## Introduction

Drought alters plant productivity [1], soil microbial biomass [2, 3] and community composition [4, 5], greenhouse gas emissions [6], and many critical biogeochemical processes [7]. Plant–microbial mutualisms mitigate plant drought response and may aid in post-drought recovery through a variety of mechanisms [8, 9]. In particular, mutualistic root-associated fungi—such as mycorrhizal fungi, which form a symbiosis with most terrestrial plant families [10]—can support plants during drought by facilitating water transport [11], soil aggregation [12], root growth [13], plant nutrient uptake [14], photosynthesis [15, 16], and stomatal conductance [15,16,17]. While mycorrhizal fungi may also influence the microbial communities that mediate nutrient cycling and other processes in drought-affected soil, multipartite plant–fungal–bacterial feedbacks remain poorly quantified [9].

The soil hyphosphere—the region that surrounds fungal hyphae—is a hotspot for fungal–bacterial interactions that influence microbial community composition [18,19,20,21,22], nutrient cycling [20,21,22,23,24,25,26,27], and plant growth [27, 28]. Interactions between soil bacteria and plant-associated hyphae are likely shaped by resource dynamics. Plants share up to 20% of photosynthates with their mycorrhizal symbionts [29, 30], which can rapidly transport these resources to surrounding bacteria [24, 31]. Because fungi may explore a volume of soil that is two orders of magnitude greater than the area explored by plant roots [32], they may exert a substantial effect on soil microbiome structure and function. Distinct bacterial communities form in proximity to different fungal lineages [18, 19, 31]. However, we have a limited understanding of how different plant-associated fungi may shape the soil microbiome’s response to environmental stress.

Investigation of fungal–bacterial interactions in hyphosphere soil is methodologically challenging because the hyphosphere is small, dynamic, and may not exert a detectable effect on soil microbial community composition or activity when assessed at a “bulk” scale [18]. Furthermore, a substantial proportion of soil DNA may represent inactive organisms or extracellular “relic” DNA [33, 34]. Stable isotope probing (SIP) coupled with 16S rRNA gene profiling can distinguish active organisms from those that are inactive by tracing isotope incorporation into newly synthesized microbial DNA [35, 36]. With quantitative SIP (qSIP), we can estimate taxon-specific growth based on shifts in DNA buoyant density caused by heavy isotope incorporation [37, 38]. This enables sensitive detection of actively growing taxa, even in samples where a substantial quantity of DNA belongs to dead or dormant organisms. In experimentally challenging environments like the hyphosphere, qSIP has the potential to provide novel insight into taxon-specific activity and microbial growth potential following experimental treatments.

In this study, we investigated how two plant-associated fungal lineages—the arbuscular mycorrhizal (AM) fungus Rhizophagus irregularis and the Sebacinales fungus Serendipita bescii—mediate bacterial growth potential following water limitation in a marginal soil planted with Panicum hallii (Hall’s panicgrass), a model species closely related to the bioenergy crop switchgrass. Both AM and Sebacinales fungi associate with a wide range of plant species, including switchgrass [39,40,41], and support plant growth and nutrition during drought [42,43,44]. However, R. irregularis has a reduced enzymatic repertoire [45, 46] and depends largely upon host-provided C and microbial transformation of nutrient sources into bioavailable forms [27, 47, 48]. In contrast, S. bescii and other Serendipita lineages are facultative symbionts with a broader enzymatic repertoire that enables direct resource acquisition from both living plants and detritus [49,50,51]. We hypothesized that both fungi would mitigate the effects of moisture limitation, but that bacterial community composition and growth potential would be distinct in the soils colonized by each fungus.

## Materials and methods

### Soil collection and characterization

Soil used for this study was a Pond Creek fine sandy loam, classified as a superactive, thermic Pachic Argiustoll [52]. Soil was collected from a pasture in Caddo County, OK (35.072417/−98.303667) where switchgrass is endemic, on traditional land of the Anadarko (Nadaco) tribe and the Caddo Nation of Oklahoma. Previous work has characterized this soil as “marginal” due to its high sand content (69%), low pH (~5), and low C, N, and P content (<0.4%, <0.04%, and <6 ppm, respectively) (ref. [53]). Surface soil (0-20 cm) was collected in May 2019, transported to Livermore, CA, sieved to 2 mm, and stored at 4 °C. A soil moisture retention curve was generated from air-dried soil using a pressure plate apparatus (WP4C, METER Environment, Pullman, WA) and the nonlinear fitting program SWRC-Fit to apply a Brooks and Corey model to the data [54].

### Fungal inoculum

Spores of R. irregularis (formerly Glomus intraradices) isolate DAOM-197198 were purchased from Premier Tech (Rivière-du-Loup, Quebec, Canada). S. bescii (sourced from the Noble Research Institute) was grown in modified Melin-Norkran’s (MMN) broth. Bentonite clay particles were mixed with MMN broth, inoculated with S. bescii, and incubated at 24 °C for 8 weeks according to methods described by Ray et al. (ref. [55]). Uninoculated bentonite clay particles were also mixed with MMN broth and incubated under the same conditions.

### Greenhouse water limitation experiment

P. hallii seeds collected from the Edwards Plateau in central Texas were grown to seed, scarified, surface sterilized, stratified at 4 °C for 1 week, and germinated in Petri plates. After 5 days, germinated seedlings were transferred into planting cones (Ray Leach Cone-tainers, Steuwe & Sons, Tangent, OR, USA) filled with double-autoclaved sand. Five days after transfer into cones, ~500 R. irregularis spores or 500 µL of S. bescii inoculum were injected into the sand to inoculate the P. hallii seedlings. A subset of seedlings was left uninoculated. Eight weeks after inoculation, seedlings were transplanted into 960 cm3 containers (Anderson Plant Bands, Steuwe & Sons) filled with a 50:50 (v:v) mixture of live soil and double-autoclaved sand (CEMEX Lapis Lustre Specialty Sand, #2/12) packed to 1.7 g mL−1 bulk density at 15% moisture. Additional inocula (consisting of 500 R. irregularis spores or 5 g bentonite clay particles coated with S. bescii) were placed near the roots of seedlings as they were transplanted. Five g of uncoated bentonite clay particles were placed near the roots of previously uninoculated plants and plants that had been inoculated with R. irregularis. Each microcosm contained a 25 µm mesh hyphal ingrowth core (2.3 cm diameter) filled with 75 g of the same live soil (no sand) mixed with 0.5 g finely milled switchgrass biomass as bait for the fungal inocula. The multi-phase fungal inoculation procedure was intended to give R. irregularis and S. bescii a colonization advantage over native fungal endophytes present in the soil.

After assembly, each microcosm was covered with 150 g double-autoclaved sand to inhibit cross-contamination of fungal inocula. All microcosms were watered with a total of 20 mL ultrapure H2O during the first week to facilitate plant and fungal establishment. Half of the microcosms were watered on a weekly or bi-weekly schedule to maintain soil moisture at approximately 15% (based on gravimetric measurements), a level that would not restrict plant growth. The other half of the microcosms were not watered for the remainder of the experiment and declined to 5% soil moisture by 3 months. In a subset of microcosms, the moisture content of the soil surrounding the roots (i.e., not within the hyphal ingrowth cores) was monitored continuously with a volumetric water probe (ECH2O EC-5, METER Group, Inc., Pullman, WA, USA). Another microcosm per treatment was weighed weekly to assess whole-microcosm gravimetric changes. Three replicate microcosms were maintained for each of the six treatment combinations (three fungal inoculum conditions × two moisture regimes). Average daytime and nighttime temperatures were 27 °C and 24 °C, respectively, with a photoperiod of 16 h. After 3 months, the microcosms were destructively harvested. Soil from hyphal ingrowth cores was homogenized, flash frozen in liquid N2, and stored at −80 °C for DNA extractions or at room temperature for qSIP assays (see Fig. 1a for experimental design).

### Soil characteristics

Soil moisture content following the 3-month harvest was determined by mass difference after drying soil for 48 h at 105 °C. Total soil C and N contents were measured with a Costech ECS 4010 Elemental Analyzer (Costech Analytical Technologies Inc., Valencia, CA).

### Soil H218O qSIP assay

To determine bacterial growth potential following water limitation, we conducted an H218O qSIP assay [35, 37, 38]. Three days after the 3-month harvest, soil from three hyphal ingrowth cores per treatment was amended with either H216O or H218O (Fig. 1a). Each soil sample was divided into three subsamples: one 3.0 (±0.2) g dry weight equivalent sample for an initial H216O qSIP timepoint (T0) and two 4.0 (±0.2) g dry weight equivalent samples for the H216O and H218O 7-day qSIP timepoint (T7). Two of the T0 samples contained ~2.0 g due to limited soil availability. This resulted in a total of 54 samples (18 T0 H216O samples, 18 T7 H216O samples, and 18 T7 H218O samples). Each subsample was air-dried to 4.7% gravimetric water content in a biosafety cabinet and then brought up to 22.1% gravimetric water content (60% field capacity) with either ultrapure water at natural isotopic abundance or water enriched with 18O (98.38 atom % H218O, Isoflex, San Francisco, CA, USA). The final estimated enrichment of the samples containing H218O was 78.76 18O atom %. We standardized the soil moisture and soil water isotopic enrichment across treatments to minimize differences across fungal treatments during the growth potential assay. The T0 subsample (amended with H216O) was immediately flash frozen in liquid N2 and stored at −80 °C. Each of the T7 soil samples was stored in the dark inside a separate 473.2 mL glass jar with a tight-fitting lid containing a septum for gas sampling. After 7 days, the soils were flash frozen in liquid N2 and stored at −80 °C.

### CO2 efflux measurement

15 mL headspace gas samples were collected at the beginning and end of the qSIP assay. CO2 concentrations were measured on a gas chromatograph equipped with a thermal conductivity detector (GC-14A, Shimadzu, Columbia, MD).

### DNA extraction and density gradient fractionation

DNA was extracted from each soil sample in quadruplicate with the DNeasy PowerSoil Pro kit (Qiagen, Germantown, MD, USA) and then pooled per sample prior to downstream analysis. To separate isotopically enriched DNA from unenriched DNA, samples were subjected to a cesium chloride density gradient formed in an ultracentrifuge as previously described [56, 57] with the following minor modifications. For each sample, 5 µg of DNA in 150 µL 1xTE buffer was mixed with 1.00 mL gradient buffer, and 4.60 mL CsCl stock (1.885 g mL−1) with a final average density of 1.730 g mL−1. Samples were loaded into 5.2 mL ultracentrifuge tubes and spun at 20 °C for 108 h at 176,284 RCFavg (equivalent to 176,284 × g) in a Beckman Coulter Optima XE-90 ultracentrifuge using a VTi65.2 rotor. Sample fractionation was automated using Lawrence Livermore National Laboratory’s high-throughput SIP pipeline, which automates the fractionation and clean-up tasks for the density gradient SIP protocol [57]. The content of each ultracentrifuge tube was fractionated into 22 fractions (~236 µL each) using an Agilent Technologies 1260 isocratic pump to deliver water at 0.25 mL min−1 through a 25 G needle inserted through the top of the ultracentrifuge tube. Each tube was mounted in a Beckman Coulter fraction recovery system with a side port needle inserted through the bottom. The side port needle was routed to an Agilent 1260 Infinity fraction collector, and fractions were collected in 96-well deep well plates. The density of each fraction was measured using a Reichart AR200 digital refractometer fitted with a prism covering to facilitate measurement from 5 µL, as previously described [58]. DNA in each fraction was purified and concentrated using a Hamilton Microlab Star liquid handling system programmed to automate glycogen/PEG precipitations [36] (GlycoBlue Coprecipitant, Invitrogen by Thermo Fisher Scientific, Waltham, MA, USA). Washed DNA pellets were suspended in 40 µL of 1xTE buffer and the DNA concentration of each fraction was quantified using a PicoGreen fluorescence assay [57].

### Quantitative PCR (qPCR) to assess bacterial and fungal abundance

We conducted qPCR with “universal” bacterial primers [59, 60] to measure 16S rRNA gene copy number in each unfractionated and SIP-fractionated DNA sample (see Table S1 and SI text for additional information). To assess the relative abundance of fungal inocula, we quantified R. irregularis and S. bescii gene copy number in unfractionated DNA (extracted in triplicate prior to the H218O qSIP assay) with qPCR primers designed specifically for each fungal lineage [55, 61] (see Table S1 and SI text for additional information). We note that DNA can be unevenly distributed throughout fungal biomass. While fungal DNA copy number may not necessarily be indicative of total biomass or activity, it is a relative indicator of inoculation success [62, 63].

### DNA sequencing

For each sample, we sequenced the 16S rRNA gene in unfractionated DNA as well as DNA fractions containing >1.0 ng µL−1, resulting in 6–11 sequenced fractions per sample. The V4 region of the 16S rRNA gene was amplified with the 515F/806R primer pair [59, 60], processed and barcoded through the Illumina V2 PE150 sample preparation kit, and sequenced on a MiSeq v2 platform in three runs (Illumina, Inc., San Diego, CA; see SI text and Tables S2, S3 for additional information about taxonomic assignment and amplicon sequence variant (ASV) recovery).

### Quantitative stable isotope probing (qSIP) analysis

We calculated taxon-specific growth potential using the H218O qSIP approach [35, 37, 38]. Briefly, the qSIP mathematical model estimates taxon-specific 18O incorporation into DNA based on the shift in DNA density following exposure to natural abundance H2O (H216O) or “heavy” H218O. We determined baseline densities for each taxon in the absence of an isotopic tracer (i.e., samples amended with H216O), because even without tracer assimilation, DNA density varies by GC content [64]. For each ASV, we calculated median 18O atom percent excess (APE) based on the difference in the buoyant density of 16S rRNA gene profiles sequenced from soils amended with H216O or H218O. We assumed linear population growth and used the change in taxon-specific 18O enrichment and 16S rRNA gene abundance over the course of the qSIP assay (i.e., at T0 and at T7) to estimate taxon-specific growth potential with the equation below. In this equation, absolute growth rate for taxon i is estimated as bi, where for each taxon i at the end of the qSIP assay (time t), NTOTALit and NLIGHTit are the total (unlabeled + 18O-labeled) and unlabeled 16 S rRNA gene abundances, respectively.

$$b_i \,=\, \frac{{N_{TOTAL_{it}} \,-\, N_{LIGHT_{it}}}}{t}$$

We estimated 16S rRNA gene abundance by multiplying the relative abundance of each ASV’s 16S rRNA gene by the total 16S rRNA gene copy number measured in each DNA fraction with qPCR. These calculations rest upon the assumptions that (a) bacterial DNA synthesis is proportional to cellular growth and division, (b) for each ASV’s population, all new bacterial DNA is synthesized with an equal proportion of 18O, and (c) bacterial taxa have an average of six 16S rRNA gene copies per cell [37]. We note that qSIP-based growth estimates are approximations due to several methodological challenges, including incomplete extraction of microbial DNA from soil [65], inter-taxonomic variation in 16S rRNA gene copies per cell [66], and amplification and sequencing biases [67, 68]. To increase confidence in these estimates, we first filtered out ASVs that were not recovered from all replicates of a particular treatment. This stringent filtering ensured that our estimates were representative of the three independent biological replicates. To quantify uncertainty around each estimate of taxon-specific 18O APE and growth, we calculated 90% confidence intervals using a bootstrapping procedure with 1000 iterations [37].

18O APE represents an integrated measure of new DNA synthesis that occurred during the qSIP assay. We note that while an increase in a taxon’s 18O APE indicates new growth, the overall population size might be increasing in abundance, remaining static, or declining depending on whether or not growth outpaces mortality. During population turnover, DNA 18O APE can increase while population-level abundance either increases (if growth outpaces mortality), remains similar (if growth matches mortality), or declines (if mortality outpaces growth). Therefore, a high 18O APE value is not necessarily indicative of a net increase in abundance. In our study, because we maintained all soils at 22.1% moisture during the qSIP assay, growth estimates represent bacterial growth potential immediately following 12 weeks of exposure to contrasting moisture regimes, rather than actual growth rates under water-limited or water-replete conditions.

We calculated a metric of growth efficiency by dividing gross bacterial growth by CO2 efflux per treatment. Microbial growth efficiency can be defined as the proportion of biomass synthesized per unit C assimilated [69] or the proportion of C allocated towards growth rather than towards other activities [70]. We consider the ratio of new DNA synthesis to CO2 efflux to be a proxy for growth efficiency since DNA synthesis is proportional to new microbial biomass production, and since CO2 efflux represents respiration from microbial activities (including both growth and non-growth processes). Based on this metric, a higher proportion of non-growth activities (such as metabolism, osmoregulation, motility, energy spilling reactions, O2 stress responses, and other maintenance activities) results in a decrease in growth efficiency.

### Statistical analyses

We performed all statistical analyses in R [71]. We assessed differences in fungal inoculum gene copy number with a nonparametric Kruskal–Wallis rank sum test. We assessed differences in soil moisture, soil C and N content, DNA extraction efficiency, Inverse Simpson’s diversity index, median ASV 18O APE, and CO2 efflux with a Tukey’s HSD test comparing the means of all treatments. First, we created a linear model with a fixed effect for treatment. If the raw data did not meet the assumptions of normality, it was log-transformed prior to further analysis.

We used the R package vegan to perform a principal coordinates analysis and permutational multivariate analysis of variance (PERMANOVA) on the weighted UniFrac distance between communities present following different moisture and fungal treatment combinations and at different qSIP timepoints [72, 73].

We used the 90% confidence interval around each ASV’s median 18O APE to identify ASVs that incorporated significant quantities of 18O into their DNA: only ASVs whose lower 90% confidence interval did not overlap with zero were considered to be actively growing. To compare taxon-specific growth potential between two treatments, we calculated pairwise ratios of 18O-based growth potential for taxa that were actively growing under both conditions. We calculated these ratios by dividing an ASV’s 18O APE following one fungal*moisture treatment by its 18O APE following the other fungal*moisture treatment. For example:

$${{{{{{{\mathrm{ASV}}}}}}}}\;^{18}{{{{{{{\mathrm{O}}}}}}}}\;{{{{{{{\mathrm{APE}}}}}}}}\;{{{{{{{\mathrm{Ratio}}}}}}}} \,=\, \frac{{{{{{{{{\mathrm{ASV}}}}}}}}\;^{18}{{{{{{{\mathrm{O}}}}}}}}\;{{{{{{{\mathrm{APE}}}}}}}}\;{{{{{{{\mathrm{in}}}}}}}}\;{{{{{{{\mathrm{water}}}}}}}}{\mbox{-}}{{{{{{{\mathrm{limited}}}}}}}}\;{{{{{{{\mathrm{soil}}}}}}}}\;{{{{{{{\mathrm{with}}}}}}}}\;R.\;irregularis}}{{{{{{{{{\mathrm{ASV}}}}}}}}\;^{18}{{{{{{{\mathrm{O}}}}}}}}\;{{{{{{{\mathrm{APE}}}}}}}}\;{{{{{{{\mathrm{in}}}}}}}}\;{{{{{{{\mathrm{water}}}}}}}}{\mbox{-}}{{{{{{{\mathrm{replete}}}}}}}}\;{{{{{{{\mathrm{soil}}}}}}}}\;{{{{{{{\mathrm{with}}}}}}}}\;R.\;irregularis}}$$

To assess whether these ratios were significantly greater than or less than one at the phylum level (indicating a significant difference between growth potential in soils maintained under different fungal or moisture conditions), we conducted a Wilcoxon signed rank test with a Benjamini–Hochberg correction for multiple comparisons on the average of the ASV 18O APE ratios within each phylum.

## Results

### Plant growth, fungal abundance, and soil characteristics following 3-month water manipulation

After 3 months of growth, P. hallii plants reached similar mass across all fungal*moisture treatments (Fig. S1; p > 0.05). R. irregularis and S. bescii abundances (measured with strain-specific qPCR primers) were higher in the hyphal ingrowth cores of microcosms inoculated with each fungus than in the ingrowth cores of microcosms that were not inoculated (Tables S4, S5, Fig. 1b, c; p < 0.001). Soil moisture within the water-replete hyphal ingrowth cores was more than three times higher than the moisture content within the water-limited hyphal ingrowth cores (Fig. 1d; p < 0.001). Matric suction was higher in water-limited soils, but did not reach a point likely to cause a loss in cell turgor [74] (Fig. S2). Neither plant biomass nor fungal inoculum abundance was associated with differences in soil moisture (Figs. S1, 1d; p > 0.05). Total soil C, N, and DNA extracted did not vary between treatments (Figs. S3, S4a; p > 0.05).

### Effects of water limitation and fungal inoculum on bacterial community structure

Moisture history and fungal inoculum shaped the structure of the bacterial communities present in hyphal ingrowth core soils (Table S6 and Fig. 2a; p < 0.001, PERMANOVA of weighted UniFrac distances). We detected only a small shift in bacterial community structure between the beginning and end of the 7-day qSIP assay in the unfractionated DNA samples (6.2% of variance explained; Table S7; p < 0.001). With this traditional 16S rRNA gene profiling of unfractionated DNA (sequenced from soils collected at the end of the qSIP assay), we found that moisture regime and fungal inoculum explained only 30.0% of the variation between “total” bacterial communities (i.e., bacteria identified in the total DNA pool, which many include dead and dormant organisms).

We used H218O qSIP to filter out non-replicating ASVs (i.e., those that did not incorporate significant quantities of 18O into their DNA) and identify the actively growing subset of the total bacterial community. In this subset, we found that moisture history, fungal inoculum, and their interactive effect explained 86% of the variation in community structure (11.9%, 49.5%, and 24.6%, respectively; Table S8 and Fig. 2b; p < 0.001, PERMANOVA). This is more than twice the effect size observed through traditional 16S rRNA gene profiling. qSIP-based filtering also highlighted differences in bacterial alpha diversity. There was no relationship between moisture history or fungal inoculum and the alpha diversity of the total communities sequenced (Fig. 2c; p > 0.05). However, for the actively growing communities, water limitation was associated with lower alpha diversity in uninoculated soils (Fig. 2d; p < 0.01), no difference in R. irregularis-inoculated soils (Fig. 2d; p > 0.05), and higher alpha diversity in S. bescii-inoculated soils (Fig. 2d; p < 0.01).

Overall, DNA from soil amended with H218O was denser than DNA from soil amended with H216O (Fig. S4b), indicating that many organisms incorporated 18O into newly synthesized DNA. Our qSIP filtering parameters removed ASVs that were not identified in all replicates of a particular treatment, resulting in 989-1,153 bacterial ASVs per treatment (Table S3). The ASVs that remained following this stringent filtering step represent 62-73% of the ASVs sequenced per treatment in unfractionated DNA and 21–25% of the ASVs sequenced per treatment in the SIP-fractionated DNA (Tables S2, S3). Of the ASVs that remained after qSIP filtering, taxon-specific 18O APE ranged from unenriched to 64.2 APE, with a median of 6.1. Between 30 and 60% of the total bacterial ASVs detected per treatment were enriched (Table S3; lower 90% CI > 0); we refer to these as actively growing taxa. We note that although the stringent filtering parameters increase confidence in the taxon-specific APE estimates, these may result in an overestimate of the proportion of actively growing taxa if less abundant taxa have lower growth rates and are disproportionately filtered from the dataset.

### Effects of water limitation and fungal inoculum on bacterial growth potential and growth efficiency

We used H218O qSIP and CO2 efflux measurements to quantify the effects of moisture history and fungal inoculum on bacterial growth potential and growth efficiency. In uninoculated soils, water limitation was associated with more than a 50% reduction in median ASV 18O APE (Fig. 3a; p < 0.05). Similarly, bacterial gross growth potential in uninoculated water-limited soil was less than one-third of gross growth potential in water-replete soil (Fig. 3b). In uninoculated soils, moisture history did not result in a difference in total CO2 efflux during the qSIP assay (Fig. 3c; p > 0.05). Because bacterial growth potential was substantially lower following water limitation, this translated into a 72% reduction in growth efficiency (gross growth potential divided by CO2 efflux) in uninoculated soils (Table S9).

In fungal-inoculated soils, bacterial communities were less affected by water limitation than in uninoculated soils. Moisture history was not associated with a difference in bacterial ASV 18O APE or potential gross growth in soil inoculated with either R. irregularis or S. bescii (Fig. 3a, b; p > 0.05 for 18O APE). CO2 efflux from S. bescii-inoculated soils was higher than from R. irregularis-inoculated soils, but the difference was only significant between the water-replete soils (Fig. 3c; p < 0.05). Compared to uninoculated soils, water limitation in fungal-inoculated soils had a less severe effect on potential bacterial growth efficiency, leading to a 17% reduction in R. irregularis-inoculated soils and a 37% reduction in S. bescii-inoculated soils (Table S9).

### Taxon-specific response to water limitation and fungal inoculum

Since compensatory dynamics may obscure taxon-specific response to different conditions [75], we calculated ratios of 18O APE for each actively growing ASV detected under two conditions. For example, for ASVs detected in both water-limited and water-replete soil, a ratio of greater than 1.0 suggests that the ASV could sustain higher growth potential in water-limited soil, a ratio of less than 1.0 suggests that the ASV could sustain higher growth potential in water-replete soil, and a ratio indistinguishable from 1.0 suggests that the ASV could maintain similar growth potential regardless of moisture history. In the absence of fungal inocula, most ASVs were suppressed by water limitation (Fig. 4a). When averaged at the phylum level, 18O APE ratios were significantly less than 1.0 for six of the ten most abundant phyla: Actinobacteria, Bacteroidetes, Chloroflexi, Gemmatimonadetes, Proteobacteria, and Verrucomicrobia (Fig. 4a; adjusted p < 0.05). In contrast, bacterial growth potential in fungal-inoculated soils was less strongly affected by moisture history: while water limitation was associated with lower growth potential for many ASVs, many others sustained similar or greater growth potential. When averaged at the phylum level, populations belonging to the phyla Actinobacteria and Gemmatimonadetes were suppressed in water-limited soil inoculated with R. irregularis (Fig. 4b; adjusted p < 0.05), while Verrucomicrobia were suppressed in water-limited soil inoculated with S. bescii (Fig. 4c; adjusted p < 0.05). Only the Acidobacteria present in R. irregularis-inoculated soil sustained higher growth potential following water limitation (Fig. 4b; adjusted p < 0.05).

To assess whether each fungus promoted or suppressed growth potential of specific bacteria in water-limited soil, we calculated 18O APE ratios for ASVs detected in both uninoculated and fungal-inoculated soils following water limitation. Most ASVs sustained higher growth potential in R. irregularis- or S. bescii-inoculated soil than in uninoculated soil (Figs. 5, S5). Averaged at the phylum level, populations belonging to the phyla Acidobacteria, Actinobacteria, Bacteroidetes, Chloroflexi, Planctomycetes, Proteobacteria, and Verrucomicrobia were significantly more active in R. irregularis-inoculated soil (Fig. 5a; adjusted p < 0.05). In S. bescii-inoculated soils, only the population belonging to the phylum Actinobacteria was significantly more active than in uninoculated soil (Fig. 5b; adjusted p < 0.05). Some taxa responded positively to both R. irregularis and S. bescii; others responded positively only to one fungal lineage (Fig. 5c–e). Overall, the magnitude of positive growth response was greater in R. irregularis- relative to S. bescii-inoculated soils.

Several bacterial ASVs incorporated significant quantities of 18O into their DNA under some conditions, but not under others, as indicated by differences in gross growth potential measured through the H218O qSIP assay (Fig. S6–S15). For any ASV that did not incorporate a significant quantity of 18O into its DNA under a particular moisture*fungal condition, it was not possible to calculate an 18O APE ratio to compare the ASV’s growth potential under that moisture*fungal condition to its growth potential under another condition. This was most apparent for the Firmicutes in uninoculated soil: none of the Firmicutes ASVs incorporated significant quantities of 18O into their DNA following water limitation (Fig. S11). Therefore, the effect of the fungal inocula on Firmicutes growth potential following water limitation is not represented in the 18O APE ratios (Fig. 5a, b; no Firmicutes bar shown). However, several Firmicutes incorporated significant quantities of 18O into their DNA in fungal-inoculated soils following moisture limitation (Fig. S11). More than double the number of Firmicutes ASVs were active in R. irregularis- compared to S. bescii-inoculated soils.

## Discussion

We found that plant-associated fungi have a protective effect on bacterial communities exposed to water limitation, and that bacterial responses to different fungal lineages are distinct. Because plant biomass was similar across all conditions investigated, we attribute differences in bacterial community composition and growth potential to direct effects of moisture history and fungal inocula, rather than to indirect effects mediated by plants. We found that H218O qSIP highlighted treatment differences that were not apparent through traditional 16S rRNA gene profiling. This demonstrates the utility of DNA qSIP for investigation of the soil hyphosphere and other systems in which it is difficult to discern a microbial signal above a complex background community.

### Soil water limitation suppresses bacterial growth potential and growth efficiency

We observed a significant decrease in bacterial growth potential following 3 months of moisture limitation, but little effect of moisture history on respiration potential. Together, these responses resulted in a substantial reduction in bacterial growth efficiency. This may reflect a trade-off between microbial stress tolerance and growth in drought-affected soil [76]. Reduced microbial growth efficiency can occur when biota prioritize essential metabolic activities over cellular growth and replication [7, 56, 76,77,78]. Slowed growth also helps bacteria persist in the presence of antibiotics [79, 80], which can accumulate in dry soils as microorganisms compete for limited resources [81].

We measured lower growth potential across a broad range of bacterial lineages present in our water-limited soils. Interestingly, this included several monoderm taxa (often referred to as Gram-positive organisms) belonging to the phyla Actinobacteria, Firmicutes, and Chloroflexi. Monoderm taxa have thick cell walls and lack an outer membrane, which can protect them against oxidative damage under dry conditions [7, 82,83,84,85]. Several studies report that monoderms maintain greater abundance and activity in dry soils compared to diderms (also known as Gram-negative organisms, such as most of the taxa belonging to the phyla Acidobacteria, Bacteroidetes, Proteobacteria, and Verrucomicrobia) [7, 84, 85]. In addition to the protection conferred by the structure of their cellular envelope, monoderms may also be poised to outcompete other organisms in drought-affected soils through their capacity to produce antibiotics [81, 84, 86] or utilize complex C substrates that remain available following water limitation [2, 81]. Our results suggest that even monoderm taxa that are considered relatively drought tolerant can be negatively affected by water limitation.

### Plant-associated fungi support bacterial resilience in drought-affected soil, but fungal–bacterial relationships are context dependent

While water limitation had a broad suppressive effect in uninoculated soils, many of the bacteria in soils inoculated with either R. irregularis or S. bescii maintained similar growth potential following cultivation under either water-replete or water-limited conditions. This protective fungal effect extended throughout the bacterial community, and affected many taxa that are often considered drought susceptible. Relative abundances of Bacteroidetes, Planctomycetes, Verrucomicrobia, and many Proteobacteria and Acidobacteria have been shown to decrease following drought [4, 5, 87]. However, we found that many ASVs belonging to these phyla sustained similar growth potential in fungal-inoculated soils, regardless of moisture treatment. This suggests that R. irregularis and S. bescii modified edaphic conditions in some way that broadly supported bacterial resilience to water limitation. Plant-associated fungi can exude plant-derived C [24, 29,30,31,32], promote biofilm formation [88], enhance soil aggregation through their interactions with other soil biota [14, 89], and facilitate bacterial transport through soil [90]. Together, these fungal-mediated processes could help maintain soil connectivity, microbial activity, and nutrient cycling under water-limited conditions, thereby preventing bacterial dormancy and death despite a substantial decline in soil moisture. By supporting bacterial function in drought-affected soil, plant-associated fungi may counteract the destabilizing effect of moisture stress [91] and improve capacity for recovery once moisture is restored.

In contrast to their synergistic effects in water-limited soils, R. irregularis and S. bescii appeared to suppress bacterial growth potential following water-replete conditions. This demonstrates that the relationship between plant-associated fungi and hyphosphere bacteria is context-dependent, and not entirely mutualistic. Plant-associated fungi are known to compete with other biota for N [92] and P [20, 21], and can suppress microbial decomposers [23, 93]. Similarly, bacteria can inhibit mycorrhizal proliferation [20, 21, 92]. Putative bacterial predators have also been found in greater abundance on extraradical mycorrhizal hyphae than in surrounding soil [18]. We did not investigate the potential mechanisms of suppressive interactions between bacteria and plant-associated fungi. However, our observation that bacterial growth potential depends on moisture history in fungal-inoculated soil indicates that there are trade-offs between fungal and bacterial growth. In this trade-off, fungi may limit bacterial growth under resource-replete conditions, but promote bacterial growth under resource-limited conditions. Similar context-dependency is well-documented in other mutualisms [94]. In our system, context-dependency may indicate a stabilizing ecological effect exerted by multipartite hyphosphere interactions.

Since R. irregularis and S. bescii were not actively associating with their plant hosts in our H218O qSIP assay, we attribute these results to fungal effects that had occurred during the preceding 3-month greenhouse experiment. Although the qSIP assay might have caused a nutrient flush from the perturbed soil and biota, CO2 efflux from the fungal-inoculated soils was not significantly greater than from the uninoculated soils. Therefore, we conclude that differences in bacterial growth potential, growth efficiency, and diversity of the actively growing community present in fungal-inoculated soil were related to preceding fungal effects on the soil environment rather than to bacterial decomposition of fungal necromass.

### Magnitude of bacterial response is fungal lineage-dependent

While both fungal lineages supported bacterial resilience, R. irregularis elicited a stronger positive response than S. bescii—both with respect to the number of ASVs and the magnitude of individual responses. Distinct microbial consortia associate with different mycorrhizal lineages [18, 19, 31, 89]. Although empirical evidence remains sparse, different mycorrhizal exudate profiles, growth habits, and other functional traits may shape the composition and activity of the surrounding microbial community [48, 89, 95, 96], a phenomenon that has been documented more extensively for root-microbe interactions [97,98,99]. Lower bacterial growth potential and growth efficiency in S. bescii- compared to R. irregularis-inoculated soils may be due to S. bescii’s wider enzymatic repertoire, which could accelerate decomposition (as indicated by higher CO2 efflux) or heighten competitive interactions with other soil biota. Additionally, higher gene copy numbers of R. irregularis compared to S. bescii detected in hyphosphere soil suggest that R. irregularis colonization levels were more robust. Greater fungal proliferation could be correlated with greater resource distribution, enhanced soil structure, or other conditions that support bacterial growth. Together, these findings suggest that diverse fungal lineages promote bacterial resilience to water limitation, but that the individual taxa and magnitude of taxon-specific response to each fungus is distinct.

## Conclusion

As global precipitation patterns change, it is important to understand how drought influences ecological functions and microbial interactions, both during and after water limitation. Plant-associated fungi are known to support plant growth and nutrition in droughted soils, but their simultaneous effects on the soil bacterial communities that mediate nutrient cycling and other critical terrestrial processes remain poorly explored. With H218O qSIP, we show that plant-associated fungi have a protective effect on bacterial communities exposed to water limitation. Both the AM fungus R. irregularis and the Sebacinales fungus S. bescii facilitated greater growth potential, growth efficiency, and diversity of actively growing hyphosphere bacteria in drought-affected soil. While these divergent fungal lineages stimulated responses of differing magnitude, the broad patterns were similar, suggesting that the dominant underlying mechanisms may be conserved across a substantial portion of the bacterial community rather than limited to interactions with a small number of bacterial taxa. Remarkably, both R. irregularis and S. bescii had a protective effect on hyphosphere bacteria exposed to water limitation in a “live” soil, which may have included functionally redundant fungal lineages. This finding is relevant for practical evaluation of fungal inoculants, whose ability to persist and elicit a positive effect in natural settings is not well-established. Additionally, our work indicates that H218O qSIP is a useful approach in challenging systems such as the terrestrial hyphosphere, where microbial dynamics may be difficult to detect with traditional 16S rRNA gene profiling. Together, our findings demonstrate that context-dependent multipartite relationships support bacterial resilience to water limitation and may promote post-drought recovery.