Fig 1.
Starch degradation of simple consortia follows an additive null biochemical model.
(A-B) All members of our six-species consortium are able to break down extracellular starch by secreting extracellular amylases. This is evidenced by the presence of a halo around colonies of all six species after Lugol stain [52]. Colonies formed on basic growth minimal media (1x bSAM) agarose plates. Extracellular Bacillus amylases bind randomly on starch chains, breaking them at random positions [53]. We propose a two-step enzymatic degradation model (Methods) that requires two cleavage reactions to turn starch into smaller oligosaccharides. (C) We fit the model (Eq 9) to the result of incubating purified B. subtilis amylase at various concentrations (from 0.4 to 100 μg/mL in 2X increments) with a 1 mg/mL starch solution for various lengths of time. In the horizontal axis, we plot z = x⋅t, where x represents the dilution of the enzyme relative to the maximum concentration (see Methods), and t is the incubation time. In the vertical axis, we plot the fraction (“Fract.”) of starch degraded for each condition. Each red dot represents a measurement for a different value of (x, t). (D) Amylolytic activity of the supernatants of individual species in monoculture. Species were grown in monoculture in 3 mL of 1x bSAM medium at 30°C for 24 hr. The vertical axis represents the fraction of starch degraded by different dilutions (x = 0.05, 0.125, 0.25, 0.5) of the filtered supernatants incubated at 30°C with 1 mg/mL starch for various time lengths (t = 3, 6, 19, 24 hr). The two-step model (Eq 9; solid black line) is fit to the data (red dots), and the fitted parameter Vi (hr−1), which corresponds to the function of the monocultures Vi0, is reported in red in each subpanel. (E) Amylolytic activity of 1:1 mixtures of all possible pairs of single-species filtered supernatants. In addition to experimental data (red points) and fit to Eq 9 (solid black line; fitting parameter Vij), we also show the fraction of starch degraded for each value of z predicted by the null biochemical model of independently acting enzymes (Eq 11; mean ± 2SE, gray region). (F) Comparison of fitted versus predicted values of Vij for the same pairs shown in (E) (error bars represent ± 2SE); red line represents perfect prediction. Vol., volume.
Fig 2.
Pairwise interactions are ubiquitous in our microbial consortia.
(A) Functional landscape predicted by the additive biochemical model (Eq 11). (B) Experimental functional landscape, where each point is the experimentally measured amylolytic rate of a consortium (error bars omitted for clarity). (C) Example of a pairwise functional landscape that is representative of communities lacking P. polymyxa (in this case, B. mojavensis and B. thuringiensis). Red lines show experimental measures (±SE), which are well predicted by the additive model (gray). (D) Functional landscape for a pairwise consortium containing P. polymyxa and B. thuringiensis. Here, the experimentally measured function of the pair (red) is not well predicted by the sum of the individual contributions (gray). The difference quantifies the pairwise interaction. (E) Pairwise interactions (
) of all possible two-member species consortia. (F) Comparison of the measured community function and the function predicted by the additive model (±2SE). Shape and color represent community size and presence of P. polymyxa, respectively. Black line represents perfect prediction. Species are designated as follows: C = B. cereus; E = B. megaterium; M = B. mojavensis; P = P. polymyxa; S = B. subtilis; and T = B. thuringiensis. This convention will be used throughout the text.
Fig 3.
Higher-order interactions in simple amylolytic consortia.
(A) Box explaining how single, pairwise, and higher-order interactions in the function of microbial consortia can be separated and quantified using ecological functional landscapes. (B) Third-order interactions ΔεABC(±SE), defined as the difference in the functional interaction () when a pair “AC” is grown alone and in the presence of a third species “B” (
), whose identity is shown in the horizontal axis. Species are designated as C, B. cereus; E, B. megaterium; M, B. mojavensis; P, P. polymyxa; S, B. subtilis; and T, B. thuringiensis. (C) Community function V (±SE) for every combinatorial consortia of P. polymyxa (“P”), B. mojavensis (“M”), and B. subtilis (“S”). In gray, we show the function predicted by adding the functions of all three taxa in isolation and all pairwise interactions (rather than, for instance, by averaging out the pairwise interactions). (D) Third-order interactions (±SE) are strongly anticorrelated (Pearson’s ρ = −0.95, P < 0.001) with the sum of all pairwise interactions (±SE in hr−1).
Fig 4.
Redundancy in the ecological facilitation of P. polymyxa explains higher-order interactions.
(A) If growth of P. polymyxa is facilitated by any and all of the other species, this facilitation may be redundant. Species are designated as C, B. cereus; E, B. megaterium; M; B. mojavensis; P, P. polymyxa; S, B. subtilis; and T, B. thuringiensis. (B) P. polymyxa grows in the presence of any of the other species (gray bars) but not in monoculture (orange bar). CFUs were determined by colony counting of serially diluted cultures after 48 hr of growth at 30°C. Error bars represent ±SE. (C) P. polymyxa grows to a comparable density as a part of a pair, trios, or higher numbers of other species in the consortium. This strongly suggests redundancy in the facilitation mechanism. (D) P. polymyxa growth (quantified by the final OD620 after 24 hr of culture, initialized at OD620 = 0; mean ± difference of two independent biological replicates) in our growth media supplemented with a 1:10 dilution of the filtered supernatant of each of the other species (Methods). (E) Function (VP(0)) of P. polymyxa monocultures shown in (D) (mean ± difference of two independent biological replicates). CFU, colony forming unit; OD, optical density at 620 nm.
Fig 5.
Including population dynamics into a null model of the structure-function landscape.
(A-B) We compare the predictive ability of the P/A null model (insets; data replotted from Fig 3F), with the revised null model that assumes that species contribution to function is proportional to their population size. For clarity, we separate consortia where P. polymyxa is absent (A) from those where it is present (B). Note the log-log scale used here for easier interpretation of the data. Red dashed line represents the identity curve where the null model perfectly predicts the function of the consortia. (C) To test whether the relation between final population size and function in P. polymyxa is well approximated by a saturating function, we grew P. polymyxa in media supplemented with different amounts of spent media from each of the other taxa. We determined the final population size in CFU/mL and the amylolytic function V (Methods, example in inset). The data were fitted using the saturating function (dashed line); 95% confidence interval (gray shading). (D) Comparison of the measured community function and the function predicted by the model assuming that function saturates with population size, for consortia where P. polymyxa is absent, and (E) for consortia in which P. polymyxa is present. All points shown ±SE. In (A), (B), (D), and (E), red dashed line represents perfect prediction. CFU, colony forming unit; Cond. supernat., Conditioning species supernatant; P/A, presence/absence.
Fig 6.
Quantifying the population (“Pop.”) dynamics and behavioral (“Behav.”) components of both pairwise and high-order interactions.
(A) Pairwise interactions (±SE) for all two-species consortia. We separate pairs by whether they include (orange) or do not include (blue) P. polymyxa. (B) Higher-order interactions in communities with three or more members. Similar to (A), we differentially color consortia by whether they include (orange) or lack (blue) P. polymyxa. Jitter in the horizontal axis (which is categorical) was added to the data for clarity of presentation.