
(BSM) and pig gastric mucin (PGM), as well as in-lab
extracted I1GM, which form thick coatings on polystyrene when
hydrated (see Supporting Information 1).
We analyzed the effects of these mucin coatings toward three
cell lines: HeLa epithelial cells, 311 fibroblasts, and C2C12
myoblasts. After 24 h, all three cell types only sparsely
colonized the mucin coatings compared to the uncoated
polystyrene surface. In addition, those cells that attached to the
mucin coatings were typically rounder in shape than the cells
on the uncoated surface. A simultaneous calcein and EthD)- I
staining suggests that the cells on the mucin coatings were still
alive (Figure IA), and that the mucin coatings were not toxic
toward the cells.
A quantification of adherent cells showed that both BSM and
native in-lab purified PGM coatings reduced adhesion of all
three cell types by nearly 90%, compared to polystyrene
(Figure 113). In contrast, the common blocking agents bovine
serum albumin (lISA) and skim milk (Figure Il) did not
significantly reduce cell adhesion. These data show that BSM
1
and in-lab purified P GM biopolymers are both excellent
candidate molecules to create cell-repelling coatings. Of note
is that commercial PGM (from Sigma-Aldrich) did not exert
the cell-repulsive effect that was observed with the two other
mucin types. We speculate that the industrial purification
which compromises their function. Indeed, as opposed to in-lab
purified PIGM, Si ma PGM have lost their ability to efficiently
gelate at low pH.
Spatially Defined Micropatterned Mucin Coatings
with Durable Cell-Repulsion. Based on these robust
antiadhesive properties, we probed the potential of mucin
coatings for cell patterning applications. For this purpose, we
adapted micromolding in capillaries . technology to generate
BSM coatings. In brief, we used a microfluidic channel that
contained vertical removable posts, which blocked mucin
adsorption to the polystyrene surface at designated regions
(Figure 2A). The mucins were flown into the channel where
they adsorbed to the unmasked regions of the polystyrene
bottom. Then the posts were removed to uncover the uncoated
surface. With this technique, patterns of various shapes and
dimensions (between tens and hundreds of microneters) could
be generated in less than 30 min. Figure 213 shows different
geometries of mucin coatings in which fluorescently labeled
mucins were used for visualization by epitluorescent microscopy (Figure 2B). When cultured with cells, the mucin-coated
areas remained distinctly cell-free, while the mucin-free areas
were densely populated by all three cell types tested (Figure
2C,D,E). The sharp transition between cell-covered and
uncovered zones suggests that tine cell-repulsive effect of the
patterned mucin coatings has high spatial accuracy. From a
practical perspective it is interesting to note that the coatings
can be dried and rehydrated without losing their cytophobic
effect (see Supporting Information 2). This is particularly
3012
142
& J, -y10
.
102iiVOi
or 'OCw,
oi1
2a01.
1'. ists 3016
Biomacromolecules
A
Figure 2. Mucin coatings for cell patterning. (A) Mucin coatings were patterned using a microfluidic device with posts masking defined areas of the
surface. (B) Patterns of mucin-free areas (in black) can be generated in different sizes and shapes. When seeded on the patterned surfaces, the
epithelial cells (C), libroblasts (D) and myoblasts () accumulated in the uncoated regions, avoiding the mucn coatings. Scale bars: 250 pm.
Figure A. Mucin coatings and their cytophobic effect are robust (A) Mucin patterns immersed in cell culture media for 14 days showed no alteration
in their shape or fluorescence intensity. (72C12 myoblasts that were cultured on the coatings for 7 days expressed troponin T, which is a marker for
early niyogenic differentiation. (Wt) Actin staining reveals that cells are confined within the pattern. A Troponin-T/Actin overlay is shown in (1"),
Scale bars: 210 pim.
important if these surfaces are to be sterilized and stored before
early marker for myotube differentiation I (Figure 3B), and
actin (Figure 311') showed that troponin T-positive myotubes
emerged after 7 days. These data suggest that in standard cell
culture conditions, mucin coatings preserve their cytophobic
effect toward C2C12 cells, while allowing for basic cellular
differentiation.
use,
To test tile stability and functionality of mucin coatings over
time, we monitored patterned fluorescent mucin coatings over
14 days in cell-free culture media. The patterns showed no
obvious signs of deterioration: both the fluorescence intensity
Mucin-Linked Glycans Contribute to Cell-Repulsion.
and the sharp edges of the patterns remained stable (Figure 3A,
A' and A"). To test for loss of cytophobicity over time, C2C12
myoblast cells were seeded otn the patterned coatings and
monitored over 7 days. Over this period, the cells remained
confined within the mucin-free regions (Figure 31'), which
To better understand the molecular origin of mucin-mediated
cell-repulsion, we focused our attention oti the features that
distinguish mucins from other proteins such as albumin that do
not generate cell-repellent coatings. Mucins are largely linear
molecules that are densely coated with O-linked glycans. These
glycans contribute to mucin's extended structure and ability to
t
form hydrogels. We hypothesized that mucin-associated
suggests that the cell-repulsion effect was maintained despite
the tendency of C2C12 cells to migrate and secrete extracellular
matrix and proteases. Immunofluorescence for Troponin T, an
3013
143
d. d,
eal
w-qi
-2,
1
o
re-a
a03,
i31
1016
Biomacromolecules
polystyrene (Figure SA). Reduced mucin adsorption on glass
was correlated with a lack of cell repulsion compared to the
coatings on polystyrene (Figure 5BB'). This finding is in
accordance with previous studies, which show that mucin
adsorbs in thick and strongly bound layers on hydrophobic
surfaces, while layers can be sparser on hydrophilic
7
substrates." " Since glass is more hydrophilic than polystyrene,
surface chemistry could be the main reason for the inefficient
mucin adsorption. If this is the case, it should then be possible
to restore efficient mucin adsorption by rendering the glass
surface hydrophobic. We coated the glass with fluoroalkyl silane
by vapor deposition and found that mucin adsorption increased
(Figure 5A). Moreover, the presence of the fluorosilane coating
restored the cytophobic effect, with cells excluded from mucin
coated areas (Figure SB").
ACKNOWLEDGMENTS
Foundation under award number DMR-819762. This work
was made possible by a Marie Curie International Outgoing
Fellowship 'BIOMUC" (to T.C.), NIH Grant 1R01GM100473
(to R.S.), NSF Grant OCE-0744641-CAREER (to R.S.), NSF
Grant
IOS-1120200
of the CMSE MRSEC facilities (to K..,
N REFERENCES
J. M.;
2008, 20, 86-100.
(2) Wan, L. 0
We have shown that mucins, the basic constituents of the
mucus barrier, are excellent biological substrates for the
generation of spatially precise and patterned coatings to
control cell adhesion and differentiation. Expanding on earlier
reports on mucins' ability to repel mammalian cells, we have
presented an analysis of the cytophobic effect of mucins against
a range of different cell lines, their use in microfluidic-based cell
patterning, and the molecular mechanism responsible for the
cytophobic effect. Mucin coatings present a fast, accessible,
cheap, and effective way of controlling cell adhesion, and
provides great flexibility when coupled to patterning
techniques. From cell biology to biomaterial and tissue
engineering studies, these coatings could be used to control
cell adhesion with high spatial resolution. Importantly,
robustness against drying will allow these coatings to be stored
and sterilized. Beyond these technological applications, these
results are intriguing from a basic science perspective as they
lead to interesting questions regarding cell-mucus interaction in
vivo. For example, our results suggest that interaction between
mammalian cells and the natural mucus matrix could be
independent of adhesion, and thus very different from
connective tissue extracellular matrices.
(to RS.), the MRSEC program of the
National Science Foundation under Award DMR-0819762.HL.
(to J.A), the Samsung Scholarship program (to JA), the MIT
startup funds (to K R.), and a Junior Faculty award for the use
(1) Anderson,
U CONCLUSIONS
*
N
We thank Michael Rubner and Robert Cohen for their valuable
comments. This work made use of the MRSEC Shared
Experiment Facilities supported by the National Science
Rodriguez, A; Chang, D. T. Semin. Immunol
Ronaldson, K.; Park, M.; Taylor, G.; Zhang, Y.;
Gimble, J. M.; Vunjak-Novakovic, G. Proc. NatL Acad. Sci. 2011, 108,
12295-12300.
(3) Popescu, D. C.; Lerns, R.; Rossi,N. A A; Yeoh, C.-T.; Loos, J.;
Holder, S.J.; Bouten, C. V. C.; Sommerdijk, N. A J. M. Adv. Marr.
2005, 17, 2324-2329.
(4) Tseng, Q; Duhemin-Pelletier, E.; Deshiere, A; Balland, M.;
Guillou, H.; Filhol, 0.; Thiry, M. Proc. NatL Acad. Sci 2012, 109,
1506-1511.
(5) Fukuda, J.; Khademlhosseini, A; Yeh, J.; Eng, G.; Cheng, J.;
Farokhzad, 0. C.; Langer, R.Biomaterials 2006, 27, 1479-1486.
(6) Krishrian, S.; Weinnian, C. J.; Ober, C. K J.Maater. Chem. 2008,
18, 3405.
(7) Morra, M.; Cassineli, C. J. Biomater. Sci Polym. Ed. 1999, 10,
1107-1124.
(8) Cao, X4 Pettit, M. F.; Conlan, S. L; Wagner, W.; Ho, A D.;
Clare, A S.; Callow, J. A; Callow, N. E.; Grnze, M.; Rosenhahn, A.
Colloids Surf, B 2009, 10, 907-91S.
(9) McLean, K. M.; Johnson, G.; Chatelier, R. C.; Beumer, G. J.;
Steele, J. G.; Griesser, H. J.Colloids Sf., B 2000, 18, 221-234.
(10) Kenausis, G. L; Vdrbs, J.; Elbert, D. L.; Huang, N.; Hofer, R;
Ruiz-Taylor, L; Textor, M.; Hubbell, J. A; Spencer, N. D. J. Phys
Chan. B 2000, 104, 3298-3309.
(11) Kenan, D. J.; Walsh, E. B.; Meyers, S. R.; O'Toole, G. A;
Carnsthers, E. G.; Lee, W. K; Zauscher, S.; Prata, C. A H.; Grinstaff,
N. W. Chem Biol 2006, 13, 695-700.
(12) Le, H.; Dellatore, S. M.; MIller, W. M.; Messersmith, P. B.
Science 2007, 318, 426-430.
(13) Brown, A A; Khan, N. S.; Steinbock, L.; Huck, W. T. S. Eur.
ASSOCIATED CONTENT
* Supporting Information
Supporting Information for this manuscript includes muon
(both PGM and BSM) adsorption kinetics on polystyrene
surface as measured by QCM-D (SI 1), measurement of cell
repulsion effect of the BSM coating after a dehydration/
rehydration cycle (S 2), and a table containing the modeled
viscoelastic parameters of the QCM-D data fitting (S] 3). This
material is available free of charge via the Internet at http://
pubs.acs.org.
Polym. J. 2005, 41, 1757-1765.
(14) Band, R.; Turner, B. S. Curr. Opin. Colloid Interface Sci. 2006,
11, 164-170.
(1S) Shogren, R; Gerken, T. A; Jentoft, N. Biochemistry (Mosc.)
1989, 28, 5525-5536.
(16) Gerken, T. A; Butenho4 K J.; Shogren, R. Biochenistry (Mosc)
1989, 28, 5536-5543.
(17) Kesimer, M.; Sheehan, J. K Glycobiology 2008, 18, 463.
(18) Shi, L; Caldwell, K D. MRS Proc. 2011, 599.
(19) Bushnak, L; Labeed, F.; Sear, R.; Keddie, J. Biofouling 2010,26,
387-397.
* AUTHOR INFORMATION
Author Contributions
(20) Shi, L; Ardehali, R.; Caldwell, K D.; Valint, P. Coloids Surf B
Bionterfaces 2000, 17, 229-239.
(21) Halthur, T. J.; Arnebrant, T.; Macakova, L; Feler, A Langmuir
2010, 26, 4901-490.
*These authors contributed equally to this work. The
manuscript was written through contributions of all authors.
All authors have given approval to the final version of the
manuscript.
(22) Bavington, C. D.; Lever, R; Mulloy, B.; Grundy, M. M.; Page, C.
P.; Richardson, N. V.; McKenzie, J. D. Cornp. Biochem. PhysioL, PartB:
Biochem. MoL Biol 2004, 139, 607-617.
(23) Sandberg, T.; Carlsson, J.; Ott, M.
70, 864-868.
Notes
The authors declare no competing financial interest.
K Microsc. Res. Tech. 2007,
(24) ShI, L Trends Glycosci. Glycotechnol 2000, 12, 229-239.
3015
145
dsdolo
R1.
5
.1021/m400m4
; Slonmokef
2013, 1,
3010-3016
Biomacromolecules
(25) Chen, C. S.; Mrksich, M.; Huang, S.; Whitesides, G. M; Ingber,
D. E. Biotechnol. Prog. 1998, 14, 356-363.
(26) Fink, J.; Thiry, M.; Azioune, A.; Dupont, R.; Chatelain, F.;
Bornens, M.; Piel, M. Lab Chip 2007, 7, 672.
(27) Dillmore, W. S.; Yousaf, M. N.; Mrksch, M. Langmuir 2004,20,
7223-7231.
(28) Kikuchi, Y.; Nakanishi, J.; Nakayana, H.; Shimizu, T.; Yoshino,
Y.; Yamaguch, K.; Yoshida, Y.; Horeike, Y. Chan. Lett 2008, 37,
1062-1063.
(29) Bilisle, J. M.; Kunik, D.; Costantino, S. Lab Chip 2009, 9,3580.
(30) Kim, E.; Xia, Y.; Whitesides, G. M. J. Am. Chem Soc. 1996,118,
5722-5731.
(31) Sandberg, T.; Blom, H.; Caldwell, K. D. j Biomed. Mater. Res. A
2009, 91A, 762-772.
(32) Lundisn, M; Sandberg, T.; Caldwell, K. D.; Blomberg, E. J.
Colloid Interface Sd 2009, 336, 30-39.
(33) Celii, J.; Gregor, B.; Turner, B.; Afdhal, N. H.; Bansil, R.;
Erramflli, S. Biomacromolecules 2005, 6, 1329-1333.
(34) Gerken, T. A; Gupta, R.; Jentoft, N. Biochemistry (Mosc) 1992,
31, 639-648.
(35) Kilcoyne, M.; Gerlach, J. Q., Farrell, M. P.; Bhavanandan, V. P.;
Joshi, L. Anal. Biochem. 2011, 416, 18-26.
(36) Voinova, MU V.; Rodahl, M; Jonson, M.; Kasemo, B. Phys. Scr.
1999, 59, 391.
(37) Dolatshahi-Pirouz, A.; Skeldal, S.; Hovgaard, M B.; Jensen, T.;
Foss, M.; Chevallier, J.; Besenbacher, F. J. Phys. Chem. C 2009, 113,
4406-4412.
(38) Weber, N.; Pesnell, A.; Bolikal, D.; Zeltinger, J.; Kohn, J.
Langmuir 2007, 23, 3298-3304.
(39) Barrantes, A.; Santos, 0.; Sotres, J.; Arnebrant, T. J. Colloid
Interface Sci. 2012, 388, 191-200.
(40) Messier, A J.; Chhatre, S. S.; Nieves, A. R.; Mabry, J. M.; Cohen,
R. E.; McKinley, G. H- Soft Matter 2011, 7, 10122.
(41) Kobevar-Nared, J.; Kristl, J.; Emid-Korbar, J. Biomaterials 1997,
18, 677-681.
(42) Katagri, T.; Yamaguchi, A.; Komaki, M.; Abe, E.; Takahashi, N.;
Ikeda, T.; Rosen, V.; Wozney, J. M.; Fujisawa-Sehara, A; Suda, T. J.
Cell Biol 1994, 127, 1755-1766.
(43) Perez-Vilar, J. In Oral Delivery of Macromolecular Drugs;
Bernkop-Schniirch, A., Ed.; Springer: New York, 2009; pp 21-4&
(44) Chen, S.; Li, L.; Zhao, C.; Zheng, J. Polymer 2010, 51, 52835293.
(45) Gunkel, G.; Weinhart, M.; Becherer, T.; Haag, R.; Huck, W. T.
S. Biomacromolecules 2011, IZ 4169-4172.
(46) Szleifer, L Biophys. J. 1997, 72, 595-612.
(47) Feld&t6, Z; Pettersson, T.; Dedinaite, A. Langinuir 2008, 24,
3348-3357.
(48) Shi, L; Caldwell, K. D. J. Colloid Interface Sci 2000, 224, 372381.
3016
146
dxdoDorg/10.1021/bm4as447z I ftimcnoffoecdes 2013, 14, 300-3016
Biomacromolecules
(25) Chen, C. S.; Mrksich, M.; Huang, S.; Whitesides, G. M.; Ingber,
D. E. Biotechnol. Prog. 1998, 14, 356-363.
(26) Fink, J.; Thiry, M.; Azioune, A.; Dupont, R.; Chatelain, F.;
Bornens, M.; Piel, M. Lab Chip 2007, 7, 672.
(27) Dillmore, W. S.; Yousaf, M. N.; Mrksich, M Langmuir2004, 20,
7223-7231.
(28) Kikuchi, Y.; Nakanishi, J.; Nakayama, I; Shimizu, T.; Yoshino,
Y.; Yamagucdi, K.; Yoshida, Y.; Horiike, Y. Chem. Left 2008, 37,
1062-1063.
(29) Bilisle, J. M.; Kunik, D.; Costantino, S. Lab Chip 2009, 9, 3580.
(30) Kim, E.; Xia, Y.; Whitesides, G. M J. Am. Chem. Soc. 1996,118,
5722-5731.
(31) Sandberg, T.; Blom, H.; Caldwel, K. D. J. Biomed. Mater. Res. A
2009, 91A 762-772.
(32) Lundin, M.; Sandberg, T.; Caldwell, K. D.; Blomberg, E. J.
Colloid Interface Sci. 2009, 336, 30-39.
(33) Celi, J.; Gregor, B.; Turner, B.; Afdha], N. H.; Bansil, R.;
Erramilli S. Biomacromolecules 2005, 6, 1329-1333.
(34) Gerken, T. A; Gupta, R.; Jentoft, N. Biochemisy (Mosc.) 1992,
31, 639-648.
(35) Kilcoyne, M; Gerlach, J. Q.; Farrell, M. P.; Bhavanandan, V. P.;
Josh, L. Anal. Biochem. 2011, 416, 18-26.
(36) Voinova, M. V.; Rodahl, M.; Jonson, M.; Kasemo, B. Phys. Scr.
1999, 59, 391.
(37) Dolatshahi-Pirouz, A; Skeldal, S.; Hovgaard, M. B.; Jensen, T.;
Foss, M.; Chevallier, J.; Besenbacher, F. J. Phys. Chem. C 2009, 113,
4406-4412.
(38) Weber, N.; Pesnell, A; Bolikal, D.; Zeltinger, J.; Kohn, J.
Langmuir 2007, 23, 3298-3304.
(39) Barrantes, A; Santos, 0.; Sotres, J.; Arnebrant, T. J. Colloid
interface Sci. 2012, 388, 191-200.
(40) Meuler, A. J.; Chhatre, S. S.; Nieves, A R.; Mabry, J. M.; Cohen,
R. E.; McKinley, G. H. Soft Matter 2011, 7, 10122.
(41) Kokevar-Nared, J.; Kristl, J.; Smid-Korbar, J. Biomaterials 1997,
18, 677-681.
(42) Katagri, T.; Yamaguchi, A.; Komaki, M.; Abe, E.; Takahashi, N.;
Ikeda, T.; Rosen, V.; Wozney, J. M.; Fujisawa-Sehara, A; Suda, T. J.
Cell Biol 1994, 127, 1755-1766.
(43) Perez-Vdlar, J. In Oral Delivery of Macromolecular Drugs;
Bemkop-Schnurch, A, Ed.; Springer: New York, 2009; pp 21-48.
(44) Chen, S.; Li, L.; Zhao, C.; Zheng, J. Polymer 2010, 51, 5283-
5293.
(45) Gunkel, G.; Weinhart, M.; Becherer, T.; Haag, R.; Huck, W. T.
S. Biomacromolecules 2011, 12, 4169-4172.
(46) Szleifer, L Biophys. J. 1997, 72, 595-612.
(47) Feld&t6, Z; Pettersson, T.; Dedinaite, A. Langmuir 2008, 24,
3348-3357.
(48) Shi, L; Caldwell, K. D. J. Colloid Interface Sci 2000, 224, 372381.
3016
146
dxedoig/10.1021/bm400447z
ISibmacromoleCdes2013, 14, 3010-3016