Active populations and growth of soil microorganisms are framed by mean annual precipitation in three California annual grasslands

Earth system models project altered precipitation regimes across much of the globe; in California, the winter wet season is predicted to extend into spring, and the summer dry period to lengthen. How these precipitation trends will affect microbial traits and soil carbon (C) cycling is a key knowledge gap. Specifically, we do not have a mechanistic understanding of the linkages between soil moisture legacy effects, microbial population dynamics and soil C persistence. Using quantitative stable isotope probing (qSIP), we compared total and growing soil microbial communities across three California annual grasslands that span a rainfall gradient yet developed on similar parent material. We also assessed multiple edaphic variables, including the radiocarbon (14C) age of soil C, and found soil C turnover time increased with annual precipitation, but that soil microbes respired recently-fixed C regardless of site rainfall history. Samples were assayed in the wet season, when we expected environmental conditions would be most similar across sites. We hypothesized that growing communities would be more compositionally similar across the gradient than the total background microbiome. We also predicted that the long-term legacy effect of soil water limitation would be reflected in a lower community growth capacity at the driest site. We found that the proportion of the total community that was detected as growing was 28%, 48% and 58% at the wet, intermediate and dry sites, respectively. The composition of growing communities strongly resembled that of total communities, and growing communities were no more similar across the gradient than total communities, indicating a strong effect of climate of the structure of growing microbial communities. Members of three phyla, Acidobacteria, Actinobacteria, and Proteobacteria, were responsible for ∼79% of the cumulative 18O assimilation and 80% of all taxa that we defined as ‘growers’. Bacterial growth rates were low at the driest site relative to the intermediate and wettest sites. Reduced growth at the driest site was observed across major phyla, including the Actinobacteria, Acidobacteria, Bacteroidetes, Gemmatimonadetes and Proteobacteria. Microbial communities at the driest site displayed phylogenetic clustering, suggesting that climate history impacts microbial growth through environmental filtering for slow growing taxa. Taxonomic identity was a strong predictor of growth, such that the growth rates of a taxon at one site predicted its growth rates at the others. This cross-site coherence in growth is likely a consequence of genetically determined physiological traits, and is consistent with the idea that evolutionary history influences growth rate.


Introduction 56
Rainfall patterns and soil water content are crucial controllers of microbial population 57 dynamics (growth and death) and Earth system models project major changes in the timing and 58 intensity of precipitation events globally (Dore, 2005;Schimel, 2018). Since microorganisms 59 mediate a wide range of ecosystem processes, including the accrual and persistence of soil 60 carbon, understanding how the climate history of a site impacts microbial communities is 61 essential. Precipitation is a key controller of ecosystems with Mediterranean climates (cool, wet 62 winters and hot, dry summers) and it plays an equally important role in shaping microbial 63 community structure and function (Angel et al., 2010;Maestre et al., 2015). Microorganisms in 64 Mediterranean ecosystems must withstand both direct physiological stress during prolonged 65 periods of low soil moisture yet be able to compete for resources when seasonal rains return and 66 plant growth resumes (Barnard et al., 2020). differently (Sokol et al., 2022). Populations that are actively turning over (growing and dying) 73 have a prominent role in biogeochemical processes since their growth requires uptake and 74 transformation of environmental substrates, while their death deposits cellular materials into the 75 soil-mineral matrix where it may be subsequently recycled into living biomass or persist as soil 76 organic matter (Liang et al., 2017). It widely accepted that microbial growth is a major factor in 77 the formation of soil carbon (Bradford et al., 2013). The factors controlling which microbial taxa 78 grow in soils and the timing of their growth are not well understood, but active communities 79 have a heightened sensitivity to changes in environmental conditions compared to the total 80 community (Angel et  influence growth as well. Repeated exposure to limited soil moisture may favor traits that confer 87 tolerance to water stress and select against traits that enhance organisms' ability to grow quickly 88 or efficiently (Evans and Wallenstein, 2014;Malik et al., 2020aMalik et al., , 2020b. For example, the 89 production of extracellular polymeric substances (EPS) can be a substantial carbon sink for 90 bacteria growing under water stress, and osmolyte regulation in response to water stress can 91 consume both nutrients and energy (Flemming and Wingender, 2001; Sandhya and Ali, 2015). In 92 bacterial isolates, moisture stress has been shown to drive a tradeoff between EPS production 93 and the length of lag phase, pointing to the relevance of growth rate in water stress acclimation 94 (Lennon et al., 2012). While such traits provide a clear advantage during chronic water stress, 95 there are likely biochemical costs involved in maintaining them. Thus, microbial taxa from soils 96 routinely exposed to low soil water potentials may have acquired traits that lead to slow growth 97 rates even in the absence of immediate water stress. At the community level, the mechanism 98 underlying changes in growth rates may include a shift in microbial community composition, a 99 change in the physiology of microbial taxa, or both (Evans and Wallenstein, 2014). 100 Bromus spp. dominate at Hopland and Sedgwick (Michael Barbour et al., 2007;Leitner et al., 147 2017 Quantitative x-ray diffraction (qXRD) was performed at Lawrence Livermore National 178 Laboratory as previously described (Fossum et al., 2021). Samples for qXRD were crushed and 179 passed through a 500 µm sieve. 3 g of soil was then ground for 5 minutes with 15 mL of 180 methanol using a MCrone mill. The ground samples were transferred to a plastic tray, air dried, 181 and homogenized for 3 min using a vortex mixer with 10 mm plastic beads (Bakker et al., 2018). 182 The soil powders were side loaded into XRD sample mounts. All powdered samples were 183 analyzed on a Bruker D8 advance XRD. The samples were scanned from 3° to 65° 2θ at 0.011° 184 steps with a 5 second per step count time. Quantitative mineralogy was determined using the 185 BGMN Rietveld refinement and the Profex interface software (Doebelin and Kleeberg, 2015 For qSIP incubations, 5 g soil from each field replicate was transferred to 15 ml Nalgene 215 flatbottom vials. Soils were lightly air dried in a laminar flow hood at room temperature for 24 216 hours prior to isotope addition. This drying is necessary to allow for sufficient enrichment of the 217 soil water with the isotope label while also maintaining soil moisture reasonably similar to field 218 moisture during the qSIP incubations. One milliliter of isotopically enriched water (98.15 at% 219 18 O-H 2 O) or natural abundance water (as a control) was pipetted onto the soil slowly and evenly 220 and then gently mixed with the pipette tip (raising soil moisture content by 3-7%). After the 221 water addition, vials were immediately sealed inside 500 ml mason jars and incubated at room 222 temperature in the dark for 8 days. At the end of the incubation, soils were frozen in liquid 223 nitrogen and then stored at -80°C. 224 DNA was extracted from all soil samples using a modified protocol adapted from 225 Barnard et al. (2015). Three replicate extractions were conducted for each sample and then 226 replicate DNA extracts were combined. For each extraction, soil (0.4 g +/-0.001 g soil) was 227 added to 2 ml Lysing Matrix E tube (MP Biomedicals) and extracted twice as follows. 500 µl 228 extraction buffer (5% CTAB, 0.5 M NaCl, 240 mM K 2 HPO 4 , pH 8.0) and 500 µl 25:24:1 229 phenol:chloroform:isoamyl alcohol were added before shaking (FastPrep24, MP Biomedicals: 30 230 s, 5.5 m s-1 ). After centrifugation (16,100 x g, 5 min), residual phenol was removed using pre-231 spun 2 ml Phase Lock Gel tubes (5 Prime, Gaithersburg, MD, USA) with an equal volume of 232 24:1 chloroform:isoamyl alcohol, mixed and centrifuged (16,100 x g, 2 min). The aqueous 233 phases from both extractions were pooled, mixed with 7 µl RNAase (10 mg/ml), mixed by 234 inverting, and incubated at 50 °C for 10 min. 335 µL 7.5 M NH4+ acetate was added, mixed by 235 inverting, incubated (4 °C, 1 h). and centrifuged (16,100 x g, 15 min. The supernatant was 236 transferred to a new 1.7 ml tube and 1 µl Glycoblue (15 mg/ml) and 1 ml 40% PEG 6000 in 1.6 237 M NaCl was added, mixed by vortex, and incubated at room temperature in the dark (2 h). After 238 centrifugation (16,100 x g, 20 min), the pellet was rinsed with 1 ml ice-cold 70% ethanol, air-239 dried, resuspended in 30 µl 1xTE and stored at -80 °C. 240 Samples were subjected to a cesium chloride density gradient formed by physical density 241 separation via ultracentrifuge as previously described with minor modifications ( Each tube was mounted in a Beckman Coulter fraction recovery system with a side port needle 252 inserted through the bottom. The side port needle was routed to an Agilent 1260 Infinity fraction 253 collector. Fractions were collected in 96-well deep well plates. The density of each fraction was 254 then measured using a Reichart AR200 digital refractometer fitted with a prism covering to 255 facilitate measurement from 5 µL, as previously described (Buckley et al., 2007). DNA in each 256 fraction was purified and concentrated using a Hamilton Microlab Star liquid handling system 257 programmed to automate previously described glycogen/PEG precipitations (Neufeld et al., 258 2007). Washed DNA pellets were suspended in 40 µL of 1xTE and the DNA concentration of 259 each fraction was quantified using a PicoGreen fluorescence assay. The fractions for each sample 260 were binned into 9 groups based on density (1.6900-1.7099 g/ml, 1.7100-1.7149 g/ml, 1.7150-261 1.7199 g/ml, 1.7200-1.7249 g/ml, 1.7250-1.7299 g/ml, 1.7300-1.7349 g/ml, 1.7350-1.7399 g/ml, 262 1.7400-1.7468 g/ml, 1.7469-1.7720 g/ml), and fractions within a binned group were combined 263 and sequenced. 264 For 16S rRNA gene amplicon sequencing, non-fractionated DNA as well as density 265 fractionated DNA was amplified in triplicate 10-uL reactions using primers 515 F and 806 R 266 was limited to taxa that occurred in at least 2 (of 3) experimental replicates and in at least 2 (of 9) 304 density fractions. These criteria were chosen to reduce the likelihood of falsely interpreting 305 spurious density shifts as growth. Technical error associated with tube-level differences in CsCl 306 density gradients was corrected as previously described (Morrissey et al., 2017). All qSIP 307 calculations are available at https://github.com/mmf289. 308 When comparing values of 18 O enrichment between sites, we chose to use the metric 309 "fraction of maximum potential enrichment" or "(FME) 18 O" to account for slight differences in 310 the enrichment of the soil water during incubation. FME 18  All statistical analyses were conducted in R (R Core Team, 2021). Soil properties, 320 relative abundances of minerals, and bulk and respired soil radiocarbon values were compared 321 between sites using one-way ANOVA and Fisher's LSD test. A Bonferroni correction was 322 applied when multiple comparisons were made. 323 The relative abundances of taxa in total communities were measured from sequencing of 324 unfractionated DNA samples. The relative abundances of taxa in growing communities were 325 computed by removing counts from taxa that were not identified as growing via qSIP from 326 unfractionated DNA samples and then re-computing relative abundance values. We computed 327 the proportion of the total community that was growing by dividing the number of ASVs 328 identified as growing by the total number of ASVs inferred from 16S rRNA sequencing at each 329 site. We then used Bray-Curtis dissimilarity to assess between-site dissimilarity for both growing 330 and total communities with vegan: Community Ecology Package (Wagner et al. 2019). Principal 331 coordinate analysis (PCoA) was used to visualize community structure. We tested for differences 332 in community composition between sites using ANOSIM on ranked Bray-Curtis dissimilarities. 333 To test if the degree of community dissimilarity between sites differed for total versus growing 334 microbial communities, we performed a t-test on mean Bray-Curtis dissimilarity for each 335 pairwise comparison of sites between total and growing communities. A Bonferroni correction 336 was applied to account for multiple comparisons. 337 We used one-way ANOVA and Fisher's LSD to test for differences between sites in 338 community, phylum, and family mean FME 18 O and only included taxa that were identified as 339 growing at each site in this analysis. A Bonferroni correction was applied when multiple 340 comparisons were made. We also compared FME 18 O of taxa that were uniquely growing at 341 Sedgwick to the FME 18 O of taxa that were growing at Sedgwick and at least one other site using 342 a t-test. We conducted NRI and NTI analyses using the Picante package to determine 343 phylogenetic relatedness of total microbial communities at each site. 344 We used linear regression to assess the relationship of taxon-specific FME 18 O between 345 sites and included only individual ASVs that were co-occurring in at least two sites. 346 Phylogenetic signal analyses were used to test whether the growth (EAF 18 O) of related 347 organisms resembled each other more than would be expected by chance alone 1 . Phylogeny was 348 built using the SILVA v128 tree using SATé-enabled phylogenetic placement (Janssen et  The Δ 14 C of bulk surface soils increased along the gradient from the wettest to the dry 389 site, indicating that the average age of bulk soil C is oldest at Angelo, the site with the highest 390 annual rainfall (Figure 1) To test the hypothesis that growing communities would be compositionally more similar 403 to each other than total communities, we analyzed the structure of growing and total 404 communities using PCoA and pairwise comparisons of Bray-Curtis dissimilarity. The proportion 405 of the total community that was detected as growing (ASVs having 95% confidence intervals for 406 EAF 18 O exceeding zero were considered growing) was 28%, 48% and 58% at the Angelo, 407 Hopland and Sedgwick sites, respectively (Supplementary Table 4 We calculated net relatedness index (NRI) and nearest taxon index (NTI) to assess 438 whether reduced microbial growth at the driest site was driven by shifts in community 439 composition (i.e. through habitat filtering for slow growing taxa). NRI and NTI analyses indicate 440 phylogenetic clustering of total microbial communities at all three sites (Supplementary Table 6). 441 We then grouped the taxa growing at Sedgwick according to whether they grew at Sedgwick 442 alone or in at least one other site. We found that taxa that were unique to Sedgwick were slower 443 growing than taxa that were growing in at least one other site (Figure 4). 444

Evolutionary constraints on microbial growth 446
To better understand the factors driving variation in growth observed in our study, we 447 used phylogenetic signal analyses to test whether taxon-specific growth was influenced by a 448 microorganism's evolutionary history. In all three grassland sites, we detected significant 449 phylogenetic signals for taxon-specific growth, as determined by Pagel's λ and Blomberg's K 450 (Supplementary Table 7 Given the evidence for a phylogenetic signal of in-situ microbial growth, we were curious 457 to know if taxa would consistently perform with high or low activity regardless of the site, so we 458 used linear regression to analyze taxon-specific growth rates across sites. For taxa that occurred 459 at more than one site and were also significantly growing, their growth rate at one site was 460 predictive of growth rates in another site ( Figure 5). We compared 18 O enrichment of bacterial 461 ASVs that were present and growing in at least two sites and found positive, linear correlations 462 of taxon-specific enrichment between sites. The growth rate of a bacterial taxon in one site 463 explained as much as 48% of the variation in that taxon's growth rate in a second site 464 (Supplementary Table 5). The amount of variation explained by each regression was the lowest 465 when comparing the growth of ASVs between Angelo and Sedgwick, the two most distant sites 466 in our gradient, that also have the largest difference in mean annual precipitation. 467

Discussion 469
In this study, we used quantitative stable isotope probing (qSIP) to assess total and 470 growing soil microbial communities across three California annual grassland ecosystems with 471 Mediterranean climates that span a rainfall gradient. We obtained samples for this study in early 472 2018 during the wet season when differences in soil moisture between our sites were minimized 473 and assessed multiple edaphic variables to characterize how variation in rainfall influences the 474 physiochemical environment experienced by soil microorganisms at each site. We also measured 475 the radiocarbon ( 14 C) age of the soils to investigate how soil C persistence varies along the 476 gradient and might be related to microbial growth potential. We hypothesized that actively 477 growing communities would be more compositionally similar across the gradient than the total 478 background microbiome due to the similarity in soil water content during the weeks preceding 479 sampling. We also predicted that the long-term legacy effect of soil water limitation would be 480 reflected in lower soil C, and lower growth capacity at the driest site even when soils were 481 controlling which taxa grow is critical for soil biogeochemistry. In ecosystems with 501 Mediterranean climates, strong seasonal changes in soil water may be an important determinant 502 of growing community structure. These patterns have clear effects on instantaneous microbial 503 growth rates, but may also have more pervasive, long-term legacy effects (Schimel, 2018). In 504 this study, we characterized the total and growing soil microbial communities in three California 505 grassland ecosystems that span a rainfall gradient during the wet season when all soils were 506 moisture replete (Supplementary Table 2). We hypothesized that growing communities would be 507 compositionally more similar to each other than total communities due to the convergence in 508 environmental conditions, namely soil moisture and an actively growing plant community, 509 during the weeks prior to sample collection. In contrast to this expectation, both growing and 510 total communities were distinct between sites, and the distinctions between sites were of the 511 same order of magnitude for total and growing communities (Figure 2, Supplementary Figure 3). Our data suggest that a site's climate history, rather than recent environmental conditions, place a 520 stronger imprint on the structure of growing soil microbial populations. This is another piece of 521 evidence suggesting that rRNA is an imperfect indicator of metabolically active populations 522 (Blazewicz et al. 2013). While heavy-water DNA qSIP quantifies growing microbial 523 populations, the tight coupling between DNA and RNA synthesis in soils suggests that the 524 majority of active taxa are likely also growing (Papp et al., 2018b). 525 We characterized microbial communities at a time of the year when soils were replete 526 with moisture and when environmental conditions should generally be favorable for microbial 527 growth. However, the growing taxa in this study were still only a subset (28-58%) of all taxa 528 detected by 16S rRNA sequencing, indicating a substantial presence of dormant organisms or 529 relic DNA (Supplementary Table 4

Persistent Effect of Climate on Microbial Growth 545
Microbial growth rate approximates a microorganism's contributions to elemental 546 cycling, so it is essential to understand whether the climate history of a site can influence 547 microbial growth rates in nature. We hypothesized that a history of repeated exposure to low soil 548 water potential at the driest site would result in lower microbial growth even when soil water was 549 not limiting. Consistent with this expectation, community and phylum mean growth was higher 550 at the wet and intermediate sites, Hopland and Angelo, than at the driest site, Sedgwick (Figure  551 3). Variation in microbial growth did not appear to be driven by pH, as we would have expected 552 increasing substrate availability and near-neutral pH levels to, on average, facilitate faster 553 bacterial growth rates at Sedgwick (Rousk et al., 2009). Microbial growth rates did not appear to 554 be driven by differences in soil texture either, as clay content (which is highest at Sedgwick) is 555 thought to support faster growth rates by providing more habitable surface area or pore space for 556 soil microorganisms 2003). While precipitation is the 557 apparent causal factor for the variation in microbial growth rates we observed, we cannot entirely 558 rule out temperature, rainfall periodicity, or plant productivity as proximal causes. 559 We suspect the variation in microbial growth in our soils reflects a legacy effect of the 560 precipitation regime, particularly at our driest site, Sedgwick. Similar findings were documented 561 for a rainfall gradient in Texas, where microbial growth decreased with lower MAP when 562 measured under controlled laboratory conditions (Leizeaga et al., 2021) Table 6). Within the community of 575 microbes that were growing at the driest site, most (155/197 ASVs) were found growing at this 576 site alone, and the ASVs that were uniquely growing at Sedgwick had slower growth, on 577 average, than taxa that occurred and grew in at least one other site (Figure 4). Thus, we infer that 578 reduced microbial growth at our driest site appears to be influenced predominantly by relatively 579 slow-growing taxa that are unique to this site. 580 Since we measured microbial growth at the ASV level using qSIP and 16S rRNA marker 581 gene sequencing, we lack direct evidence regarding the traits and genomic potential of the 582 microbial taxa that exhibited faster growth at our driest site. However, under the conditions 583 evaluated here (wet season, moisture replete soils), the consistent arrangement of growth rates 584 among taxa from site to site ( Figure 5) most likely reflects the influence of genetic constraints on 585 growth. This finding is consistent with the idea that evolutionary history influences growth rate 586 and suggests that the phenotype of in situ growth rate for a given taxon spans a range constrained 587 by its genetics and physiology, constraints that persist across rainfall gradients, edaphic 588 properties, and biological communities (Morrissey et al., 2019). Repeated exposure to low soil 589 water potential at Sedgwick may favor the success of microbial taxa whose physiological traits 590 confer resistance to water stress and contribute to slow growth. While not statistically significant 591 In our study, microbial growth, and presumably necromass production, also increased with 605 increased MAP (Figure 3). If necromass-C is preferentially stabilized in these soils, this could 606 result in more persistent soil C over time, leading to an older average age for bulk soil C. In 607 addition, greater rates of vertical flushing of water at our highest MAP site could also facilitate 608 weathering and the development of secondary clay, aluminum, and iron oxide minerals that are 609 key contributors to soil C storage and persistence (Rasmussen et  soil C quality in this study, did not vary between our sites (Table 2). However, soil C:N is a 619 broad proxy and a more targeted molecular characterization of soil C is needed to conclude 620 whether microbial processing of C increases with MAP along the gradient studied here. 621 622

Conclusions 623
Soil water is an important controller of microbial growth and Earth system models project 624 significant changes in the timing and intensity of precipitation events globally (Dore, 2005;625 Schimel, 2018). Seasonal changes in soil moisture have clear effects on instantaneous microbial 626 growth rates and our study demonstrates that soil water also has a more pervasive, long-term 627 legacy effect on microorganisms. Our data show that a site's historical climate regime, rather 628 than recent environmental conditions, strongly influences the structure of growing microbial 629 communities. Our study also demonstrates that a history of exposure to low soil moisture reduces 630 microbial growth, likely through shifts in community composition. Since changes in microbial 631 community structure operate over long time scales, ecosystems may experience a pervasive 632 legacy effect of climate history as precipitation regimes change. Lastly, our study demonstrates 633 that taxon-specific growth rates are remarkably consistent between three grassland sites from a 634 wide precipitation gradient, consistent with the idea that evolutionary history influences 635 microbial growth and points to the genetic and physiological constraints on growth rate in nature.