Biofilm formation, including adherence to surfaces and secretion of extracellular matrix, is common in the microbial world, but we often do not know how interaction at the cellular spatial scale translates to higher-order biofilm community ecology. Here we explore an especially understudied element of biofilm ecology, namely predation by the bacterium Bdellovibrio bacteriovorus. This predator can kill and consume many different Gram-negative bacteria, including Vibrio cholerae and Escherichia coli. V. cholerae can protect itself from predation within densely packed biofilm structures that it creates, whereas E. coli biofilms are highly susceptible to B. bacteriovorus. We explore how predator-prey dynamics change when V. cholerae and E. coli are growing in biofilms together. We find that in dual-species prey biofilms, E. coli survival under B. bacteriovorus predation increases, whereas V. cholerae survival decreases. E. coli benefits from predator protection when it becomes embedded within expanding groups of highly packed V. cholerae. But we also find that the ordered, highly packed, and clonal biofilm structure of V. cholerae can be disrupted if V. cholerae cells are directly adjacent to E. coli cells at the start of biofilm growth. When this occurs, the two species become intermixed, and the resulting disordered cell groups do not block predator entry. Because biofilm cell group structure depends on initial cell distributions at the start of prey biofilm growth, the surface colonization dynamics have a dramatic impact on the eventual multispecies biofilm architecture, which in turn determines to what extent both species survive exposure to B. bacteriovorus.

Most organisms do not naturally live in isolated monocultures but rather in communities composed of many species, and microbes are no exception (1)(2)(3)(4). Bacterial communities are present ubiquitously including sinking detritus particles in aquatic environments, deep-sea hydrothermal vents, rhizosphere habitats, animal digestive tracts, fouling surfaces across human industry, and many types of chronic infections (5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15)(16)(17)(18)(19). In many of these contexts, surface attachment and growth in large cell groups, or biofilm formation, are important strategies for sequestering limited space and nutrients, as well as protecting against common biotic and abiotic threats (20)(21)(22)(23)(24)(25). Biofilms are small-scale ecosystems encased in a wide range of secreted polymeric substances that control cell-cell and cellsurface engagement (26)(27)(28)(29)(30)(31)(32). The benefits of biofilm formation have been examined in many contexts, showing their capacity for public goods sequestration, exclusion of newly arriving competitors, predation protection, and antibiotic tolerance (33)(34)(35)(36)(37). The precise functions of biofilm formation vary across species, as do the multicellular architectures that emerge from the combination of cell growth, matrix secretion, and environmental feedbacks (23,27,29,(38)(39)(40)(41)(42)(43)(44)(45)(46). In what few cases have been examined, multispecies biofilms create structures that may be distinct from those found in single species biofilms of the community constituents in isolation (47)(48)(49). Understanding the connections between biofilm architecture and microbial community ecology at different scales remains an important area for ongoing work on numerous topics, including cross-feeding relationships, diffusible and contact-mediated toxin antagonism, resilience against antibiotics in therapeutic contexts, industrial and medical surface degradation, and others (50)(51)(52)(53)(54)(55)(56).

Predation within biofilms is a broad subclass of microbial ecology that has received relatively little attention with high-resolution imaging and analysis. While many predators, such as phages, can usually attack only a small number of prey species, the ubiquitous predator Bdellovibrio bacteriovorus can be considered a generalist (36,(57)(58)(59)(60). It is not known to rely on a specific receptor for cell entry and can prey on a variety of proteobacteria (8,59,(61)(62)(63). However, the extent of its predation on biofilm-dwelling target cells appears to vary widely between prey species, and the mechanisms underlying this variability remain mostly unknown.

Studies using macroscopic measurement techniques have reported the susceptibility of several biofilm-producing species to B. bacteriovorus predation (57). For example,

Bacteria live in multispecies, spatially structured communities ubiquitously in the natural world. These communities, or biofilms, have a strong impact on microbial ecology, but we often do not know how cellular scale interactions determine overall biofilm structure and community dynamics. Here we explore this problem in the context of predator-prey interaction, with two prey species-Vibrio cholerae and Escherichia coli-being attacked by the bacterial predator Bdellovibrio bacteriovorus. We find that when V. cholerae and E. coli grow together in biofilms, the architectures that they produce change in ways that cannot be predicted from looking at each prey species alone, and that these changes in cell group structure impact the community dynamics of predator-prey interaction in biofilms.

Escherichia coli and Pseudomonas fluorescens can be largely consumed by B. bacteriovorus in laboratory biofilm culture (36). We recently documented a different outcome in Vibrio cholerae biofilms, which can protect themselves from predator exposure via their highly packed cell arrangements that occur after the prey cell groups grow beyond several hundred cells (35). Given that we have examples of prey species whose biofilm structure protects them from B. bacteriovorus predation, and other prey whose biofilms confer little protection, we were curious as to what would happen to predator-prey dynamics in multispecies prey biofilm contexts. How often and to what extent does the cellular architecture of the two species in co-culture depart from what is normally observed in monoculture? How do any of these changes influence the susceptibility of different prey species to Bdellovibrio predation and overall predator-prey population dynamics?

We chose two proteobacteria-V. cholerae and E. coli-to study these questions. Both species have been isolated from biofilms in the same environments in proximity to humans, and they can coinfect hosts (64)(65)(66)(67)(68)(69). Their respective biofilm formation mechanics have also been very well characterized. V. cholerae forms dense cell groups whose structure depends on the matrix proteins RbmA, RbmC, and Bap1, as well as vibrio polysaccharide (VPS) (27,28,40,(70)(71)(72). This highly packed cell group architecture is critical for protection from phages and from B. bacteriovorus (35,37). The cell packing required for predator protection does not occur immediately as biofilm growth begins, though, leaving smaller V. cholerae cell groups open to predation. E. coli by contrast forms biofilms with many different matrix components including cellulose, polyglycolic acid, colanic acid, Type 1 fimbriae, flagellar filaments, and curli fiber proteins (73). E. coli matrix architecture and curli protein, in particular, have been shown to confer protection against phage exposure, but other prior work has indicated that E. coli biofilms are not protected from B. bacteriovorus (57,74). Here we use single-cell resolution microscopy to examine the structure of these two prey species in monoculture and dual-culture biofilms, finding that the multispecies context causes unexpected changes in biofilm architecture that in turn alter predator-prey interaction and the overall population dynamics.

Effects for E. coli and V. cholerae. We engineered V. cholerae N16961, E. coli AR3110, and B. bacteriovorus 109 J to constitutively produce the fluorescent proteins mKate2, mKO-κ, and GFP, respectively, so that they could be distinguished by live confocal microscopy. N16961 is naturally repressed for Type VI Secretion System activity in many conditions, and it does not kill E. coli via this mechanism in our experiments (75)(76)(77). Overnight cultures of the two prey species were both normalized to OD = 1.0 before inoculating them at a 1:1 ratio into poly-dimethylxilosane microfluidic flow devices bonded to coverslip glass (see Materials and Methods). In parallel, we performed monoculture experiments in which V. cholerae and E. coli were introduced to chambers on their own. Cells were allowed to colonize the underlying glass surface in stationary conditions for 1 h, after which M9 minimal media with 0.5% glucose was introduced into the chambers at 0.2 μL/min (average flow velocity = 90 μm/s). After 48 h of growth in co-culture, groups with varying composition of both species could be found distributed throughout the chambers (SI Appendix, Fig. S1A), and prior to exposure to B. bacteriovorus, the two prey species equilibrated at frequencies of ~90% V. cholerae and ~10% E. coli (Fig. 1 and SI Appendix, Fig. S1A). Following 48 h of prey biofilm growth, we introduced B. bacteriovorus under continuous flow for 1 h, followed by a return to influx of sterile M9 media for both the dual-culture chambers and the monoculture controls. When exposed to B. bacteriovorus in a mono-species context, V. cholerae survives within cell groups that have reached high cell packing, as we have shown previously (35). 48 h following predator exposure, monoculture V. cholerae biofilms maintain net positive growth (Fig. 1 A andB). By contrast, and consistent with prior reports, E. coli biofilms in monoculture exhibit little survival in the presence of B. bacteriovorus, with viable prey biomass (i.e., E. coli cells without B. bacteriovorus inside or attached to them) falling nearly to zero 48 h after predator introduction (Fig. 1 A and B and SI Appendix, Fig. S1B) (36).

If B. bacteriovorus predates V. cholerae and E. coli independently of one another in dual prey species biofilms, we would expect residual V. cholerae biomass and elimination of most E. coli subpopulations, as seen in the prey monoculture experiments above. However, examining the population dynamics quantitatively, we found a significant increase in E. coli survival and a modest decrease in V. cholerae survival following predator introduction in the dual-species condition relative to the single species controls (Fig. 1). These results imply interactions between V. cholerae, E. coli, and B. bacteriovorus, which alter prey population dynamics in a manner that the mono-species controls cannot predict. Below we explore why E. coli fares better in dual-species biofilms against predator exposure, and why V. cholerae fares worse, relative to their respective single species prey conditions.

within V. cholerae Cell Groups. Prior to introduction of B. bacteriovorus predators, we noticed that the spatial distributions of V. cholerae and E. coli in dual-culture prey biofilms were heterogeneous on scales of 10 to 100 μm, and in some locations, there were groups of E. coli cells that had been enveloped by along the basal layers of expanding colonies of highly packed V. cholerae. 48 h following introduction of predators, most surviving E. coli were those embedded along the bottom of packed V. cholerae biofilms in this manner (Fig. 2A and SI Appendix, Figs. S2 andS4D). As previously documented, V. cholerae biofilm clusters reach a cell packing threshold that blocks predator entry and allows the interior cells to survive (35). E. coli cells enveloped within V. cholerae groups that have reached this threshold appear to gain the predation protection conferred by V. cholerae biofilm structure. To assess this point in more detail, we quantified the extent of E. coli predation and V. cholerae fluorescence in proximity to E. coli in these images; E. coli with high V. cholerae fluorescence in close proximity clearly also experienced the least predation (Fig. 2 B andC; this result is clarified with further quantitative detail in the next section). Additionally, protection of E. coli within V. cholerae cell groups was dependent on the high-density structure that wild-type V. cholerae creates. When the experiment was repeated with a mutant strain of V. cholerae that cannot produce RbmA-a matrix component required for the tight cell packing found in mature biofilms-B. bacteriovorus could freely enter and access both V. cholerae and E. coli (SI Appendix, Fig. S3 A andB). In this case, almost all cells of both species were killed off by the predators (SI Appendix, Fig. S3 C andD). In a parallel study illustrating the generality of these results, we show that E. coli embedded within expanding V. cholerae biofilm clusters are also protected from exposure to the obligate lytic phage T7 and temperate phage λ in a manner dependent on the high-packing structure of V. cholerae biofilms (78).

Although E. coli cells embedded within V. cholerae biofilms gain predator protection, it was not clear if they remain viable, or if they are able to subsequently disperse to colonize downstream locations. If not, then the protection E. coli gains from B. bacteriovorus exposure within V. cholerae cell groups would not necessarily translate to meaningful fitness gains on longer timescales. We explored this question by inducing a disturbance regimen following B. bacteriovorus predation of co-culture prey biofilms, allowing the V. cholerae and E. coli to colonize new microfluidic devices from the efluent exiting the initial chambers. We could show that almost all E. coli outside the periphery of highly packed V. cholerae colonies were killed by B. bacteriovorus exposure, and that the remaining E. coli that were protected while embedded within V. cholerae colonies remained viable. When these surviving E. coli cells were dislodged by disturbance, they could successfully colonize new locations downstream (SI Appendix, Fig. S4). The predation protection gained by E. coli cell groups embedded within highly packed V. cholerae biofilm clusters explains the increase in E. coli survivorship that we originally observed in dual culture. We note however that the V. cholerae cells within these cell groups are also protected from B. bacteriovorus predation (Fig. 2), so the results so far do not yet explain why V. cholerae survivorship declines in co-culture with E. coli under predation pressure.

Co-Culture with E. coli Can Lead to Breakdown of V. cholerae Biofilm Architecture. In the previous section, we made note of highly packed V. cholerae biofilms into which E. coli had become embedded and gained protection from B. bacteriovorus exposure. These colonies appear to behave in much the same way as monospecies V. cholerae biofilms, as they contain large continuous groups of V. cholerae in their ordered radial alignment and tight packing, albeit with pockets of E. coli along the glass substratum included as well (Fig. 3 A andB). This growth pattern was not the only kind that emerged in prey co-culture experiments, however: There was a second, qualitatively distinct colony architecture in which V. cholerae and E. coli were homogenously mixed together (Fig. 3 A andC). These colonies were disordered in comparison with the radial alignment in ordered V. cholerae cell groups, with visibly reduced cell packing density. B. bacteriovorus could enter throughout these disordered cell groups, gaining access to and killing most V. cholerae and E. coli cells within them (Fig. 3 A andC).

To study the differences between the ordered and disordered colony structure in more detail, we isolated examples of each for further analysis (Fig. 3 B andC). The relative abundance of V. cholerae looked to be somewhat lower in disordered colonies (Fig. 3 B and C, see also next section for quantitative detail). The combined cell packing of the two species together was indeed reduced in disordered colonies (Fig. 3D), falling below the threshold necessary for predation protection in ordered V. cholerae cell groups (35). An alternative but not mutually exclusive factor for predation susceptibility for V. cholerae could be abundance of E. coli in their immediate vicinity. However, measurements of the relationship between V. cholerae predation and local abundance of E. coli in the two colony types clearly show that overall colony structure is the dominant factor controlling the extent of predation by B. bacteriovorus (Fig. 3E). A similar reciprocal analysis of E. coli predation as a function of colony type and local abundance of V. cholerae gave the same outcome (SI Appendix, Fig. S5). The shift in architecture between ordered and disordered cell groups also appeared to be driven entirely by the loss of V. cholerae's ability to produce highly packed clusters, as it normally does on its own. Examining the cell packing of groups formed after 48 h supports this interpretation, as V. cholerae within-species cell packing with respect to itself (by contrast with joint cell packing of the two species together noted in Fig. 3D) shifted between ordered versus disordered biofilm types (Fig. 3F). E. coli within-species cell packing, on the other hand, was not statistically different in ordered versus disordered cell groups (Fig. 3F). On longer timescales, it should be noted, E. coli cell groups that have been enveloped within ordered V. cholerae colonies can also be driven to high within-species cell packing by the confinement imposed by V. cholerae biofilm architecture (78).

Our results thus far indicate that in mixed prey biofilms of V. cholerae and E. coli, a fraction of the E. coli along the basal surface becomes engulfed within V. cholerae cell groups with ordered high-packing structure that protects both species from B. bacteriovorus. This observation explains the improvement in E. coli predation survival in dual culture relative to monoculture biofilms. On the other hand, a fraction of the V. cholerae population becomes entangled with E. coli in well-mixed colonies that fail to develop V. cholerae's normal packing structure, instead growing into disordered, loosely assembled groups that are fully susceptible to predation. This observation clarifies why V. cholerae predation protection declines in biofilm co-culture with E. coli on a population-wide scale.

on Surface Colonization Conditions. To better understand how the highly packed versus disordered dual-species colonies originate, we ran new experiments tracking biofilm growth at 1 h intervals from the initial stages of surface colonization to clusters containing hundreds of cells at 48 h. Examples of the two cell group types were found by visual inspection at 48 h and then traced back to their initial conditions corresponding to 9 h after the start of incubation (Fig. 4 A andC). On larger spatial scales than shown in Fig. 4, both types of colony morphology could be found in close proximity (Fig. 3A and SI Appendix, Fig. S6). The two colony types could be reliably distinguished by their combined cell packing, which was systematically lower in the core regions of disordered colonies (Fig. 4 B and D, see below for temporally resolved detail). The two colony structures were also consistently different in total biovolume and species composition at 48 h, with ordered high-packing cell groups growing to larger population sizes and containing higher relative abundance of V. cholerae compared to E. coli (Fig. 4 E andF).

Though the two biofilm architectures could be quantitatively separated in several respects after they had grown to several hundred cells in size, the mean distance between V. cholerae and E. coli cells was the only factor we could discern at early time points that clearly differentiated cell groups that would later become ordered, highly packed biofilms versus disordered cell groups (Fig. 4G and SI Appendix, Fig. S7). The two cell group types in fact begin and remain on different trajectories with respect to V. cholerae-E. coli distance across all replicate colony time series acquisitions. V. cholerae that produced ordered, highly packed architecture were on average 8 μm away from the nearest E. coli cell at early time points, while those that became disordered clusters were only 1 μm from the nearest E. coli cell at early time points (Fig. 4 G andH). This observation suggests that the colonization conditions at the beginning of biofilm growth, as surface-attached cells are just starting to divide in place, are crucial for the eventual consolidation of V. cholerae packing architecture.

Reviewing the data obtained from time series of the two dual-species colony structures, we also noted an important transition that occurs in ordered high-packing V. cholerae cell groups. In our culture conditions at ~19 to 26 h, a core region of highly packed V. cholerae cells is nucleated (Fig. 4I). This process creates a secondary front of structural consolidation that lags behind the outermost, less densely packed growth front that represents the interface between the growing colony and the surrounding liquid medium (Fig. 4B). This secondary front bounding the central core biofilm region has been observed previously (28,29,79,80), and it corresponds to the portion of the cell group with suficiently high cell density to provide predator protection (35). In disordered colonies containing well-mixed V. cholerae and E. coli, the nucleation of this high-density core never occurs. This difference between the two cell group types can be tracked quantitatively via the cell packing of the inner core of each (Fig. 4J). In ordered colonies, once the highly packed core was initiated, it was stable over time, and interruption of this core nucleation process only occurred if V. cholerae and E. coli cells happened to begin growing in close proximity from the start of biofilm formation. Allowing V. cholerae to grow on its own for 48 h, followed by invasion of E. coli into the biofilm environment, never led to any observable disruption of highly packed V. cholerae groups. Introduced planktonic E. coli cells could not invade V. cholerae biofilms and were completely susceptible to predation if B. bacteriovorus was later added to the system (SI Appendix, Fig. S8).

Our results here highlight critical points for the production of V. cholerae biofilm structure with its characteristic packing and cell alignment architecture (27-29, 34, 79, 80), which in turn are necessary for protection from B. bacteriovorus (35). If V. cholerae cells are suficiently isolated during early biofilm formation, they can grow, divide, and secrete biofilm matrix components that effectively coordinate their normal architecture. However, if V. cholerae cells are too close to E. coli at the start of biofilm formation, the two species become entangled in the process of growth and division in a manner that interrupts the longer-term production of cell group architecture that V. cholerae normally produces on its own. The disruption of the structure that V. cholerae typically produces suggests that if E. coli is in close enough proximity at early stages of growth, V. cholerae either does not produce biofilm matrix normally, or it does produce matrix, but with a disruption of the cell-cell orientation and matrix localization that V. cholerae would obtain on its own.

To begin parsing these possibilities, we performed additional biofilm co-culture experiments in which V. cholerae produced a FLAG-labeled version of one of its primary matrix proteins RbmA, RbmC, or Bap1. These experiments confirmed that V. cholerae is indeed producing biofilm matrix in direct proximity to E. coli (SI Appendix, Fig. S9 A-C). We also found, surprisingly, that RbmC localizes substantially above background around E. coli cells (SI Appendix, Fig. S9 B andD). This result was supported by experiments in which cell-free V. cholerae supernatant containing RbmC-FLAG was added to E. coli biofilms growing in monoculture (SI Appendix, Fig. S10). RbmC is a diffusible matrix protein that participates in the internal architecture of V. cholerae clonal cell groups but also, importantly, contributes to binding of the cell groups to the underlying surface (27,29,71). The mechanical details of binding between the group and surface are important for the transition from lateral expansion of flat monolayers to the extension of V. cholerae biofilms into 3D space and subsequent packing architecture (29,81). Accumulation of RbmC around E. coli does not appear to diminish the amount of RbmC that V. cholerae accumulates around itself relative to monoculture conditions (SI Appendix, Fig. S9E). Nevertheless, the localization of this matrix protein around E. coli in close proximity to V. cholerae may potentially contribute to the disruption of normal V. cholerae cell group architecture when the two species begin biofilm growth directly adjacent to each other. The precise physical and biochemical details of how multispecies biofilm architecture quantitatively and qualitatively departs from clonal biofilm architecture will be an important area for future work.

via its Influence on Biofilm Architecture. Our results above suggest that the average distance between V. cholerae and E. coli cells at the start of biofilm growth should directly determine the relative occurrence of ordered, highly packed V. cholerae groups that envelope pockets of surface-attached E. coli as they expand, as opposed to disordered dual-species cell groups. As highly packed V. cholerae cell groups are protected from B. bacteriovorus, while disordered colonies are not, our observations lead to the ecological prediction that initial surface colonization conditions, via their impact on the relative proportion of highly packed versus disordered dual-species colonies, can substantially change the population dynamics of predator-prey interaction. In other words, the initial surface coverage should indirectly determine the overall impact of predation on survival of both species via its direct impact on colony structure development. We tested this prediction by inoculating two sets of two-species chambers with relatively low or high surface colonization density (20% versus 60% surface coverage, respectively, SI Appendix, Fig. S11A). Low or high initial density alters the distributions of distance between V. cholerae and E. coli cells (SI Appendix, Fig. S11B), in turn leading to large amounts of highly packed V. cholerae colonies containing small numbers of E. coli (low initial density), or mostly disordered, mixed colonies containing more evenly distributed V. cholerae and E. coli (high initial density) (Fig. 5 A andB).

As anticipated, low versus high initial surface occupation led to distinct survival outcomes for both prey species after the introduction of B. bacteriovorus. In the low initial density condition, the majority of V. cholerae, as well as the smaller groups of E. coli embedded in their packed biofilms, survive predator exposure (Fig. 5 A, C, andE). By contrast, in the high initial density conditions predominated by disordered groups of V. cholerae and E. coli, a far greater fraction of the dual-species prey community is killed off by B. bacteriovorus (Fig. 5 B, D andE ). Taken together, these data are consistent with our prediction that surface colonization conditions determine the relative amounts of ordered versus disordered biofilm cell groups of E. coli and V. cholerae, which in turn govern the survival of both prey species under B. bacteriovorus predation.

Exploration of multispecies biofilm communities using live high-resolution imaging is crucial to understanding microbial ecology at the spatial scale on which cell-cell interactions occur (11,20,23,(82)(83)(84)(85)(86)(87)(88)(89)(90)(91). Here we tracked the spatial population dynamics of the bacterial predator B. bacteriovorus in dual-species prey biofilms of V. cholerae and E. coli, finding that the survival rates of both prey species are altered, but in opposite directions, when they are growing together. V. cholerae produces biofilm cell clusters that reach a cell packing density threshold past which B. bacteriovorus cannot enter, protecting the prey within. E. coli can become enveloped along the basal layers of these highly packed structures, co-opting predator protection from V. cholerae and increasing E. coli survival relative to when growing on its own. By contrast, in dual-species biofilms, a fraction of V. cholerae becomes entangled with E. coli early during biofilm growth, leading to an alternate structure that is more homogeneously mixed, disordered, and loosely packed. These disordered cell groups do not block predator cell entry, and all prey within them are killed by B. bacteriovorus. As a result of these biofilm structural dynamics, V. cholerae survival decreases in co-culture with E. coli relative to when growing on its own. At any given location, which of these two alternative cell group structures emerge depends on the initial distance between V. cholerae and E. coli cells that have attached to the underlying surface. Surface colonization patterns therefore determine the relative occurrence of predation-protected cell groups versus susceptible cell groups and the overall rates of B. bacteriovorus predator survival for each prey species.

This study makes explicit that the cellular arrangement and tightly packed structure of clonal V. cholerae groups can operate as a type of public good (39,92) that confers predator protection to the cells within [among many other benefits (30, 33-35, 55, 74, 93-97)]. Other species-here, E. coli, whose mono-species biofilms are susceptible to B. bacteriovorus-can take advantage of this protective architecture when small groups of them become enveloped by expanding, highly packed biofilms of V. cholerae. By contrast, if too many E. coli cells are present in close enough proximity to V. cholerae at the start of biofilm growth, then V. cholerae cannot initiate its normal cell group structure, and the public good benefit of predation protection completely breaks down in that location. It is notable that the spatial architecture of biofilm-producing bacteria can manifest as a public good that is exploitable across species in this manner. In this case, the stability of V. cholerae cooperative architecture depends on the initial surface population density, which determines whether V. cholerae cell lineages have enough space to nucleate the highly packed core regions of expanding biofilm clusters before encountering cells of other species. Though distinct in mechanistic detail, this example should fall under related social evolution principles as other kinds of microbial cooperation that provide benefits in a distance-dependent manner. Recent work has highlighted in detail how the population dynamics and evolutionary stability of this class of cooperative behavior depend on the spatial range of cooperative sharing, the population/ community composition, and spatial cell arrangements during early biofilm growth (33,34,39,93,96,98).

The interplay of V. cholerae, E. coli, and B. bacteriovorus in co-culture emphasizes that the population dynamics of different species in a community can depend quite strongly on the cellular resolution details of biofilm structure, which in turn can differ in unexpected ways between mono-species and multispecies systems. In recent years microbiologists have made tremendous strides in understanding the cellular and molecular nuances of biofilm architecture and their relationship to microbial ecology and evolution. By necessity for tractability in many cases, much of this work has focused on one species at a time. Our experiments here highlight how new and interesting questions about the drivers of biofilm structure, and the relationship between biofilm structure and community ecology, can arise from modest increases in complexity with multispecies systems.

Here, it appears as though E. coli-if adjacent to V. cholerae at the start of biofilm growth-may interfere with normal localization of at least one component of the V. cholerae matrix, which could then contribute to the qualitative differences in ordered versus disordered architectures that appear later during biofilm growth.

The connection between initial surface coverage and multispecies prey biofilm architecture, and the additional connection between biofilm architecture and predator exposure, together lead to an interesting dependence between early biofilm growth conditions and predator-prey ecology. It would be fruitful to explore how and when these relationships generalize to other species combinations and biofilm environmental growth conditions with increasing ecological realism. Where prior studies have analyzed multispecies biofilms at high resolution, they have also indicated important consequences for community structure and environmental impacts (11,12,22,47,48,(99)(100)(101)(102)(103). A notable recent example examined the detailed structure of multispecies biofilm communities growing as plaque in dental caries (48). Kim et al. showed that Streptococcus mutans forms consistent spatial arrangements in biofilm co-culture with other oral microbiota species. In this case, S. mutans consistently produces core clonal regions, around which form layers of non-mutans streptococci followed by non-streptococci. The metabolic activity of S. mutans within the inner regions of these multispecies biofilms caused low local pH that could recapitulate the rapid demineralization of enamel that occurs during development of caries in vivo.

Our work here highlights how the details of early surface colonization conditions can cascade into qualitative differences in subsequent biofilm architecture and ecological dynamics. This result points to several goals for future work. Any phenotypes that alter surface exploration or settling patterns, including gliding and twitching motility, as well as any positive or negative interactions within and between species-for example, via shared adhesin production, metabolite trophic interaction, or toxin secretion-could also cascade to major differences in biofilm spatial architecture. Differences in environmental topography and the orientation of nutrient supply, which may often derive in natural environments from the underlying surface rather surrounding liquid, should also be pursued to gain a fuller picture of how the subtleties of biofilm growth in realistic environments impact community structure. Novel experimental systems that implement these increases in ecological realism while maintaining access by high-resolution imaging will be important platforms for further study.

V. cholerae strain N16961, E. coli strain AR3110, B. bacteriovorus strain 109J, and their fluorescent derivatives were cultivated using standard microbiology techniques.All biofilm experiments were performed with syringe pump-driven microfluidic devices fabricated using polydimethylsiloxane bonded to glass coverslips. Image data were gathered using Zeiss 880 and 980 point-scanning confocal microscopes and analyzed using the freely available BiofilmQ framework (104).A detailed description of all methods is provided in the SI Appendix, Materials and Methods. Data, Materials, and Software Availability. Raw numerical data corresponding to all main text and SI figures are included in SI Appendix, Data S1.