(C) PLOS One
This story was originally published by PLOS One and is unaltered.
. . . . . . . . . .



Oxidative stress changes interactions between 2 bacterial species from competitive to facilitative [1]

['Rita Di Martino', 'Department Of Fundamental Microbiology', 'University Of Lausanne', 'Lausanne', 'Aurore Picot', 'Center For Interdisciplinary Research In Biology', 'Cirb', 'Collège De France', 'Cnrs', 'Inserm']

Date: 2024-02

Knowing how species interact within microbial communities is crucial to predicting and controlling community dynamics, but interactions can depend on environmental conditions. The stress-gradient hypothesis (SGH) predicts that species are more likely to facilitate each other in harsher environments. Even if the SGH gives some intuition, quantitative modeling of the context-dependency of interactions requires understanding the mechanisms behind the SGH. In this study, we show with both experiments and a theoretical analysis that varying the concentration of a single compound, linoleic acid (LA), modifies the interaction between 2 bacterial species, Agrobacterium tumefaciens and Comamonas testosteroni, from competitive at a low concentration, to facilitative at higher concentrations where LA becomes toxic for one of the 2 species. We demonstrate that the mechanism behind facilitation is that one species is able to reduce reactive oxygen species (ROS) that are produced spontaneously at higher concentrations of LA, allowing for short-term rescue of the species that is sensitive to ROS and longer coexistence in serial transfers. In our system, competition and facilitation between species can occur simultaneously, and changing the concentration of a single compound can alter the balance between the two.

We found that the toxicity of LA for At was due to the spontaneous accumulation of reactive oxygen species (ROS). Using the models, we show how to drive the interactions in the two-species co-culture towards more competition or facilitation by simply reducing or increasing the initial LA concentration, respectively, in line with SGH. Moreover, by artificially reducing environmental toxicity (by adding an antioxidant) at high LA concentration, we were also able to revert back to competition, hence mechanistically showing how the SGH works in our system. We then used our model to predict how interactions would affect long-term coexistence and found that toxicity extends the duration of coexistence in the short term, which was experimentally validated. Overall, the simplicity of our system allowed us to identify the mechanism behind the interactions between our bacteria and to shape them just by manipulating toxicity through the concentration of a single chemical compound.

In this study, we tested the SGH in a simpler system, combining experiments and mathematical models to study species interactions and coexistence in environmentally defined conditions. Our system consists of 2 bacterial species originally isolated from MWF, Agrobacterium tumefaciens (henceforth At) and Comamonas testosteroni (Ct), growing in a defined medium containing linoleic acid (LA), a fatty acid commonly present in many MWFs [ 36 ] as the sole carbon source. We chose this carbon source as it becomes toxic for At—but not Ct—at high concentration, making it suitable to explore the relationship between toxicity and interactions, and to mechanistically test the SGH and how facilitation affects coexistence.

One example where context-dependency has been proposed is the stress-gradient hypothesis (SGH), which states that positive interactions between species are likely to increase with environmental stress. The SGH was originally described in plants [ 33 , 34 ], but has also been observed in bacterial communities [ 16 , 21 , 23 , 35 ]. In our previous work [ 23 ], we measured pairwise interactions between 4 bacterial species that were isolated from polluting industrial liquids called metalworking fluids (MWFs). We showed that all pairwise interactions were positive in toxic MWF, but became more competitive when the environment was made more benign. Although this result supports the SGH, the complexity of the MWF medium and its unknown chemical composition prevented us from identifying a mechanism to explain facilitation in this toxic environment.

Ultimately, our understanding of interspecies interactions should allow us to predict the long-term composition of microbial communities, i.e., which species are likely to coexist. Coexistence theory usually considers competitive interactions, although recent approaches also integrate facilitation and mutualism [ 24 ]. In particular, Modern Coexistence Theory [ 25 , 26 ] focuses on species–species interactions often with a general Lotka–Volterra framework, while Contemporary Niche Theory [ 27 , 28 ] sees interspecific interactions as mediated by an environment from a consumer–resource perspective [ 29 ], which takes the dependency of interactions on environmental conditions into account [ 11 , 30 ]. Indeed, coexistence outcomes can critically depend on the mechanistic details of interactions [ 24 , 31 , 32 ].

Whether or not a species improves the environment or makes it more difficult for another species to grow will depend on the properties of the environment itself, suggesting that interactions should be context dependent [ 12 , 15 ]. Indeed, several studies have demonstrated that changes to the chemical composition of the environment can alter interaction sign [ 16 – 23 ].

Here, we define interactions between species as the effect of one species on the growth and death of another and focus on environmentally mediated interactions [ 11 ]. A species may negatively affect another by making the environment worse for it, for example, by depleting a common resource or releasing a toxic waste product. Alternatively, a species may affect another positively by improving the environment for the other, or for itself, accidentally benefiting the other [ 11 , 12 ]. Such positive effects can happen through detoxification, cross-feeding, or the release of public goods such as siderophores [ 13 , 14 ].

Multispecies microbial communities colonize almost every environment. Despite their small individual size, microbes can form very large populations that significantly affect their surroundings. For example, they can greatly influence the health and behavior of their living hosts [ 1 ]—for better or worse—or alter the physical and chemical properties of surfaces they are living on [ 2 ]. How these effects play out depends on a community’s species composition and how it changes over time, which in turn depends on how the different species interact: who restricts or enhances whose growth. But how we expect microbial species to interact with one another, what drives their interactions, and how interactions shape long-term coexistence remain matters of debate [ 3 – 10 ].

Although the time-scale of our experiments only allowed us to explore coexistence in the short term, it is still important to understand whether the 2 species would be able to coexist in the long term. We use our final model to explore this by simulating the outcome of mono- and co-cultures of At and Ct in a gradient of initial LA concentrations. We found no conditions where the long-term stable coexistence of the 2 species was possible, which is consistent with the idea that 2 species coexisting on a single resource is unlikely [ 47 ]. Nevertheless, to understand whether Ct had an effect on the extinction time of At, we measured the time (transfer) at which At went extinct when alone and compared it to its extinction time when in co-culture with Ct. At lower initial LA concentrations, At survives indefinitely in the long term in mono-culture ( Fig 4C ). As it eventually goes extinct in co-culture at all concentrations, the absence of coexistence at low concentration must be due to competitive exclusion by Ct ( Fig 4C ). In contrast, at higher initial LA concentrations, even though At quickly goes extinct in mono-culture, it survives for longer in the presence of Ct, meaning that Ct facilitates At’s survival. We then recapitulate the SGH in the long term: Ct competitively excludes At at low LA concentrations, but facilitates it by lengthening its survival at high LA concentrations.

(A) Prediction of short-term coexistence over 5 simulated serial transfers of the co-culture of At and Ct according to model 1 (left) and model 2 (right): Both models allow for short-term coexistence, but the parameter space in which this is possible is larger in model 2 (panel B, larger area representing coexistence compared to Ct surviving alone). White circles indicate the conditions in which experiments were run, as shown in panel B. (B) 5-Transfer experiment of At and Ct in mono- and co-culture at both 0.1% (left) and 0.75% LA (right). We show the population size in the initial culture (“transfer” 0), which is then quantified at each transfer (every 72 h). We illustrate the 100-fold dilution at each transfer, although this is not explicitly quantified. All 3 technical replicates per condition are shown. At mono-culture goes extinct as expected in 0.75% LA, but the 2 species coexist at both LA concentrations, as correctly predicted by model 2 in panel A. (C) Model 2 predicts that At should survive indefinitely in mono-culture at a 100-fold dilution rate up to 0.75% LA. In co-culture, Ct excludes it after a few transfers at low LA concentrations, but as the concentration increases, At can survive for longer in co- compared to mono-culture, meaning that Ct facilitates At’s survival by extending its duration (see also S3 Fig ). The data underlying this figure can be found at https://zenodo.org/records/8033845 . LA, linoleic acid.

In the first version of the model, toxicity depended on LA concentration, while in the second, it emerged on ROS accumulation, while LA acted exclusively as a nutrient. While this difference may seem like an implementation detail, we know from early theoretical work that 1 resource and 1 inhibitor are predicted to allow 2 species to coexist, while a single compound should only do so under very restrictive conditions [ 31 , 32 ]. We sought to verify this prediction and explore whether coexistence between At and Ct would be possible in either version of the model and in the experiments. We first extended the models to simulate a transfer experiment, where cultures were grown in batch for 72 h and then diluted 100-fold into fresh medium and regrown, and asked how long coexistence between the 2 species was possible. The second version of the model including ROS predicted a larger parameter range in which the 2 species could coexist in the short term (Figs 4A and S3 ). We tested this prediction experimentally by transferring 1% of the At and Ct mono-cultures and the co-culture in respective tubes every 72 h in both 0.1% and 0.75% LA. After 5 transfers, we found a significant variation for Ct co-culture in 0.1% LA compared to the beginning of the transfer experiment (first versus last transfer Ct co-culture 0.1% LA, t test, P = 0.0029), but there was no significant change for At in the same condition (first versus last transfer At co-culture 0.1% LA, t test, P = 0.30). Moreover, in 0.1% LA, competition was still evident from Ct to At, as At in mono-culture maintained a significantly higher population size than in the presence of Ct (mono-culture At 0.1% LA versus co-culture At 0.1% LA, t test P<0.0001, Fig 4B , left). This suggests that coexistence between At and Ct is possible in the short term despite the presence of negative interactions and provides further support for our updated model. In 0.75% LA, we observed the extinction of At mono-culture as we expected, but we found no significant variation neither for At nor for Ct in the co-culture (co-culture At 0.75%, t test P = 0.91, co-culture Ct 0.75%, t test P = 0.2, Fig 4B , right).

We used this second model to explore what would happen if we removed the toxicity by setting the production of ROS to 0 and leaving LA strictly as a nutrient. This led to 2 predictions: in absence of toxicity, (i) at 0.75% LA, At should survive even in mono-culture and reach a higher population size compared to 0.1% LA; and (ii) we should observe competition from Ct to At even at 0.75% LA ( Fig 3D , left). To test these predictions, we first added an antioxidant molecule to the cell-free LA medium to verify whether its presence would decrease ROS concentration. We chose to use tert-butylhydroquinone (TBHQ) for its antioxidant properties [ 46 ] at a concentration of 1.5 μm that did not inhibit bacterial growth ( S2 Fig ). In addition to the cell-free medium, we also added TBHQ to At and Ct mono-cultures, and the co-culture at the beginning of the growth assay and every 24 h to have a regular input of fresh antioxidant, mimicking continuous ROS neutralization by Ct. We found that TBHQ successfully decreased ROS concentration in all tested culture conditions compared to their value in standard 0.75% LA ( Fig 3B and 3E ). In support of prediction (i), adding TBHQ rescued At in mono-culture ( Fig 3D , center), allowing it to reach a significantly higher population size in 0.75% LA + TBHQ compared to 0.1% (AUC of At mono-culture in 0.1% LA versus 0.75% LA + TBHQ, t test P = 0.013, compare solid lines in Fig 3D , center and Fig 2C ). We suppose that in this ROS-free condition At could exploit the greater availability of LA as a nutrient. And as per prediction (ii), in 0.75% LA + TBHQ At grew to significantly lower population sizes in the presence of Ct than in mono-culture ( Fig 3D , center, AUC of At mono-culture in 0.75% LA + TBHQ: 3.7×10 9 ±1.3×10 9 versus co-culture: 7.9×10 8 ±1.5×10 8 , df = 5, t test P = 0.019), meaning that the interaction from Ct to At switched from facilitation to competition in the absence of toxicity. Overall, both model predictions were confirmed, demonstrating how we can shape the interaction from Ct to At by manipulating the level of toxicity in the environment.

The lack of ROS accumulation whenever Ct was present led us to hypothesize that Ct neutralizes ROS, reducing environmental toxicity and rescuing At in co-culture, allowing it to survive and grow ( Fig 3A shows an experimental repeat of Fig 2D and 2F ). These results are in line with the idea that toxicity was not caused by the increase of LA concentration itself, but rather by the accumulation of ROS. We adapted our model accordingly, leaving LA as a nutrient exclusively and adding ROS as an additional toxic compound that is generated through spontaneous LA oxidation. The co-culture growth prediction is qualitatively captured in this updated version of the model, in particular, the growth of At in high LA concentration is not underpredicted as it was in model 1 ( Fig 3A compared to Fig 2B ).

(A, D) Population sizes of both species in mono- and co-culture at 0.75% LA (A) or 0.75% LA with the antioxidant TBHQ added daily (D) (see Methods ). Panel (A) shows a biological repeat of the experiment shown in Fig 2 , but where ROS was measured simultaneously (panel B). The mathematical model in the left panels of (A) and (D) is the second implementation of the model, where ROS is explicitly modeled as a separate chemical from LA. In model 2, LA is only a resource (see Methods ). Model parameters were estimated by fitting to the mono-culture data (see Methods ). (B, E) A proxy for ROS concentration (MDA-TBA2, see Methods for assay) over time in the different experimental treatments and in model 2 (left panel) in the 0.75% LA medium (B) or where TBHQ was added (E). The cell-free and co-culture data are identical in the At and Ct subpanels (in black), as they come from the same samples. For experimental data, all 3 technical replicates per condition are shown. In the model panels, the cell-free lines are hidden by the green ones for At, as they are identical. (C, F) A diagrammatic representation of the relationships between chemicals (in squares) and bacterial species (circles): (C) LA is consumed as a resource by both species and generates ROS. ROS inhibits At but can be reduced by Ct. (F) TBHQ inhibits ROS, making ROS removal by Ct superfluous. ROS removal by TBHQ or by Ct in co-cultures, rescues At. This translates into facilitation of At by Ct in the top row and competition in the bottom row, where TBHQ removes ROS. The data underlying this figure can be found at https://zenodo.org/records/8033845 . LA, linoleic acid; ROS, reactive oxygen species; TBHQ, tert-butylhydroquinone.

Based on the literature, we hypothesized that spontaneous oxidation of LA might release ROS [ 41 – 44 ] even in the absence of bacteria. Accordingly, we used the thiobarbituric acid reactive substances (TBARSs) assay (see Methods ) to test for the presence of ROS in our system [ 45 ]. Indeed, at both LA concentrations of the cell-free medium, ROS accumulated over time ( Fig 3B , cell-free), supporting the idea of a chemical reaction leading to ROS production, such as spontaneous oxidation due to exposure of light and air [ 42 , 44 ]. ROS were significantly less abundant in Ct mono-culture and in the co-culture than in the mono-culture of At (Ct mono-culture versus At mono-culture, Tukey’s test for multiple comparisons P = 0.0002; Ct co-culture versus At mono-culture, P = 0.0002, Fig 3B ).

Although these results matched the model predictions qualitatively, At’s growth in co-culture was greatly underestimated by the model ( Fig 2B and 2D , green dashed lines). Furthermore, using the parameters estimated from all mono-cultures, the model does not correctly predict the hump-shaped growth of At at 0.75% LA ( Fig 2D ), even though it already assumes the accumulation of toxicity. This suggests that estimating model parameters where both growth and death are caused by a single compound is challenging. We next focused on exploring the mechanism behind this possible increase in toxicity, in particular, if we can decouple toxicity from the LA itself.

The model predicted that increasing the concentration of LA in a co-culture of the 2 species could change the interaction sign from negative to positive. More specifically, at low concentration, both species compete for the sole nutrient source LA. Once the concentration becomes high enough to kill At, however, we expect to observe facilitation, as Ct consumes LA and reduces its concentration, making the environment less toxic for At ( Fig 2A and 2B ). We tested this prediction in the lab by growing At and Ct in mono- and co-culture in 0.1% and 0.75% LA as low and high LA concentrations, respectively. The results were in line with the predictions of the model: at low LA concentration (0.1%), At grew to significantly smaller population sizes in the co-culture compared to mono-culture (area under the growth curve (AUC) of At at 0.1% LA in mono-culture: 9×10 8 ±4×10 7 versus co-culture: 7×10 7 ±9×10 6 , t test P<0.001, Fig 2C ), showing that there was competition for LA that negatively affected the growth of At. At high LA concentration (0.75%), the presence of Ct in the co-culture rescued At, allowing it to grow to population sizes that were orders of magnitude greater than alone (AUC of At at 0.75% LA in mono-culture: 4×10 7 ±2×10 7 versus co-culture: 2×10 9 ±10 9 , t test P = 0.03, Fig 2D ). The growth of Ct was not significantly affected by At in either condition (AUC of Ct in 0.1% LA in mono-culture: 9×10 8 ±5×10 7 versus co-culture: 9×10 8 ±10 8 , t test P = 0.98; AUC of Ct in 0.75% LA in mono-culture: 10 10 ±8×10 9 versus co-culture: 2×10 10 ±2×10 10 , t test P = 0.42, Fig 2E and 2F ). Taken together, we classify the interactions between the 2 species as “ammensalism” at 0.1% LA and “commensalism” at 0.75% LA, with only 1 species being negatively or positively affected by the other, respectively [ 40 ].

(A, B) Predictions of model 1, where LA is a nutrient but becomes increasingly toxic over time. Model parameters were estimated by fitting to the mono-culture data (see Methods ). (C–F) Experiments showing the growth of At (C, D) and Ct (E, F) at 0.1% LA (C, E) or 0.75% LA (D, F). Both species survive in mono-culture at 0.1% LA, whereas At grows and then dies in mono-culture at 0.75%. At 0.1% LA, At suffers from the presence of Ct in co-culture, while Ct’s growth is not significantly affected. At 0.75% LA, At is rescued by Ct in the co-culture, and Ct’s growth is not significantly affected. The model does a reasonable job at capturing the overall dynamics, but underestimates At’s growth in co-culture at 0.75% LA. For experimental data, all 3 technical replicates per condition are shown. See main text for statistics. The data underlying this figure can be found at https://zenodo.org/records/8033845 . LA, linoleic acid.

(A) Growth of A. tumefaciens (At) and C. testosteroni (Ct) in presence of different compounds at a range of concentrations in triplicates (each replicate is one colored square in the rectangle). The tested compound is the only carbon source added to the medium (see Methods ). Heatmaps show the fold change between the AUC of each sample replicate and the AUC of the mean of the 3 control replicates where no compound was added. Blue shades represent negative fold change (bacteria grew to significantly smaller populations than the control) and orange shades represent positive fold change (bacteria grew to greater population sizes than the control). Statistical significance is determined by t tests comparing N = 3 AUCs at e.g., [C] = 0.05% to N = 3 AUCs at [C] = 0 (*: P<0.05,**: P<0.01,***: P<0.001). (B) Growth curves that generated the LA data in panel A. We observe some growth for both species in the absence of any carbon source (0%) but little competition when co-cultured ( S1 Fig ). While we were unable to find the source of this growth, similar observations have been previously reported [ 37 – 39 ]. All 3 technical replicates per condition are shown. The bottom panel shows population sizes at day 8 for a clearer comparison. The data underlying this figure can be found at https://zenodo.org/records/8033845 . AUC, area under the growth curve; LA, linoleic acid.

To build on our previous work [ 23 ] and track how pair-wise interactions change with varying compound concentration and varying toxicity, we designed a simpler medium containing a single MWF compound at a time. We selected 10 compounds commonly found in MWF and tested their effect on At and Ct in mono-culture at different concentrations ( Fig 1A ). As Ct was facilitating At in MWF [ 23 ], we were looking for compounds on which Ct could grow but that would be challenging for At. Our goal here is not to explain what we observed in MWF, but rather to design a comparable model chemical environment in which to study interactions in a more controlled way. LA was a good candidate, as it acted as a nutrient source for both species at low concentration, while at high concentration, it was toxic for At even though Ct could still grow well ( Fig 1B ). To generate quantitative predictions on the behavior of these 2 species in mono- and co-culture, we developed a mathematical model (ordinary differential equations (ODEs) describing dynamics of species and substrate abundances, see Methods ) with parameters (e.g., growth and death rates) that best fit the mono-culture growth curves ( Fig 1B ). To capture the response of At to LA in the model, LA acted as a nutrient source that could also cause death at increasing concentrations (see Methods ). We then used the model to predict how the 2 species are expected to interact if co-cultured at different LA concentrations ( Fig 2A and 2B ).

Discussion

Most current research on interactions between microbes has explored positive interactions through cross-feeding [5,10,14,48]. Our study instead focuses on positive interactions through detoxification [11,23] and shows how they can occur mechanistically. Facilitation through detoxification relates to the SGH. Here, with the help of our mathematical model, we show how the SGH can predict population dynamics: one species can provide a benefit to another by removing the stress that would otherwise drive the second species extinct, i.e., “niche facilitation” [24]. Experimentally removing the stress from the co-culture of the 2 species eliminates facilitation and restores competition (Fig 3D). This mechanistic understanding of the SGH increases its potential as a guiding principle for interspecies interactions [33,35].

How common is this phenomenon and is it specific to the compound we focused on, linoleic acid? Fig 1A shows that different compounds can be toxic in a species- and concentration-dependent way, and if species are also able to remove their toxic effect, they should behave similarly to the compound we chose to study. We also know that all aerobic microbes suffer from oxidative stress to some extent, as ROS is generated as a metabolic byproduct [49], and many frequently encounter ROS in their natural habitats [50]. In fact, ROS accumulation is a main plant stress response upon exposure to pathogens [51,52], and both organisms used in this study have been associated with plants [53,54]. One reason why the role of stress in mediating microbial interactions may not receive much attention, is that lab studies are typically designed to optimize bacterial growth, which a natural environment does not.

An important gap in our study is what type of ROS is accumulating, and how Ct is reducing the concentration of our ROS proxy: Is ROS being taken up and neutralized intracellularly or is Ct secreting extracellular enzymes that eliminate ROS? Answering this question is not as simple as analyzing Ct’s genome, since all aerobic bacteria are capable of dealing with oxidative stress and we find relevant enzymes in both Ct and At [55] (S1 Table). What we do know is that adding an antioxidant to the medium mimics the effects of Ct on At. Future experiments could explore the role of different types of ROS and different knockout mutants of Ct on the interaction between these 2 species. Spent-media experiments in a medium containing a simpler carbon source in addition to an ROS could also help to distinguish intra- and extracellular ROS elimination by Ct.

Our second main message is that even in very simple systems like the one we have studied here, several layers of interactions can play out simultaneously. Here, we find 2 layers: competition for the resource, linoleic acid, and facilitation through the removal of the toxin, ROS. Competition is present at all LA concentrations; if we remove environmental toxicity exogenously at high LA, the underlying resource competition is revealed. What we measure under given environmental conditions, then, is the net effect of species on each other, taking into account all the chemical substrates they may be competing for, feeding to one another or removing to facilitate each other’s growth. Because ROS generation is proportional to LA concentration, modifying the initial concentration of just that one compound alters the balance between resource abundance and toxicity, and thereby the dominance of competition or facilitation. This mechanism supports the intuition of the SGH (Fig 1 in [56]) that these interactions really are context dependent and do not simply emerge from confounding effects or methodological biases [57]. In this system, we have uncovered 2 mediators or environmental factors (LA and ROS); however, it is possible that there may be more that were not salient enough to be observed. More generally, microbes can construct a surprising number of new niches that could potentially mediate interactions with others [58,59].

Another important biological question is whether the underlying interactions can help to predict long-term coexistence. According to early theoretical work, coexistence on a single resource is only possible in small parameter regions in fluctuating environments, when metabolic trade-offs exist [47]. Intuitively, therefore, we did not expect coexistence in the competitive scenario at low LA concentration, but over the short time-scale of 5 transfers, our model predicts coexistence, which we recapitulate with the experiment.

It was less clear whether positively interacting species would coexist. Species that interact positively through cross-feeding are expected to coexist if the positive feedback is not too strong and leading to chaotic dynamics [24,30], but according to our model, 2 species that are competing for a single resource but facilitating each other through detoxification should not. Eventually, the stronger competitor should dominate. Nevertheless, we show that facilitation by detoxification can increase the duration of coexistence—again in line with the SGH—which may allow the weaker species to survive a limited period of harsh conditions [60,61]. It has been argued that the “expected time to extinction” is an appropriate measure of fitness [62]. Nevertheless, according to the model, in longer co-culture experiments, At should go extinct at all concentrations (Fig 4C), which would be an interesting hypothesis to test.

One indication that our model may be underestimating the potential for coexistence is that in Fig 4B, At does not appear to be close to extinction in either condition at transfer 5, while our model predicts that it should go extinct between transfers 6 and 10 (S3 Fig). This discrepancy leads us to suspect that despite our efforts to quantitatively predict outcomes of these co-cultures, our model is likely still missing further interaction layers between the 2 species like cross-feeding that were not observed in this study. Detecting cross-feeding in our system was challenging, as we found LA to not be easily amenable to standard chemical analysis methods. Understanding the coupling between detoxification and cross-feeding and their effect on long-term coexistence is left for future work.

This brings us to an important discussion point: How much detail is needed to construct a predictive model of a community like the one we have studied here? While we cannot give a definitive answer to this question, our study shows that details of the model matter. First, a consumer–resource (CR) model was needed to capture the context-dependency of interactions, compared to implicitly defining interactions as in a Lotka–Volterra framework [11,63,64]. Second, within the CR framework, whether we model a single compound whose effect changes or 2 compounds—one resource and one toxin—likely increases the parameter space under which the 2 species coexist (Fig 4A). This is in line with predictions that more environmental factors favor coexistence [28], and while we suspect that our model is still missing some of these factors to make it predictive in the long term, it nevertheless served to understand our system, how to control it, and to generate the testable hypothesis that toxicity may extend the duration of coexistence. On the one hand, this may be seen as bad news on fitting models to data: without a mechanistic understanding, it will be difficult to build a good predictive model. On the other hand, some underlying knowledge on the chemistry of the growth medium can go a long way. As long as parameterization is possible, the advantage of even a partial mechanistic understanding is that it can generate more realistic hypotheses on the addition of further species or changes in resource concentrations, compared to an uninformed or abstract exploration of the parameter space of a model. Making sure enough interaction layers are included in the model to describe the observed experimental pattern requires a tight back and forth between models and data.

As we and others have shown, chemical compounds mediate positive or negative interactions between species, such that even in simple laboratory microcosms, several interaction layers may emerge whose balance and net effect will depend on initial conditions and how the chemical environment changes over time. Whether or not species will coexist in the long term will not only depend on the sign of these interactions but also the mechanism behind them: Are they removing stress for one another or feeding one another? Further verifying the model in this study and understanding how the layering of interactions plays out in the long term are next important steps.

[END]
---
[1] Url: https://journals.plos.org/plosbiology/article?id=10.1371/journal.pbio.3002482

Published and (C) by PLOS One
Content appears here under this condition or license: Creative Commons - Attribution BY 4.0.

via Magical.Fish Gopher News Feeds:
gopher://magical.fish/1/feeds/news/plosone/

You are viewing proxied material from magical.fish. The copyright of proxied material belongs to its original authors. Any comments or complaints in relation to proxied material should be directed to the original authors of the content concerned. Please see the disclaimer for more details.