Abstract
The recycling of particulate organic matter (POM) by microbes is a key part of the global carbon cycle, one which is mediated by the extracellular hydrolysis of polysaccharides and the production of public goods that can trigger social behaviors in bacteria. Despite the potential importance of these microbial interactions, their role in regulating of ecosystem function remains unclear. In this study, we developed a computational and experimental model system to address this challenge and studied how POM depolymerization rate and its uptake efficiency –two main ecosystem function parameters– depended on social interactions and spatial self-organization on particle surfaces. We found an emergent trade-off between rate and efficiency resulting from the competition between oligosaccharide diffusion and cellular uptake, with low rate and high efficiency being achieved through cell-to-cell cooperation between degraders. Bacteria cooperated by aggregating in cell-clusters of ~10-20μm, where cells were able to share public goods. This phenomenon, which was independent of any explicit group-level regulation, led to the emergence of critical cell concentrations below which degradation did not occur, despite all resources being available in excess. By contrast, when particles were labile and turnover rates were high, aggregation promoted competition and decreased the efficiency of carbon utilization. Our study shows how social interactions and cell aggregation determine the rate and efficiency of particulate carbon turnover in environmentally relevant scenarios.
Significance Statement Microorganisms can cooperate by secreting public goods that benefit local neighbors, however, the impact of cooperation on ecosystem functions remains poorly constrained. We here pair computation and experiment to show that bacterial cooperation mediates the degradation of polysaccharide particles recalcitrant to hydrolysis in aquatic environments. On particle surfaces, cooperation emerges through the self-organization of cells into ~10-20μm clusters that promote cooperative uptake of hydrolysis products. The transition between cooperation and competition in aggregates is mitigated by individual cell behaviors such as motility and chemotaxis, that promote reorganization on the particle surface. When cooperation is required, the degradation of recalcitrant biopolymers can only take place when degraders exceed a critical cell concentration, underscoring the importance of microbial interactions for ecosystem function.
Introduction
The microbial breakdown of complex polysaccharides is a key ecosystem process that enables the recycling of carbon from plant and animal detritus into global biogeochemical cycles and is a relevant process in all heterotrophic microbial ecosystems, from animal guts (1–3) to soils (4, 5) and oceans (6–8). A key feature of these polysaccharides is their insoluble nature: a large fraction is found in particles at the scale of 100 μm, that require both surface colonization and extracellular hydrolysis to be degraded (9, 10). On particle surfaces cells can attach and grow in close proximity, increasing the opportunity for microbial interactions to impact the ecosystem process. One particularly relevant type of interaction in this context is cell-cell cooperation mediated by the sharing of public goods such extracellular enzymes and hydrolysis products (11). However, the extent to which these interactions take place and impact bacterial growth in the environment remains unclear. Previous work on cooperative interactions has largely focused on the opportunity that public goods open for exploitative populations to invade (so-called cheaters), and less on the environmental and physiological conditions that enable cooperation to take place or the potential role of cooperative behavior on ecosystem processes. From the perspective of ecosystem modeling, efforts to incorporate the role of microbes in organic matter degradation equate microbial activity with enzymatic activity without considering the role of population-level phenomena such as cooperation. In this paper we reveal how microbial social interactions can impact relevant ecosystem parameters.
The extent to which social interactions mediated by public goods play a relevant role in ecosystem function is highly dependent on how public goods diffuse (12, 13). In a three-dimensional aqueous environment like the ocean, if cells are too far apart only a minuscule fraction of the public goods are recovered by neighbors, while the rest is lost to the environment. In contrast, if cells are sufficiently proximal to each other and the resource is limiting, growth kinetics can be cooperative, meaning that the per-capita growth rate is positively dependent on the density of degrader cells (14). This logic suggests that cooperation should be accompanied by the emergence of spatial patterns, such as cell patches. If the cooperative effects in these patches are strong, critical population density thresholds might emerge below which degradation cannot support population growth (14, 15). Less recognized is the contribution of individual cell behaviors, such as surface attachment, chemotaxis, and biofilm formation, on the ability of cells to find those critical densities by aggregating into cell patches. Therefore, in order to begin to understand the role of social interactions in natural systems, we need to take into account the physical constraints of the micro-environment and how populations interact with these constraints through their behavior.
To quantify the impact of bacterial social interactions and spatial behavior on ecosystem function, we focus on two main parameters: the speed at which polymers are hydrolyzed and converted to soluble oligosaccharides, that is, the turnover rate (16, 17), and the POM uptake efficiency, which is the fraction of the dissolved oligosaccharide that can be taken up by cells and converted into biomass. To study the role of social interactions and spatial behavior on ecosystem function, we developed a computational and experimental model of the colonization of insoluble particulate polysaccharides by marine heterotrophic bacteria. The individual-based model (18, 19) simulates the functional traits of individual cells: chemotactic movement, particle attachment and detachment, the secretion of enzymes, oligosaccharide uptake and growth. The experimental system validates computational predictions in a chitin-degrading bacterial strain isolated from the coastal ocean, and clarifies the role of physiological parameters on social interactions (17). We leveraged the computational model to study the relationship between degradation rate and POM uptake efficiency and how emergent bacterial behaviors influence their ability to degrade recalcitrant particles, and we tested some of our predictions using our experimental model of chitin colonization. Our work demonstrates that cell-cell cooperation is critical for the degradation of complex biomaterials, implying that the degradation of recalcitrant polysaccharides can be bacteria-limited. Moreover, cell-density thresholds that determine the onset of cooperative growth depend strongly on individual cell behavior, in particular those behaviors that regulate the residence time of bacteria on particles.
Results
We modeled the dynamics of cell colonization, enzyme secretion and growth (Figure 1A) using an individual based model to describe cells, coupled to a reaction-diffusion framework to describe enzymes and oligosaccharides. In the model, bacterial cells that attached to the surface of a polysaccharide particle broadcast enzymes that reacted with the surface of the particle, releasing oligosaccharides to which non-attached cells could chemotax. Cellular uptake of oligosaccharides followed Monod kinetics (20) and cells were allowed to divide after a certain quota of oligosaccharide is consumed (19) (see Methods and Supplementary Information for a detailed description and Table S1 for the parameters). This individual-based approach allowed us to modulate traits such as chemotaxis or particle-attachment rate, and measure their impact on the carbon uptake rate on a cell by cell basis.
A crucial parameter of our model was the “particle lability”, Kp, which defined how many grams of oligosaccharide were released per gram of enzyme acting on the polysaccharide surface per unit of time. Kp was a compound parameter that resulted from the product of the catalytic activity of the enzyme, kcat and the recalcitrance of the substrate. This parameter played a central role because it determined the maximum degradation rate and controlled the nutrient supply rate of bacteria. A survey of hydrolysis rate values reported in the literature revealed that the particle lability, Kp can exhibit significant variation across natural environments and microbial enzymes. Kp varied by more than 6 orders of magnitude within glycosyl hydrolase families- a trend that held true among different substrate types such as chitin, alginate and starch (Figure 1B). This led us to ask how variation in particle lability, Kp affected population growth dynamics and the rate – efficiency relation of POM depolymerization.
Our results revealed that there is an emergent negative relationship between the rates of depolymerization and growth, and the POM uptake efficiency of the particle-associated bacterial population (Figure 1C) (21, 22). This emergent rate – efficiency trade-off was a consequence of the diffusion of oligosaccharide in a three dimensional environment where soluble products that were not taken up by cells in the vicinity of the particle were lost. At high values of Kp, oligosaccharides are produced in excess of Ks, the half-saturation constant of the Monod growth function, and therefore cells approached their maximum growth rate. However, high Kp also led to ~99% loss of oligosaccharide to diffusion (~1% recovery), which reduced the theoretical biomass yield of the population and the POM uptake efficiency. For comparison, if the system was closed, as in a laboratory reactor, POM uptake efficiency could theoretically reach 100% because dissolved oligosaccharides would accumulate (Figure S1). However, natural environments are rarely, if ever, closed and diffusive losses are likely to limit POM uptake efficiency in nature, given low particle densities, (Table S3). Moreover, adding the movement of fluids around particles by convective flow to the model (23–25), further increased the loss rate of oligosaccharide to the bulk environment, and reduced POM uptake efficiency from 10% to 2% (Figure S2–S5). Although the exact value of POM uptake efficiency could depend on substrate affinity (1/Ks) and on cell numbers (the more cells that can capture oligosaccharides the higher uptake efficiency), the trade-off between rate and efficiency held for different physiological parameters (Figure S6). Taken together, these results suggest that in natural environments most public goods are lost and that the competition between diffusion and uptake should lead to a tradeoff between POM uptake efficiency and its turnover rate. Quantifying the uptake efficiency for natural marine particles revealed that the maximum efficiency barely exceeded 7% at the optimum particle lability (Kp ~100hr−1) for the highest of particle-associated cell density observed (~2.32×107) (Table S3).
Surprisingly, we found that the high POM uptake efficiency observed at low Kp (recalcitrant particles/low enzymatic activity per cell) was mediated by the aggregation of cells into micro-scale patches on the particle surface, a phenomenon that was not hardcoded in the model but emerged from the interplay between diffusion, cell behavior and growth (Figure 1C and Figure S7). Within these patches, cells grew cooperatively by sharing oligosaccharides that would otherwise be lost to diffusion, which increased the per capita growth rate and POM uptake efficiency up to a density of 0.3 cells/μm2 (Figure S8). To characterize the spatial density dependence, we performed simulations to quantify particle depolymerization and mean growth rates as a function of the inter-cell distance (Figure 1D). Our analysis showed that dense spacing (a nearest neighbor distance of 8 μm) promoted cooperation by sharing of oligosaccharides, but only when particles were recalcitrant and the oligosaccharide production rate was slow (Kp<100hr−1) (Figure 1E). More precisely, when the amount of oligosaccharide available to cells fell near Ks, the half-saturation of the Monod growth curve, an increase in the local concentration of oligosaccharide due to cell-cell aggregation increased the per capita growth rate. In contrast, at high Kp (~2000 hr−1), oligosaccharides quickly accumulated and the uptake rate was decoupled from the spatial organization of the cells on the particle, since there were enough resources for cells to grow at their maximal rate ([C]≫Ks) (Figure 1E). Under these conditions, there is no benefit to aggregation and even cells spaced 22 μm apart reached their maximum oligosaccharide uptake rate (Figure 1E).
In our model, cell detachment and reattachment from the particle surface was a critical behavior that enabled the formation of patches and the degradation of recalcitrant particles. On recalcitrant particles (Kp=10-100 hr-1) 1% detachment significantly increased the particle degradation rate and its uptake efficiency (Figure 2A), and also increased the mean carbon uptake rate by a factor of 5 (Figure S9A), compared to a non-detaching population. This allowed populations to survive on recalcitrant particles that might otherwise not sustain growth and drive the population to extinction (Figure S9B). Without chemotaxis, random motility alone still allowed detaching populations to grow on more recalcitrant particles than non-detaching populations, but at ~1/6 the POM uptake efficiency (Figure 2A) and ~1/10 the rate of biomass accumulation (Figure S9B). This was due to the fact that with chemotactic motility most cells had access to the same of hydrolysis products emanating from cell patches, with the distribution of carbon uptake rates for individual cells displaying a tight peak near the maximum uptake rate (μ~0.8μmax) (Figure S9C). Our model thus suggests that detachment and chemotaxis enhance POM uptake efficiency under nutrient-limited conditions ([C]~ Ks) by enabling the formation of patches where cells cooperate by sharing public goods. This also implies that the onset of cooperation is dependent on the individual strain physiology. Organisms that had either a high affinity for oligosaccharides (low Ks) or a high hydrolytic activity saturated their growth at low oligosaccharide concentrations (Figure 2B), circumventing the need to cooperate Figure (2C-D). In contrast, organisms with a low uptake affinity for the public good, or organisms with a low per-cell rate of hydrolysis, such as those that tether enzymes to their membrane, had a higher need to cooperate with other cells (Figure 2D). Therefore, although there was a general trend to increase cooperation as particles became harder to degrade (Figure 2D), traits such as motility, surface detachment rates, substrate affinity or enzyme localization determined determine the exact onset of cooperation for each population.
To experimentally validate our prediction that cell-cell cooperation drives the degradation of hard-to-degrade polysaccharides, we turned our attention to Psychromonas sp., psych6C06, a marine isolate that had previously been enriched from coastal seawater on model chitin particles(10). The strain readily degrades chitin hydrogel in ~30 hours (17) and encodes at least eight predicted chitinases, or glycosyl hydrolase family 18 and 19 homologs, but no other families of glycosyl hydrolases, leading us to conclude that the strain is representative of a chitin specialist. We reasoned that if cooperative growth kinetics played a role in this system, we would observe a strong dependency between the initial number of cells that can colonize the particle and the growth of the population. In particular, we would expect a critical cell density below which the population is unable to form the patches required to degrade particles, revealing that the degradation process is bacteria-limited.
In agreement with this prediction, psych6C06 displayed a strong density dependence when growing on hydrogel chitin beads, in the form of a critical cell density below which degradation never occurred (Figure 3A). Interestingly, we observed that colonization involved the formation of cell patches, in agreement with the model results (Figure 3B). At concentrations just below the threshold critical cell density, we saw that cells that initially colonized the particle were not able to persist. Populations that persisted did so by forming cell patches (Figure 3C). We artificially increased Kp by adding exogenous chitinase to supply 776 μg/h GlcNAc. Consistent with individual-based model results (Figure 2D), the addition of the exogenous enzyme activity lowered the cell density-dependent threshold for colonization of the chitin hydrogel beads (Figure 3D). In addition, the broadcast chitinases led to a more uniform distribution of psych6C06 cells on the chitin hydrogel bead, a state that was morphologically distinct from the patchy colonization achieved at 24 h psych6C06 without enzyme (Figure 3B, Figure S10C) and similar to the simulation results at high Kp (Figure 2D).
To obtain a more mechanistic understanding of the critical threshold phenomenon we took a bottom-up approach by predicting the threshold cell density based on measurements of the relevant physiological and behavioral parameters of psych6C06. We calculated particle attachment and detachment rates by quantifying cell density on particles (Figure 4A-B). Our measurements revealed rapid attachment and detachment rates (attachment rate 0.03 h−1, detachment rate 0.26 h−1) that were proportional to the density of cells off and on the particle, respectively suggesting that the population undergoes frequent rearrangement on the surface of the particle. These observations echo the rearrangement observed in the individual based simulations (Figure 2A). We measured Kp for psych6C06, using the fluorescent substrate 4-Methylumbelliferyl (MUF)-N-acetyl-β-D-glucosaminide, which detects the release of GlcNAc from chitin. We noted that very little chitinase activity was detected in the culture supernatant of psych6C06, but robust activity was associated with the cells themselves (Figure 4C, Figure S10A) indicating that enzymes were membrane bound. Using MUF-conjugated substrates with different cleavage specificities, we determined that most of psych6C06 chitinase activity was derived from exochitinase, which releases GlcNAc as a product (Figure S10B). Thus, we assessed the GlcNAc to biomass conversion factor (the biomass yield) of this strain by direct measurement of sugar consumption and cell density in exponentially growing cultures (Figure 4D). We measured of the growth rate of psych6C06 on a range of GlcNAc concentrations, and used these substrate-limited growth measurements to derive μmax and Ks from a fit of the Monod growth equation(20) (Figure 4E). We used the diffusive loss of oligosaccharides predicted by our 3D simulations to estimate GlcNAc loss, and assumed that all cells on the surface experienced the same concentration (no local gradients). Using these measurements, we parameterized a simple version of the individual based model written to describe population-level growth dynamics on a surface where cells can attach, detach, and grow as a function of the hydrolyzed product concentration (see Methods for a full description of the analytical bottom-up model). Using this simplified model, with no free parameters, we studied how the initial cell density determines the population colonization rate and the growth of bacteria on the particle surface. We found a remarkable quantitative agreement between model and experiments, with critical thresholds predicted between initial densities of 5*106 and 107 cells/mL (Figure 4F). Our analysis thus shows that individual cell level physiology and behavior together with diffusion regulate the onset of social degradation of POM.
Discussion
Despite the key regulatory role of microbes in carbon cycling (17, 26, 27), linking the micro-scale physiology and behavior of bacteria to carbon flux models remains elusive. Here, we show that in conditions where diffusion limits oligosaccharide accumulation, like in the ocean, the breakdown of particulate polysaccharides is subject to population density-dependent effects. These effects are driven by three key physiological parameters: the affinity of cells for hydrolyzed oligosaccharide, the rate of polysaccharide hydrolysis, and the amount of exchange on/off the particle surface that define a tradeoff between the rate of polysaccharide degradation and biomass yield. These features contrast with a common assumption of carbon flux models: that model heterotrophic cells consume nutrients at rates comparable to the consumption of simple dissolved substrates by laboratory-adapted model organisms in a well-mixed systems (27–29). In particular, our computational and experimental results highlight that the frequency at which cells exchange on and off the polysaccharide surface has a large impact on cell carbon uptake. This effect is emerged not only by setting the threshold population density that is achieved by the initial population without growth, but by reconfiguring the arrangement of surface-associated cells in ways that maximize oligosaccharide uptake and growth. In addition, we show that the emergence of microbial aggregates on recalcitrant particles increases the chance of survival for populations of bacteria by enhancing the local dissolved carbon production rate and cell uptake rates. The benefit of aggregate formation is dependent on size: structures that are too large or dense to support the maximum uptake rate of individual cells promote competition rather than cooperation (30). In natural ecosystems, aggregate formation and dispersal is the rule rather than an exception (31–34) and cells are likely to carry adaptations to enhance aggregation/dispersal on particle surfaces, for instance by regulating chemotactic movement or the expression of biofilm components such as adhesins and matrix proteins (35, 36). Thus, micro-scale interactions could significantly affect the rates of POM turnover in the environment, underscoring the need to incorporate these interactions into models of carbon cycling.
Our computational and experimental system also highlights the importance of social collective behavior and spatial self-organization on community fitness and survival. Previous studies have shown that secretion of public goods (enzyme) favors the formation of patchy microbial aggregates by enhancing cooperative behavior (11, 14, 15, 37). Notably, many bacteria actively regulate enzyme secretion and other group behaviors at the level of transcription (38–40), and cell-density-dependent transcription factors such as quorum signal receptors are capable of sensing both changes to the environment and to cell density (41). While our simulations reveal that physiology alone is sufficient to explain patch formation, further studies are required to evaluate the contribution of such regulation on the group behaviors that may facilitate patch formation and dispersal by psych6C06. Our results also suggest that the benefit of clustering is not universal and is instead dependent on physiology: the POM uptake efficiency and growth rate of cells with high affinity for substrates suffers in the context of an aggregate, while aggregation is optimal for strains with lower substrate affinity and lower rates of polysaccharide hydrolysis (11). This observation highlights the fact that in systems with potential for spatial organization – that is, most systems outside the lab, the balance between cooperation and competition can be delicate and modulated by the intersection of physical processes with microbial physiology.
Methods
Individual based model of individual cell behavior and physiology
The mathematical model represents metabolism, surface interaction and flagellar motility of individual cells in 3D space in the presence of chemical gradients. We introduce an individual-based model (42, 43) to quantify single-cell interactions with organic particles by abstracting the structural heterogeneities of natural POM into a mathematically simpler spherical shape, while preserving some key physical and chemical processes associated with POM degradation. A spherical organic particle of 200 μm radius is simulated such that it remains static in the middle of an aqueous volume (~1mm3). While natural organic matter aggregates may show various shapes and chemical compositions, we modeled particles as perfect spheres made of a single type of insoluble linear polysaccharides such as chitin, alginate, or cellulose. This computational model is inspired by experimental model systems used to study community assembly on marine POM (10, 17). The particle’s size and its surface chemistry are assumed to be unchanged during particle degradation: only the particle density changes over time to satisfy mass conservation. This assumption is consistent with experimental observations that have shown no significant change in organic particle size during microbial degradation until the final stages of collapse(17). We simulated a scenario where an isogenic population of cells is allowed to colonize and degrade a particle with a defined volume. The simulations were started with zero oligosaccharides and the particle was considered to be the sole carbon source.
To take into account the fact that cells might regulate their enzymatic activity, the model limits enzyme secretion to two scenarios: when cells adhere to the particle surface or when the rate of oligosaccharide supply exceeds the maintenance threshold. Importantly, our simulations ensure mass conservation between total carbon uptake, growth and loss of oligosaccharides. Individual cells are initialized as a uniform random distribution in the aqueous volume, and are allowed to disperse following gradients of chemo-attractant (in this case, oligosaccharide). The cells can consume the oligosaccharide, grow, and divide to new daughter cells and experience a range of local conditions. A full derivation of the mathematical expressions and steps used for modeling of microbial growth, dispersal and enzyme secretion can be found in the Supplemental Information.
Experimental methods
Strain psych6C06 was previously isolated from an enrichment of nearshore coastal seawater (Nahant, MA, USA) for surface-associated chitin degrading microbial communities (10, 17). The strain was maintained as colonies on Marine Broth 2216 (Difco 279110) with 1.5% agar (BD 214010). To establish exponential growth without hysteresis, we modified a culturing protocol previously developed for Escherichia coli K12(44), and grew cells on a defined seawater medium with the N-acetyl-D-glucosamine (GlcNAc) at concentrations indicated. Chitin hydrogel beads (NEB) were washed and diluted to 200-250 particles per mL with size range from 40 to 100 μm in diameter. The beads were rotated end over end at 21-25 °C. The density of inoculated cells was set to be at an A600 of 0.01, diluted from 20 mM GlcNAc minimal medium cultures prepared as described above. To visualize particles and their surface-associated bacteria, 200 μl subsamples were stained with the DNA-intercalating dye SYTO9 (Thermo Fisher, S34854) at a 1:285 dilution of the stock in 96-well plates with optically clear plastic bottoms (VWR 10062-900).
Cell density measurements (Absorbance at 600 nm, A600) of exponentially-growing cells were used to measure the maximum cellular growth rate, and plating was used to measure growth under GlcNAc limitation, from which we derived the half-saturation constant. GlcNAc depletion was measured during growth using the dintrosalicylic acid reagent method(45), and the depletion rate was used to calculate the biomass yield (see Supplemental Information)
Chitinase activity was quantified using Methylumbelliferyl(MUF)-conjugated substrates N,N′-diacetyl-β-D-chitobioside, N-acetyl-β-D-glucosaminide, and β-D-N,N′,N″-triacetylchitotriose (Sigma CS1030). Microscopy was performed on micro-confocal high-content imaging system (ImageXpress Micro Confocal, Molecular Devices), using the 60 μm pinhole spinning disk mode. Fluorescent signal was visualized with a LED light cube (Lumencore Spectra X light engine), and bandpass filters (ex 482/35 nm em 538/40 nm dichroic 506 nm), with a 40x objective (Nikon Ph 2 S Plan Fluor ELWD ADM 0.60 NA cc 0-2 mm, correction collar set to 1.1), and a sCMOS detector (Andor Zyla). Image analysis was performed in MATLAB (release 2018a). Briefly, image stacks were split in half and a maximum intensity projection was obtained for each half. The low level of fluorescent signal associated with free dye in the hydrogel particles was used to define an intensity threshold suitable to create a binary mask for the particle projections. A mask of the cells within the beads was then defined using their brighter fluorescence intensity. We used this segmentation to quantify the total surface area occupied by the cells on the bead, and to quantify the total surface area occupied by patches (areas where cells contact other cells >10 μm2).
Acknowledgements
We thank Lu Lu for technical assistance, and all members of the Cordero lab for their support and critical feedback. This project was supported by Simons Early Career Award 410104 and the Simons Collaboration: Principles of Microbial Ecosystems (PriME), award number 542395. A.E. acknowledges funding from Swiss National Science Foundation Grant P2EZP2 175128.