Extended Data Fig. 5: NMF (Non-negative matrix factorization) reveals low-dimensional shifts in growth at the phylum level.
From: Functional regimes define soil microbiome response to environmental change
(a) Initial community composition (T0) of CAF soils at the phylum level. The x-axis indicates soils with different native pH levels (n = 10 soils: Soil 3, 5, 6, 9, 11, 12, 14, 15, 16, 17, see Table S1 for their properties). The y-axis represents the relative abundance (summed to 1) of the top 10 phyla out of 40, with the cumulative abundance of the remaining phyla depicted in grey as ‘Other phyla’. (b) For each soil, at each pH perturbed condition (n = 130 conditions, 10 soils × 13 perturbations) we computed the growth fold at the phylum level by summing absolute abundance of all ASVs within that phylum in both CHL+ and CHL- and computing AbsCHL−/AbsCHL+ (y-axis). Each panel shows growth folds for one phylum as labeled in the title. Each line corresponds to a soil with color indicating native pH (color bar to right). y-axis values above 1 indicate growth in that phylum at that pH. (c) To systematically identify the underlying lower-dimensional growth response to pH, we used non-negative matrix factorization (NMF) on the growth fold values to decompose the growth response of all phyla into two modes (Axis #1 and Axis #2, see Methods). The growth folds of each phylum are the linear combination of two modes whose weights are plotted on the left panel (points are colored by phylum as in (a)). Mode #2 is composed of the Bacillota phylum (Regime III). Mode #1 is composed of Pseudomonadota and Bacteroidota (Regime II). (d-f) The same analysis was performed with growth fold values of phyla in the pH perturbation experiments on the additional soils from IL, IN, and CA, USA (Table S1) (n = 55 conditions, 5 soils × 11 perturbations). (g) Ammonium dynamics in CHL− and CHL+ samples were measured for 20 CAF soils using the Salicylate-hypochlorite assay64, plots only showing data for Soils 1, 5, 12, and 17 out of 20 CAF soils. \({{\rm{NH}}}_{4}^{+}\) accumulation (3–50% of 2 mM \({{\rm{NO}}}_{3}^{-}\)) in Regime III suggests activation of the dissimilatory nitrate to ammonia (DNRA) pathway by Regime III strains. The NaOH concentration in perturbed samples also impacted \({{\rm{NH}}}_{4}^{+}\) measurements, because the Salicylate-hypochlorite assay includes a step where OCl− reacts with the N-H moiety, resulting in N-Cl and OH−. Higher NaOH addition results in slightly lower detection of chloramphenicol in the CHL+ samples. We used the constant \({{\rm{NH}}}_{4}^{+}\) levels in the controls without 2 mM \({{\rm{NO}}}_{3}^{-}\) (No-Nitrate controls) in the CHL+ conditions for each soil to offset the NaOH effect in the CHL- samples (by computing the conversion factor ratio of \({{\rm{NH}}}_{4}^{+}\) levels of No-Nitrate controls in CHL+ conditions to the initial \({{\rm{NH}}}_{4}^{+}\) levels of each condition with different NaOH additions in CHL+ samples) and subtracted \({{\rm{NH}}}_{4}^{+}\) levels caused by chloramphenicol in CHL+ samples. pH in each panel indicates the endpoint pH of the experiment.
