Extended Data Fig. 8: Impact of native pH on regime boundaries, community composition, and soil physicochemical properties.

(a) Cartoon illustrating how native pH impacts the Regime I-II boundary. Identical acid concentrations (ΔH⁺) cause larger pH changes in neutral soils (ΔpH^neutral) compared to acidic soils (ΔpH^acidic), as explained by their positions on the titration curve in (b). (b) Titration curves showing endpoint pH (y-axis) versus acid/base added (x-axis) for neutral (dark brown) and acidic (light brown) soils. The dashed vertical line at 0 marks the unperturbed pH. For the same ΔH⁺, neutral soils undergo greater pH changes (shaded regions). This suggests acidic soils transition from Regime II to I after smaller pH perturbations (blue points, (e)). (c) Cartoon showing how native pH affects the Regime II-III boundary. Fixed NaOH additions release identical carbon (ΔC(0), Fig. 5), with larger C(0) driving transitions to Regime III. Neutral soils reach higher pH values for the same NaOH additions due to titration curves (d). (d) Soil pH titration curves of soils from varying native pH levels (full titration curve shown in (n)). The vertical dashed line shows NaOH added to transition from Regime II to III. Neutral soils (darker colors) achieve higher pH values, aligning with the increasing Regime II-III boundary pH (purple points, (e)). (e) Regime transition pH levels (y-axis) for soils with different native pH levels (x-axis). Blue points show Regime II-I boundaries, and purple points show Regime II-III. Boundary pH is the midpoint between the last Regime I/II sample’s pH and the first Regime II/III sample’s pH. Error bars represent pH differences between these samples. The dashed blue line (Regime I-II boundary) and dashed purple line (Regime II-III boundary) are weighted least squares fits, with the weights inversely proportional to the error of each point. The dashed black line (slope 1) indicates a constant change in pH from native to the regime boundary for all soils. Slopes differing from 1 show that this pH change depends on native soil pH. The blue dashed line has a slope of 0.7 (95% CI: [0.44, 0.97]). (f-g) Using sequencing data, we examined whether taxa adaptation to native pH influences regime boundary pH levels. (f) To analyze the transition to the Acidic death regime (Regime I), we calculated survival folds for Pseudomonadota and Bacteroidota across perturbed pH levels as \(Ab{s}_{CHL+}/Ab{s}_{{T}_{0}}\) (endpoint abundance in CHL+ vs. initial time point, T₀). Using a uniform survival fold threshold, we determined the pH where survival folds fell below the threshold, defined as either (1) “dying” (< 1) or (2) “dead” (→ 0). For Resurgent growth regime (Regime III), growth folds for Bacillota were calculated as Abs_CHL−/Abs_CHL+, with a growth fold threshold of 3 to find the boundary pH. (f) Boundaries using the “dying” definition show a Regime I-II slope (0.56 ± 0.09), indicating neutral soils tolerate larger ΔpH. (g) The “dead” definition for Pseudomonadota and Bacteroidota shows Regime II-I pH transition points with a slope (0.11 ± 0.08) near 0, suggesting death occurs at a fixed pH (~3.5). Error bars reflect pH differences around threshold crossings; points represent midpoint pH. Linear fits were computed using least squares (blue/purple dashed lines). The grey dashed line represents y = x. (h-l) Taxonomy of growing strains in Regime III varies with soil native pH (see Methods). (h) Predicted vs. observed soil pH using LASSO-regularized regression based on ASV presence/absence (threshold: relative abundance > 0.005) to predict soil pH. Left: in-sample predictions; right: ‘Leave-one-soil-out’ (Loso, Methods). Prediction quality (R²) is calculated from mean predicted vs. observed pH. (i) Bar plots of non-zero ASV regression coefficients from in-sample predictions in (h). (j) R² of regression by taxonomic level for presence/absence. Predictions fail above the genus level. Bars: dark = in-sample, grey = Loso. (k) Impact of relative abundance threshold on presence/absence definition. (l) A permutation test was conducted to evaluate the significance of ASV-level predictions. Native pH values were permuted 1000 times to generate R² distribution. The observed R² in (h) (black arrow) has p = 0.012. (m) The community composition of our soils at T0 samples effectively spans the pH gradient of the global topsoil microbiome¹⁹. To compute the compositional difference between our soils used for the experiment and global topsoils (n = 237 global samples)¹⁹, we performed non-metric multidimensional scaling (NMDS; Vegan v2.5.7) on the Bray-Curtis distance matrix at the family level (k=3, stress=0.11). Points were colored by soil pH. Global: global topsoil data, CAF: Cook Agronomy Farm, IL: LaBagh in Illinois, IN: Pinhook in Indiana, CA: Sedgwick in California (see Table S1). (n-q) pH titration curves and physicochemical properties across soils (see SI). Fitting a logistic function (o) to soil pH titration curves (n) for 20 CAF soils across native pH levels. (n) Soil pH titration curves of CAF soils (colored by native pH). The y-axis shows endpoint pH after 4 days of incubation following pH perturbation (n = 457 endpoint samples). (o) Logistic function and parameters. Titrations with H⁺ and OH⁻ were unified on an OH⁻ x-axis by shifting curves 0.2 m mol right, aligning them at 0 m mol OH⁻. (p) Summary of correlations between soil physicochemical properties and titration curve characteristics. Native soil pH shifts the titration curves horizontally, moving X_mid. (q) Correlations supporting diagram (p) with R² values (n = 20 CAF soils; Soil 1–18 used for X_mid values).