Bacterial species with different nanocolony morphologies have distinct flow-dependent colonization behaviors

Significance

Bacterial populations interact with dynamic fluid flows in their natural environments. Reports from bacterial heart valve infections or urinary tract infections present a surprising observation: Bacterial populations robustly colonize areas with the highest shear rates that would naively be assumed to wash away bacteria. Here, we recapitulate this counterintuitive colonization in microfluidic environments for two pathogens, Staphylococcus aureus and Enterococcus faecalis. Flow-dependent colonization by S. aureus and E. faecalis is mediated by distinct mechanisms that depend on each species’ multicellular nanocolony morphologies: transport of dispersal signaling molecules for clustered S. aureus and mechanical forces for linear chains of E. faecalis. These results suggest that nanocolony morphologies have previously unappreciated costs and benefits in different environments, like those introduced by flow.

Abstract

Fluid flows are dominant features of many bacterial environments, and flow can often impact bacterial behaviors in unexpected ways. For example, the most common type of cardiovascular infection is heart valve colonization by gram-positive bacteria like Staphylococcus aureus and Enterococcus faecalis (endocarditis). This behavior is counterintuitive because heart valves experience high shear rates that would naively be expected to reduce colonization. To determine whether these bacteria preferentially colonize higher shear rate environments, we developed a microfluidic system to quantify the effect of flow conditions on the colonization of S. aureus and E. faecalis. We find that the preferential colonization in high flow of both species is not specific to heart valves and can be found in simple configurations lacking any host factors. This behavior enables bacteria that are outcompeted in low flow to dominate in high flow. Surprisingly, experimental and computational studies reveal that the two species achieve this behavior via distinct mechanisms. S. aureus grows in cell clusters and produces a dispersal signal whose transport is affected by shear rate. Meanwhile, E. faecalis grows in linear chains whose mechanical properties result in less dispersal in the presence of higher shear force. In addition to establishing two divergent mechanisms by which these bacteria each preferentially colonize high-flow environments, our findings highlight the importance of understanding bacterial behaviors at the level of collective interactions among cells. These results suggest that distinct multicellular nanocolony morphologies have previously unappreciated costs and benefits in different environments, like those introduced by fluid flow.

Bacterial populations experience a range of complex environments in their natural settings (1). For example, within a host, bacteria must contend with fluid flows such as those found in the cardiovascular system (2). Fluid flow can significantly impact how bacteria interact with one another or with the surfaces that they colonize (3–11). Recent studies have begun to examine the effects of flow on single-cell bacterial behaviors like growth (12, 13) and adhesion (14–19). For example, catch bonds can enable individual bacterial cells to adhere to epithelial surfaces more strongly in flow (20, 21). However, these single-cell mechanisms often reflect specific interactions between a bacterial adhesin and a host ligand and are not widely conserved. Beyond the single-cell level, but before reaching larger communities of 100 s of cells (typically referred to as “microcolonies”), bacteria form smaller multicellular collectives that we term “nanocolonies”. These nanocolonies are clonal groups of multiple cells that can have different geometries like chains, clusters, or rosettes. Bacterial nanocolonies have been observed in physiological settings like human infections, but the effects of flow at this scale are underexplored.

One example of a counterintuitive bacterial behavior in the presence of flow comes from clinical reports of infective endocarditis. Infective endocarditis is an infection in the heart, where bacteria are most commonly reported to colonize heart valves (18, 22–25). These infections are difficult to treat and even with the use of antibiotics surgery is often needed to replace infected valves (26). From the perspective of fluid mechanics, the valves have the narrowest cross-sectional areas in the heart, causing them to have the fastest flow speed (2). Other examples include reports from urinary tract infections where bacteria colonize well despite the presence of strong shear forces. These observations raise an important question: Higher flow speeds would naively be thought to reduce surface colonization, so how might bacteria preferentially colonize host environments with high flow conditions?

Here, we sought to directly examine the effects of flow on the surface colonization of two bacterial species that often colonize high shear host environments. Specifically, we studied two different bacterial species commonly implicated in endocarditis infections, Methicillin-resistant Staphylococcus aureus (MRSA) and Enterococcus faecalis, and imaged their surface colonization within microfluidic devices. Our results demonstrate that these bacteria preferentially colonize surfaces in high shear conditions, even when the surface is abiotic, suggesting that this counterintuitive behavior does not require specific host factors and is driven by the bacteria themselves. Elucidating how the two species preferentially adhere in high shear environments reveals two different mechanisms that rely on their distinct nanocolony morphologies and are driven by either transport of signaling molecules or mechanical responses to flow.

Results

Bacterial Populations Exhibit Counterintuitive Adhesion Behaviors in Microfluidic Studies.

To assess the impact of flow on bacteria in the absence of host factors, we examined the colonization of glass coverslips for both S. aureus (USA300 MRSA) and E. faecalis (OG1RF) in untreated microfluidic channels (Fig. 1A). For all experiments described below, cells were seeded into the channel and allowed to adhere to the glass coverslip before we flowed in sterile media, such that no new cells were introduced to the system once flow began. In any given flow system, there is a flow rate (volume/time) and a typical speed (flow rate divided by the cross-sectional area of the flow). The fluid speed is zero at the boundaries, so cells at the surface typically experience a flow (or velocity) profile that increases with distance from the wall. Hence, the cells attached to a wall experience a shear rate (units 1/s). For a given configuration, higher flow rates correspond to higher shear rates. The shear rate in heart valves is estimated to be roughly 100/s (estimated from (27, 28) as detailed in SI Appendix). For each species, we thus initially examined two shear rates: a low shear rate of 40/s (low flow) and a high shear rate of 400/s (high flow). Despite the naive expectation that higher shear rate should hinder colonization, by the end of our experiments we observed that more bacteria colonized the surface for both S. aureus and E. faecalis (Fig. 1). Performing similar experiments on other bacterial species like Escherichia coli, and Streptococcus pneumoniae revealed that these species do not increase colonization in high flow, suggesting that this behavior is specific to certain species (SI Appendix, Fig. S1 and Movies S1–S4).

Fig. 1.

Counterintuitive final cell population in low versus high flow. (A) Schematic of microfluidic chamber used for all flow experiments. A PDMS microfluidic channel, channel dimensions shown, is bonded to a glass coverslip. Following cell loading, a syringe pump is turned on, flowing fresh media through the channel. Cells attached to the coverslip are imaged. (B) Analysis of doubling times for each S. aureus MRSA and E. faecalis in low (red) and high (blue) flow conditions. (C) Representative phase contrast images of S. aureus at 0, 3, and 6 h of low flow (shear rate 40/s). (D) Representative phase contrast images of S. aureus at 0, 3, and 6 h of high flow (shear rate 400/s). (E) Fold change of percent area covered of both low (red) and high (blue) flow experiments for S. aureus. (F) Representative phase contrast images of E. faecalis at 0, 3, and 6 h of low flow (shear rate 40/s). (G) Representative phase contrast images of E. faecalis at 0, 3, and 6 h of high flow (shear rate 400/s). (H) Fold change of percent area covered of both low (red) and high (blue) flow experiments for E. faecalis. Image (Scale bars, 10 μm.) Error bars are SEM.

Closer examination of our timelapse movies of S. aureus and E. faecalis indicated that the attachment of new cells to the surface was relatively rare in both high flow and low flow for both species (http://www.pnas.org/lookup/doi/10.1073/pnas.2419899122#supplementary-materialsMovies S5–S8). Furthermore, analysis of single-cell doubling time showed no significant difference between the low and high flow conditions for either species, indicating that the differences in colonization cannot be explained by differences in growth rate (Fig. 1B and SI Appendix, Table S1). Since the differences between high and low flow could not be explained by differences in attachment or growth, we focused our subsequent studies on understanding mechanisms driving differences in detachment.

To better understand flow-dependent S. aureus colonization, we examined colonization dynamics and found that for the first several hours surface colonization increased at a similar rate in both low and high flow. During these early timepoints the bacteria formed small nanocolonies with a clustered morphology. After roughly 3 h, however, a striking difference between the low and high flow conditions began to emerge. At these later timepoints, S. aureus continued to colonize more and more of the surface in the high flow conditions. But in the low flow condition, S. aureus began to disperse from the surface, leading to a decrease in coverage and smaller nanocolony clusters. Image analysis of the fraction of the surface area covered by bacteria confirmed that in high flow colonization proceeded to monotonically increase, but that in low flow, colonization increased for the first 3 h and then exhibited increasing dispersal for the rest of the experiment (Fig. 1 C–E).

The dynamics of E. faecalis colonization revealed a different pattern from that of S. aureus. Throughout the experiments, we found that E. faecalis grew as nanocolonies with a linear chained morphology (Fig. 1 F and G). In both low and high flow, E. faecalis surface colonization increased mostly monotonically, but the rate of the increased surface colonization was consistently higher in high flow than in low flow conditions (Fig. 1 F and H).

The glass surface used in the experiments above is informative in that it clearly lacks host-derived factors, but leaves open the question of whether the behaviors we observed occur on physiologically relevant surfaces. To address this question we reexamined flow-dependent S. aureus and E. faecalis colonization after coating the surfaces of our channels with fibrinogen, a major component of the extracellular matrix found on mammalian cell surfaces (29). Both S. aureus and E. faecalis appeared to adhere better to fibrinogen-coated surfaces than to glass. But for both species, the flow dependence of the colonization was the same regardless of surface composition. Specifically, S. aureus exhibited a peaked colonization profile in low flow but monotonic profile in high flow and E. faecalis exhibited monotonic profiles for both low and high flow with a higher rate of colonization in high flow (SI Appendix, Fig. S2). Together, these experiments suggest that the surface colonization of both E. faecalis and S. aureus is counterintuitively greater in high flow on two different surfaces. While changes in detachment rate appear to underlie both behaviors, the two species respond to flow differently, as they exhibit distinct colonization dynamics.

Flow-Dependent Competition Dynamics Between S. aureus and P. aeruginosa.

Could preferential surface colonization provide a competitive advantage in flow? S. aureus and Pseudomonas aeruginosa are two bacterial pathogens that are often found together in polymicrobial infections. Previous studies have established that in the absence of flow P. aeruginosa can coexist with S. aureus, though their dynamics often lead to preferential growth of P. aeruginosa (30, 31). However, the interactions between these species have not been previously examined in higher-flow environments. We found that unlike S. aureus, P. aeruginosa exhibited the intuitive response to flow in that increased flow led to decreased surface colonization (S. aureus, Movie S9, P. aeruginosa, Movies S10 and S11). We thus mixed S. aureus and P. aeruginosa and examined their surface colonization dynamics in our system. We first studied their dynamics in a minimal flow condition (shear rate of 2/s). This minimal flow rate is the lowest flow rate we found sufficient to clear unattached cells, enabling us to image the system, provide fresh media, and limit new cell attachment with minimal impacts from flow. In these minimal-flow cocultures P. aeruginosa preferentially colonized over S. aureus and quickly came to dominate the channel (Fig. 2 A and B and Movie S12), suggesting that the preferential growth of P. aeruginosa dominates. In contrast, observing the cocultures in high flow conditions (shear rate of 400/s) showed the opposite effect. In these conditions, S. aureus preferentially colonized the surface compared to P. aeruginosa, leading to S. aureus ultimately dominating the coculture (Fig. 2 C and D and Movie S13) and reducing P. aeruginosa levels even lower than those found in high flow in monoculture (Movie S11). Thus, even using the same bacterial strains in the same media, one species can dominate in minimal flow while a different species dominates in high flow. These studies provide proof-of-concept that flow-dependent colonization may provide an adaptive benefit for S. aureus, enabling it to successfully compete with bacteria that can dominate S. aureus in other conditions.

Fig. 2.

Flow-dependent colonization dynamics in cocultures of S. aureus and P. aeruginosa. (A) Representative images showing merged phase contrast, red, and green channels at 0, 3, and 6 h from a coculture experiment of S. aureus (green fluorescence) and P. aeruginosa (red fluorescence) in the minimal flow condition. (B) Population fraction over the 6 h experiment run for the minimal flow condition for S. aureus (green) and P. aeruginosa (red). (C) Representative images showing merged phase contrast, red, and green channels at 0, 3, and 6 h from a coculture experiment of S. aureus (green fluorescence) and P. aeruginosa (red fluorescence) in the high flow condition. (D) Population fraction over the 6 h experiment run for the high flow condition for S. aureus (green) and P. aeruginosa (red). (Scale bars for flow images are 10 μm.) Error bars on graphs are SEM.

The S. aureus Flow Response Is Mediated By Signaling Molecule Transport.

To gain insight into the mechanisms by which flow might affect colonization, we first considered the possibility that the bacterial cells might be responding to changes in their chemical environments. Bacteria are known to produce and sense various molecules whose concentration changes can activate or repress signaling pathways that can alter the expression of factors like adhesins (32) that influence colonization (23). To test whether such chemical changes could explain the behaviors we observed for S. aureus and E. faecalis, we generated “conditioned” media where each bacterial species was grown overnight, sterile-filtered to remove intact bacterial cells, and diluted 1:1 with fresh media to ensure access to the nutrients necessary for growth. We then examined bacterial colonization in the presence of high shear rate conditions (high flow) as above, but switched the media halfway through the experiment from fresh to conditioned media. In the presence of high flow, conditioned media caused opposite effects on S. aureus and E. faecalis. Specifically, conditioned media decreased colonization for S. aureus, leading to colonization in high flow that more closely resembled S. aureus colonization in low flow (Fig. 3 A and B). Meanwhile, for E. faecalis, conditioned media increased colonization, exacerbating the difference between high and low flow (Fig. 3 C and D). We found consistently similar results when repeating the experiments with conditioned media present throughout the experiment (SI Appendix, Fig. S3). Thus, though both S. aureus and E. faecalis better colonize surfaces at higher flow rates in the absence of conditioned media, the addition of conditioned media has different impacts on the flow-dependent colonization of these two species.

Fig. 3.

High flow transports signaling molecules S. aureus uses for dispersal. (A) Phase contrast images of S. aureus high shear rate experiment at 0, 3, and 6 h. At 3 h, the flow was switched to conditioned media. (B) Fold changes of S. aureus percent area covered of high flow + conditioned media (green) experiments, with low (red) and high (blue) flow experiments shown for comparison. (C) Representative phase contrast images of E. faecalis high shear rate experiment at 0, 3, and 6 h, with the switch to conditioned media at 3 h. (D) Fold changes of E. faecalis percent area covered of high flow + conditioned media (green) experiments. Low (red) and high (blue) flow experiments shown for comparison. (E) Equations for the transport-dependent colonization model and schematic of the effects of high flow transport of signaling molecules. The first equation describes how the number of cells, n, changes over time t as a function of the growth rate g and the signaling constant β, concentration of autoinducer c, and half maximal concentration of the autoinducer, K. We use a Hill function to describe the QS dynamics with a hill coefficient, h. The second equation describes autoinducer concentration, c, over time as a function of the production rate per cell, p, and transport q, representing our flow. Depending on the sign of the β term, the response under flow differs. (F) Model prediction of dispersal in low flow with a negative β term. (G) Model prediction of increased cell number in low flow with a positive β term. (H) Phase contrast images of S. aureus at 0, 3, and 6 h of high flow with media supplemented with 50 nM AIP for the entire experiment. (I) Fold change of S. aureus percent area covered of high + 50 nM AIP (light blue) flow experiments, with low (red) and high (blue) flow experiments shown for comparison. (J) Phase contrast images of E. faecalis high shear rate (400/s) experiment at 0, 3, and 6 h with media supplemented with 50 nM GBAP for the entire experiment. (K) Fold change of E. faecalis percent area covered of high + 50 nM GBAP (light blue) flow experiments, with low flow (red) and high flow (blue) experiments shown for comparison. Image (Scale bars, 10 μm.) Error bars are SEM.

One way in which bacteria are known to respond to conditioned media is through quorum sensing (QS). In QS, bacteria both secrete and sense signaling molecules called autoinducers. In well-mixed environments, autoinducer concentration correlates with bacterial cell density, thereby enabling bacteria to drive collective behaviors. Flow can influence QS and other signaling systems by affecting small molecule transport, here defined as the movement of signaling molecule into or out of the bacterial biofilm by diffusion or advection by flow (33–36). To determine whether the interactions between flow and QS could explain the flow-dependent surface colonization of S. aureus and E. faecalis, we developed a model in which the number of cells (n) on a surface changes as a function of cell growth and autoinducer signaling and autoinducer concentration (c) changes as a function of the number of cells and transport (Fig. 3E). In this model autoinducer concentration can drive cell detachment and dispersal (if the QS parameter β is negative) or promote increased cell adhesion, (if β is positive). Further, autoinducer concentration does not affect growth because we observed no changes in growth rate in our experiments (Fig. 1B). Equations and further discussion of the transport-dependent colonization model can be found in supplemental materials.

Exploring the parameters of our transport-dependent colonization model revealed that whether QS positively or negatively influences if a cell is attached to the surface (the sign of the QS constant, β) was sufficient to explain some, but not all, of the dynamics we observed experimentally. In all cases, the system starts with low cell numbers, such that the effect of QS becomes more pronounced at later times. Regardless of the sign of β, increased shear rate led to more autoinducer transport (represented by the transport term, q) and thus less autoinducer accumulation. When β is negative and shear rate is low, growth rate initially dominates so cell numbers increase at early timepoints, but as cell density increases, autoinducer concentration, c, increases and our QS term decreases, describing an experimental case where cell attachment is inhibited to an extent that eventually overwhelms growth and decreases cell numbers. In contrast, when β is negative and shear rate is high, autoinducer is transported from the cells. This leads to low autoinducer concentrations such that QS-induced dispersal never overtakes growth and cell number steadily increases throughout the experiment (Fig. 3F). These results suggest that the effects of shear rate on QS can explain the behaviors we observe for S. aureus in which colonization initially increases in both high and low flow but then later diverges, continuing to increase in high flow but decreasing in low flow. To further support this model we examined S. aureus colonization in intermediate flow rates and found good agreement between our model predictions and experimental findings (SI Appendix, Fig. S4).

We also sought to determine whether our transport-dependent colonization model can explain the colonization behavior of E. faecalis. Since E. faecalis colonization is not seen to decrease over time in any conditions, the negative β model that works for S. aureus does not work for E. faecalis. QS is known to stimulate E. faecalis adhesion and matrix production (37) and we observed that conditioned media enhanced colonization, which could arise from E. faecalis QS stimulating adhesion (Fig. 3 C and D). We thus explored colonization dynamics in our model in the regime where β is positive such that an increase in autoinducer concentration would increase the number of cells. In this case, we found that cell number always increased over time but did so more rapidly in low flow than in high flow because autoinducer accumulates more with a low flow, and thus a lower transport rate q. This leads to a larger QS term in the ODE describing cell number and an overall increase in total number of cells in low flow compared to high flow (Fig. 3G). In contrast, our experimental results demonstrated that E. faecalis cells colonized more in high flow than in low flow (Fig. 3 F–H). These results suggest that the transport-dependent colonization model cannot explain the flow dependence of the colonization behavior of E. faecalis.

To experimentally test our predictions that QS can explain the effects of flow on the colonization of S. aureus but not E. faecalis, we synthesized the primary autoinducer signaling molecule from each species: autoinducing peptide (AIP), which activates agr QS in S. aureus, and gelatinase biosynthesis–activating pheromone (GBAP), which activates fsr QS in E. faecalis. Addition of AIP to S. aureus decreased colonization and caused cell coverage in high flow to more closely resemble that in low flow (Fig. 3 H and I and SI Appendix, Fig. S5). These results are consistent with previous reports that AIP signaling downregulates adhesin expression in S. aureus (38–40). Furthermore, genetically disrupting AIP production with an agrB mutant (41, 42) eliminated S. aureus dispersal in low flow (SI Appendix, Fig. S6). By contrast, GBAP addition to E. faecalis under flow conditions increased colonization, expanding the colonization difference between the low and high flow conditions (Fig. 3 J and K and SI Appendix, Fig. S5). Together, our findings indicate that the effects of flow on AIP QS explain the paradoxical colonization behavior of S. aureus but that a different mechanism must be present in E. faecalis.

Mechanics of Chained Cells in Flow Drive E. faecalis Colonization Dynamics.

How could high flow enhance E. faecalis colonization independently of QS? One striking difference between S. aureus and E. faecalis is that they have distinct growth patterns that result in different nanocolony morphologies. S. aureus cells divide in alternating division planes (43), resulting in rounded nanocolonies of closely clustered cells. By contrast, E. faecalis cells divide along a single plane (44), resulting in linear microcolonies of cell chains. The mechanical effects of flow on this chained nanocolony morphology could conceivably help to explain the flow dependence of E. faecalis. Specifically, in the presence of flow, a cell chain that is anchored to the surface at one end will experience a torque produced by the shear force from the flow. The higher the flow rate, the larger this shear force becomes, increasing the torque and pushing the chain closer to the surface.

We thus sought to test whether E. faecalis colonization depends on shear force. Shear force can be increased independently of shear rate by increasing the viscosity of the fluid, so we increased the viscosity of the bacterial media roughly fivefold by supplementation with 10% ficoll. Direct measurement of doubling time in the flow channel established that ficoll addition did not significantly affect cell doubling times (SI Appendix, Table S1). Meanwhile, increasing viscosity did significantly enhance E. faecalis colonization at equivalent shear rates, indicating that E. faecalis colonization does depend on shear force (Fig. 4 A and B). We also examined whether increased shear force pushes E. faecalis chains more toward the surface (as illustrated in Fig. 4C). We observed that cell chains in high shear force conditions were more sharply in focus in our images, indicating that they were closer to the focal plane positioned at the surface (Fig. 4D). To quantify this effect, we measured the fraction of total cells in focus (SI Appendix, Fig. S7). Relative to low flow, increasing shear force by either increasing shear rate or viscosity resulted in a significant increase in the fraction of cells in focus (Fig. 4E). We also noted that the fraction of cells in focus increased over time during each experiment (Fig. 4E). This result is consistent with our hypothesis, as cell chains elongate during the experiment and longer chains experience higher torque.

Fig. 4.

Mechanical dynamics of chains in flow lead to more cells attached and in focus in E. faecalis. (A) Phase contrast images of an E. faecalis low shear rate experiment at 0, 3, and 6 h with 10% ficoll added to the media. (B) Fold changes of E. faecalis percent area covered of low flow experiments with 10% ficoll (purple). Low flow with 0% ficoll (red) is shown for comparison. For flow experiment images, (Scale bars, 10 μm.) (C) Schematic demonstrating chain dynamics and focus plane for low versus high flow. (D) Example cropped phase contrast images of chains in low (Top) and high (Bottom) flow at t = 2 h, demonstrating differences in focus. (E) Fold change of fraction in focus for E. faecalis experiments of low flow (red), low flow with 10% ficoll (purple), and high flow (blue). (Scale bars in A are 10 μm and in D are 1 μm.) For graphs, error bars are SEM.

To mechanistically explore the mechanical responses of E. faecalis chains to flow, we developed a mechanics-dependent colonization model (Fig. 5A). This is a 2D agent-based model in which cells are agents that can be connected to one another to form chains. Chains are allowed to grow, as growth continues during our experiments, and each cell can independently attach or detach from the surface. The attachment probability depends on a cell’s distance from the surface, while detachment probability increases if neighboring cells are detached. Moreover, the connection between two neighboring cells in a chain can break in a manner dependent upon a tension term that describes the spring stretching. The spring stretches depending on the forces acting upon the cells (such as the drag force from flow, described below) and with a large enough tension the spring will break. A chain is considered to be attached to the surface if any of its cells are attached. To mimic the dynamics observed experimentally, chains begin with an origin cell attached to the surface and subsequent cells are attached at the end of the chain by a stiff spring with a bending energy that keeps the cells in a relatively straight line. An additional force acting on the chains is a drag force that acts in the horizontal direction to recapitulate the shear force from flow. We found that in our simulations, increased flow rate increased the number of cells that attached to the surface (Fig. 5B), recapitulating our experimental results with E. faecalis. Our simulations also revealed that the effect of flow on E. faecalis colonization could be explained by the impact of flow on the outcomes of chain breakage events. When a chain breaks, the part of the chain proximal to the ancestral origin cell has the same probability of remaining attached regardless of the flow rate (Fig. 5C). This portion of the chain has a high probability of being attached in all cases (Fig. 5D). In contrast, the part of the chain distal to the origin cell has a higher probability of remaining attached in high flow than in low flow (Fig. 5C). This results from the higher drag force pushing the cell chain closer to the surface and thus increasing the probability that cells in the distal part of the chain will be attached to the surface (Fig. 5D).

Fig. 5.

Mechanics-dependent colonization model and E. faecalis ΔatlA mutant demonstrate increased colonization in higher flow. (A) Schematic of the forces and attachment/detachment dynamics implemented in our mechanics-dependent colonization model. (B) Comparison of total cell coverage from the model in low and high flow (red and blue, solid line) compared to experimental E. faecalis runs (red and blue, dashed line). (C) Schematic depicting possible chain dynamics following a breakage event in low (Upper two) and high (Lower two) flow. (D) Mechanics model prediction of chain breakage dynamics in low (red) and high (blue) flow. After breakage, the origin and distal attachments were recorded for 100 simulated chains. (E) Phase contrast images depicting breakage events from low flow (Top) – only the origin cell is still attached, and high flow (Bottom) – both the distal and origin cells are attached. Images shown are two consecutive time points (Δt = 5 min). (F) Fraction attached for both origin and distal in high and low flow experiments. Flow videos were analyzed for breakage events throughout the 6 h experiment. For n = 30 chains in both high and low flow, the outcome after a breakage—anchor or distal attachment—was counted. We note that chains could break and have both origin and distal attach or could completely detach from the surface, leading to neither origin nor distal attachment. (G) Representative phase contrast images of a low flow E. faecalis OG1RF ΔatlA experiment at 0, 3, and 6 h. (H) Representative phase contrast images of a high flow E. faecalis OG1RF ΔatlA experiment at 0, 3, and 6 h. (I) Fold change of total cells for our mechanics-dependent colonization model. Original model runs for low and high flow are shown in dashed lines for comparison. The breakage tension in the model was increased to explore longer chain dynamics, mimicking the ΔatlA mutant and leading to more cells seen in both low (red) and high (blue) flow model runs. (J) Fold change of the percent area covered of both our WT E. faecalis (low and high flow, red and blue dashed) and a mutant ΔatlA strain (low and high flow, red and blue) that alters septum cleavage leading to an even longer chained morphology. Both low and high flow for the mutant lead to a larger area covered. (K) Chain breakage analysis and comparison of WT (low and high flow, pale red and blue) to a ΔatlA mutant (low and high flow, red and blue). P value was determined using a paired t test. * = P < 0.05, *** = P < 0.005. (Scale bars in E are 1 μm and in G and H are 10 μm.) For graphs, error bars are SEM.

To validate the mechanics-dependent colonization model’s prediction that chain breakages should have flow-dependent outcomes, we analyzed chain breakage events from our experiments in both high and low flow. Specifically, we quantified the fraction of breakages where the origin-proximal cells or distal cells remained adhered after the break (Fig. 5E). As predicted, there was no significant difference between the fraction of origin cells adhered following a breakage between low and high flow conditions (Fig. 5F). However, there was a significant difference in distal cell adherence, with more distal cells adhering in high flow than in low flow after a breakage event (Fig. 5F). Examples of the differences in chain breakage outcomes in high and low flow are shown in Movies S14– S17.

Our mechanical model also predicted that strains that formed longer chains should increase colonization in both low and high flow (Fig. 5I). To experimentally test this prediction we examined colonization dynamics in an E. faecalis OG1RF ΔatlA mutant (45), that has longer cell chains due to the lack of the peptidoglycan hydrolase AtlA. The ΔatlA mutant displayed increased surface colonization in both low and high flow conditions (Fig. 5 G, H, and J). ΔatlA mutant cells increased distal attachment following breakages in low flow (Fig. 5K) but showed no significance difference between the origin-proximal cells attached after a breakage (Fig. 5K). Together, these results establish the mechanism of E. faecalis flow-dependent colonization: High flow rates mechanically push E. faecalis chains toward the surface, resulting in increased colonization by increasing attachment and thereby reducing cell dispersal upon chain breakage.

Discussion

Fluid flow introduces complexities to bacterial dynamics that can result in unexpected emergent behaviors. Here, we elucidate two different mechanisms by which two species preferentially colonize surfaces in higher flow environments. Previous studies of bacterial colonization in flow have primarily focused on the mechanisms by which single cells adhere to surfaces, such as the catch bonds formed between Uropathogenic E. coli and the mannosylated surface of epithelial cells (20) or the mechanosensitive Type IV pili of Pseudomonas aeruginosa (46, 47). By contrast, here we find that increased flow stimulates colonization of both S. aureus and E. faecalis via distinct mechanisms that are surface-independent and function at the multicellular scale. For S. aureus, but not E. faecalis, colonization is driven by QS autoinducer transport. High flow conditions increase autoinducer transport, thus reducing QS signaling. Low flow conditions allow accumulation of QS autoinducers, downregulating adhesins, and leading to cell detachment (35, 38, 48). Importantly, whereas previous studies on QS in flow suggested that flow negates the physiological impact of QS in many contexts, our findings suggest that bacteria could harness the inhibition of QS in flow as a beneficial adaptation to promote colonization. The potential adaptive benefit of flow-dependent colonization is supported by competition experiments between S. aureus and P. aeruginosa, another pathogen that is often found with S. aureus. Specifically, whereas P. aeruginosa grows faster and dominates S. aureus in minimal flow, S. aureus colonizes better and dominates P. aeruginosa in high flow. Additional studies inspired by this work will help to determine the importance of interactions between multicellular signaling and flow in more complex environments.

In contrast to S. aureus, we found that mechanical effects on chained-cell nanocolonies were sufficient to explain how flow stimulates the colonization of E. faecalis. Altering fluid viscosity indicated that the E. faecalis colonization effect depends upon shear force rather than solely the shear rate, and agreement between predictions generated by a biophysical mechanics-dependent colonization model and experimental observations confirmed this hypothesis. Specifically, we found that the linear chains of cells produced by E. faecalis nanocolonies experience a torque that is proportional to viscous effects. Consequently, higher flow pushes the cells more toward the surface, leading to increased attachment, and a mutant that increases chain length enhances colonization. Our model suggests that chained colony morphology could be a general strategy that could promote flow-dependent colonization of a wide range of pathogens. E. faecalis chained nanocolony morphologies have been reported in clinical patient samples (49).

Interestingly, while S. aureus and E. faecalis achieve flow-dependent colonization in different ways, both mechanisms do not operate on single cells but rather depend on the collective morphologies of each species’ nanocolonies. The clustered multicellular morphology in which S. aureus cells grow facilitates the local 3D signaling interactions that mediate QS. Meanwhile, the linear multicellular morphology of E. faecalis produces a lever-like effect that promotes the torque required for flow-dependent attachment. Many studies of single-cell bacterial morphology have noted the importance of individual cell shape (50–52), but our findings suggest that even when individual bacteria have cells of the same shape, there can be adaptive benefits to the collective morphologies that form in their multicellular nanocolonies. Similarly, previous studies have examined the effects of mechanics on larger-scale biofilms (9, 34, 53), but our findings demonstrate the importance of considering these interactions at the mesoscale of nanocolonies.

Materials and Methods

Bacterial Strains and Growth Conditions.

S. aureus experiments used the strain USA300 MRSA. The S. aureus agrB transposon mutant was provided by the Network on Antimicrobial Resistance in Staphylococcus aureus for distribution through BEI Resources, NIAID, NIH: Nebraska Transposon Mutant Library Screening Array, NR-48501(41, 42). Constitutive green fluorescent S. aureus USA300 used for coculture experiments was from (54)(generously provided by the Torres Lab at NYU). E. faecalis studies were done using the strain OG1RF, fluorescently labeled with the plasmid pBSU101-GFP (55) (generously provided by the Wood Lab at the University of Michigan). The E. faecalis OG1RF ΔatlA mutant used was from (45). Constitutive red fluorescent P. aeruginosa PA01 was from (46). Wild type species used for supplemental studies in low and high flow were E. coli MG1655 and S. pneumoniae TIG4 (generously provided by the Veening Lab at the University of Lausanne).

Media were prepared according to manufacturer recommendation. Overnight cultures were grown in floor shakers at 37C in Luria-Bertani Miller (LB) Broth (BD Biosciences) for S. aureus, P. aeruginosa, and E. coli or Brain Heart Infusion (BHI) Broth (BD Biosciences) for E. faecalis and S. pneumoniae following isolation of a single colony from LB or BHI agar plates, respectively. The green fluorescent S. aureus was grown with 10 µg/mL chloramphenicol (Sigma) in plates and overnight cultures. For E. faecalis OG1RF, the plasmid pBSU101-GFP was maintained with the addition of 120 μg/mL of the antibiotic Spectinomycin (Sigma) in the plates as well as growth media. The E. faecalis OG1RF ΔatlA mutant was grown with the addition of 3.5 μg/mL of Tetracycline (Thermo Fisher) in plates as well as growth media.

Following overnight growth, cells were diluted in fresh media to an optical density (OD) between 0.40 and 0.45 and loaded into plastic syringes with a 27G needle in preparation for microfluidic chip experiments. For mixed coculture experiments, both S. aureus and P. aeruginosa were diluted in fresh LB to an OD between 0.40 and 0.45 before mixing together 1:1. This mixture was then loaded into plastic syringes with a 27G needle in preparation for microfluidic chip experiments.

For conditioned media experiments, overnight cultures were centrifuged at 9,000×g for 5 min to pellet the cells. The supernatant was removed and filtered using 0.2 μm filter disks (Pall Corporation) to ensure all cells had been removed. This filter-sterilized overnight medium was then mixed 1:1 with fresh LB (S. aureus) or BHI (E. faecalis) to create our conditioned media. For conditioned media experiments, we switched the syringes to conditioned media syringes from fresh media following 3 h of the flow run.

For S. aureus AIP experiments, synthesized AIP-I (obtained as described below) was added to fresh LB to a final concentration of 50 nM. The LB supplemented with AIP was used for the entire flow experiment. Similarly, for E. faecalis GBAP runs, synthesized GBAP was added to fresh BHI to a final concentration of 50 nM. Again, the BHI supplemented with GBAP was used for the entire 6 h flow experiment.

For E. faecalis viscosity and shear force experiments, 10% ficoll (Sigma) was added to BHI. This BHI + 10% ficoll medium was used for the entire 6 h experiment.

AIP and GBAP Synthesis.

AIP.

AIP-I was synthesized via a protocol adapted from Zhao et al. 2022 (56). Briefly, we performed standard automated Fmoc-SPPS on a Liberty Blue microwave-assisted peptide synthesizer (CEM), utilizing a DIC-Oxyma coupling strategy on a hydrazine derivatized Cl-TCP(Cl)-ProTide resin (CEM). After SPPS, cleavage of the resin yielded a linear peptide with a C-terminal hydrazide. This hydrazide was oxidized with 15 eq. of NaNO₂ in a 6 M guanidinium hydrochloride buffer solution at pH 3. A peptide thioester was formed upon the addition of 15 eq. MESNa at pH 6.2. The MESNa thioester was purified by RP-HPLC and lyophilized. To generate the cyclized final product, the peptide thioester was solubilized in PBS with 5 mM TCEP at pH 7. AIP-I was purified again by RP-HPLC, lyophilized, and taken up in DMSO. Concentration was determined by NMR relative to known standards. (SI Appendix, Fig. S8).

GBAP.

GBAP was synthesized via a protocol adapted from McBrayer et al. 2017 (57). Briefly, GBAP was generated by Fmoc-SPPS using a HATU coupling strategy extending the chain from Gln9 of the peptide. The unprotected side chain of Fmoc-Glu-Oall was coupled to a H-Rink amide ChemMatrix® resin (Matrix Innovations). The peptide was extended to Asn2 by standard Fmoc-SPPS. To the N-terminus was coupled Boc-Gln-OH to supply Gln1 and cap this chain of the peptide. To add Met9 and Trp10 and form the ester linkage of the peptide lactone, the O-trityl protecting group of Ser3 was first removed with 1% TFA. Fmoc-Met-OH (10 eq) was coupled to the Ser3 side chain by a Steglich-type esterification, utilizing DIC (9.9 eq), DIEA (10 eq), and DMAP (0.5 eq). After validating the formation of the Met9 ester by LC–MS, Trp10 was added by standard Fmoc-SPPS. The C-terminal Oalloc protecting group on Gln9 was removed by treatment with 10 eq 1,3-dimethylbarbituric acid and 0.3 eq tetrakis (triphenylphosphine) palladium (0) in DCM. After repeating the Alloc deprotection, the Pd catalyst was removed by washing with 1% sodium diethylthiocarbamate. On-resin cyclization was performed with PyOxim (1.25 eq) and DIEA (2.5 eq). Cyclized product was cleaved from the resin and purified by RP-HPLC. Pure products were lyophilized, taken up in DMSO, and quantified by NMR. (SI Appendix, Fig. S8).

Activity of both AIP and GBAP was confirmed using quantitative PCR: Following addition of the respective QS molecules, S. aureus and E. faecalis known QS genes were upregulated compared to controls.

Microfluidic Devices.

Flow channels were designed in Blender and a 3D mold was printed by Protolabs (Maple Plain, MN). Each chip had six channels that were 2 cm long × 500 μm wide × 100 μm tall with separate inlet and outlet ports for each channel. Briefly, Polydimethylsiloxane (PDMS) was poured onto the 3D mold, baked overnight, and then plasma-bonded to 22 × 44 mm #1.5 glass coverslips (Avantor). Following this, the completed chips were baked overnight at 65 °C to complete the bonding before use.

Flow Experiments.

Microfluidic chips were prepared for flow experiments by adding inlet and outlet tubing (BD Intramedic Polyethylene Tubing 0.015” ID 0.043” OD) and loading the channel with media before addition of our diluted cells. Following cell addition, a plastic media-filled syringe with 27G needle was connected to the inlet tubing and mounted on a syringe pump (New Era Pump Systems, Inc.) and the outlet tubing was placed into a waste container to collect the outflow. We could not immediately begin flow following cell loading as we needed to connect the media syringes and set positions and focus for microscopy. Following 15 to 20 min of initial attachment time, flow was turned on at one of two rates. For high flow conditions corresponding to a shear rate 400/s, a flow rate of 20 μL/min was used. For low flow conditions corresponding to a shear rate 40/s, a flow rate of 2 μL/min was used.

For S. aureus and P. aeruginosa coculture experiments, the high flow conditions were as described above. For minimal flow, following initial attachment of the channels, flow was turned on at a flow rate of 35 μL/min for 5 min to clear the channels of excess, unattached cells. For minimal flow experimental runs, flow rate was lowered to 0.1 μL/min, a shear rate of 2/s, to keep the channel clear and provide fresh media throughout the run.

For fibrinogen coating experiments, we followed a modified protocol from (58). Briefly, prior to loading with media and cells, a solution of 4 μg/mL of fibrinogen (Millipore Sigma) in Phosphate Buffered Saline (PBS, MP Biomedicals) was prepared. This solution was then loaded into our microfluidic channel and the channel was incubated at 25 °C for 2 h. Following incubation, the channel was rinsed with PBS and then media. The coated channels were then used in experiments following the high and low shear conditions as described above.

Phase Contrast and Fluorescence Microscopy.

For all flow experiments, a Nikon 90i inverted microscope equipped with a 100× 1.4 N.A. objective and with an incubator held at 37 °C was used. Microscope control and image acquisition was done using NIS Elements (Nikon, version 4.60.00). Five positions spread along each channel were chosen and the image plane was focused on the glass coverslip. Flow was turned on following the first set of images. Images were then taken every 5 min for a total of 6 h for both species. Phase contrast images were taken for all species. For E. faecalis, fluorescent images were taken using a GFP filter. To obtain doubling time information (Fig. 1B), the same flow setup and microscopy filters were used, but images were taken every 2 min for better doubling time resolution.

Quantification of Cell Coverage.

Cell coverage analysis was completed using a custom image analysis macro in Fiji and MATLAB code (software available upon request, Mathworks, Natick, MA). Briefly, for S. aureus, cell edges were found and contrast between the edges and background was enhanced using a CLAHE filter. Images were thresholded to remove background, binarized, and edges were filled in. Total cell coverage was determined by summing the total pixel count of the binarized images. For E. faecalis, analysis was done using fluorescent images to better capture the chains. Images were run through the CLAHE filter before having the background subtracted and binarized. Once again, total cell coverage was then determined by summing the total pixel count of the binarized images. The measured cell coverages were then loaded into MATLAB. The total coverage was normalized to the first image for each experiment position, and then the normalized total coverage for all positions of each experimental condition were averaged together and plotted over the experimental time. Error bars on these averages are reported as SEM.

Population fractions of the mixed cocultures of S. aureus and P. aeruginosa were again analyzed using custom image analysis pipelines using Fiji and MATLAB. Briefly, fluorescent images for each species were background subtracted, enhanced using a CLAHE filter and binarized. Cell coverage for each species was determined by summing total pixel count and these measurements were loaded into MATLAB. Population fraction was found by dividing the individual species coverage by the total coverage of both species at each time point. The population fractions from multiple experimental runs were then averaged together for each experimental condition (minimal or high flow). Error bars are reported as SEM.

Cell Doubling Analysis.

Doubling time estimates reported in Fig. 1B were performed for S. aureus and E. faecalis experiments by counting the events by hand using experimental data with phase images taken every 2 min. Cells were noted immediately following a division and were subsequently monitored frame-by-frame until a division occurred. Events were recorded throughout the time course and for different positions along the channel, to a total of 30 cellular divisions for each condition (S. aureus Low Flow, E. faecalis High Flow, etc.). Doubling time was found by averaging together all observations for each condition and error bars are reported as SEM. See SI Appendix, Table S1 for the doubling time values.

Focus Analysis.

E. faecalis focus analysis was again completed with a custom pipeline in Fiji and MATLAB (software available upon request). Analysis began in Fiji, with the phase contrast images CLAHE filtered followed by finding the edges of the cells. These edged images were then loaded into MATLAB. The edges of out-of-focus cells were dim and could be excluded by thresholding the images. The same threshold was used for all conditions. For each time point in an experiment, the thresholded and nonthresholded images were binarized and the total pixel value of each binarized image was calculated. The fraction in focus was then calculated by dividing the thresholded image total by the nonthresholded image. These results were then normalized to the initial time point for each position and all positions for a specific condition were averaged together. Error bars are again reported as SEM. Representative phases’ images of low and high flow E. faecalis experiments and an example schematic of the focus analysis process can be found in SI Appendix, Fig. S7.

Chain Breakage Analysis.

For analysis of chain breakage outcomes, thirty breakage events were calculated by hand for low and high flow. Following a breakage, origin attachment and/or distal attachment was recorded. The fraction of events where origin or distal attachment was recorded was calculated by dividing the total number of times the breakage outcome was recorded by our total number of breakage observations (thirty). Error bars are SEM. Significance was determined used a paired t test.

For simulations, anytime a breakage event was occurred, the outcome (origin and/or distal attachment) was recorded. Following the end of a chain simulation, these outcomes were recorded. For 100 chains in each simulation set, representing a “position” along a simulated flow channel, the total outcomes of each were calculated and again the fractions were calculated, error bars are SEM, and significance was determined using a paired t test.

Supplementary Material

Appendix 01 (PDF)

Movie S1.

E. coli low flow response. Representative E. coli experimental run in the low flow condition. Experiments ran for six hours and images were taken every five minutes.

Movie S2.

E. coli high flow response. Representative E. coli experimental run in the high flow condition. Experiments ran for six hours and images were taken every five minutes.

Movie S3.

S. pneumoniae low flow response. Representative S. pneumoniae experimental run in the low flow condition. Experiments ran for six hours and images were taken every five minutes.

Movie S4.

S. pneumoniae high flow response. Representative S. pneumoniae experimental run in the high flow condition. Experiments ran for six hours and images were taken every five minutes.

Movie S5.

S. aureus low flow response. Representative S. aureus experimental run in the low flow condition. Experiments ran for six hours and images were taken every five minutes.

Movie S6.

S. aureus high flow response. Representative S. aureus experimental run in the high flow condition. Experiments ran for six hours and images were taken every five minutes.

Movie S7.

E. faecalis low flow response. Representative E. faecalis experimental run in the low flow condition. Experiments ran for six hours and images were taken every five minutes.

Movie S8.

E. faecalis high flow response. Representative E. faecalis experimental run in the high flow condition. Experiments ran for six hours and images were taken every five minutes.

Movie S9.

S. aureus monoculture six-hour minimal flow experiment (shear rate 2/s). Experiments ran for six hours and images were taken every five minutes. Cells disperse and due to the low shear are seen out of focus in the channel as the run progresses.

Movie S10.

P. aeruginosa monoculture six-hour minimal flow experiment (shear rate 2/s). Experiments ran for six hours and images were taken every five minutes.

Movie S11.

P. aeruginosa monoculture six-hour high flow experiment. Experiments ran for six hours and images were taken every five minutes.

Movie S12.

Surface colonization of a co-culture of P. aeruginosa and S. aureus in minimal flow. Representative movie of P. aeruginosa (red fluorescence) and S. aureus (green fluorescence) coculture in the minimal flow condition (shear rate 2/s) for the standard six-hour experimental run.

Movie S13.

Surface colonization of a co-culture of P. aeruginosa and S. aureus in high flow. Representative movie of P. aeruginosa (red fluorescence) and S. aureus (green fluorescence) coculture in the high flow condition for the standard six-hour experimental run.

Movie S14.

E. faecalis breakage event in low flow (example 1). An excerpt from the experimentallow flow E. faecalis run, demonstrating a breakage event of a chain.

Movie S15.

E. faecalis breakage event in low flow (example 2). A second excerpt from the experimental low flow E. faecalis run, demonstrating a breakage event of a chain.

Movie S16.

E. faecalis breakage event in high flow (example 1). An excerpt from the experimental high flow E. faecalis run, demonstrating a breakage event of a chain.

Movie S17.

E. faecalis breakage event in high flow (example 2). A second excerpt from the experimental high flow E. faecalis run, demonstrating a breakage event of a chain.

Movie S18.

S. aureus spontaneous detachment or attachment. Representative movie of a five-hour low flow experiment of S. aureus with a minimal media with no glucose to observe the spontaneous detachment and attachment (from upstream attachment).

Data, Materials, and Software Availability

The data supporting the findings of the study are available in this article and its SI Appendix. Additionally, the raw data from microfluidic studies that support the findings of this study are freely available from the corresponding author upon request. Custom Fiji Macros and MATLAB code used for cell quantification and focus analysis of the microscopy data are also freely available from the corresponding author upon request. Simulation code has been deposited on Github at the following link: https://github.com/khallinen/BacterialMorpholgiesFlow_Models (59).