Environmental connectivity controls diversity in soil microbial communities
Soil resident microbesWe chose sand as a realistic source of a mixed microbial community (which is referred to as the sand community or SC). Because the sand community cannot be preserved as a whole by freezing, we collected fresh material for better consistency for each experiment from the same spot in St. Sulpice near Lake Geneva (GPS coordinates: 46.508032N, 6.544050 E) as described in Moreno et al.51. Sampled sand at different seasons thus likely carried slightly different starting communities and cell densities. The sand was sieved through 2 mm2 pores to remove large particles. The sieved sand was stored at room temperature and used within 7 days for extraction of resident microbial cells.Microbial cells were extracted from four aliquots of 200 g of sand. Each 200 g aliquot was transferred into a 1-l conical flask and submerged in 400 ml of 21 C minimal media salts (MMS) (containing, per litre: 1 g NH4Cl, 3.49 g Na2HPO4·2H2O, 2.77 g KH2PO4, pH 6.8)21. Flasks were incubated at 25 °C under rotary shaking at 120 rpm for 1 h. The sand was allowed to settle and the supernatant was decanted into a set of 50 ml Falcon tubes, which were centrifuged at 800 rpm with an A-4-81 rotor and a 5810R centrifuge (Eppendorf AG.) for 10 min to precipitate heavy soil particles. Supernatants were decanted into clean 50 ml Falcon tubes and centrifuged at 4000 rpm for 30 min to pellet cells. The supernatants were carefully discarded and the cell pellets were resuspended and pooled from the four aliquots (i.e., from the initial 800 g of sand) in one tube using 5 ml of MMS. The pooled liquid suspension was further sieved through a 40 µm Falcon cell strainer (Corning Inc.) in order to remove any particles and large eukaryotic cells that may obstruct flow cytometry analysis (see below). A small proportion of the sieved liquid suspension was used to quantify the numbers of recovered cells (see below); the remainder was used within 12 h for bead encapsulation or for mixed liquid suspended growth (see below). With this gentle method, we extracted approximately 3 × 105 cells g−1 of sand.Flow cytometry cell countingCell numbers in extracted soil communities and in the mixed liquid suspended growth experiments were counted by flow cytometry. SC-suspensions were diluted 100 times in MMS and stained in 200 µl aliquots with 2 µl of diluted SYBR Green I solution (1:100 in DMSO; Molecular Probes) in the dark for 30 min at room temperature. In some experiments, cells were additionally stained with 2 µl propidium iodide solution (10 µg ml–1, Molecular Probes). Aliquots of 20 µl were aspired at 14 µl min–1 on a Novocyte flow cytometer with absolute volumetric cell counting (ACEA Biosciences, USA). Cells were thresholded above a forward scatter signal (FSC-H) of 20 and further gated for propidium iodide-staining (excited at 535 nm and its fluorescence was collected at 617 ± 30 nm) and for SYBR Green I (excitation 488 nm, 530 ± 30 nm band-pass filter; channel voltage at 441 V) above values of 1000 (Supplementary Fig. 5).Cell samples from the mixed liquid suspension growth experiments were diluted to approximately 106 ml−1 and subsampled to aliquots of 100 µl. The subsamples were then mixed with 100 µl of 8 g l–1 sodium azide in phosphate buffered saline and incubated for 1 h at 4 °C to arrest cell respiration and growth. Samples were then stained with SYBR Green I as above and quantified by flow cytometry using the same thresholds and gates as describe above.Bacterial strains and pre-culturing proceduresP. veronii 1YdBTEX2 is a toluene, benzene, m-xylene and p-xylene degrading bacterium isolated from contaminated soil20. The strain was tagged with a single-copy chromosomally inserted mini-Tn7 transposon carrying a Ptac–mCherry cassette (Pve, strain 3433) as described in the ref. 52. A single P. veronii colony from a selective plate with toluene as the sole carbon substrate after 48 h incubation at 30 °C was inoculated into 10 ml of liquid MMS containing 5 mM sodium succinate as the sole carbon source and grown for 24 h at 30 °C with rotary shaking at 180 rpm21. After 24 h, the cells were harvested and washed for bead encapsulation or for comparative liquid mixed suspension growth, as described below.Agarose bead encapsulationSC cell suspensions containing between 2 × 107 to 108 cells ml–1 were encapsulated in agarose using rapid mixing with pluronic acid in dimethylpolysiloxane and subsequent cooling, followed by sieving to achieve beads with a diameter range of 40–70 µm53. The entire procedure was carried at room temperature and near a gas flame to maintain antiseptic conditions. 1% (w/v) low melting agarose (GEPAGA04-62, Eurobio ingen, France) was prepared in PBS solution (PBS contains per L H2O: 8 g NaCl, 0.2 g KCl, 1.44 g Na2HPO4, 0.24 g KH2PO4, pH 7.4) and dissolved by heating in a microwave. The molten agarose solution was cooled down and equilibrated in a 37 °C water bath. Separately, 15 ml of dimethylpolysiloxane (Sigma-Aldrich, DMPS5X-500G) was poured in a 30 ml glass test tube. 1 ml of the 37 °C-agarose solution was mixed with 30 µl of pluronic acid (10% Pluronic® F-68, Gibco, Life Technologies) by vortexing at the highest speed (Vortex-Genie 2, Scientific Industries, Inc.) for a minute. Into this mixture of agarose and pluronic acid, 200 µl of prepared SC cell suspension at 0.2–1.0 × 108 cells ml–1 was pipetted and vortexed again at the highest speed for another minute. Five hundred microliter of this mixture was added drop-wise into the glass tube with dimethylpolysiloxane that was being vortexed at maximum speed. Vortexing was continued for 2 min. The tube was then immediately plunged into crushed ice and allowed to stand for a minimum of 10 min. After this, the total content of the tube was transferred into a 50 ml Falcon tube. The tube was centrifuged for 10 min at 2000 rpm using an A-4-81 swinging-bucket rotor (Eppendorf). The oil was carefully decanted while retaining the beads pellet. Fifteen milliliter of sterile PBS was added to the pellet and the beads were resuspended by vortexing at a speed set to 5. The tubes were again centrifuged at 2000 rpm for 10 min and any visible oil phase on the top was removed using a pipette. The process was repeated once more to remove any visible oil phase. Beads of diameter between 40 and 70 µm were then recovered by passing the PBS-resuspended bead content of the tube first over a 70-µm cell strainer (Corning Inc.). A further 5 ml of PBS was added to the cell strainer to flush remaining beads ( 0.25). Finally, we tested further interaction terms that influenced attributed growth rates. The initial carbon concentration was set to 50 mg ml–1, which allowed similar community development in terms of size (i.e., cell numbers) as in the experiments.Growth was simulated for 120 time steps, corresponding to 60 h in the experiments, at which point the substrate is depleted and cells stop dividing (stationary phase). Based on the attributed growth rates to every OTU (i.e., every cell and genotype of the vector or double vector), the model calculates per time step how much substrate is converted into biomass (we allow continuous biomass formation) and lost in form of CO2, which is subtracted to calculate the remaining substrate concentration for the next time step. When the overall substrate concentration is lower than ({S}_{{min }}) = 3 × 10–6 g ml–1, growth stops. The production of cell biomass is converted to cell numbers, which is then subsampled at the last time step per OTU (to an equivalent of 50,000 sequence reads), per single or pair (to an equivalent of 5000 beads) to calculate developed microcolony sizes, diversity measures, and interaction effects (as in Figs. 5 and 6).The following interaction scenarios were simulated. Although we allowed growth penalties and interspecific interactions to influence attributed OTU growth rates, a threshold of ~0.6 h–1 was imposed as the maximum individual OTU growth rate in all simulations.High connectivity random vs. OTU-abundance growth rates. OTU-specific growth rates (µmax,sp) were drawn randomly between 0.01 and 0.6 h–1. Alternatively, growth rates were attributed to the vector of OTUs according to the probability distribution function reflecting their measured log10 empiric abundance at t = 0 h (Supplementary methods, Section 2).Low connectivity single founder cell growth penalty. We contrasted simulations with single founder cells growing according to their OTU-proportional attributed growth rate and those in which that growth rate was multiplied by a penalty, composed by a factor equal to the inverse proportion of the initially attributed µmax,sp per OTU. The assumptiton is that the slower the inherent growth rate, the more likely that OTU is penalized when it is alone (Supplementary methods, Sections 2 and 3). This was combined with testing the effect of random or biased death on the starting community.$${{rm{mu }}}_{{{rm{max}} },{{rm{sp}}}}=frac{1.2}{{{log }}_{10}({{rm{mu }}}_{{{rm{max}} },{{rm{sp}}}})}$$
(3)
Low connectivity paired interspecific interaction effects. We further tested different assumptions on the nature of interspecific interactions and simulated how these affected community growth rates and diversity outcomes. These effects directly influenced the OTU attributed growth rates in the doubles (Supplementary methods, Sections 3.1–3.3). In the bimodal scenario (Supplementary methods, Section 3.4.1), we assumed that the community is composed of two underlying distributions; rare and abundant members (the threshold being placed at log10 measured OTU relative abundance = 2.8), with abundant members having a higher probability to be positively influenced in pairs. The probability is drawn from a bimodal interaction curve that attributes an interaction factor (between 0.01 and 2.2), which is multiplied with the assigned OTU growth rate at start.In the biased positive model (Supplementary methods, Section 3.4.2) we allowed a 40% chance for an interaction term imposed independently on each founder cell in a pair to lower the attributed OTU growth rate (factor range 0.4–0.6), and 60% chance for a factor in between 0.6 and 1.4 to modulate or increase the growth rate.In the positive on slow model (Supplementary methods, Section 3.4.3), the attributed OTU growth rates on each founder cell in a pair had a chance of 40% to become improved inversely proportionally to its initial growth rate, thereby favoring slow growers$${{rm{mu }}}_{{max },{{rm{sp}}}}={{mbox{-}}}{rm{ln}}({{rm{mu }}}_{{max },{{rm{sp}}}}),times {{rm{mu }}}_{{max },{{rm{sp}}}}$$
(4)
In the biased negative model (Supplemtary methods, Section 3.4.4), we attributed OTU-abundance proportional growth rates to each partner of the founder pair, but penalized faster growers (µ > 0.15) at 20% chance and the others at 40% chance that their growth rate would be multiplied by a negative interaction factor (range 0.01–0.1).Finally, in a random model (Supplementary methods, Section 3.4.5), we allowed OTU-abundance proportional growth rates in pairs to be multiplied with a factor randomly drawn in the range of 0.01–1.25, independently for each partner in a pair. The models were contrasted to those without any assumed interspecific interactions, and without or with assumed random or fast-growing genotype biased cell death at start (Supplementary methods, Section 2.3.1).All simulations were run five times from the beginning, independently producing five derived parameter values for alpha-diversity, OTU- and microcolony size distributions in stationary phase and partner interactions.Statistics and reproducibilityLiquid suspension growth experiments were carried out in biological quadruplates and all bead experiments were carried out in biological triplicates. Total numbers of analyzed bead and those of beads with single or double occupancy are reported. Derived community growth rates and P. veronii-normalized yields were compared using t-tests (n as reported, two-sided test, unequal variance). Normalized PBP bin-size distributions were globally compared using Fisher’s exact test implemented in R (2000 replicates). Median and 75th percentile aggregate PBPs across different experiments were compared using the non-parametric Wilcoxon signed-rank test. Correlations between simulated species abundance distributions and empirical OTU relative abundances were calculated by bootstrapping (n = 1000) in MATLAB. Correlation coefficients from five independent simulations were compared using t-tests. The proportion correctly predicted OTU abundances by simulation was calculated as the ratio to observed values within a two-fold or four-fold range, and compared by two-sided t-tests on five independent simulations. Simulated and observed microcolony size distributions for single or paired founder cells among different models were compared by principal component analysis in MATLAB (pca), and by Spearman rank correlation (spear) from five independent simulations. Single and paired productivity was then compared between each other using two-sided t-tests of the 75th percentiles of microcolony size distribution (n = 5). Simulated and observed paired growth (excluding pairs with dead cells) was categorized and counted in a grid of 12 × 12 (each bin covering 0.5 log10-distance) using MATLAB’s hist3d function, and then compared by pca from five independent simulations. Confidence intervals on ratios of paired simulated microcolony sizes (excluding those pairs with a non-growing or dead partner) were determined by subsampling (n = 1000) from mean ratio distributions, which were then used to calculate the fractions deviating from experimentally observed paired size ratios.Reporting summaryFurther information on research design is available in the Nature Research Reporting Summary linked to this article. More
