Cell morphology drives spatial patterning in microbial communities William P. J. Smitha, Yohan Davitb, James M. Osbornec, Wook Kimd,1, Kevin R. Fosterd,2, and Joe M. Pitt-Francisa,2

aDepartment of Computer Science, University of Oxford, Oxford OX1 3QD, United Kingdom; bInstitut de Mecanique´ des Fluides de Toulouse (IMFT), Universite´ de Toulouse, Centre National de la Recherche Scientifique (CNRS), Institut National Polytechnique de Toulouse (INPT), Universite´ Paul Sabatier (UPS), F-31400 Toulouse, France; cSchool of Mathematics and Statistics, University of Melbourne, VIC 3010, Australia; and dDepartment of Zoology, University of Oxford, Oxford OX1 3PS, United Kingdom

Edited by Roman Stocker, Eidgenossische¨ Technische Hochschule Zurich,¨ Zurich, Switzerland, and accepted by Editorial Board Member Edward F. DeLong November 15, 2016 (received for review August 5, 2016) The clearest phenotypic characteristic of microbial cells is their cells and surfaces (20, 21). The resulting orientational order shape, but we do not understand how cell shape affects the affects how cell groups expand in microfluidic channels and dense communities, known as biofilms, where many microbes enables motile cells to swarm together in raft-like collectives live. Here, we use individual-based modeling to systematically (22, 23). Aligned cells are also subject to buckling interactions, vary cell shape and study its impact in simulated communities. which fold neighboring cell groups into one another to form We compete cells with different cell morphologies under a range fractal-like interdigitations (21, 24), and differences in cell sizes of conditions and ask how shape affects the patterning and evo- may drive depletion effects that lead to genetic demixing (25). lutionary fitness of cells within a community. Our models predict These studies suggest that, by influencing biomechanical inter- that cell shape will strongly influence the fate of a cell lineage: actions between microbes, shape may have far-reaching con- we describe a mechanism through which coccal (round) cells rise sequences for the properties and prospects of a cell within a to the upper surface of a community, leading to a strong spatial community. structuring that can be critical for fitness. We test our predictions Individual-based modeling has emerged as a powerful way to experimentally using strains of Escherichia coli that grow at a sim- study biofilms. These models serve as a testing ground to study ilar rate but differ in cell shape due to single amino acid changes how phenotypes, including adhesion, antibiotics, and extracel- in the actin homolog MreB. As predicted by our model, cell types lular polymeric substances (EPSs), affect individual strains and strongly sort by shape, with round cells at the top of the colony biofilms as a whole (26–31). However, the majority of individual- and rod cells dominating the basal surface and edges. Our work based models do not allow cell shape to be altered (32). We have suggests that cell morphology has a strong impact within micro- therefore developed a flexible simulation framework that allows bial communities and may offer new ways to engineer the struc- us to incorporate cell shape alongside cell division, physical inter- ture of synthetic communities. actions, and metabolic interactions via nutrient consumption. Our analyses identify a mechanism by which different cell shapes biofilms | cell morphology | biophysics | self-organization | can self-organize into layered structures, thereby providing par- synthetic biology ticular genotypes with preferential access to favorable positions in the biofilm. We test our model predictions with experiments ingle-celled microorganisms such as bacteria display signif- in which mutant Escherichia coli strains of different shapes are Sicant morphological diversity, ranging from the simple to the complex and exotic (1–3). Phylogenetic studies indicate that Significance particular morphologies have evolved independently multiple times, suggesting that the myriad shapes of modern bacteria may be adaptations to particular environments (4–6). Microbes can Microbial communities contain cells of different shapes, and also actively change their morphology in response to environ- yet we know little about how these shapes affect community mental stimuli, such as changes to nutrient levels or predation biology. We have developed a computational model to study (7, 8). However, understanding when and why particular cell the effects of microbial shape in communities. Our model shapes offer a competitive edge remains an unresolved question predicts that shape will have strong effects on cells’ posi- in microbiology. tioning, and, consequently, their survival and reproduction. Previous studies have characterized selective pressures favor- Rod-shaped cells are better at colonizing the base of the com- ing particular shapes (7, 9–11): for example, highly viscous envi- munity and its expanding edges, whereas round cells dom- ronments may select for the helical cell morphologies observed inate the upper surface. We show that the same patterns in spirochete bacteria (12). Thus far, these studies have predom- occur in colonies of Escherichia coli, using strains with dif- inantly focused on selective pressures acting at the level of the ferent shapes. Our work suggests that cell shape is a major individual cell. However, many species live in dense, surface- determinant of patterning and evolutionary fitness within associated communities known as biofilms, which are fundamen- microbial communities. tal to the biology of microbes and how they affect us—playing major roles in the human microbiome, chronic diseases, antibi- Author contributions: W.P.J.S., Y.D., J.M.O., K.R.F., and J.M.P.-F. designed research; W.P.J.S. and W.K. performed research; W.P.J.S., Y.D., J.M.O., W.K., K.R.F., and J.M.P.-F. analyzed otic resistance, biofouling, and waste-water treatment (13–17). data; and W.P.J.S., Y.D., J.M.O., W.K., K.R.F., and J.M.P.-F. wrote the paper. As a result, there has been an intensive effort in recent years to The authors declare no conflict of interest. understand how the biofilm mode of growth affects microbes and This article is a PNAS Direct Submission. R.S. is a Guest Editor invited by the Editorial their evolution (18, 19), but we know very little of the importance Board. of cell shape for biofilm biology. 1Present address: Department of Biological Sciences, Duquesne University, Pittsburgh, PA In biofilms, microbial cells are often in close physical con- 15282. tact, making mechanical interactions between neighboring cells 2To whom correspondence may be addressed. Email: [email protected] or particularly significant. Recent studies have suggested that rod- [email protected]. shaped cells can drive collective behaviors in microbial groups This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. because of their tendency to align their orientations with nearby 1073/pnas.1613007114/-/DCSupplemental.

E280–E286 | PNAS | Published online December 30, 2016 www.pnas.org/cgi/doi/10.1073/pnas.1613007114 Downloaded by guest on September 30, 2021 Downloaded by guest on September 30, 2021 mt tal. et Smith h ulcHat mg irr 3)adRcyMuti aoaois(35). Laboratories Mountain by Rocky provided and Micrographs (34) t2). Library (time Image biofilm Health mature Public a the to inocu- these t1) initial (time to an cells from according of structure colony lum configuration of (pro- cell development the dimensions the simulate three we updating between in rules, sequentially interactions colony the By repulsive of 3). and expansion cess 2), the (process to lead divide cells and growing grow (pro- to surroundings 1) their from cess nutrients absorb cells (“Dynamics”): processes depict Micrographs (“Cells”). coli capsules Escherichia (S) short and (L) are cells length Left variable (Top Bacterial of model. capsules elastic hybrid rigid mixed-morphotype using represented a use we surface, hard 1. Fig. approximate geometries whose capsules, (L) of long those or cells (S) length short All segment “Dynamics”). the of 1, 1, causing capsules (Fig. (Fig. nutri- divide, as space diffusing and represented in absorb are grow expand hard Cells to to a t2). surroundings community and on their t1 cells from times elongating ents 1, of (Fig. collections surface as communities bial Modeling. Individual-Based and devised it. subsequently test we to that performed experiments the describe to cul- on are bacteria shaped plates. agar differently on framework together which (IbM) tured in model hybrid experiments individual-based and the an within simula- with shape computer approaches: tions cell two used bacterial we of environment, biofilm consequences the explore To fitness Results and architecture spatial both differences communities. to microbial that central within shows are shape work cell Our in colonies. in together cultured mmmdlfrivsiaigteiflecso elsaein shape cell of influences the investigating for min- model appropriate only an imum conceptually therefore, is and, model complex mathematical our more now shapes. incrementally detailed, underlying more can cell the be we to requires nonspherical formulation that extension containing except this (26–28), communities Although cells study spherical also on based els communities bacterial in together 7). found (4, frequently ubiq- are and morphologies these uitous simple, Although 33). (24, “Cells”) ee eitouetemdladispeitos eoegoing before predictions, its and model the introduce we Here, vrl,ormdli iia nsrcuet rvosmod- previous to structure in similar is model our Overall, osmlt h rwho ie-pce atra ooyo a on colony bacterial mixed-species a of growth the simulate To ;cca n ailrclsaeapoiae epcieyuiglong using respectively approximated are cells bacillar and coccal ); ;bcla n oclcl hpsaemdldusing modeled are shapes cell coccal and bacillar Top); tpyoocsaureus Staphylococcus atra ifimgot sdie ytreprimary three by driven is growth Biofilm bacteria. (Bottom) u b rmwr osdr micro- considers framework IbM Our and epciey(i.1, (Fig. respectively coli, E. .aureus S. l l n xdradius fixed and n radius and and (Top) r r o Lclne odvlplyrdsrcue;b otat the contrast, by structures; layered tendency develop the to quantifies colonies distribution SL This for the volume. along colony distribution total biovolume the to epo itgasof histograms plot we 2B, Fig. cell In side-by-side. whereas growing 2A, Fig. colonies in shown are colonies 2D modeling previous in 32). reported (24, as studies groups, cell adjacent interdigitations mixtures fractal-like between developing third SS adja- patterning, a produced spatial contrast, between colonies of LL By type boundaries layering. no cells. vertical with L groups, strain smooth, blue cent mix- LL produce atop cell and to lying SL groups cells tended SS, with present. S (SL, structure, were red layered 2 of that L a Fig. developed shapes and emerged, spontaneously cell of structures tures S the top spatial of on different the that depending combinations at found 1:1 shown We mixtures). three as using shapes, cell colonies grew we model to assumption shapes. homoge- this cell different remove spatially between we and interactions study, competitive complete this is in biofilm Later a neous. the at in exponentially perfusion grow ent to rate cells growth all maximal allowed we full in boundary approximation, allow described domain as and and are grow conditions constraint to treatments Initial this forced movement. simulations, lift were of 2D cells simulations freedom that the 3D so In monolayer; domains a dimensions. thin very three biofilm used and vertical we of two growth in the simulate sections to framework Colonies. model Within our Patterning Spatial Drives Shape Cell are framework the on in details provided Further communities. microbial case. per simulations 20 were There 2D. is in shown layering the those this to to confined simulations, similar longer control no LL are S and entations with SS colonies, In 3D (C and cells. absent. 2D L both above in sitting structure cells layered morphol- a L) develop (long, mixtures cell rod-like SL or of S) histograms depending (short, Volume-weighted coccal patterns (B) a ogy. spatial have different strains the form mixtures whether strains 1:1 on from blue-labeled grown and biofilms red- 2D of of Sections (A) growth. biofilm nential 2. Fig. o hs he ae,rpeettv n-tt npht of snapshots end-state representative cases, three these For growth, biofilm on shape cell of impact the investigate To z oriae o h w tan:here, strains: two the for coordinates elmrhlg fet efognzto nsmltoso expo- of simulations in self-organization affects morphology Cell he-iesoa iuain,i hc elpstosadori- and positions cell which in simulations, Three-dimensional ) IAppendix. SI PNAS µ max ersnigaseai hr nutri- where scenario a representing , | ulse nieDcme 0 2016 30, December online Published safirst a As Methods. and Materials xz ln,pouesailpatterns spatial produce plane, z oriae (P coordinates z oi S1 Movie xs omlzdby normalized axis, P (z ) efis used first We corresponds (z hw these shows )so that show )) | E281

MICROBIOLOGY BIOPHYSICS AND PNAS PLUS COMPUTATIONAL BIOLOGY Fig. 3. Cell morphology influences strain fitness within colonies. (A) Competition between different cell shapes is simulated by growing biofilm sections under nutrient-limited conditions. Nutrients are supplied either from above the colony (first row) or from below (second row). Cells on the left side of the colony are colored by their growth rate, showing that rapid cellular growth is limited to a thin layer at the top or at the base of the colony, in the two scenarios, respectively. Contours on the nutrient field (background) correspond to u/u0 = 0.25, 0.50 and 0.75. (B) Volume-weighted histograms (P(z)) of cell heights in nutrient-limited SL colonies show similar layering effects to those created under homogeneous growth conditions, with S cells (red) growing atop L cells (blue). (C) Bar plots of strain fitness show that coccal morphologies outcompete rod-like morphologies in scenarios where nutrients are delivered from above the colony, but the reverse is true when nutrients are delivered from below. There were 10 simulations per case.

vertical strain distributions seen in SS and LL control colonies Shape-Driven Layering Alters Strain Fitness. Next, we examined are identical, confirming the absence of layering. This effect can the consequences of layering phenomena on strain competition be reproduced in 3D simulations, where cell centers and axes are within the biofilm environment. Because biofilms are typically not confined to the xz plane (Fig. 2 B and C). characterized as highly heterogeneous environments, systematic In SI Appendix, we discuss further simulations carried differences in the positioning of bacterial strains could lead to out to interrogate the mechanism of the layering effect (SI significant differences in their fitness, offering a competitive Appendix, Figs. S1–S7 and Movie S2). We hypothesize that advantage to particular cell morphotypes (39–41). layering occurs because groups of rod-shaped cells form wedge- To examine this hypothesis, we repeated our 2D and 3D like structures, which, guided by the basal surface, collectively biofilm growth simulations under nutrient-limited conditions. burrow beneath groups of coccal cells. The necessary group Instead of allowing all cells to grow at the same rate as before, structure is, in turn, produced by steric interactions between we coupled each cell’s growth rate µi to the local concentration neighboring cells and the basal surface, which produce nematic u of a rate-limiting nutrient field, evaluated at that cell’s centroid ordering in rod-shaped cells but not in coccal cells. pi . As in previous studies (26, 28, 42, 43), we model this coupling We also report tests that verify that the layering effect is robust using the Monod equation µi = µmaxu(pi )/(K + u(pi )), where with respect to changes in simulation boundary conditions (SI µmax and K are the maximum specific cell growth rate and the Appendix, Fig. S9), relative cell size (SI Appendix, Fig. S7A), uptake saturation constant, respectively. By imposing different simulation parameters such as the rate of division noise injec- boundary conditions on the nutrient reaction-diffusion equation, tion (SI Appendix, Fig. S7B), and changes in initial cell align- we created opposing perfusion scenarios: one in which nutrient ment (SI Appendix, Fig. S7B). However, we find that layering is was supplied from above the colony and another in which sup- completely dependent on the presence of the basal plane, con- ply came from below. In each simulation, we measured the fit- firming its importance in the layering mechanism (SI Appendix, ness of each strain, ωstrain, by computing the number of division Fig. S7B). events per unit time, ωstrain = log2(Vstrain(tend)/Vstrain(0))/tend. Overall, these observations suggest that individual cell shape Here, Vstrain(tend) refers to the total cell volume of a strain at is capable of driving significant structural changes in the the endpoint of the simulation tend. The results of these simu- wider biofilm environment—and, in particular, that combina- lations are shown in Fig. 3. Nutrient field parameters are listed tions of differently shaped cells can drive the formation of lay- in SI Appendix, Table S3; these are chosen so as to create thin ered structures, as reported previously in multispecies biofilms growth layers of similar thicknesses in the two perfusion sce- (36–38). narios. In both cases, the Damkohler¨ number D (the ratio of

E282 | www.pnas.org/cgi/doi/10.1073/pnas.1613007114 Smith et al. Downloaded by guest on September 30, 2021 Downloaded by guest on September 30, 2021 fteetresrisehbtma setrto f44,35,and 4, 3.55, 4.44, Fig. of (44); ratios respectively aspect 2.50 mean exhibit strains three strains these mutant of and WT) as here REL606mreB to (referred different strain REL606 three colonies WT of bacterial combinations grew binary and proteins using fluorescent with strains the REL606mreB mutant the for 0.015 mt tal. et Smith set in together cultured rate be growth specific to the which shapes example, [For rates, with- different proportions. growth ratio, of cell aspect cells exponential cell allows in affecting changes significantly stable out and mutations these substantial Critically, cod- MreB. cause gene protein the cytoskeletal are in the library mutation for this point ing in a for Strains except impasse. identical this genetically to the solution a However, provide dif- effects. (44) confounding physiologically coli introducing be E. well, typically as will experi- ferent predictions strains because our challenging, shaped test is differently empirically to shape sought cell then Studying composition mentally. we and model, development our community with affect can shape cell Patterning. Shape-Driven of Tests Experimental from supply nutrient limited with growing above. cells, L and S of colony nutri- limiting. the becomes when this supply only ent that differences but fitness into availability, translates nutrient patterning of irrespective occurs layering in availability where and reversed. S cases are of mixtures In fates SL the 3B. in colony, Fig. cells the L in below as from enhanced, histograms delivered is height are colony nutrients the cell of the layer by upper number the the shown the result, in a present to As cells relative exists. S layering boost of no fitness where a cases LL gain and and SS afforded rate correspondingly at are growth access cells increased S nutrient an below. preferential cells L receive of from cells expense delivered the S are into colony, nutrients the translate When above conditions, fitness. strain idealized in under differences can, effects layering cells. L of favor in time this reduced fitness, Next, is strain differential is case. fitness strain SL red the the the and to of way, at ratio same maximum aspect the the in a cases, at increased to LL increases and up SL cells 1.5, cells, the S 1.1, between L of values of fitness the expense the through the aspect Correspondingly, increased birth 3.0. is the and strain cases, 2.0, blue SL and the mix- SS of strain the ratio Between cell ratios. used. aspect LL shapes 3B cell and cell Fig. intermediate of SL, of row SS, top mixtures The Here, using of also scenarios. only but competition not tures nutrient colonies two grow the we in fitness strain way. same the in prepared mixtures, LL and SS for histograms and lay- con- a nutrient-limited adopt under In still scenar- ditions. together strains in two grown cell heights the when L cell structure in and of ered S thin colony, histograms that the plot a showing of we colonies, to 3B, base these growth Fig. limited the In their at is respectively. by or ios growth top colored the cellular are at colony rapid layer the that of but indicating before, face as rate, type left by the colored are on bulk the cells in Cells after growth. mixtures of cell h SL 12 1:1 from grown colonies showing scenarios, 0.01. to set rate; diffusion to rate uptake nutrient ots h rdcin formdli ir,w labeled we vitro, in model our of predictions the test To nbt ae,w xmndteefcso ayn h nutrient the varying of effects the examined we cases, both In strain that hypothesis the confirm simulations our Importantly, and shape cell between relationship the explore we 3C, Fig. In 3A Fig. .We h uretcniin r eesd otoi the is too so reversed, are conditions nutrient the When ∼0. h uat eetyietfidaddsrbdb od tal. et Monds by described and identified recently mutants −1 epoiecrepnigimages corresponding provide we S8, Fig. Appendix, SI o h neta W)sri and strain (WT) ancestral the for rvdseapesasoso ohsimulation both of snapshots example provides eosrtn that demonstrating S11 , and S10 Figs. Appendix, SI 5S(S n REL606 mreB and (AS) A53S rvdsapcoilrpeetto fthe of representation pictorial a provides oi S3 Movie Inset 5T(44).] A53T hw niiulclso each of cells individual shows was Methods) and Materials hw neapeo 2D a of example an shows 0.516 aigsonthat shown Having 5K(K.Cells (AK). A53K .coli E. ± µ tan:the strains: is 0.007 0. 519 h −1 ± natfc rdcdb h ursetlbln scheme. labeling fluorescent the by not is produced layering artifact that an demonstrating 5B), (Fig. pro- second result labels same the color the the by duces Reversing predicted 3. as Fig. in base, over- shown colony an scenario the modeling see at also cells we rod colony, of the growth below nutrients agar Because the colony. from the supplied in are depths increasing at taken projec- side slices from both confocal visible is of effect tions This colony. show from the (red) we cells of 5A, AK base Fig. shorter the displace In blue) cells. in longer (shown cells of WT top that on lying structures, cells layered of shorter emergence with the in result can together strains composi- edge. the colony affect the can of com- tion shape In shape cell other dimension. that for images demonstrating fractal the colony binations, lower additional in show a as we S12, with patterns Fig. fractal-like but generating case, interme- mixing, WT an of produces again column) level center column), 4, diate (Fig. LL interme- shape The cell 2, (24). AS studies diate (Fig. previous of model findings the the recapitulating interstrain in smooth seen show fractal-like interdigitations the colonies reproduce to mixtures AK-AK WT-WT Similarly whereas column), boundaries, row). (SS (bottom con- sections 2 shows and edge 4 Fig. row) colony Fig. of (top ratio. images colonies aspect focal complete cell of (red-blue) on images interstrain dependent epifluorescent (AK-AK, of strongly morphotype degree is that mixing one showed only WT-WT) of AS-AS, composed colonies in Firstly, Patterning Drives coli Morphology are Cell details Further blue. in shown in cells sig- provided fluorescence GFP-labeled the with from microscopy. pseudocolor, nal in scanning are imaged shown laser being images confocal The before and markers, flu- epifluorescence GFP red using or with labeled (RFP) strains, using protein three grown orescent these were of Colonies pairing possible comparison. each visual for strains these eetknatr4 fgot nec ae Saebr:tprw mm; 1 row, top 44. bars: ref. (Scale from case. permission with each taken in 20 growth of row, Images h bottom colonies. behavior, 48 WT after in intermediate than taken display dimension were fractal colonies lower of AS extending patterning colony. cell showing boundaries blue the group and simulations: smoother through red show our vertically between colonies (bot- of AK mixing projections predictions whereas fractal-like groups, orthogonal complex, the show corroborate alongside colonies images shown WT These are row). sec- colony edge tom each of images of confocal Inset representative (Center tions magnification, ratio higher aspect At cell rate. in differ WT which of (AK), composed colonies compares coli morphotype- row E. of top test The experimental an relationships. as patterning culture, solid in together grown are i.4. Fig. u xeiet lovrf htmxn ogadsotcell short and long mixing that verify also experiments Our ooisrpoue eea e rdcin formodel. our of predictions key several reproduced colonies ihtoeo ci ooo mutants homolog actin of those with loecnl labeled Fluorescently aeil n Methods. and Materials . hs-otatcl mgs(fwdh3.5 width (of images cell Phase-contrast µm.) o o)adfo horizontal from and row) top 5A, (Fig. z-stacks PNAS .coli E. | ulse nieDcme 0 2016 30, December online Published tan ihdfeetcl morphologies cell different with strains .coli E. mreB u rwa h same the at grow but ) A53S Colonies. A)and (AS) IAppendix, SI )were µm) mreB The | E283 A53K E.

MICROBIOLOGY BIOPHYSICS AND PNAS PLUS COMPUTATIONAL BIOLOGY Fig. 5. Mixed-morphotype colonies reproduce the layering phenomena predicted by simulations. (A) Pseudocolor confocal images show that GFP-labeled WT E. coli (shown in blue) are able to burrow beneath shorter, wider AK mutants (red), leading to the WT cells’ enrichment at the base of the colony. (B) The same is true when fluorescent labels are reversed. Images correspond to vertical colony sections (A and B, top row) and horizontal slices taken at successive depths (A and B, bottom row). (C) We quantify layering effects by measuring the volume fraction of GFP-labeled cells as a function of depth in the colony, using automated image segmentation (Materials and Methods). (D) The gradients of these traces, corresponding to the percentage gain (or loss) of GFP signal per unit depth, are compared for various binary shape combinations. Cell micrographs in A, B, and D are taken with permission from ref. 44. (Scale bars: 20 µm.) Confocal images and data taken after 48 h of growth.

In Fig. 5C, we use automated image analysis to quantify the ing of the functional value of cell shape, both as a competitive layering effect. By counting the number of blue pixels in each phenotype in the biofilm context and as a means for bacteria to layer of the stack (Materials and Methods), we estimate the vol- influence their environment through collective action. ume fraction of AK (top row) or WT (bottom row) cells as a From an evolutionary perspective, our work suggests that the function of depth in the colony. These traces show an approxi- biofilm environment may select for particular cell shapes in spe- mately linear reduction in the AK cell fraction (and correspond- cific environments because of the ways in which they collec- ing increase in WT fraction) within the first 15 µm below the tively influence biofilm architecture. Given the predominance of colony surface, which is in agreement with the cell arrangements the biofilm environment for microbial life (18, 19), any selective predicted in Figs. 2 and 3. effect produced might be expected to have a strong impact on the Finally, we repeated this analysis with images taken from evolution of bacterial morphology. In natural biofilms, rapid cell colonies of other morphotype combinations, including the con- growth is often limited to the upper regions of a biofilm (28, 41), trol colonies shown in Fig. 4. In each case, we quantified inter- which will select for coccal morphologies. Indeed, this selective strain layering using the average slope of the GFP fraction traces, pressure provides a rationale for morphological transitions asso- such as those shown in Fig. 5C, using linear fitting. Example ciated with biofilm development (46–48) and may partly explain image segmentations, along with the complete set of GFP frac- the ubiquity of the coccal morphology despite disadvantages, tion traces, are provided in SI Appendix, Fig. S13. The degree of such as decreased nutrient absorption area (7). layering for each type of mixture is shown in Fig. 5D. As pre- In reality, microbial communities are far more intricate dicted by the model, we observe a strong relationship between than the description used in our simulations. Like all mod- differences in cell shape and the degree of layering in the colony els, our framework makes simplifying assumptions: we deliber- across these different genotypes. Our experiments, therefore, ately neglect the role of cell motility, detachment, and shear support multiple predictions of our model (Figs. 4 and 5). More- forces from the liquid surrounding the colony (42, 43, 49). over, this support came without needing to tune conditions; Our representation of EPS secretions—a hallmark of microbial we performed the experimental tests after the modeling. This biofilms—is purely implicit, whereas the inclusion of explicit EPS observation suggests that the effects we report in this study particles has been shown to influence colony structure in previ- are robust to experimental conditions and may be common to ous studies (25, 41). Our model also neglects adhesive interac- many existing experimental systems that use mixtures of different tions, which could inhibit the layering effect if coccal cells were cell shapes. irreversibly attached to the basal surface. However, these omissions allow a degree of realism to be Discussion traded for additional control and tractability. Rather than We have used simulations and experiments to show that cell attempting to reproduce the exact dynamics of colony growth, shape can be a determinant of both spatial patterning and com- our simulations instead predict the rich dynamics that can position within a microbial community. In particular, we find emerge when nonspherical cell shapes are introduced to existing that mixtures of rod-shaped and coccal cells can produce layered modeling paradigms (28, 50). The fact that these predictions are colony structures, as observed previously in both biotic and abi- corroborated by our experiments demonstrates the usefulness of otic environments (36, 45). This result extends our understand- this approach.

E284 | www.pnas.org/cgi/doi/10.1073/pnas.1613007114 Smith et al. Downloaded by guest on September 30, 2021 Downloaded by guest on September 30, 2021 mt tal. et Smith 27 G DR Cewrsaingahc ad o iuain without simulations for cards graphics workstation PCIe GDDR5 6GB for C2075 labels 1:1 color a blue of and axis red consisted cell with inoculum with marked tracking. Each strains, domain, lineage case. bacterial the 3D two of of the base mixture in the randomly on drawn randomly vectors arranged cells of length base lum of plane a to dimensions confined were capsules the 3D L simulations, of 2D axes are In and demanding. which centroids more conditions, computationally boundary but periodic realistic of more of effects sections the surrounding reproduces and imation surface repul- basal In approximate biofilm. the to the from walls forces hard mechanical by surrounded sive were domains cases, both Simulations. IbM Methods and Materials evolu- communities. and microbial patterning cell within of that fitness determinant suggests tionary neglected work major patterns a Our same is colonies. shape the bacterial documented real in also emerging have We micro- morphology. by driven bial mecha- patterning spatial examined and and self-organization described, of nisms predicted, have we different incorporating shapes, competi- framework cell simulation and the a architecture building study biofilm By to as on tion. experiments morphology known cell and communities, of modeling impact diverse used have together genetically We and grow biofilms. dense commonly their shapes in different of Microorganisms Conclusions and strains (14). commu- between species interactions for the selecting affect that by productive architectures consortia, more nity microbial engineer synthetic to stable (28). help or stable therefore evolutionarily could mixing findings more genetic Our become as less such may strategies selective secretion produce cooperative enzyme control cells, instance, cells rod-shaped to For of coccal mixtures strategies. used than of social be mixtures certain turn against because in or for within could mixing pressures strain which of degree colony, the results affect the our can shape to Furthermore, cell experimenters that structures. show by colony used desirable shape- be that promote could suggests layered which self-organization a 53–55), with mediated (36–38, associated biofilms often in are structure Metabolic species relationships. between cooperative discuss exchanges more we influence mechanisms also patterning could the microbes, between tions cancer and tissues developing 52). of (51, tumors morphogenesis the microbes of for an collectives also for in be but just anisotropy to not growth prove parameter physical may and important shape shape systems—ultimately, for work biological roles our other new generally, suggest More also communities. may microbial in of mor- influences shape the of cell characterize inclusion fully the help will by con- variability altered phological other are of studies isolation theoretical because of in predictions previous process, the varied how this investigating be Likewise, to variables. to founding instrumental shape an be cellular of will allow instead they (44) by variable al. developed those et physical as more Monds such a strains in as Bacterial attribute. colonies shape incidental cell mixed-shape treating examine detail, to studies theoretical .VsP ta.(01 egysMna fSseai atrooy h Fir- The Bacteriology: shape. cell Systematic Bacterial of (2005) C Manual Jacobs-Wagner MT, Bergey’s Cabeen 3. (2011) al. et P, mor- selected Vos of history 2. natural The diversity: Bacterial (1965) VBD Skerman MP, Starr 1. x ntotal, In interac- antagonistic only considers study our although Finally, and empirical further for need the highlight predictions Our iue Srne cec uiesMda odeh,TeNetherlands), The Dordrecht, Media, 601–610. Business & Science 3. Vol (Springer micutes bacteria. unusual phologically = 300 ,wees3 iuain sdacbia oanwt base with domain cuboidal a used simulations 3D whereas µm, ,0 oe iuain eepromdo w VDATesla NVIDIA two on performed were simulations model ≈3,000 L x eso httehr onayapprox- boundary hard the that show we S9, Fig. Appendix, SI = iuain eepromduig2 n Ddmis In domains. 3D and 2D using performed were Simulations L y = 40 .Smltoswr ntaie sn ninocu- an using initialized were Simulations µm. nuRvMicrobiol Rev Annu 19:407–454. a e Microbiol Rev Nat 3(8): eevdfnigfo h alv eerhCnr o vlto n Human Oxford). and Evolution College, for (Magdalen K.R.F. Centre Science and Research W.P.J.S. Calleva Sci- Both the 242670. from Physical sup- funding Grant received is and Council K.R.F. Research and Engineering European EP/G03706X/1), by an (Grant ported on studentship by Council comments supported Research helpful is ences for W.P.J.S. reviewers Schluter, and manuscript. Jonas anonymous Durham, the cards two Mack and and Insti- graphics Newman, CellModeller at with Dianne C2075 resources help their computing Tesla for the Murphy two to M access de the providing tut for donating Quintard NVIDIA Michel thank also mutant of Otsu’s ACKNOWLEDGMENTS. using intensi- layer with individual data each channel for green computed the (58). threshold, by method in of estimated a pixels each was above of cells In ties GFP-labeled number six outright. of than the analysis number fewer counting the the number with images, from stacks sufficient remaining confocal removed the a on, were later ensure remaining manually fitting To images chosen curve accuracy. for were segmentation points thresholds data These the data. optimize of channel 12% to green combin- than by less and created intensities red images had grayscale removing ing pixels composite by of in 1% measured excluded the maximum, than were the more using which images segmented in each dark layer then For excessively any procedure. were or following Images incomplete the sig- to normalizing stack, range. according by intensity plugin, and bioformats 1% full Matlab to the saturation to pixel setting nal by FIJI in enhanced Analysis. Image Confocal scan- mixed- laser inoculation, 57). confocal (41, following and described previously epifluorescence days as by microscopy Two ning imaged out. were from colonies carried reverse using culture are also experiments combina- sur- control study were each labels, the fluorescent this For labeling and glu- on volume. strains in cell between 1-µL volumes of reported difference at tion 10-µL data significant spotted or plates supplementation; no 1- glucose-supplemented observed lactose either We or plates. in cose DM spotted of then face and ratio, 1:1 plasmids. maintain 37 g 15 with (Thermo-Fisher (175 LB supplemented in (57), rpm medium 250 (DM) at 37 shaking cul- at grown routine were Scientific) For cells microscopy. liquid, confocal in and turing epifluorescence (44). for time strains phase three lag in the is marginally share A53S only strains differ and and (44). three rate morphology, protein all growth cellular Importantly, specific MreB same A53. widest the with of and compared residue shortest intermediate, acid the amino nonsynony- exhibits 53rd respective A53K the the at with mutation REL606 mous from constructed previously (REL606mreB were A53K A53. as here to referred Plasmids. and Strains Bacterial are assumptions, and equations model including in framework, provided details IbM Further 2015a. the Matlab using on out carried was results of postprocessing operation. norm and where o iuain nldn hm oacutfrtesohsi os terms criterion convergence noise the stochastic card using converged, the graphics Appendix measurements for GFX (SI account 4GB model To our K5000 in them. Quadro including NVIDIA simulations two for on and fields nutrient .Ksl T adc M acm D rnY 21)Dvriytkssae Under- shape: takes Diversity (2016) YV Brun PD, Caccamo AM, Bring- Randich DT, (2001) M Kysela Vicente A, 6. Valencia J, Mingorance M, Gonzalez-Moreno morphology. J, bacterial Tamames of mapping 5. Phylogenetic (1998) GE Fox JL, Siefert 4. vrih utrswr eilydltdi B,mxdtgte na in together mixed PBS, in diluted serially were cultures Overnight eitoue lsispaRPo mxF AaaLna noall into (Amaxa/Lonza) pmaxGFP or pmaxRFP plasmids introduced We the and 4.3.1, Paraview using checked and visualized were Simulations ogy tnigtemcaitcadaatv ai fbceilmorphology. bacterial of basis adaptive and 14(10):e1002565. mechanistic the standing shape. bacterial into order gene ing ◦  .Almdawr upeetdwt aayi t50 at kanamycin with supplemented were media All C. µg TOL 4 P 10):2803–2808. (Pt 144 hXi · sacnegnetlrnestt 5%; to set tolerance convergence a is cnqedsFudsd olue etakTmRdeadNiall and Rudge Tim thank We Toulouse. de Fluides des ecanique ´ mL N .coli E. IAppendix SI −1 eoe envco fmaueet o apesize sample a for measurements of vector mean a denotes rlcoe(210 lactose or ) ◦ .Frpaeclue,clswr rw nDvsminimal Davis on grown were cells cultures, plate For C. ||hXi el n ewnHagfrpoiigtesris We strains. the providing for Huang Kerwyn and cells N etakCrln rpn o ugsigteuse the suggesting for Tropini Carolina thank We PNAS . hXi − olwn olcin ofcliaesak were stacks image confocal collection, Following ,w nrae iuainsml ie ni our until sizes sample simulation increased we ), | N .coli E. −1 µg ulse nieDcme 0 2016 30, December online Published rnsGenet Trends || · l 2 mL E66i h aet(T tan(56), strain (WT) parent the is REL606 / ||hXi −1 speiul ecie 4)at (44) described previously as ) A53K N · ||·|| || 17(3):124–126. n 5S(REL606mreB A53S and ) L l − 2 l 1 2  < ersnsteEuclidean the represents gradete glucose either and agar TOL , µg · Microbiol- mL LSBiol PLoS | − A53S E285 1 [1] N to ) ,

MICROBIOLOGY BIOPHYSICS AND PNAS PLUS COMPUTATIONAL BIOLOGY 7. Young KD (2006) The selective value of bacterial shape. Microbiol Mol Biol Rev 34. Matthew J, Arduino JC (2007) Staphylococcus aureus SEM (image 6486). Accessed July 70(3):660–703. 7, 2016. 8. Justice SS, Hunstad Da, Cegelski L, Hultgren SJ (2008) Morphological plasticity as a 35. Rocky Mountain Laboratories, NIAID, NIH (2006) Escherichia coli SEM. Accessed July bacterial survival strategy. Nat Rev Microbiol 6(2):162–168. 7, 2016. 9. Young KD (2004) Bacterial shape. Mol Microbiol 49(3):571–580. 36. Christensen BB, Haagensen JAJ, Heydorn AJ, Molin S (2002) Metabolic commensal- 10. Young KD (2007) Bacterial morphology: Why have different shapes? Curr Opin Micro- ism and competition in a two-species microbial consortium. Appl Environ Microbiol biol 10(6):596–600. 68(5):2495–2502. 11. Young KD (2010) Bacterial shape: Two-dimensional questions and possibilities. Annu 37. Hansen SK, Rainey PB, Haagensen JAJ, Molin S (2007) Evolution of species interactions Rev Microbiol 64:223–240. in a biofilm community. Nature 445(7127):533–536. 12. Nakamura S, Adachi Y, Goto T, Magariyama Y (2006) Improvement in motion effi- 38. Elias S, Banin E (2012) Multi-species biofilms: living with friendly neighbors. FEMS ciency of the spirochete Brachyspira pilosicoli in viscous environments. Biophys J Microbiol Rev 36(5):990–1004. 90(8):3019–3026. 39. Wimpenny J, Manz W, Szewzyk U (2000) Heterogeneity in biofilms. FEMS Microbiol 13. McLoughlin K, Schluter J, Rakoff-Nahoum S, Smith AL, Foster KR (2016) Host selection Rev 24(5):661–671. of microbiota via differential adhesion. Cell Host Microbe 19(4):550–559. 40. Nadell CD, Xavier JB, Foster KR (2009) The sociobiology of biofilms. FEMS Microbiol 14. Coyte KZ, Schluter J, Foster KR (2015) The ecology of the microbiome: Networks, Rev 33(1):206–224. competition, and stability. Science 350(6261):663–666. 41. Kim W, Racimo F, Schluter J, Levy SB, Foster KR (2014) Importance of positioning for 15. Wolcott RD, Ehrlich GD (2008) Biofilms and chronic infections. JAMA 299(22): microbial evolution. Proc Natl Acad Sci USA 111(16):E1639–E16347. 2682–2684. 42. Picioreanu C, Van Loosdrecht MCM, Heijnen JJ (2000) Effect of diffusive and convec- 16. Stewart PS, Costerton JW (2001) Antibiotic resistance of bacteria in biofilms. Lancet tive substrate transport on biofilm structure formation: a two-dimensional modeling 358(9276):135–138. study. Biotechnol Bioeng 69(5):504–515. 17. Klapper I, Dockery J (2010) Mathematical description of microbial biofilms. SIAM Rev 43. Kreft JU, Wimpenny JWT (2001) Effect of EPS on biofilm structure and function 52(2):221–265. as revealed by an individual-based model of biofilm growth. Water Sci Technol 18. Costerton JW, Lewandowski Z, Caldwell DE, Korber DR, Lappin-Scott HM (1995) 43(6):135–141. Microbial biofilms. Annu Rev Microbiol 49:711–745. 44. Monds RD, et al. (2014) Systematic perturbation of cytoskeletal function reveals 19. Hall-Stoodley L, Costerton JW, Stoodley P (2004) Bacterial biofilms: From the natural a linear scaling relationship between cell geometry and fitness. Cell Rep 9(4): environment to infectious diseases. Nat Rev Microbiol 2(2):95–108. 1528–1537. 20. Volfson D, Cookson S, Hasty J, Tsimring LS (2008) Biomechanical ordering of dense 45. Nava GM, Friedrichsen HJ, Stappenbeck TS (2011) Spatial organization of intestinal cell populations. Proc Natl Acad Sci USA 105(40):15346–15351. microbiota in the mouse ascending colon. ISME J 5(4):627–638. 21. Boyer D, et al. (2011) Buckling instability in ordered bacterial colonies. Phys Biol 46. Serra DO, Richter AM, Klauck G, Mika F, Hengge R (2013) Microanatomy at cellular 8(2):026008. resolution and spatial order of physiological differentiation in a bacterial biofilm. 22. Cho H, et al. (2007) Self-organization in high-density bacterial colonies: Efficient MBio 4(2):e00103–e00113. crowd control. PLoS Biol 5(11):e302. 47. Bergkessel M, Basta DW, Newman DK (2016) The physiology of growth arrest: Uniting 23. Copeland MF, Weibel DB (2009) Bacterial swarming: A model system for studying molecular and environmental microbiology. Nat Rev Microbiol 14(9):549–562. dynamic self-assembly. Soft Matter 5(6):1174–1187. 48. Hendrickx L, Hausner M, Wuertz S (2003) Natural genetic transformation in monocul- 24. Rudge TJ, Federici F, Steiner PJ, Kan A, Haseloff J (2013) Cell polarity-driven instabil- ture Acinetobacter sp. strain BD413 biofilms. Appl Environ Microbiol 69(3):1721–1727. ity generates self-organized, fractal patterning of cell layers. ACS Synth Biol 2(12): 49. Picioreanu C, et al. (2007) Microbial motility involvement in biofilm structure forma- 705–714. tion - a 3D modelling study. Water Sci Technol 55(8-9):337–343. 25. Ghosh P, Mondal J, Ben-Jacob E, Levine H (2015) Mechanically-driven phase separa- 50. Xavier JB, Foster KR (2007) Cooperation and conflict in microbial biofilms. Proc Natl tion in a growing bacterial colony. Proc Natl Acad Sci USA 112(17):E2166–E2173. Acad Sci USA 104(3):876–881. 26. Picioreanu C, van Loosdrecht MCM, Heijnen JJ (1998) Mathematical modeling of 51. Heisenberg CP, Bella¨ıche Y (2013) Forces in tissue morphogenesis and patterning. Cell biofilm structure with a hybrid differential-discrete cellular automaton approach. 153(5):948–962. Biotechnol Bioeng 58(1):101–116. 52. Basan M, Elgeti J, Hannezo E, Rappel W-J, Levine H (2013) Alignment of cellular motil- 27. Kreft J-U (2004) Biofilms promote altruism. Microbiology 150(Pt 8):2751–2760. ity forces with tissue flow as a mechanism for efficient wound healing. Proc Natl Acad 28. Nadell CD, Foster KR, Xavier JB (2010) Emergence of spatial structure in cell groups Sci USA 110(7):2452–2459. and the evolution of cooperation. PLoS Comput Biol 6(3):e1000716. 53. Sekiguchi Y, Kamagata Y, Nakamura K, Ohashi A, Harada H (1999) Fluorescence in 29. Schluter J, Foster KR (2012) The evolution of mutualism in gut microbiota via host situ hybridization using 16S rRNA-targeted oligonucleotides reveals localization of epithelial selection. PLoS Biol 10(11):e1001424. methanogens and selected uncultured bacteria in mesophilic and thermophilic sludge 30. Nadell CD, et al. (2013) Cutting through the complexity of cell collectives. Proc Biol granules. Appl Environ Microbiol 65(3):1280–1288. Sci 280(1755):20122770. 54. Wilmes P, et al. (2009) Natural acidophilic biofilm communities reflect distinct organ- 31. Estrela S, Brown SP (2013) Metabolic and demographic feedbacks shape the emer- ismal and functional organization. ISME J 3(2):266–270. gent spatial structure and function of microbial communities. PLoS Comput Biol 55. Brenner K, Arnold FH (2011) Self-organization, layered structure, and aggregation 9(12):e1003398. enhance persistence of a synthetic biofilm consortium. PLoS One 6(2):e16791. 32. Storck T, Picioreanu C, Virdis B, Batstone DJ (2014) Variable cell morphology 56. Lenski RE, Rose MR, Simpson SC, Tadler SC (2016) Long-term experimental evolution approach for individual-based modeling of microbial communities. Biophys J 106(9): in Escherichia coli. I. Adaptation and divergence during 2,000 generations. Am Nat 2037–2048. 138(6):1315–1341. 33. Nakayama M, Shigemune N, Tsugukuni T, Tokuda H, Miyamoto T (2011) Difference 57. Kim W, Levy SB, Foster KR (2016) Rapid radiation in bacteria leads to a division of of EGCg adhesion on cell surface between Staphylococcus aureus and Escherichia coli labour. Nat Commun 7:10508. visualized by electron microscopy after novel indirect staining with cerium chloride. J 58. Otsu N (1975) A threshold selection method from gray-level histograms. Automatica Microbiol Methods 86(1):97–103. 11(285-296):23–27.

E286 | www.pnas.org/cgi/doi/10.1073/pnas.1613007114 Smith et al. Downloaded by guest on September 30, 2021