Transcriptional and metabolic stasis define desiccation-induced dormancy in the soil bacterium Arthrobacter sp. AZCC_0090 until water vapor initiates resuscitation

19 Microbes inhabiting soils experience periodic water deprivation. The effects of desiccation on 20 DNA, protein, and membrane integrity are well-described. However, the effects of drying and 21 rehydration on the composition of cellular RNA and metabolites are still poorly understood. 22 Here, we describe how slow drying and rehydration with water vapor influence the composition 23 of RNAs and metabolites in a soil Arthrobacter. While drying reduced cultivability relative to 24 hydrated controls, water vapor rehydration fully restored it. Ribosomal RNA proportions 25 remained constant throughout all treatments, and mRNA profiles showed stable composition 26 during desiccation—changing only during transitions into and out of desiccation-induced 27 dormancy. Six transcriptional modules displayed distinct expression patterns in desiccated-28 rehydrated samples relative to hydrated controls, including desiccation-rehydration responsive 29 and rehydration-specific profiles. Targeted intracellular metabolomics revealed similarly static 30 profiles during desiccation, with a cluster of ribonucleosides and nucleobases increasing in 31 response to desiccation and returning to baseline levels upon rehydration with water vapor. 32 These findings demonstrate that both mRNA and metabolite profiles remain essentially frozen in 33 desiccated Arthrobacter, with dynamics changes occurring only during state transitions. These 34

have important implications in environments with frequent drying cycles where stable 35 mRNA in dormant cells combined with intracellular RNA recycling may obscure interpretations 36 of RNA-based environmental analyses that use RNA as a marker of microbial activity. Our 37

suggest that RNA-based activity assessments in periodically dry environments require 38 careful consideration of dormancy-associated molecular preservation. 39 40 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 16, 2025. ; https://doi.org/10.1101/2025.10.16.681527doi: bioRxiv preprint 2 SIGNIFICANCE STATEMENT 41 Metabolic activity quickly ceases in drying bacteria as they enter desiccation-induced 42 dormancy. We show mRNA and metabolite profiles were variable during drying and rewetting 43 but did not change while desiccated. Additionally, water vapor stimulated the shift from the 44 static to active state when exiting desiccation-induced dormancy. These shifts coincided with 45 increased cultivability, indicating water vapor resuscitated dry cells. Because RNAs are transient, 46 labile molecules that are turned over rapidly in growing bacteria, the presence of RNA in the 47 environment is used as a marker for microbial activity. Our research shows this assumption may 48 not hold for desiccated cells, indicating reliance on RNA as a marker of activity in environments 49 that experience drying may obscure estimates of in situ microbial activity. 50 51

52 Water is required for life as a substrate in essential biological reactions and as a solvent to 53 transport nutrients and waste (1). As such, the lack of water stresses nearly all organisms. Future 54 climate models predict changes in global precipitation patterns, further restricting precipitation in 55 drylands and introducing drought to regions that have not previously experienced it (2–4). The 56 impacts of drought can accelerate agricultural crop losses and food insecurity (5, 6), reduce 57 freshwater connectivity and services (7), and alter soil biogeochemistry (8–11). Soil 58 microbiomes are restructured during drought (8, 12, 13) and can influence crop drought tolerance 59 (14–17). Despite the impacts of drying on soil microbes and their critical roles in global carbon 60 dynamics and food production, basic aspects of how soil microbes persist in the dry state remain 61 unknown. 62 As soil dries, the solutes in pore water become concentrated, reducing water potential and 63 limiting water availability. Microbes equilibrate rapidly to the water potential of their 64 surroundings because of their small size (13, 18). Eventually, growth substrates precipitate, 65 rendering them unavailable for microbial metabolism. Thus, microbial activity slows with 66 dehydration until cellular respiration stops (13, 19). Macromolecules in dehydrated cells are 67 prone to structural damage. Proteins fold improperly, denature, or are oxidized (20–22). DNA 68 can also be oxidized by reactive oxygen species (ROS) (18, 23). And membrane integrity can be 69 compromised in the desiccated state or upon rehydration, where lysis may occur due to rapid 70 turgor pressure changes that come with the influx of water (24, 25). 71 Some bacteria endure water scarcity through dormancy, a reversible state of reduced metabolic 72 activity (26). Spore-forming microbes like Bacillus and Streptomyces undergo dormancy via cell 73 differentiation to produce ultra-resistant spores in response to nutrient starvation (26–28). Yet, 74 many—perhaps most—soil microbes cannot sporulate (29, 30), and the systems regulating their 75 desiccation tolerance remain poorly understood. Non-spore formers respond to water loss by 76 mitigating cellular damage through multiple coordinated mechanisms. These include 77 accumulating organic osmolytes like trehalose that lower intracellular solute potential and may 78 stabilize membranes and macromolecules (13, 31–35); remodeling membrane lipids to resist 79 disruption (36, 37); producing exopolysaccharides and biofilms that slow water loss (38, 39); 80 upregulating oxidative stress protection (40–42); and enriching transcripts for molecular 81 chaperones and DNA repair proteins (40, 41, 43). 82 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 16, 2025. ; https://doi.org/10.1101/2025.10.16.681527doi: bioRxiv preprint 3 Yet, the fate of RNA and intracellular metabolites during desiccation-induced dormancy remain 83 understudied, particularly how the transcriptional and metabolic profiles change in dry cells and 84 upon rehydration. While endospores show a dramatic reduction in RNA content during 85 dormancy (44), the RNA dynamics in desiccation-tolerant non-spore-forming bacteria are 86 underexplored. This question is important to resolve for environmental microbiology, where 87 RNA abundances—both ribosomal and messenger RNA—serve as a key indicator of microbial 88 activity across diverse environments (45–48), including soils (49–51). If RNAs are detectable in 89 inactive cells, RNA may not reliably indicate microbial activity. Here, we demonstrate that 90 desiccation induces a dormant state in a soil Arthrobacter characterized by: 1) reduced but 91 recoverable cultivability, with water vapor alone sufficient for resuscitation; 2) stable mRNA and 92 intracellular metabolite profiles during desiccation, despite dynamic changes during transitions 93 into and out of the desiccated state; and 3) ribosomal RNAs that occupy a constant fraction of the 94 total RNA pool irrespective of hydration (and hence, activity) status. By identifying distinct 95 temporal patterns in mRNAs and metabolites across the desiccation-rehydration cycle, we reveal 96 implications for how RNA-based approaches may misrepresent microbial activity in 97 environments that experience periodic drying. 98 99

& DISCUSSION 100 Experimental design. To investigate cellular responses to desiccation stress, we studied 101 Arthrobacter sp. strain AZCC_0090, a soil actinobacterium isolated from semiarid soil in 102 Southern Arizona that is closely related to the Arthrobacter phylotypes found across the United 103 States (29, 52, 53). We developed a controlled desiccation system using two humidity chambers 104 in which the relative humidity (RH) of the atmosphere was controlled with saturated salts (Fig. 1 105 a,b). The chambers consisted of a “control” chamber maintained at 100% RH and a “treatment” 106 chamber where RH was gradually reduced from 100% to 26% over 14 days to desiccate cells, 107 corresponding to an atmospheric water potential change from 0 to -183 MPa (Fig. 1c). After 14 108 days of desiccation, we restored the 100% RH atmosphere of the treatment chamber to rehydrate 109 dried cells for two additional days (Fig. 1c). The degree of desiccation extended far below the 110 water potentials that inhibit growth in both E. coli (-4.6 MPa) and Arthrobacter spp. (-17 MPa) 111 (18). Cells were collected on non-hygroscopic polycarbonate filters and placed into these 112 chambers. Replicate filters were collected from the chambers at four timepoints to analyze cell 113 cultivability, intracellular metabolite profiles, and gene expression (Fig. 1c). 114 This experimental design incorporated three unique aspects relative to previous studies (40, 41). 115 First, hydrated control cells were maintained alongside treatments, allowing us to specifically 116 identify the responses to desiccation and rehydration. Second, dehydration occurred gradually 117 over 14 days rather than rapidly. Finally, rehydration was achieved through water vapor alone. 118 This approach enabled control over cellular water status while minimizing confounding 119 variables. 120 Dry cells are dormant. We defined cultivability as the fraction of filtered cells that were 121 culturable on solid media. Initial cultivability immediately after filtering was 36.3 ± 24.5% 122 (mean ± SD, n=4; Fig. 2A), with variability reflecting the technical challenges of quantitatively 123 recovering cells from filters. Cells maintained in hydrated control conditions showed stable 124 c ultivability throughout the experiment, with no significant variation over time (Kruskal Wallis p 125 = 0.580; Fig. 2A). In contrast, cells subjected to desiccation showed significant temporal 126 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 16, 2025. ; https://doi.org/10.1101/2025.10.16.681527doi: bioRxiv preprint 4 variation in cultivability (Kruskal Wallis p = 0.022), with mean values decreasing by an order of 127 magnitude from 24.0% at day 2 (100% RH) to 2.46% at day 8 (65% RH; Fig. 2A). Desiccated 128 cells showed significantly reduced cultivability compared to hydrated controls at both day 8 and 129 day 14 (Mann-Whitney p = 0.028 for both timepoints; Fig. 2A), indicating that reduced 130 cultivability was due to desiccation. 131 Water vapor restored cultivability of desiccated cells, demonstrating dry cells were dormant, not 132 dead. When cells from day 14 (26% RH) were exposed to 100% RH for two days, their mean 133 cultivability increased to 14.4%—statistically indistinguishable from constantly hydrated 134 controls at day 16 (Mann-Whitney P = 0.49; Fig. 2A). Although the AZCC_0090 genome lacks 135 genes associated with sporulation (53), we tested whether the observed recovery was due to 136 spore formation. No growth was observed after treating cells with ethanol, confirming 137 AZCC_0090 survives desiccation without sporulating. 138 We further calculated the relative effects of desiccation and rehydration as the cultivability ratio 139 of treatment to control cells at each timepoint (Fig. 2B). This confirmed significant variation 140 across the experiment (Kruskal-Wallis p = 1.76e-8), with days 8 and 14 showing significantly 141 lower relative cultivability compared to both pre-desiccation (day 2) and post-rehydration (day 142 16) timepoints (Fig. 2B; Dunn's test, Bonferroni-adjusted P ≤ 0.05). These results demonstrate 143 that desiccation caused reduced cultivability and water vapor restored cultivability in desiccated 144 Arthrobacter. 145 Effect of desiccation and rehydration on RNA content and composition. Initial total RNA 146 yields from washed cells were 150 ± 29.1 ng µl-¹ (mean ± SD, n=3; Fig 3a). After 2 days at 147 100% RH, total RNA decreased in both the control and treatment groups relative to the washed 148 cells, with no significant difference between the conditions, likely due to nutrient starvation 149 leading to ribosome degradation (54, 55). Subsequently, RNA amounts in the control condition 150 varied significantly over time (Kruskal-Wallis p = 0.033 between days 2 and 16; Fig. 3A), while 151 desiccated and rehydrated cells showed no temporal variation (p = 0.305; Fig. 3A). Because 152 rRNA constitutes most of the total RNA, these temporal differences likely represent ribosomal 153 turnover in metabolically active control cells versus desiccated cells undergoing metabolic arrest. 154 However, no significant differences in total RNA content were detected between control and 155 treatment groups at any individual timepoint (Wilcoxon rank-sum tests, all p ≥ 0.08). 156 We investigated whether the fraction of total RNA occupied by the main cellular rRNAs (16S 157 and 23S rRNA) varied during the experiment. To do this, we quantified 16S and 23S rRNA 158 transcript abundances and normalized them to the amount of total extracted RNA (Fig. 3 B,C). 159 The proportions of both rRNA species remained stable across all time points in both treatment 160 and control conditions (Kruskal-Wallis 16S rRNA treatment p=0.312 control p=0.541; 23S 161 rRNA treatment p=0.644, control p=0.384). Pairwise comparisons revealed no significant 162 differences between control and treatment groups at any individual timepoint for either 16S 163 rRNA or 23S rRNA (Wilcoxon rank-sum tests, p ≥ 0.1). These results indicate that the 164 contributions of 16S and 23S rRNAs to the total RNA pool were maintained despite changes in 165 total RNA content. 166 In contrast to the effects on rRNA, drying and rehydrating reshaped mRNA transcription 167 patterns, but transcription profiles were stable in desiccated cells. Analysis of mRNA 168 transcriptional responses revealed significant effects of both time point and treatment, with each 169 factor explaining ~20% of the observed variation in gene expression (PERMANOVA time: 170 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 16, 2025. ; https://doi.org/10.1101/2025.10.16.681527doi: bioRxiv preprint 5 21.9%, p=0.001; treatment: 19.1%, p≤ 0.001; Supplemental Fig. 1). The interaction between time 171 and treatment explained an additional 16.8% variation (p=0.009), indicating that temporal gene 172 expression patterns differed between control and treatment conditions. The control samples 173 showed changes over time, though none of the consecutive timepoint shifts were significant 174 (PERMANOVA p >0.05 between consecutive timepoints). In contrast, the treatment samples 175 displayed a significant shift in gene expression upon desiccation (day 2 to 8: PERMANOVA 176 p=0.023, R²=0.53). Upon rehydration, although the time variable explained a large proportion of 177 variance in gene expression profiles (R²=0.56), this shift was not significant (day 14 to 16: 178 PERMANOVA p=0.1). Yet, the transcriptional profiles were indistinguishable in the desiccated 179 state (day 8 to 14 PERMANOVA p=0.73, R2= 0.15). These findings show the composition of 180 mRNA profiles were variable during drying and rehydration but did not change over time in dry 181 Arthrobacter cells. 182 Analysis of the temporal patterns in gene expression supported our interpretation that gene 183 expression profiles remained static during desiccation while hydrated cells were transcriptionally 184 dynamic. We detected 1,409 genes with significantly different temporal expression patterns 185 across treatment and control conditions (ImpulseDE2 FDR-corrected p ≤ 0.05; Supplementary 186 Table 1). These genes clustered into six distinct gene co-expression modules based on their 187 combined temporal patterns across both conditions (Fig. 4, Supplementary Table 1). Across all 188 transcriptional modules, we observed further evidence of a frozen transcriptional state in 189 desiccated cells, as no substantial change in mean gene expression patterns in the treatment 190 samples between days 8 and 14 was apparent (bold brown lines, Fig. 4). However, transcriptional 191 profiles were dynamic in treatment samples during drying and/or rehydration in Transcriptome 192 Modules 2, 3, 4, 5 and 6 (Fig. 4 B-F). In contrast, these gene co-expression modules exhibited 193 distinct variable responses throughout the experiment in control samples, including between days 194 8 and 14, suggesting hydrated cells were transcriptionally active and dynamically responded to 195 starvation. 196 Two main desiccation/rehydration-specific gene expression patterns emerged in the treatment 197 samples. First, Transcriptome Modules 5 (143 genes) and 6 (80 genes) exhibited similar 198 temporal responses where expression increased during drying, remained elevated throughout 199 desiccation, and returned to baseline after rehydration (Fig. 4 E,F and Supplementary Table 1). 200 While these transcriptional modules showed similar expression patterns in treatment samples, 201 their expression in hydrated controls was distinct, suggesting that although their roles are similar 202 during desiccation, they were expressed independently during starvation. 203 The annotations of genes expressed in Transcriptome Modules 5 and 6 largely align with 204 previously described responses to desiccation in non-spore forming microbes (Fig. 4 E,F; 205 Supplementary Table 1). For example, Transcriptome Module 5 included genes coding for fatty 206 acid metabolism, the osmolyte trimethyl glycine (betaine) synthesis and transport, and purine 207 catabolism. The coordinated upregulation of choline-to-betaine conversion machinery alongside 208 osmolyte transporters suggests AZCC_0090 depends on the conversion of external choline to 209 betaine for osmoregulation during drying. The upregulation of compatible solute synthesis and 210 transport machinery is a common response to drying (42). Additionally, Transcriptome Module 5 211 contained regulatory and repair systems including numerous transcriptional regulators, DNA 212 repair proteins, and ROS protection enzymes—all hallmarks of desiccation stress responses in 213 microbes (42). Finally, a large-conductance mechanosensitive channel was also identified that 214 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 16, 2025. ; https://doi.org/10.1101/2025.10.16.681527doi: bioRxiv preprint 6 may prime cells to cope with the stresses associated with rehydration (56). However, we were 215 surprised by the presence of genes in Transcriptome Module 5 that code for protein synthesis 216 machinery including several ribosomal proteins and translation factors. This was unexpected 217 given the apparent metabolic slowdown during drying, though similar findings have been 218 previously reported (35, 57). Transcriptome Module 6 focused on distinct cellular processes 219 including carbohydrate processing, protein quality control through proteases, aromatic 220 compound and lipid metabolism, and nucleotide processing via a putative NUDIX family 221 pyrophosphohydrolase. The genes in Transcriptome Modules 5 and 6 largely align with 222 previously reported bacterial desiccation responses, including expected mechanisms for ROS 223 mitigation, osmoregulation, putative membrane remodeling, and activation of DNA and protein 224 repair systems to address oxidized biomolecules (35, 42, 58). Notably, 33% of Transcriptome 225 Module 5 and 38% of Transcriptome Module 6 genes lacked meaningful annotations, suggesting 226 novel mechanisms and genes may contribute to desiccation tolerance. The expression patterns of 227 both modules—increased relative abundance during drying and decreased relative abundance 228 upon rehydration—suggest these transcripts are likely recycled upon rewetting. 229 The second desiccation-rehydration specific gene expression pattern was the dramatic increased 230 abundance of 290 genes across Transcriptome Modules 3 & 4 during rehydration with water 231 vapor, suggesting key roles in cellular resuscitation (Fig. 4 C,D). These genes were expressed at 232 low levels in the treatment condition prior to rehydration though their expression was distinct in 233 controls (Fig. 4 C,D and Supplementary Table 1). Transcriptome Module 3 contained genes 234 encoding fatty acid beta-oxidation and aromatic compound degradation pathways, alongside 235 abundant transcriptional and translational machinery indicating active protein synthesis during 236 rehydration. Stress response signatures included multiple chaperones and catalase, suggesting 237 mitigation of oxidative damage, while the high density of transcriptional regulators indicates 238 extensive gene expression reprogramming during rehydration. Transcriptome Module 4 contains 239 genes encoding osmoprotectant transport (distinct from Module 5), a xylose metabolism operon, 240 histidine catabolism, DNA repair and replication machinery, and energy metabolism genes. The 241 xylose pathway and associated sugar phosphate enzymes may shuttle 5-carbon sugars toward 242 phosphoribosyl pyrophosphate (PRPP) for nucleic acid repair or synthesis. Histidine degradation 243 may serve dual functions: chelating divalent cations during desiccation to reduce ROS 244 generation, then providing carbon and nitrogen upon rehydration (59). InterProScan analysis 245 revealed seven 'hypothetical proteins' (HNP00_000350, HNP00_000836, HNP00_001555, 246 HNP00_001771, HNP00_002967, HNP00_002968, and HNP00_003348) containing disorder 247 domains with polyampholyte subdomains—signatures of eukaryotic anhydrins with chaperone-248 like roles during desiccation (60, 61). Two of these (HNP00_001555 and HNP00_001771) also 249 contained HNH-nuclease-like domains, suggesting dual roles in nucleic acid metabolism and 250 protein stability. Together, Transcriptome Modules 3 and 4 represent coordinated cellular 251 machinery activated during rehydration: Module 3 mobilizes lipids for energy while reactivating 252 transcription, translation, and stress management, while Module 4 manages osmotic stress, 253 alternative carbon utilization, DNA repair, and resource acquisition essential for growth 254 resumption. 255 Desiccacted cells are in metabolic stasis. To identify metabolic changes that correspond to 256 survival during drying and rehydration we also conducted targeted and untargeted intracellular 257 metabolite analysis alongside the transcriptomes. Targeted metabolite analysis revealed that 258 treatment was the dominant factor shaping metabolic composition (PERMANOVA R² = 24.3%, 259 p < 0.001), followed by temporal dynamics (PERMANOVA R² = 17.8%, p < 0.001) and their 260 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 16, 2025. ; https://doi.org/10.1101/2025.10.16.681527doi: bioRxiv preprint 7 interaction (PERMANOVA R² = 15.9%, p = 0.002). Collectively, these factors explained 58% of 261 total targeted metabolomic variance in the experiment. Like the results from the transcriptome 262 analysis, we observed a treatment-specific response characterized by metabolic restructuring 263 during drying (between days 2 and 8; PERMANOVA R² = 63.8%, p = 0.01) and rehydration 264 (between days 14 and 16; PERMANOVA R² = 33.3%, p = 0.01), with no change in the dry state 265 (between days 8 and 14; PERMANOVA R² = 7.3%, p = 0.71). In contrast, control samples 266 displayed a muted temporal drift (Day 2-8: PERMANOVA R² = 26.6%, p = 0.01; Day 8-14: 267 PERMANOVA R² = 21.5%, p = 0.05; Day 14-16: PERMANOVA R² = 10.8%, p = 0.43). These 268

demonstrate that like the observed compositional stasis in the dry state transcriptomes, 269 Arthrobacter also undergo metabolic dormancy during desiccation with subsequent recovery 270 toward profiles like the pre-desiccated state. 271 To identify metabolites with correlated temporal profiles, we clustered targeted metabolite 272 profiles based on the shape of their normalized peak height variation over time. Of the 105 273 targeted metabolites analyzed, 46 (43%) showed significantly distinct temporal patterns across 274 the treatment and control samples (modified ImpulseDE2 FDR-corrected p ≤ 0.05; 275 Supplementary Table 2). Clustering of significantly different metabolites across the treatment 276 and control conditions revealed four Targeted Metabolite Modules (Fig. 5; Supplementary Table 277 2). The largest Targeted Metabolite Module (module 1, 21 metabolites) showed a similar profile 278 during desiccation to that observed for Transcriptome Modules 5 and 6 (Figs 4 E,F), 279 characterized by increasing peak area during drying, an elevated level when dry, and a return to 280 baseline after rehydration (Fig. 5A). Of these metabolites, 12 (57%) were either directly related 281 to nucleic acid metabolism, including all ribonucleosides, purine and pyrimidine bases, or 282 nucleotide degradation products. These nucleotide degradation products likely accumulate 283 because transcription ceases in the transition to the dry state. We speculate that existing 284 nucleotide monophosphates are progressively dephosphorylated, and their N-glycosidic bonds 285 cleaved, releasing the ribonucleosides and nitrogenous bases we observed. This observation is 286 consistent with internal mRNA recycling across the desiccation-rehydration continuum. Also 287 notable were the presence of potential antioxidant compounds including B-vitamins (B1 & B3), 288 L-gulonolactone (an intermediate in the ascorbic acid biosynthetic pathway), and N-acetyl L-289 glutamic acid (62–65). Finally, both L-methionine and methylthioadenosine (MTA) were 290 elevated during desiccation. These metabolites may work as part of a methionine-methionine 291 sulfoxide antioxidant cycling system where methionine serves as a ROS scavenger, with MTA 292 indicating ongoing methionine recycling, SAM production, and/or polyamine synthesis during 293 desiccation (66). 294 The remaining Targeted Metabolite Modules showed metabolite depletion in treatment samples 295 from days 2-14, potentially due to active catabolism, abiotic degradation, non-enzymatic 296 transformations, or residual enzymatic activity—possibilities our data cannot distinguish. 297 Betaine, a known osmolyte, decreased in both conditions (Targeted Metabolite Module 4; Fig. 298 5D) but more steeply in controls. In desiccated cells, betaine synthesis genes were more 299 abundant (Transcriptome Module 5; Fig 4E) despite declining betaine concentrations, suggesting 300 consumption or export during osmotic adaptation. In hydrated controls, betaine catabolism genes 301 (Transcriptome Module 1; Fig 4A) increased in abundance alongside steeper betaine depletion, 302 consistent with its use as a carbon or nitrogen source during starvation. Supporting betaine 303 catabolism during combined stress, genes encoding sarcosine oxidase, glycine 304 hydroxymethyltransferase, and serine dehydratase (betaine-to-pyruvate pathway) were enriched 305 in Transcriptome Module 1 and more abundant in controls (Fig. 4A). Finally, Targeted 306 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 16, 2025. ; https://doi.org/10.1101/2025.10.16.681527doi: bioRxiv preprint 8 Metabolite Module 3 compounds (Fig. 5C) increased throughout the experiment in starving 307 hydrated cells and may represent critical survival compounds, including N-acetyl-alpha-D-308 galactosamine, vanillic acid, 3-hydroxyphenylacetic acid, riboflavin, pyridoxine, thymine, 2,4-309 dihydroxypyrimidine-5-carboxylic acid, and kynurenic acid. 310 We further analyzed 3,349 untargeted metabolite features from combined positive and negative 311 ion modes (1,424 negative ion mode features and 1,925 positive ion mode features). Temporal 312 analysis revealed that 133 features (4%) exhibited significant temporal trajectories across the 313 experimental conditions (Supplementary Table 3). These features were clustered into three 314 modules based on their combined treatment and control patterns, mirroring the approach used in 315 the targeted metabolomics analysis. Untargeted Metabolite Module 1 (Supplementary Fig. 3A) 316 exhibited a temporal profile like Targeted Metabolome Module 1 (Fig. 5A): features increased 317 during desiccation, returned to baseline upon rehydration, and remained stable under control 318 conditions. Untargeted Metabolite Module 2 is comprised of features that remained stable in the 319 treatment condition across days 2, 8, and 14 but decreased in intensity upon rehydration 320 (Supplementary Fig. 3B). These same features were progressively depleted in controls, 321 consistent with their loss in metabolically active hydrated cells while persisting in metabolically 322 arrested desiccated cells. Untargeted Metabolite Module 3 features increased under control 323 conditions but remained low in treated cells on days 2, 8, and 14 (Supplementary Fig. 3C), then 324 increased markedly upon rehydration—consistent with their production or accumulation in 325 metabolically active hydrated cells. 326 Many untargeted features could not be assigned specific molecular formulas. Among those that 327 were assignable, several showed consistent patterns across both targeted and untargeted analyses. 328 For example, gluconic acid, uridine, and hypoxanthine exhibited similar temporal trajectories in 329 both datasets (Fig 5A and Supplementary Fig. 3A). In contrast, several metabolites that showed 330 strong temporal signals in the targeted analysis—with increased intensity upon desiccation and 331 return to baseline upon rehydration (Fig. 5A)—exhibited more muted responses in the untargeted 332 analysis (Untargeted Metabolite Module 2; Supplementary Fig. 3B), characterized by stable 333 intensity during desiccation rather than increases, and reduced intensity in hydrated cells. These 334 metabolites included adenosine, deoxyuridine, nicotinamide, adenine, xanthine, and 335 methylthioadenosine. The differences between analytical approaches likely reflect 336 methodological factors: targeted analyses use compound-specific optimization and isotope-337 labeled internal standards for each metabolite, providing greater sensitivity and accuracy for 338 known compounds, while untargeted approaches employ generic parameters optimized for 339 detecting the broadest possible range of features, often at the expense of sensitivity for specific 340 metabolites. The targeted data are therefore more reliable for quantifying these specific 341 compounds, while the untargeted data provide complementary discovery of unanticipated 342 metabolites. Importantly, despite quantitative differences, the directional trends—depletion in 343 active cells and persistence in dry cells—were generally consistent between approaches. 344

345 Dormancy is defined as a reversible state of low to no metabolic activity (67). Our results 346 demonstrate that desiccation induces a dormant state in Arthrobacter sp. AZCC_0090, 347 characterized by compositionally stable mRNA and intracellular metabolite profiles in the 348 desiccated state (days 8-14), despite dramatic restructuring during the transitions into and out of 349 desiccation. Water vapor was sufficient to resuscitate dormant cells, inducing shifts in both 350 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 16, 2025. ; https://doi.org/10.1101/2025.10.16.681527doi: bioRxiv preprint 9 mRNA and metabolites that prime cells for regrowth. We identified the accumulation of 351 ribonucleosides and nucleobases during desiccation that persisted when mRNA profiles were 352 stable. The source of these nucleotide degradation products during a period of apparent 353 transcriptional stability remains unclear, but may represent degradation of damaged RNAs, 354 turnover of non-coding RNAs not captured in our analysis, or accumulation of RNA turnover 355 products generated during the drying transition. More focused experiments measuring RNA 356 turnover rates directly are needed to resolve this apparent paradox. 357 The signatures we observed in desiccation-induced dormancy differ markedly from other 358 dormancy states such as endospores. While most RNAs in dormant endospores are unstable and 359 degrade to serve as nucleotide reservoirs for germination (44, 68), Arthrobacter maintained 360 stable RNA profiles throughout desiccation. This pattern more closely resembles RNA dynamics 361 in the non-spore forming actinobacterium, Curtobacterium, and biocrust communities where 362 RNA profiles remained stable over extended dry periods (43, 69). 363 Our findings challenge fundamental assumptions underlying RNA-based assessments of 364 microbial activity. Though controversial (48, 70), the presence of RNA is thought to identify 365 active microbes in situ. Methods like rRNA gene sequencing or rRNA:rDNA gene abundance 366 ratios (51, 71) have been used as indicators of microbial activity and metatranscriptomic analyses 367 have used transcript presence as evidence of ongoing transcription (72, 73). Our data 368 demonstrate that: 1) RNA is extractable and sufficiently intact for quantification and sequencing 369 from desiccated cells, 2) some of these RNAs remain abundant during desiccation-induced 370 dormancy, and 3) proportional transcript changes occur primarily during transitions into or out of 371 desiccation, but not during the dormant desiccated state. Consistent with this, recent studies show 372 community-level transcription follows rewetting, suggesting that increased transcription is 373 associated with the transition out of the dry state (35, 74). Collectively, the ability to extract 374 compositionally stable RNAs from desiccated cells suggests that RNA-based activity 375 assessments cannot distinguish between metabolically active and dormant populations in 376 environments experiencing desiccation. This has important implications for interpreting RNA-377 based studies in dryland soils, where the presence of RNA may reflect a mixture of active cells 378 and dormant cells retaining stable transcripts, potentially leading to overestimates of in situ 379 microbial activity. 380 There are several important limitations of this work. First, we examined only one strain of 381 Arthrobacter, a genus known for its desiccation tolerance. Thus, it is unclear whether RNA 382 stability is a specific feature of Arthrobacter's desiccation tolerance mechanisms or represents a 383 more general response to desiccation stress. However, the similar findings in Curtobacterium 384 suggest this phenomenon may be widespread among desiccation-tolerant bacteria (69). 385 Additionally, these experiments were performed under controlled laboratory conditions, and their 386 relevance field conditions remains yet to be determined. Unfortunately, methods for studying 387 RNA stability in soil microbial communities are currently lacking. Finally, while we 388 demonstrated water vapor was sufficient for resuscitation, the mechanisms by which desiccated 389 cells sense and respond to water vapor remain unknown. Some metabolites accumulated during 390 desiccation may exhibit humectant-like properties, but the biophysical and molecular basis of 391 vapor-phase rehydration requires further investigation. 392 The persistence of RNA in desiccated cells has implications extending far beyond arid soil 393 m icrobiology, with relevance for understanding microbial survival and dispersal in extreme 394 environments, astrobiology applications, pathogen dispersal across climate regimes, and the 395 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 16, 2025. ; https://doi.org/10.1101/2025.10.16.681527doi: bioRxiv preprint 10 function of airborne microbiomes. The ability of desiccated cells to maintain RNA while 396 remaining dormant until water vapor triggers resuscitation suggests environmental microbes may 397 be more resilient than previously recognized, potentially reshaping our understanding of 398 microbial biogeography and ecosystem function in water-limited systems on Earth and beyond. 399

400 Strain source and propagation: Arthrobacter sp. strain AZCC_0090 was isolated as described 401 elsewhere (52) and its genome sequence is available (53). All growth experiments were 402 conducted on Yeast Mannitol media (YM), consisting of (per liter) 1.0 g yeast extract, 10.0 g 403 mannitol, 0.5 g dipotassium phosphate, 0.2 g magnesium sulfate, 0.1 g sodium chloride, and 1.0 404 g calcium carbonate. Solid media was prepared with 2% Noble agar. Cell washes and 405 resuspensions were conducted in YM salts (per liter): 0.5 dipotassium phosphate, 0.2 g 406 magnesium sulfate, and 0.1 g sodium chloride. All incubations were at 25ºC. 407 Cell preparation for desiccation: We pelleted a 1.0 L overnight (~18 h) culture of AZCC_0090 408 (27,500 rpm for 30 minutes) and washed once with YM salts before resuspending in 1.0 L YM 409 salts. We incubated resuspended cells overnight at 25ºC with shaking. The following day, 10 ml 410 aliquots of the cell suspension (~5-6 /i1108 cell ml-1) were vacuum-filtered through individual 0.2 411 µm pore-size hydrophilic non-hygroscopic polycarbonate membrane filters (GTTP-type; 25 mm 412 diameter) to collect cells. Each filter was placed into a vented petri dish. 413 Humidity chamber construction and humidity control: Humidity chambers were constructed 414 from plastic storage containers. Several 1/4” diameter ports were drilled in each container to 415 accommodate humidity sensor wires and to suspend a small fan (SEPA 12V 0.03A). Access 416 ports were sealed with tape or silicone. The relative humidity was controlled using a series of 417 supersaturated salt solutions as described previously (75). The following saturated salts were 418 prepared with Nanopure water (nH2O): nH2O (no salt added), KCl, NaCl, NaBr, K2CO3, MgCl2, 419 and CH3CO2K. The salts were considered saturated by the appearance of precipitated salt at 420 25ºC. Saturated salt solutions were contained in an open beaker sitting underneath a fan blowing 421 toward the surface of the solution. The fans were wired to a power supply (TekPower TP3005T) 422 set to 12V 0.08A. Temperature and relative humidity were continuously monitored with a 423 Sensirion SEK-SHT35 digital humidity sensor placed inside an empty, closed vented petri dish. 424 The Sensirion sensor was connected to a laptop via a Sensiron SEK-SensorBridge running the 425 Sensiron Control Center. 426 Experimental Design: We randomized filtered cells to treatment or control humidity chambers. 427 The RH in the control chamber was maintained at 99-100% for the duration of the 428 experiment. The RH in the treatment chamber was slowly dehydrated from 100% RH to ~25% 429 RH over the course of two weeks (48 h at each humidity level). After 14 days the treatment 430 chamber was rehydrated with water vapor for 48 h. Replicate filters were selected randomly 431 from the treatment and control chambers for cultivability assays, metabolomics, or 432 transcriptomics at days 2, 8, 14, and 16 after the start of the experiment. The chambers were only 433 opened to change the saturated salt solutions or for sampling. 434 Cultivability and spore formation assays: Cells were resuspended from filters in YM salts by 435 vortexing for 10 minutes at maximum speed. 200 µl of resuspended cells were fixed with 5% 436 formaldehyde (vol/vol) and stained with SYBR green I (1:40 dilution of commercial stock in 437 Tris-EDTA) for 24 hours before counting with an EMD-Millipore Guava Technologies flow 438 cytometer (Millipore, Billerica, MA, USA), as described previously (52). Culturable cells were 439 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 16, 2025. ; https://doi.org/10.1101/2025.10.16.681527doi: bioRxiv preprint 11 enumerated by plate counts by serially diluting resuspended cells in YM salts and plating on 440 solid YM. We defined percent cultivability as the percent of resuspended cells enumerated with 441 flow cytometry that formed colonies on solid media. We used a 1 h treatment with 70% ethanol 442 to determine if Arthrobacter produced ethanol-resistant spores. In brief, filters containing dried 443 cells were floated on 70% ethanol for 1 h before spotting on solid YM and incubating. No 444 growth was observed after ethanol treatment. 445 RNA extraction, sequencing and mapping: Replicate filters were placed into microcentrifuge 446 tubes containing glass disruptor beads, and frozen at -80ºC until extraction. RNA was extracted 447 using a Qiagen RNEasy mini kit, per the manufacturer’s instructions. DNA was removed with 448 Ambion TURBO DNase per the manufacturer’s instructions. RNA was quantified with a Qbit 449 using hs or br kits, as appropriate based on RNA concentration. Samples were sequenced at 450 SeqCenter, LLC (Pittsburgh, PA), using their standard protocols. In brief, RNAs were DNAse 451 treated a second time with Invitrogen DNAse (RNAse free). Library preparation was performed 452 using Illumina’s Stranded Total RNA Prep Ligation with Ribo-Zero Plus kit and 10bp IDT for 453 Illumina indices. Samples were sequenced on a NextSeq2000 giving 2x51bp reads. 454 Demultiplexing, quality control, and adapter trimming was performed with BCL Convert v3.9.3. 455 The quality of generated raw reads was assessed by FastQC v0.73 via Galaxy (galaxy0) (76, 77). 456 The reads were mapped to Arthrobacter sp. AZCC_0090 genome using Bowtie2 v2.4.1 (78) with 457 default parameters in paired-end mode. Counts were generated with the featureCounts function 458 in the Rsubread v2.12.3 R Bioconductor package (79). 459 rRNA qPCR: RNA extracts were used as template for RT-qPCR using a SuperScript™ III One-460 Step RT-PCR System with Platinum™ Taq DNA Polymerase kit, per the manufacturer’s 461 instructions. 16S rRNA RT-qPCR was performed using 515F 5′ -GTGCCA 462 GCMGCCGCGGTAA-3′ and 806R 5′ -GGACTACHVGGGTWTCTAAT-3′ 16S rRNA primers 463 (80). 23s rRNA RT-qPCR was performed using 111F 5’-ATGTCCGAATGGGGAAACCC-3’ 464 and 557R 5’-CACGGTACTGGTCCGCTATC-3’ 23S rRNA primers. Cycle conditions for both 465 16S and 23S rRNAs: 48ºC 30 minutes, 95ºC for 10 minutes, followed by 40 cycles of 95ºC 15 466 seconds, and 60ºC for 1 minute. All qPCRs were run in triplicate and quantified against a 467 standard curve prepared with Arthrobacter AZCC_0090 gDNA. 468 Temporal analysis of RNA abundances. Differential expression analysis was performed using 469 ImpulseDE2 (81) to identify genes with distinct expression across the experimental conditions. 470 ImpulseDE2 was selected for its ability to model complete temporal trajectories and detect 471 impulse-like expression patterns where genes temporarily change expression before returning to 472 baseline levels (81). The analysis was conducted in case-control mode to compare treatment 473 (case) and control conditions across all time points simultaneously, using count data for genes 474 that had at least 10 reads in three or more samples (3709 of 4653 genes). Genes with adjusted p-475 values ≤ 0.05 were considered significantly differentially expressed. 476 Clustering for temporal transcriptome dynamics: We used DESeq2 size factor normalized 477 expression values for significantly differentially expressed genes from the ImpulseDE2 analysis 478 for clustering analysis. Size factor-normalized expression values were standardized using Z-479 score normalization to ensure genes were grouped by expression patterns rather than magnitude. 480 Mean expression values were calculated for each gene at each time point within treatment and 481 control conditions. To determine the optimal number of gene expression clusters, we used 482 hierarchical clustering using correlation-based distances (1 - Pearson correlation) and Ward's 483 linkage. The number of clusters was determined using the elbow method applied to within-484 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 16, 2025. ; https://doi.org/10.1101/2025.10.16.681527doi: bioRxiv preprint 12 cluster sum of squares, with a minimum threshold of 6 modules to ensure adequate resolution of 485 expression patterns. 486 Metabolite extraction: Replicate filters were frozen at -80ºC until metabolite extraction. 487 Intracellular metabolites were extracted from filtered cells by direct sonication in methanol. 488 Briefly, we added 1 mL of cold methanol to tubes containing filtered cells on ice and sonicated 489 with a probe-style sonicator (Fisher Scientific Model FB120 with CL-18 probe) for 5 minutes at 490 an amplitude of 60%. Filter debris was separated from supernatant by centrifugation. The 491 supernatant was transferred to a new tube and dried to completeness in an Eppendorf Vaccufuge 492 plus at 45ºC for 40 min. We extracted metabolites from several filters without cells and methanol 493 only as negative controls. 494 Metabolomics: Analyses were performed at the Joint Genome Institute using standard 495 procedures. Dried extracts were resuspended in methanol containing internal standards (13C-15N 496 labeled amino acids and 2-amino-3-bromo-5-methylbenzoic acid), filtered, and analyzed by LC-497 MS using an Agilent 1290 Infinity LC system (Agilent, Santa Clara, CA) coupled to a Thermo 498 QExactive HF orbitrap mass spectrometer (Thermo Scientific, San Jose, CA). Polar metabolites 499 were separated using HILIC chromatography on a Poroshell 120 HILIC-Z column with a water-500 acetonitrile gradient. Full MS spectra (m/z 70-1050) and MS/MS fragmentation data were 501 acquired in both positive and negative ion modes. Sample injection order was randomized with 502 methanol blanks between samples. Complete analytical conditions are provided in 503 Supplementary Methods. 504 For targeted metabolite identification, experimental mass spectra were compared to compound 505 standards using custom Python code (82). Spectral features were assigned confidence scores (0-506 3) according to Metabolomics Standards Initiative levels (83), with "level 1" identifications 507 requiring mass accuracy and retention time matches to pure standards, and highest confidence 508 identifications also requiring matching MS/MS fragmentation patterns. Complete analytical and 509 identification parameters are provided in Supplementary Methods. 510 For untargeted features, experimental mass spectra of unassigned features with MS/MS 511 fragmentation data were compared against the Global Natural Products Social Molecular 512 Networking (GNPS) (84) that performs molecular networking and grouping partially and fully 513 identified metabolites based on their MS/MS fragmentation tree, leading to the discovery of 514 related metabolites/features based on common/similar functional groups; and SIRIUS 4 (85) 515 which uses isotope patterns in MS spectra and fragmentation profiles in MS/MS spectra to 516 further assist in molecular formula assignment and potential structural elucidation. CSI:FingerID 517 was used for molecular structure annotation through database searches and CANOPUS for de 518 novo compound class prediction. This integrated workflow supports comprehensive 519 identification of both known and novel compounds from tandem mass spectrometry data in 520 complex matrices. 521 Volume adjustment and Ion mode selection: We experienced sample volume loss during 522 metabolite extraction for nine samples before drying: T0C rep 4 (100 μ L collected), T0C rep 1 523 (200 μ L), T0E rep 5 (200 μ L), T0E rep 4 (250 μ L), T0E rep 3 (300 μ L), T1C rep 3 (450 μ L), 524 T1E rep 3 (450 μ L), T1E rep 1 (300 μ L), and T2E rep 1 (400 μ L). To account for this volume 525 loss, we adjusted metabolite peak heights by multiplying by the ratio of expected volume (500 526 μ L) to actual volume collected: Adjusted peak height = Original peak height × (500/Actual 527 volume collected). For targeted metabolites identified in both positive and negative ion modes, 528 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 16, 2025. ; https://doi.org/10.1101/2025.10.16.681527doi: bioRxiv preprint 13 we selected the polarity with higher mean peak heights for subsequent analyses. For untargeted 529 features, duplicate metabolites in the negative and positive ion mode were identified. Features 530 with identical predicted molecular formulas and retention times within 5 seconds across positive 531 and negative ion modes were considered duplicates. For each duplicate pair, the metabolite 532 measured in the ion mode with higher mean intensity across all samples was retained. This 533 deduplication process removed 32 duplicated features. For both targeted and untargeted 534 metabolomics, a small value was added to missing and zero values (10% of the smallest positive 535 peak area) to facilitate Log transformation in downstream analyses. 536 Temporal analysis of metabolites: ImpulseDE2 was designed for RNA-seq count data analysis 537 but was adapted for metabolomics peak intensity data. To do this, we Log2-transformed the peak 538 heights, multiplied them by 100, and rounded to create pseudo-count values. Results were 539 extracted and clustered using the same framework as transcriptomics analysis described above. 540 Significantly different metabolites were defined as those with Impulse2DE adjusted p-values ≤ 541 0.05. 542 MDS plots and statistics of transcriptome and metabolome data: We used multidimensional 543 scaling (MDS) plots and permutational multivariate analysis of variance (PERMANOVA) to 544 compare transcriptome and metabolome profiles over time across conditions. Raw transcriptome 545 count data were processed using DESeq2 (86) with variance stabilizing transformation for 546 visualization and multivariate analyses. We Log10-transformed the metabolome data. For both 547 datasets, multidimensional scaling was used to visualize sample relationships based on Euclidean 548 distances. PERMANOVAs were implemented through the adonis2 function in the vegan R 549 package (87), employing a factorial design (~day + treatment + day:treatment) with 999 550 permutations. To resolve temporal dynamics within each condition, we conducted separate 551 pairwise PERMANOVA comparisons between time point combinations for control and 552 treatment groups independently. 553 554 DATA AVAILIBILITY: 555 Processed RNA-seq data generated in this study have been deposited in the NCBI Gene 556 Expression Omnibus (GEO) under accession number GSE309279. Individual samples are found 557 in accessions GSM9264299-GSM9264323. Raw metabolome data is available in the massIVE 558 repository at the Global Natural Product Social Molecular Networking (GNPS) site under the 559 accession MSV00009260. Code used in analyses and metabolite sample metadata are available 560 at 10.6084/m9.figshare.30370339. 561

562 The authors would like to thank Bradley Schlottman, and Caitlin Tribelhorn for assistance with 563 equipment and experiments. Funding support is from the National Science Foundation in a 564 Graduate Research Fellowship Grant award to IAV DGE-2137419 and IOS-2141605 to PC. PC, 565 LM and PM were supported by the University of Arizona Research, Innovation & Impact (RII) 566 and Technology Research Initiative Fund/Water, Environmental, and Energy Solutions. The 567 work (proposal:https://doi.org/10.46936/10.25585/60001177) conducted by the U.S. Department 568 of Energy Joint Genome Institute (https://ror.org/04xm1d337), a DOE Office of Science User 569 Facility, is supported by the Office of Science of the U.S. Department of Energy operated under 570 Contract No. DE-AC02-05CH11231. 571 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 16, 2025. ; https://doi.org/10.1101/2025.10.16.681527doi: bioRxiv preprint 14 FIGURE LEGENDS: 572 Figure 1: Experimental system for controlled desiccation and rehydration. (A) Custom 573 humidity chambers housed within a temperature-controlled incubator. (B) Interior chamber 574 configuration showing saturated salt solution for humidity control and real-time 575 temperature/humidity sensor. (C) Relative humidity profiles over the 16-day experiment, with 576 sampling points (purple circles and vertical dashed lines) and water/salt solutions used at each 577 stage. Abbreviations: K-Ac: Potassium Acetate; dH2O: deionized water. 578 Figure 2: Desiccation-induced growth inhibition is reversible with water vapor. (A) Cell 579 cultivability throughout the 16-day experiment with points showing individual replicate filters 580 (n=4) for initial washed cells (no fill), controls (blue), and treatment samples (brown). Brackets 581 with asterisk indicate significant differences between treatment and control (Mann-Whitney p ≤ 582 0.05). (B) Relative cultivability (treatment/control ratio) showing desiccation-induced decreases 583 in cultivability and recovery post rehydration. Points are all-by-all pairwise cultivability ratios. 584 The dashed line in (B) indicates 100% relative cultivability (treatment equals control). Box plots 585 with shared letters are not significantly different (Dunn's test with Bonferroni correction, p > 586 0.05). C=Control, T=Treatment, given in terms of percent RH. 587 Figure 3: rRNA composition during desiccation and rehydration is constant. (A) Total RNA 588 yield from filtered cells over the 16-day experiment. Individual points represent RNA 589 concentrations from replicate filters: initial samples (unfilled), hydrated controls (blue), and 590 desiccation-rehydration-treated samples (brown). (B,C) Normalized abundance of rRNA 591 transcripts (copies ng total RNA/i4 ¹) for (B) 16S rRNA and (C) 23S rRNA. Box plots show 592 median (horizontal line), interquartile range (box), and 1.5 × interquartile range (whiskers). 593 Kruskal-Wallis test p-values shown for control (C) and treatment (T) groups across all 594 timepoints (Days 2-16). 595 Figure 4: Temporal expression modules of differentially expressed genes. Six distinct co-596 expression modules (A-F) identified through correlation-based hierarchical clustering of genes 597 with significantly distinct temporal responses across treatment and control conditions 598 (ImpulseDE2, FDR-corrected p ≤ 0.05). Each panel displays mean z-score normalized gene 599 expression value (±SEM) for control (bold blue) and treatment (bold brown) conditions, with 600 individual gene profiles shown as thin lines. Sample sizes (n) indicate the number of genes 601 within each module. 602 Figure 5: Temporal modules of differentially abundant targeted metabolites. Four 603 metabolite modules (A-D) were identified through correlation-based hierarchical clustering of 604 targeted metabolites with significantly distinct temporal responses across treatment and control 605 conditions (modified ImpulseDE2, FDR-corrected p ≤ 0.05). Each panel displays mean z-score 606 normalized targeted metabolite peak height value (±SEM) for control (bold blue) and treatment 607 (bold brown) conditions, with individual metabolite profiles shown as thin lines. Sample sizes 608 (n) indicate the number of targeted metabolites within each module. 609 610

611 1. R. Tecon, D. Or, Biophysical processes supporting the diversity of microbial life in soil. 612 FEMS Microbiol. Rev. 41, 599–623 (2017). 613 2. S. Sherwood, Q. Fu, Climate change. A drier future? Science 343, 737–739 (2014). 614 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 16, 2025. ; https://doi.org/10.1101/2025.10.16.681527doi: bioRxiv preprint 15 3. R. Seager, et al., Whither the 100th Meridian? The Once and Future Physical and Human 615 Geography of America’s Arid–Humid Divide. Part I: The Story So Far. Earth Interact. 22, 616 1–22 (2018). 617 4. B. I. Cook, T. R. Ault, J. E. Smerdon, Unprecedented 21st century drought risk in the 618 American Southwest and Central Plains. Sci Adv 1, e1400082 (2015). 619 5. C. Lesk, P. Rowhani, N. Ramankutty, Influence of extreme weather disasters on global crop 620 production. Nature 529, 84–87 (2016). 621 6. X. He, et al., Integrated approaches to understanding and reducing drought impact on food 622 security across scales. Current Opinion in Environmental Sustainability 40, 43–54 (2019). 623 7. P. S. Lake, Ecological effects of perturbation by drought in flowing waters. Freshw. Biol. 624 48, 1161–1172 (2003). 625 8. S. E. Evans, M. D. Wallenstein, Climate change alters ecological strategies of soil bacteria. 626 Ecol. Lett. 17, 155–164 (2014). 627 9. S. E. Evans, I. C. Burke, Carbon and Nitrogen Decoupling Under an 11-Year Drought in the 628 Shortgrass Steppe. Ecosystems 16, 20–33 (2013). 629 10. M. Wallenstein, S. Evans, Microbial adaptations to envrionmental change: a moving target 630 for global change ecology. Nature Precedings (2011). 631 https://doi.org/10.1038/npre.2011.5510.1. 632 11. A. T. Austin, et al., Water pulses and biogeochemical cycles in arid and semiarid 633 ecosystems. Oecologia 141, 221–235 (2004). 634 12. T. Roy Chowdhury, et al., Metaphenomic Responses of a Native Prairie Soil Microbiome to 635 Moisture Perturbations. mSystems 4 (2019). 636 13. J. P. Schimel, Life in dry soils: Effects of drought on soil microbial communities and 637 processes. Annu. Rev. Ecol. Evol. Syst. 49, 409–432 (2018). 638 14. N. Xi, C. Chu, J. M. G. Bloor, Plant drought resistance is mediated by soil microbial 639 community structure and soil-plant feedbacks in a savanna tree species. E nviron. Exp. Bot. 640 155, 695–701 (2018). 641 15. E. J. Sayer, et al., Adaptation to chronic drought modifies soil microbial community 642 responses to phytohormones. Commun Biol 4, 516 (2021). 643 16. D. Naylor, D. Coleman-Derr, Drought Stress and Root-Associated Bacterial Communities. 644 Front. Plant Sci. 8, 2223 (2017). 645 17. C. Santos-Medellín, J. Edwards, Z. Liechty, B. Nguyen, V. Sundaresan, Drought Stress 646

in a Compartment-Specific Restructuring of the Rice Root-Associated 647 Microbiomes. MBio 8 (2017). 648 18. M. Potts, Desiccation tolerance of prokaryotes. Microbiol. Rev. 58, 755–805 (1994). 649 19. J. T. Lennon, Z. T. Aanderud, B. K. Lehmkuhl, D. R. Schoolmaster Jr, Mapping the niche 650 space of soil microorganisms using taxonomy and traits. Ecology 93, 1867–1879 (2012). 651 20. M. Potts, Mechanisms of desiccation tolerance in cyanobacteria. Eur. J. Phycol. 34, 319–652 328 (1999). 653 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 16, 2025. ; https://doi.org/10.1101/2025.10.16.681527doi: bioRxiv preprint 16 21. M. Potts, Desiccation tolerance: a simple process? Trends Microbiol. 9, 553–559 (2001). 654 22. J. K. Fredrickson, et al., Protein oxidation: key to bacterial desiccation resistance? ISME J. 655 2, 393–403 (2008). 656 23. K. Dose, A. Bieger-Dose, M. Labusch, M. Gill, Survival in extreme dryness and DNA-657 single-strand breaks. Adv. Space Res. 12, 221–229 (1992). 658 24. Y. Fang, L. M. McMullen, M. G. Gänzle, Effect of drying on oxidation of membrane lipids 659 and expression of genes encoded by the Shiga toxin prophage in Escherichia coli. Food 660 Microbiol. 86, 103332 (2020). 661 25. P. A. Hingston, M. J. Piercey, L. Truelstrup Hansen, Genes Associated with Desiccation 662 and Osmotic Stress in Listeria monocytogenes as Revealed by Insertional Mutagenesis. 663 Appl. Environ. Microbiol. 81, 5350–5362 (2015). 664 26. J. T. Lennon, S. E. Jones, Microbial seed banks: the ecological and evolutionary 665 implications of dormancy. Nat. Rev. Microbiol. 9, 119–130 (2011). 666 27. J. W. Lengeler, G. Drews, H. G. Schlegel, Biology of the Prokaryotes (Georg Thieme 667 Verlag, 1999). 668 28. M. T. Madigan, K. S. Bender, D. H. Buckley, W. M. Sattley, D. A. Stahl, Brock Biology of 669 Microorganisms. 15th Global Edition. Boston, US: Benjamin Cummins (2018). 670 29. T. E. Brewer, et al., Ecological and Genomic Attributes of Novel Bacterial Taxa That 671 Thrive in Subsurface Soil Horizons. MBio 10 (2019). 672 30. P. Beskrovnaya, et al., No endospore formation confirmed in members of the phylum 673 Proteobacteria. Appl. Environ. Microbiol. 87 (2021). 674 31. L. N. Csonka, Physiological and genetic responses of bacteria to osmotic stress. Microbiol. 675 Rev. 53, 121–147 (1989). 676 32. J. M. Wood, Bacterial responses to osmotic challenges. J. Gen. Physiol. 145, 381–388 677 (2015). 678 33. K. Killham, M. K. Firestone, Salt stress control of intracellular solutes in streptomycetes 679 in digenous to saline soils. Appl. Environ. Microbiol. 47, 301–306 (1984). 680 34. J. H. Crowe, J. F. Carpenter, L. M. Crowe, THE ROLE OF VITRIFICATION IN 681 ANHYDROBIOSIS. (2003). https://doi.org/10.1146/annurev.physiol.60.1.73. 682 35. L. K. Honeker, et al., Drought re-routes soil microbial carbon metabolism towards emission 683 of volatile metabolites in an artificial tropical rainforest. Nat. Microbiol. 8, 1480–1494 684 (2023). 685 36. G. R. Brown, I. C. Sutcliffe, D. Bendell, S. P. Cummings, The modification of the 686 membrane of Oceanomonas baumanniiT when subjected to both osmotic and organic 687 solvent stress. FEMS Microbiol. Lett. 189, 149–154 (2000). 688 37. M. van de Mortel, L. J. Halverson, Cell envelope components contributing to biofilm 689 growth and survival of Pseudomonas putida in low-water-content habitats: Adaptation of P. 690 putida to matric water stress. Mol. Microbiol. 52, 735–750 (2004). 691 38. E. B. Roberson, M. K. Firestone, Relationship between Desiccation and Exopolysaccharide 692 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 16, 2025. ; https://doi.org/10.1101/2025.10.16.681527doi: bioRxiv preprint 17 Production in a Soil Pseudomonas sp. Appl. Environ. Microbiol. 58, 1284–1291 (1992). 693 39. X.-J. Du, X.-Y. Wang, X. Dong, P. Li, S. Wang, Characterization of the Desiccation 694 Tolerance of Cronobacter sakazakii Strains. Front. Microbiol. 9, 2867 (2018). 695 40. E. J. Cytryn, et al., Transcriptional and physiological responses of Bradyrhizobium 696 japonicum to desiccation-induced stress. J. Bacteriol. 189, 6751–6762 (2007). 697 41. J. C. LeBlanc, E. R. Gonçalves, W. W. Mohn, Global response to desiccation stress in the 698 soil actinomycete Rhodococcus jostii RHA1. Appl. Environ. Microbiol. 74, 2627–2636 699 (2008). 700 42. P. H. Lebre, P. De Maayer, D. A. Cowan, Xerotolerant bacteria: surviving through a dry 701 spell. Nat. Rev. Microbiol. 15, 285–296 (2017). 702 43. L. Rajeev, et al., Dynamic cyanobacterial response to hydration and dehydration in a desert 703 biological soil crust. ISME J. 7, 2178–2191 (2013). 704 44. E. Segev, Y. Smith, S. Ben-Yehuda, RNA dynamics in aging bacterial spores. Cell 148, 705 139–149 (2012). 706 45. Sheridan G. E. C., Masters C. I., Shallcross J. A., Mackey B. M., Detection of mRNA by 707 Reverse Transcription-PCR as an Indicator of Viability in Escherichia coliCells. Appl. 708 Environ. Microbiol. 64, 1313–1318 (1998). 709 46. T. J. Hellyer, L. E. DesJardin, G. L. Hehman, M. D. Cave, K. D. Eisenach, Quantitative 710 analysis of mRNA as a marker for viability of Mycobacterium tuberculosis. J. Clin. 711 Microbiol. 37, 290–295 (1999). 712 47. R. I. Adams, et al., Microbes and associated soluble and volatile chemicals on periodically 713 wet household surfaces. Microbiome 5, 128 (2017). 714 48. S. J. Blazewicz, R. L. Barnard, R. A. Daly, M. K. Firestone, Evaluating rRNA as an 715 indicator of microbial activity in environmental communities: limitations and uses. ISME J. 716 7, 2061–2068 (2013). 717 49 . M. Wutkowska, A. Vader, S. Mundra, E. J. Cooper, P. B. Eidesen, Dead or Alive; or Does 718 It Really Matter? Level of Congruency Between Trophic Modes in Total and Active Fungal 719 Communities in High Arctic Soil. Front. Microbiol. 9, 3243 (2018). 720 50. M. Schostag, et al., Distinct summer and winter bacterial communities in the active layer of 721 Svalbard permafrost revealed by DNA- and RNA-based analyses. Front. Microbiol. 6, 399 722 (2015). 723 51. A. W. Bowsher, P. J. Kearns, A. Shade, 16S rRNA/rRNA Gene Ratios and Cell Activity 724 Staining Reveal Consistent Patterns of Microbial Activity in Plant-Associated Soil. 725 mSystems 4 (2019). 726 52. R. P. Bartelme, et al., Influence of Substrate Concentration on the Culturability of 727 Heterotrophic Soil Microbes Isolated by High-Throughput Dilution-to-Extinction 728 Cultivation. mSphere 5, e00024-20 (2020). 729 53. M. R. Kridler, et al., Draft genome sequences of Arthrobacter sp. AZCC_0090 and 730 Mycobacterium sp. AZCC_0083 isolated from oligotrophic subsurface forest soil in the 731 Santa Catalina mountains of Southern Arizona. Microbiol Resour Announc e0108923 732 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 16, 2025. ; https://doi.org/10.1101/2025.10.16.681527doi: bioRxiv preprint 18 (2024). 733 54. J. G. Kramer, F. L. Singleton, Variations in rRNA content of marine Vibrio spp. during 734 starvation-survival and recovery. Appl. Environ. Microbiol. 58, 201–207 (1992). 735 55. Y. Himeoka, B. Gummesson, M. A. Sørensen, S. L. Svenningsen, N. Mitarai, Distinct 736 survival, growth lag, and rRNA degradation kinetics during long-term starvation for carbon 737 or phosphate. mSphere 7, e0100621 (2022). 738 56. P. Blount, I. Iscla, Life with bacterial mechanosensitive channels, from discovery to 739 physiology to pharmacological target. Microbiol. Mol. Biol. Rev. 84, e00055-19 (2020). 740 57. N. Gruzdev, et al., Global transcriptional analysis of dehydrated Salmonella enterica 741 serovar Typhimurium. Appl. Environ. Microbiol. 78, 7866–7875 (2012). 742 58. S. P. Couvillion, et al., Rapid remodeling of the soil lipidome in response to a drying-743 rewetting event. Microbiome 11, 34 (2023). 744 59. R. J. Sundberg, R. B. Martin, Interactions of histidine and other imidazole derivatives with 745 transition metal ions in chemical and biological systems. Chem. Rev. 74, 471–517 (1974). 746 60. S. Chakrabortee, et al., Catalytic and chaperone-like functions in an intrinsically disordered 747 protein associated with desiccation tolerance. Proc. Natl. Acad. Sci. U. S. A. 107, 16084–748 16089 (2010). 749 61. K. Goyal, L. J. Walton, J. A. Browne, A. M. Burnell, A. Tunnacliffe, Molecular 750 anhydrobiology: identifying molecules implicated in invertebrate anhydrobiosis. Integr. 751 Comp. Biol. 45, 702–709 (2005). 752 62. I. L. Jung, I. G. Kim, Thiamine protects against paraquat-induced damage: scavenging 753 activity of reactive oxygen species. Environ. Toxicol. Pharmacol. 15, 19–26 (2003). 754 63. Y. Okai, K. Higashi-Okai, E. F Sato, R. Konaka, M. Inoue, Potent radical-scavenging 755 activities of thiamin and thiamin diphosphate. J. Clin. Biochem. Nutr. 40, 42–48 (2007). 756 64. A. G ę gotek, E. Skrzydlewska, Ascorbic acid as antioxidant. Vitam. Horm. 121, 247–270 757 (2023). 758 65. T. Hirakawa, S. Tanno, K. Ohara, N-acetylglutamic acid alleviates oxidative stress based on 759 histone acetylation in plants. Front. Plant Sci. 14, 1165646 (2023). 760 66. E. Albers, Metabolic characteristics and importance of the universal methionine salvage 761 pathway recycling methionine from 5â/i4 2-methylthioadenosine. IUBMB Life 61, 1132–1142 762 (2009). 763 67. D. B. Roszak, R. R. Colwell, Survival strategies of bacteria in the natural environment. 764 Microbiol. Rev. 51, 365–379 (1987). 765 68. P. Setlow, G. Christie, Bacterial Spore mRNA - What’s Up With That? Front. Microbiol. 766 11, 596092 (2020). 767 69. N. C. Scales, K. T. Huynh, C. Weihe, J. B. H. Martiny, Desiccation induces varied 768 responses within a soil bacterial genus. Environ. Microbiol. 25, 3075–3086 (2023). 769 70. Y. Wang, et al., RNA-based amplicon sequencing is ineffective in measuring metabolic 770 activity in environmental microbial communities. Microbiome 11, 131 (2023). 771 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 16, 2025. ; https://doi.org/10.1101/2025.10.16.681527doi: bioRxiv preprint 19 71. E. Blagodatskaya, Y. Kuzyakov, Active microorganisms in soil: Critical review of 772 estimation criteria and approaches. Soil Biol. Biochem. 67, 192–211 (2013). 773 72. K. Tamrakar, P. W. Miller, M. C. Dolan, A. J. Wijeratne, Assessment of rhizosphere 774 microbial activity using optimized RNA extraction coupled with universal ribosomal RNA 775 depletion techniques. BMC Methods 2, 1–10 (2025). 776 73. T. Bang-Andreasen, et al., Total RNA sequencing reveals multilevel microbial community 777 changes and functional responses to wood ash application in agricultural and forest soil. 778 FEMS Microbiol. Ecol. 96, fiaa016 (2020). 779 74. P. F. Chuckran, et al., Codon bias, nucleotide selection, and genome size predict in situ 780 bacterial growth rate and transcription in rewetted soil. Proc. Natl. Acad. Sci. U. S. A. 122, 781 e2413032122 (2025). 782 75. L. Greenspan, Humidity fixed points of binary saturated aqueous solutions. J. Res. Natl. 783 Bur. Stand. A Phys. Chem. 81A, 89 (1977). 784 76. S. Andrews, FastQC A Quality Control tool for High Throughput Sequence Data. (2012). 785 Available at: https://www.bioinformatics.babraham.ac.uk/projects/fastqc/ [Accessed 9 May 786 2023]. 787 77. D. Blankenberg, et al., Galaxy: a web-based genome analysis tool for experimentalists. 788 Curr. Protoc. Mol. Biol. Chapter 19, Unit 19.10.1-21 (2010). 789 78. B. Langmead, S. L. Salzberg, Fast gapped-read alignment with Bowtie 2. Nat. Methods 9, 790 357–359 (2012). 791 79. Y. Liao, G. K. Smyth, W. Shi, The R package Rsubread is easier, faster, cheaper and better 792 for alignment and quantification of RNA sequencing reads. Nucleic Acids Res. 47, e47 793 (2019). 794 80. A. Apprill, S. McNally, R. Parsons, L. Weber, Minor revision to V4 region SSU rRNA 795 806R gene primer greatly increases detection of SAR11 bacterioplankton. Aquat. Microb. 796 Ecol. 75, 129–137 (2015). 797 81. D. S. Fischer, F. J. Theis, N. Yosef, Impul se model-based differential expression analysis of 798 time course sequencing data. Nucleic Acids Res. 46, e1 19 (2018). 799 82. Y. Yao, et al. , Analysis of Metabolomics Datasets with High-Performance Computing and 800 Metabolite Atlases. Metabolites 5, 431–442 (2015). 801 83. L. W. Sumner, et al., Proposed minimum reporting standards for chemical analysis 802 Chemical Analysis Working Group (CAWG) Metabolomics Standards Initiative (MSI). 803 Metabolomics 3, 211–221 (2007). 804 84. T. F. Leao, et al., Quick-start infrastructure for untargeted metabolomics analysis in GNPS. 805 Nat. Metab. 3, 880–882 (2021). 806 85. K. Dührkop, et al., SIRIUS 4: a rapid tool for turning tandem mass spectra into metabolite 807 structure information. Nat. Methods 16, 299–302 (2019). 808 86. M. I. Love, W. Huber, S. Anders, Moderated estimation of fold change and dispersion for 809 RNA-seq data with DESeq2. Genome Biol. 15, 550 (2014). 810 87. Community Ecology Package [R package vegan version 2.7-1]. Comprehensive R Archive 811 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 16, 2025. ; https://doi.org/10.1101/2025.10.16.681527doi: bioRxiv preprint 20 Network (CRAN) (2025). Available at: https://CRAN.R-project.org/package=vegan 812 [Accessed 11 September 2025]. 813 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 16, 2025. ; https://doi.org/10.1101/2025.10.16.681527doi: bioRxiv preprint Relative humidity (% at 25ºC) 0 10 20 30 40 50 60 70 80 90 100 Time elapsed (days) 0 2 4 6 8 10 12 14 16 Treatment Control Cultivability Transcriptomics Metabolomics A B C dH2O KCl NaCl NaBr K2CO3 MgCl2 K-Ac dH2O .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 16, 2025. ; https://doi.org/10.1101/2025.10.16.681527doi: bioRxiv preprint Washed Day 2 C: 100% T: 100% Day 8 C: 100% T: 65% Day 14 C: 100% T: 26% Day 16 C: 100% T:100% * BA Day 2 C: 100% T: 100% Day 8 C: 100% T: 65% Day 14 C: 100% T: 26% Day 16 C: 100% T:100% Cultivability (%)1 10 100 * Relative cultivability (% of hydrated control) 0.01 0.1 1 10 Kruskal-Wallis P= 1.76e-8 ba b a .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 16, 2025. ; https://doi.org/10.1101/2025.10.16.681527doi: bioRxiv preprint C: KW p = 0.033 T: KW p = 0.305 C: KW p = 0.384 C: KW p = 0.541 T: KW p = 0.644 T: KW p = 0.312 A B C 23S rRNA copies ng RNA-1 106 2 106 106 107 Total RNA concentration (ng µl-1) 0 50 100 150 200 5 105 2 105 Washed Day 2 Day 8 Day 14 Day 16 16S rRNA copies ng RNA-1 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 16, 2025. ; https://doi.org/10.1101/2025.10.16.681527doi: bioRxiv preprint Condition Control T reatment A −1 0 1 2 2 8 14 16 Days Z−score Normalized Expression T ranscriptome module 1 (n=571) B −1 0 1 2 2 8 14 16 Days Z−score Normalized Expression T ranscriptome module 2 (n=325) C −1 0 1 2 2 8 14 16 Days Z−score Normalized Expression T ranscriptome module 3 (n=173) d −1 0 1 2 2 8 14 16 Days Z−score Normalized Expression T ranscriptome module 4 (n=117) E −1 0 1 2 2 8 14 16 Days Z−score Normalized Expression T ranscriptome module 5 (n=143) F −1 0 1 2 2 8 14 16 Days Z−score Normalized Expression T ranscriptome module 6 (n=80) .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 16, 2025. ; https://doi.org/10.1101/2025.10.16.681527doi: bioRxiv preprint A −1 0 1 4 8 12 16 Time Z−score Normalized Intensity Metabolite cluster 1 (n=21) B −1 0 1 4 8 12 16 Time Z−score Normalized Intensity Metabolite cluster 2 (n=14) C −1.5 −1.0 −0.5 0.0 0.5 1.0 4 8 12 16 Time Z−score Normalized Intensity Metabolite cluster 3 (n=8) D −1.5 −1.0 −0.5 0.0 0.5 1.0 4 8 12 16 Time Z−score Normalized Intensity Metabolite cluster 4 (n=3) Condition Control Treatment .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 16, 2025. ; https://doi.org/10.1101/2025.10.16.681527doi: bioRxiv preprint