Community composition and physiological plasticity control microbial carbon storage across natural and experimental soil fertility gradients

Many microorganisms synthesise carbon (C)-rich compounds under resource deprivation. Such compounds likely serve as intracellular C-storage pools that sustain the activities of microorganisms growing on stoichiometrically imbalanced substrates, making them potentially vital to the function of ecosystems on infertile soils. We examined the dynamics and drivers of three putative C-storage compounds (neutral lipid fatty acids [NLFAs], polyhydroxybutyrate [PHB], and trehalose) across a natural gradient of soil fertility in eastern Australia. Together, NLFAs, PHB, and trehalose corresponded to 8.5–40% of microbial C and 0.06–0.6% of soil organic C. When scaled to “structural” microbial biomass (indexed by polar lipid fatty acids; PLFAs), NLFA and PHB allocation was 2–3-times greater in infertile soils derived from ironstone and sandstone than in comparatively fertile basalt- and shale-derived soils. PHB allocation was positively correlated with belowground biological phosphorus (P)-demand, while NLFA allocation was positively correlated with fungal PLFA : bacterial PLFA ratios. A complementary incubation revealed positive responses of respiration, storage, and fungal PLFAs to glucose, while bacterial PLFAs responded positively to PO43-. By comparing these results to a model of microbial C-allocation, we reason that NLFA primarily served the “reserve” storage mode for C-limited taxa (i.e., fungi), while the variable portion of PHB likely served as “surplus” C-storage for P-limited bacteria. Thus, our findings reveal a convergence of community-level processes (i.e., changes in taxonomic composition that underpin reserve-mode storage dynamics) and intracellular mechanisms (e.g., physiological plasticity of surplus-mode storage) that drives strong, predictable community-level microbial C-storage dynamics across gradients of soil fertility and substrate stoichiometry.


Statistical analyses
We first tested the significance of soil type and variables related to soil fertility and microbial C and P demand as predictors of NLFA, PHB, and trehalose, while using PLFA-C as a covariate to account for collinearity between microbial storage-C and microbial biomass.We used the 'lme4' package [12] to construct nested linear mixed effects models of each storage compound.Models containing the predictor of interest were compared to equivalent, nested null models that lacked the predictor of interest using F-tests in which degrees of freedom were approximated by the Kenward-Roger method [13].Base null models consisted solely of a random intercept term for sampling site in order to account for non-independence of samples collected from the same site.The significance of total microbial PLFA-C as a covariate (total bacterial PLFA-C in the case of PHB, which is not synthesised by eukaryotes) was tested against these null models.PLFA-C was a significant, positive predictor of each putative storage compound (P < 0.05 in all cases).Thus, predictors of interest (soil type and continuous variables related to soil C and P availability and microbial C and P demand) were subsequently tested against models containing terms for sampling site as a random intercept and PLFA-C as a covariate.This approach enables conservative estimation of the effects of soil C and P availability and soil microbial C and P demand on the tendency for soil microorganisms to allocate C to storage compounds while accounting for the overall influence of site and the absolute quantity of microbial biomass C. The use of PLFA-C as a covariate is conceptually similar, but statistically preferable [14][15][16], to the use of neutral lipid : polar lipid ratios employed in studies of fungal allocation of C to storage structures [17,18].
Next, to understand better the potential influence of microbial community composition as a factor influencing the dynamics microbial C-storage across our soil fertility gradient, we used an equivalent modelling approach to test the significance of fungal PLFA : bacterial PLFA ratios (molar basis) as predictors of NLFA and trehalose allocation, and Gram positive (Firmicutes + Actinomycetes) bacterial PLFA : Gram negative bacterial PLFA ratios as a predictor of PHB allocation.We subsequently compared fungal PLFA : bacterial PLFA ratios among soil types using linear mixed effect models, along with overall microbial PLFA composition using non-metric multidimensional scaling analysis (based on two dimensions and Bray-Curtis dissimilarities) and permutational analysis of variance (PERMANOVA) using 'vegan' [19].Molar quantities of each microbial PLFA were standardised (i.e., converted to relative molar quantities) and log-transformed (loge[x+1]) prior to these multivariate analyses, which were carried out on the level of individual soil samples (n = 6 for each soil type; 24 observations in total).
The effects of incubation treatments on total MBC and the molar contents of microbial, bacterial, and fungal PLFA biomarkers after ten days of incubation were evaluated using factorial ANOVAs.The effect of incubation treatments on NLFA and PHB were likewise assessed using factorial ANOVAs with an additional covariate term for total microbial PLFA-C in the case of NLFA and total bacterial PLFA-C in the case of PHB.The response of soil respiration was analysed with a factorial ANOVA that included a term for incubation day as a numeric variable.There was a significant three-way interaction between soil type, amendment, and incubation day, due to marked differences in the temporal dynamics of CO2 respiration under glucose-addition between the basalt-derived and sandstone-derived soil.Thus, to simplify interpretation, we re-analysed respiration for each soil type separately, and we focus on those results hereafter.

Stoichiometric model of microbial carbon allocation
Full details of this model are provided in Manzoni, et al. (2021 [ref. 19]).In the model, C is partitioned between growth and storage based on the two different storage use modes.In the reserve storage mode, C that is taken up by micro-organisms is converted to new biomass or storage according to a fixed proportion, and C in excess of stoichiometric requirements is respired (overflow respiration); stored C is remobilized according to prescribed kinetics.In the surplus mode, C that is taken up is preferentially allocated to storage when it is in stoichiometric excess with respect to P; stored C is then remobilized when the external supply of organic C is lower than the stoichiometric requirements.In both modes, P is only immobilized from the inorganic P compartment and P in excess is mineralized.Microbial respiration includes both growth and maintenance components (the latter is neglected here for simplicity), along with overflow respiration to compensate stoichiometric imbalances.
The main assumption of this model is that microbial biomass must grow at a fixed 'structural' C:P ratio, but it can vary its overall biomass composition by means of intracellular C storage.In the model, by balancing C and P allocation so that the 'structural' biomass C:P is fixed, all the C and P transfer rates can be calculated as a function of the substrate C and P contents.For the purpose of this work, we consider an organic C substrate and inorganic P, whose relative proportions are changed to explore how adding organic C or inorganic P alters microbial C allocation.Moreover, we focus on storage of C only (i.e., potential P-storage is not considered).The model is only used to describe the rates of C and P transfer between the substrate and microbial compartments for given compartment sizes.In this way we simulate how a perturbation in the substrate elemental composition affects the fluxes in the short-term time span of a laboratory experiment.

Figure S2 .
Figure S2.Estimated marginal mean values (±95% confidence intervals) of (a) microbial biomass carbon (C), (b) total microbial polar lipid fatty acids (PLFAs), (c) bacterial PLFAs, (d) fungal PLFAs, (e) neutral lipid fatty acid (NLFA)-derived C for a given value of PLFA-C (which serves as a proxy of non-storage microbial biomass C), and (f) polyhydroxybutyrate (PHB)-derived C for a given value of PLFA-C after ten days of incubation with added glucose or PO4 3-(added as NaH2PO4) for basalt-and sandstone-derived soils (P-values show significant sources of variation in log-transformed response variables according to linear models; n = 5 for each treatment combination; 30 observations in total), and (g) proportional and (h) absolute levels of storage compound-derived C in basalt and sandstone soils under the respective incubation treatments.Note that y-axes on panels 'a'-'f' are on a log scale.Legend in panel 'b' applies to panels 'a'-'f', legend in panel 'h' applies to panels 'g' and 'h'.

Figure S3 .Figure S4 .
Figure S3.Two-dimensional non-metric multidimensional scaling analysis of polar lipid fatty acid (PLFA) composition (based on relative molar concentrations) of individual samples of soils used in an incubation experiment in which a sandstone-derived soil and a basaltderived soil were subjected to additions of carbon (as glucose; +C) and phosphorus (as NaH2PO4; +P).Soil samples were collected from West Head, NSW, Australia in March 2022.Permutational analysis of variance (PERMANOVA) P values indicate the significance of soil type and incubation treatment as sources of variation in PLFA composition (n = 5 for each soil type × treatment combination).PLFA composition provides insight into the taxonomic composition of the living soil microbial biomass, with PLFA nomenclature and taxonomic assignations following Joergensen (2022).

Table S1 .
Model parameters (see detailed equations and explanations in Manzoni et al,.2021[ref.19])and measured compartment contents and C:P ratios (averages of two sites for each parent material, from Table1).