Sarbecovirus –associated gut microbiome instability in a natural bat reservoir

microbiome, Rhinolophus, diet, infection, Anna Karenine Principle, coronavirus 23 24

25 Sarbecoviruses, a subgenus of Betacoronavirus, display both respiratory and gastrointestinal tropism, 26 suggesting potential interactions with host gut microbial communities. However, ecological signatures 27 of infection in wild bats remain poorly understood. We investigated associations between Sarbecovirus 28 infection status, gut microbiome structure, and diet composition in Rhinolophus shameli roosting in 29 northeastern Cambodia. Fecal samples collected across dry and wet seasons (2023–2024) were subject 30 to full-length 16S rRNA gene sequencing and arthropod DNA metabarcoding. Sarbecovirus–positive 31 bats exhibited stable alpha diversity but consistent shifts in gut community composition and increased 32 interindividual variability consistent with the Anna Karenina Principle, suggesting infection–associated 33 destabilization of community assembly rather than diversity erosion . Infection status was associated 34 with enrichment of Shigella and Escherichia species, taxa linked to inflammatory or epithelial stress 35 states in bats. In contrast, dietary composition showed no strong global structuring by infection status 36 and weak coupling with bacterial community structure, suggesting that trophic ecology is unlikely to 37 be the main driver of the infection–associated microbiome signal. Although causal directionality cannot 38 be inferred, our results reveal measurable and consistent microbiome restructuring associated with 39 Sarbecovirus detection in a natural reservoir host and highlight the potential of microbiome profiling 40 for monitoring wildlife disease processes. 41 42

43 Coronaviruses are globally distributed RNA -viruses infecting a wide range of species, including 44 humans, and causing a broad spectrum of diseases. Research into the origins of SARS -CoV-2 and 45 continuing interest in Coronavirus ecology and evolution have highlighted the value of wild bat 46 surveillance. Bats harbour ancestral lineages of betacoronaviruses from which several viruses of major 47 public health concern have emerged, including SARS -CoV and SARS-CoV-2, both belonging to the 48 subgenus Sarbecovirus [1–2]. Despite variability in tropisms of Sarbecovirus across bat species [3], 49 intense testing of bats worldwide for betacoronaviruses showed increased detection in rectal or fecal 50 samples compared to other sample types [4]. The key protein responsible for SARS-CoV-2 viral entry 51 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 27, 2026. ; https://doi.org/10.64898/2026.03.26.714368doi: bioRxiv preprint into the host cell is glycoprotein S, known as the spike protein. This viral spike protein binds to 52 angiotensin–converting enzyme 2 (ACE2), a cell surface receptor [5] expressed in lung tissue as well 53 as in the oesophageal and intestinal epithelium. As documented in 2020 [6], a variable range of gastro–54 intestinal (GI) symptoms are observed during SARS -CoV-2 infection in humans. Those symptoms 55 range from diarrhea, nausea, anorexia, abdominal pain and belching [7]. However, the vast majority of 56 studies evaluating the bat host response to Coronavirus infection have been performed in cell lines [8], 57 and infected animals did not exhibit evident clinical signs of infection [2, 9]. We could expect that such 58 an infection with GI tract tropism would have consequences on the gut microbiota. 59 Across wildlife, domestic animals, and humans, reduced microbiome diversity is often associated with 60 pathogenic overgrowth, loss of rare taxa, weakened colonization resistance and increased co-infection 61 risk [10]. Experimental work further suggests that bat-associated gut microbiota may contribute to viral 62 tolerance, as mice transplanted with fecal microbiota from Asian insectivorous bats showed reduced 63 mortality, symptoms, and viral loads following H1N1 infection compared to controls [11]. 64 Host biological traits also shape pathogen dynamics at multiple scales. At the individual level, 65 physiological states such as gestation and lactation involve immunological shifts toward anti -66 inflammatory responses that may transiently increase viral susceptibility and contribute to sex - and 67 season-specific seroprevalence patterns [12–14]. At the population level, seasonal pulses of viral 68 circulation are largely driven by the synchronous introduction of immunologically naïve juveniles, 69 leading to rapid declines in population) level immunity and increased transmission [12, 15]. Such 70 immune fluctuations may interact with microbial communities , alter ing community structure and 71 influencing colonization resistance and tolerance mechanisms [16]. 72 Patterns of microbiome variability under infection have led to the application of the Anna Karenina 73 Principle (AKP) to host–associated microbial communities [17]. Under AKP, stressors are expected to 74 induce stochastic, host –specific disruptions of microbial communities rather than a shared, 75 deterministic shift, such that healthy microbiomes are relatively similar to one another, whereas 76 dysbiotic microbiomes diverge among hosts. This manifests as greater β–diversity dispersion rather 77 than a consistent shift in mean composition. In humans with SARS -CoV-2 infection, increased fecal 78 microbiome dispersion correlates with infection status and symptom severity [18, 19]. Evidence in bats 79 remains scarce. A recent investigation in Artibeus jamaicensis demonstrated that enteric astrovirus 80 infection altered gut microbial richness in an age ‑dependent manner yet did not lead to increased 81 among‑individual community dispersion [20]. In contrast, infection by Hibecovirus in Hipposideros 82 caffer reduced bacterial richness and increased interindividual dispersion consistent with AKP 83 predictions [21]. Although causality cannot be confirmed, the short gut transit time of bats and the 84 detectability of viral RNA via qPCR support a potential link between active infection and microbial 85 perturbation [21, 22]. 86 Experimental induction of inflammation in Rousettus aegyptiacus also generated AKP-like dispersion 87 and revealed inflammation–associated microbial markers, particularly Escherichia spp., consistent with 88 immune–driven alterations of the gut environment [23]. Identifying similar microbial indicators in free–89 ranging bats may provide valuable, non –invasive insights into subclinical infection or immune 90 activation [9). Beyond pathogens, environmental and ecological factors also influence bat gut 91 microbiomes. Dietary shifts [24–26], habitat disturbance [27–28] and reproductive state [29] can all 92 alter microbial diversity and stability, with potential implications for pathogen dynamics [11). While 93 many bat diets remain to be fully described [30], few studies to date investigated viral shedding with 94 diet composition. Falvo et al. [31] and Vanalli et al. [32] tested experimentally the impact of diet change 95 on influenza viral charge in the Jamaican fruit bat and observed less viral shedding under optimal diet 96 despite not reaching statistical significance in every case explored. 97 Overall, the literature supports a tight interconnection between bacterial diversity, viral prevalence, host 98 biology, and environmental context. Both causal directions remain plausible: viral infection and 99 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 27, 2026. ; https://doi.org/10.64898/2026.03.26.714368doi: bioRxiv preprint associated immune responses may disrupt the microbiome, but pre –existing microbial configurations 100 shaped by diet and seasonality may also influence susceptibility, tolerance, or shedding. Building on 101 this framework, we investigated associations between Sarbecovirus infection status, gut microbiome 102 structure and dietary composition in Rhinolophus shameli. This Southeast Asian bat species occupies a 103 wide range of forest habitats [33–34] and shows high prevalence of Sarbecoviruses in northern 104 Cambodia [35–36] Yet its ecology is poorly documented, with diet data limited to microscopic analyses 105 indicating predominance of Lepidoptera and Coleoptera [37]. 106 Given the GI tropism of Sarbecoviruses, we hypothesized that Sarbecovirus infection would be 107 associated with increased gut microbiome instability consistent with AKP expectations while 108 uninfected individuals would exhibit higher microbial diversity and more constrained community 109 structure. We further hypothesized that diet would act as a major driver of the gut microbiome structure 110 and potentially modulate infection–associated patterns. By comparing bacterial diversity, community 111 composition, and bacterial –diet co -occurrences between Sarbecovirus–positive and –negative 112 individuals, we aimed to (i) assess whether infection status is associated with gut microbial imbalance, 113 (ii) evaluate the extent to which such patterns are independent of trophic ecology, and (iii) identify 114 microbial taxa consistently associated with inflammatory or infection-related states in wild bats, without 115 presuming causal directionality. 116 117

118 1. Bat capture and sampling 119 Bat sampling was conducted in Stung Treng province (northern Cambodia) between March 2023 and 120 December 2024 as part of a longitudinal investigation following repeated detections of SARS–CoV–2–121 related viruses in Rhinolophus bats [36, 38]. Nine field sessions were scheduled according to the 122 reproductive phenology of Rhinolophus spp., when Coronavirus (CoVs) circulation is typically 123 elevated [12]. Sessions occurred every 6–8 weeks from March to August, with an additional late–season 124 survey in 2024. Three previously surveyed hills (Phnom Chhgnauk, Phnom Kar Ngoark and Phnom 125 Chab Pleurng, Figure 1) were repeatedly sampled [36, 38]. Each session lasted six nights, with each 126 site visited twice on non –consecutive nights. Mist nets and harp traps were set at cave entrances and 127 adjacent forested areas. 128 Captured bats were individually placed in cotton bag and transported to a mobile field laboratory for 129 species identification and morphometric measurements [39]. Standardized samples were collected for 130 each animal, including rectal swabs (one in viral transport medium, one in TRIzol) and feces. Fecal 131 pellets were preserved in 70% Ethanol and stored on ice until transfer to the Institut Pasteur du 132 Cambodge (IPC). All other samples were kept on ice, transferred to liquid nitrogen within 24 h, and 133 stored at −80 °C at the IPC. 134 2. Sarbecovirus screening 135 Sarbecovirus detection followed the protocol described in Guillebaud et al. [36] Total RNA was 136 extracted from TRIzol–preserved rectal swabs using the Direct-zol RNA MiniPrep kit (Zymo Research, 137 USA). A duplex one -step real -time PCR targeting the E and N genes was performed using the 138 Superscript III one –step RT –PCR system with Platinum Taq Polymerase (Invitrogen, Darmstadt, 139 Germany). Samples with Ct < 40 were subjected to Sanger sequencing (Macrogen, Inc., Seoul, Republic 140 of Korea) in both forward and reverse directions using pan-CoV nested-PCR [40] targeting the RdRp 141 gene. The sequences obtained were verified by similarity using the NCBI BLASTn search. A total of 142 43 bats were Sarbeco–positive, 12 CoV–positive and 174 negative individuals (Table 1, Figure 1), and 143 Sarbecovirus detected belong to the lineage “group 1,” with representative strains such as RshSTT039 144 and RshSTT671, related to the earlier RshSTT182/200 viruses [36]. 145 3. DNA processing 146 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 27, 2026. ; https://doi.org/10.64898/2026.03.26.714368doi: bioRxiv preprint DNA extraction from feces preserved in ethanol was conducted using the QIAamp Fast DNA Stool 147 Mini Kit from Qiagen following the manufacturer’s protocol. An extraction control was added to each 148 extraction batch (n = 9). Four mock communities from bacterial DNA (HM –782D, BEI Resources) 149 were added to control sequencing accuracy. In PCR 1, we amplified the full 16SrRNA gene using one 150 primer pair (27F–1492R [41]). Following DNA purification, quantification and pooling at equimolarity, 151 DNA library construction and sequencing were conducted at the University of Liège GIGA Genomics 152 platform using a a total of 217 samples, nine extraction controls, ten PCR 1 controls and four mock 153 communities samples were sequenced on an GridION (Oxford Nanopore Technologies, Oxford, UK) 154 using a Ligation Sequencing Kit (SQK–LSK112) with 100,000 reads per sample as target. 155 We used the same DNA extract for the diet of the bats that corresponds to the same samples extracted 156 for microbiome data generation. Briefly, similar to [42] we amplified the mitochondrial cytochrome c 157 oxidase subunit I (COI) using two primer pairs [43–44] Further details on data processing can be found 158 in Supp. material. Following the whole filtering process, our dataset consisted of 1716 distinct 159 biological entities across 153 samples. Due to the low level of taxonomical resolution, we calculated 160 the frequency of occurrence per order for statistical analysis. 161 4. Bioinformatics treatments 162 After basecalling and barcode demultiplexing using Dorado v1.0.2 and discarding unclassified reads, 163 we controlled the quality of these demultiplexed reads and used cutadapt [45] to trim primers for all 164 sequence reads. For the identification of bacteria at species level, fastq files containing full length 16S 165 rRNA gene amplicons were uploaded to the EPI2ME desktop agent 16S workflow (version 2020.2.10, 166 ONT) in which each file was classified using the NCBI 16S rRNA gene blast database through 167 Minimap2 [46] sequence alignment program , yielding closed–reference taxonomic profiles without 168 ASV or OTU inference. Exclusion criteria for nanopore reads were an alignment count accuracy <95%, 169 query cover <90%, quality score (QC) score <10 and read length between 1000 and 2000 bp. The 170 decontam package in R was used to detect true contaminants present in extraction and PCR1 controls, 171 resulting in 48 true bacterial contaminants. Mock communities’ samples were checked for accurate 172 taxonomic affiliation (Supp. Figure 1). Control samples were removed, and taxa not belonging to the 173 kingdom Bacteria were also removed leaving a total of 3253 bacterial taxa. Finally, due to great 174 variation in sampling depth between samples with a mean of reads by sample of 74210 (SD = 46880), 175 we rarefied data to 10,000 reads per sample based on rarefaction curves, resulting in a final dataset of 176 199 samples and 2053 bacterial taxa. 177 5. Statistical analysis 178 a. Diet analysis 179 We analyzed the structure of the bat dietary assemblages based on the presence –absence of arthropod 180 orders detected in each fecal sample. Non –metric multidimensional scaling (NMDS) was performed 181 using the Jaccard distance to visualize differences in dietary composition among samples. The number 182 of dimensions was selected to minimize stress while retaining ecological interpretability. To identify 183 diet order associated with dietary variation, we fitted diet orders onto the NMDS ordination using the 184 envfit function (vegan package). In addition, we computed Spearman rank correlations (with 185 Benjamini–Hochberg procedure) between NMDS axis scores and the relative frequencies of each 186 arthropod order to quantify their direction and strength of association with the ordination gradients. We 187 further tested whether Sarbecovirus infection status explained differences in dietary composition using 188 a PERMANOVA (permutations n = 9999) based on the same Jaccard distance matrix. To evaluate 189 temporal trends in dietary composition and potential effects of Sarbecovirus infection, we modeled the 190 NMDS axis scores as smooth functions of sampling date. Specifically, Generalized Additive Models 191 (GAMs) were fitted separately for the first two NMDS axes (details in Supp. material). 192 b. Microbiome analysis 193 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 27, 2026. ; https://doi.org/10.64898/2026.03.26.714368doi: bioRxiv preprint Based on collinearity investigations (Supp. material), we retained observed species richness (reflecting 194 taxonomic diversity), Simpson index (reflecting evenness) and Faith’s phylogenetic diversity 195 (phylogenetic breadth) for subsequent analyses of bacterial diversity. For each index, we fitted linear 196 mixed–effects models including age group (Adult, immature and juvenile) , sex, site of capture , 197 Sarbecovirus infection status, and dietary composition (represented by NMDS1 and NMDS2 scores) as 198 fixed effects, while accounting for season and year as a random intercept to control for temporal 199 dependence among sampling sessions. Models were inspected for residual normality and 200 homoscedasticity, and statistical significance of fixed effects was evaluated using Satterthwaite’s 201 approximation (lmerTest package). To evaluate whether Sarbecovirus infection was associated with 202 shifts in bacterial community composition (i.e., dysbiosis, AKP), community dissimilarity among 203 samples was quantified using three complementary distance metrics: Bray–Curtis, unweighted UniFrac, 204 and weighted UniFrac. Ordinations were visualized using Principal Coordinates Analysis (PCoA) to 205 illustrate patterns of beta diversity. We tested the effects of host and environmental variables on 206 microbiome composition using a PERMANOVA for each distance metric following the model: 207 Distance ∼ Age_Group+Sarbeco_status+Sex+Site. Season_Year was used as a stratification variable 208 in the permutation procedure and permutation tests were performed with 9,999 iterations. We also 209 evaluated multivariate homogeneity of variance using the betadisper function. 210 Following methods detailed in Melville et al. [21), We examined the effect of Sarbecovirus infection 211 status on log–transformed body condition using a linear mixed–effects model with sex and site as fixed 212 effects and season as a random intercept . We additionally used a binomial generalized linear mixed –213 effects model (logit link) to assess associations between Sarbecovirus infection status, microbiome 214 alpha diversity indices, and diet composition (NMDS axes values) , including capture periods as a 215 random intercept. 216 Differences in bacterial taxon abundance between Sarbecovirus infection statuses were tested using the 217 ANCOM–BC2 framework on non –rarefied reads . The model was fitted at the species level with 218 infection status as a fixed effect and no random effect. Taxa with prevalence below 10% or total library 219 counts below 500 reads were excluded prior to testing. Significance was assessed at α = 0.05 after false 220 discovery rate (FDR) adjustment of p–values. 221 c. HMSC model 222 We analysed the joint responses of insect taxa and bacterial species using Hierarchical Modelling of 223 Species Communities (HMSC), a Bayesian joint species distribution modelling framework 224 implemented in the Hmsc R package [47–48]. We modelled two distinct biological communities 225 measured across the same 139 samples: (1) insect occurrence data (15 taxa; presence/absence), and (2) 226 bacterial abundance data (19 taxa showing differential abundance from ANCOM -BC2 analysis; 227 abundance data). The two response matrices were combined column–wise into a single response matrix. 228 Because the taxa belong to different biological groups and require different statistical error models, we 229 assigned each column a specific response distribution: a probit distribution for insect presence/absence 230 data and a lognormal–Poisson distribution for bacterial abundance data. We ran MCMC chains with a 231 burn-in period and sampling length sufficient to achieve convergence, evaluated using trace plots, 232 effective sample sizes, and potential scale reduction factors (PSRF). Posterior summaries were obtained 233 for model parameters, including species –specific environmental responses, residual species 234 associations, and latent factor loadings. Model performance was assessed using posterior predictive 235 checks, explained variance, and species –specific predictive power measured via Tjur’s R² (for binary 236 responses) and log -normal pseudo -R² (for abundance data). The residual correlation matrix among 237 species was used to infer potential ecological associations. 238 239

240 1. Diet variation and viral status 241 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 27, 2026. ; https://doi.org/10.64898/2026.03.26.714368doi: bioRxiv preprint The overall diet of R. shameli was dominated by Coleoptera (Frequency Of Occurrence, FOO = 0.451), 242 followed by Hemiptera (FOO = 0.379), Lepidoptera (FOO = 0.353) and Diptera (FOO = 0.333), with 243 lower detection of Diptera and Lepidoptera in 2023 compared to 2024 (Figure 2A). Despite low level 244 of taxonomical resolution, we detected high FOO for the families of Cicadellidae and 245 Rhyparochromidae (Hemiptera, 0.163 and 0.111), Scabaridae (Coleoptera, 0.098), Blattelidae and 246 Termitidae (Blattodea, 0.084 and 0.071), Lecithoceridae, Geometridae and Crambidae (Lepidoptera, 247 0.065) as well as Limoniidae (Diptera, 0.063). Among taxa detected at the species level, the 248 Anatrachyntis simplex moth was found in 2.6% of samples which is a common pest of cotton, maize, 249 banana, pomegranate [49]. 250 The NMDS ordination based on Jaccard distances (stress = 0.095) provided a robust two–dimensional 251 representation of dietary composition among R. shameli individuals (Figure 2B). Samples displayed 252 moderate overlap, suggesting partial differentiation in diet profiles across individuals. The envfit 253 analysis confirmed that dietary composition was significantly structured by the relative occurrence of 254 eight insect orders (all q < 0.05). The strongest correlates were Hymenoptera (R² = 0.37, p < 0.001) and 255 Hemiptera (R² = 0.19, p < 0.001), indicating that variation along the ordination axes primarily reflected 256 contrasts among these dominant prey groups. Correlation strengths were consistent with the envfit 257

(Figure 2C), highlighting a strong positive association between NMDS1 and Coleoptera (ρ = 258 0.62, q 0.30, q 0.45, q < 0.001), 260 whereas Hemiptera showed a significant negative correlation ( ρ = –0.38, q < 0.001). These patterns 261 confirm that variation in insect order composition primarily reflects contrasts between beetle – and 262 hymenopteran–dominated diets versus those dominated by dipteran or hemipteran prey. 263 In contrast, Sarbecovirus infection status had no detectable effect on overall diet composition 264 (PERMANOVA: R² = 0.01, p = 0.14), and homogeneity of dispersion did not differ between groups 265 (betadisper: p = 0.27). (GAMs were used to assess whether diet composition varied seasonally or in 266 relation to Sarbecovirus infection. For the first diet NMDS axis, neither the cyclic smoother for day of 267 year (edf = 0.68, F = 0.12, p = 0.277) nor Sarbecovirus infection status ( t = -0.57, p = 0.567) were 268 significant predictors. The model explained only 1.4% of the deviance (adjusted R² = 0.014). In contrast, 269 the second diet NMDS axis displayed a significant seasonal pattern ( edf = 3.30, F = 2.26, p = 0.0038) 270 and was negatively associated with Sarbecovirus infection (Estimate = -0.43 ± 0.16, t = -2.92, p = 271 0.0036). The model explained 14.9% of deviance (adjusted R² = 0.124), suggesting that both temporal 272 variation and infection status moderately contributed to shaping dietary composition along this axis. 273 Specifically, infected bats tended to occupy lower NMDS2 scores, corresponding to a dietary shift 274 towards Hemiptera prey, relative to non–infected individuals (Figure 2D). 275 2. Alpha–, beta– diversity in microbiome and viral status 276 The gut microbial community was dominated by Pseudomonadota (78%; formerly known as 277 Proteobacteria) represented mainly by members of bacterial class Gammaproteobacteria 278 (Enterobacteriaceae, Pasteurellaceae), and Bacilliota (18.9%; formerly known as Firmicutes), by and 279 large ascribed to Bacilli (Streptococcaceae, Listeriaceae, Bacilliaceae , Figure 3). Non–infected and 280 Sarbeco–positive bats shared only 26% of bacterial species in common (543 species), representing 92% 281 of the overall microbiome in read abundance (Supp Figure 2). 282 Linear mixed–effects models were used to explore ecological and host predictors of bacterial alpha 283 diversity. Across all models, the random effect of seasonality explained little variance (SD = 0. 04–284 12.3), indicating limited temporal influence on bacterial diversity. For species richness, none of the 285 predictors were statistically significant (p > 0.05, Supp. Table 1). For Simpson diversity, age group had 286 a modest effect (p = 0.002), with immature bats exhibiting lower bacterial evenness (Estimate = -0.09 287 ± 0. 03) compared to adults. No other covariates, including infection status or diet composition, 288 significantly explained variation in Simpson diversity ( p > 0.24 for all). For Faith’s phylogenetic 289 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 27, 2026. ; https://doi.org/10.64898/2026.03.26.714368doi: bioRxiv preprint diversity (PD), the same pattern was found as Simpson index for immature bats (Estimate = -1.29± 0.3, 290 p = 0.02), while none of the other tested variables were significant (p > 0.10). Overall, bacterial alpha 291 diversity was relatively stable across sex, site, and infection status, with only slight evidence for lower 292 diversity in younger individuals. 293 Multivariate analyses revealed that Sarbecovirus infection was significantly linked to distinct bacterial 294 community composition in bats (Supp . Table 2). In fection status explained a small but significant 295 portion of variation under Bray –Curtis (F = 1.97, R² = 0.01, p = 0.006; Fig ure 4A) and unweighted 296 UniFrac distances ( F = 1.59, R² = 0.008, p = 0.016), indicating detectable infection –associated 297 dysbiosis. Dispersion tests showed higher within–group variability for infected individuals under Bray–298 Curtis F = 3.75, p = 0.05) and weighted UniFrac ( F = 4.81, p = 0.029), suggesting concurrent 299 community instability (AKP, Figure 4B). No consistent effects of age, sex, or site were detected across 300 models (Supp. Table 1). The ANCOM –BC2 analysis identified 19 bacterial species whose relative 301 abundances differed significantly between Sarbeco–positive and Sarbeco–negative R. shameli (FDR < 302 0.05; Figure 5A). Among the most discriminant taxa, Shigella dysenteriae, S. flexneri, Escherichia coli, 303 E. fergusonii, and E. marmotae were significantly enriched in infected individuals (log–fold change > 304 1). In contrast, a broad suite of Enterobacteriaceae symbionts as well as Enterococcus casseliflavus 305 and Serratia ureilytica were significantly depleted in infected bats (Figure 5B). Together, these results 306 indicate that Sarbecovirus infection in R. shameli is associated less with a reduction in within –host 307 bacterial diversity than with a destabilization and reorganization of between–host community structure. 308 3. Viral status as response 309 Among non–pregnant adults (N=113), log body condition did not differ significantly by Sarbecovirus 310 infection or sex or site (Supp. Table 3). The random effect of seasonality accounted for a small portion 311 of the variance (SD = 0.026), indicating that body condition remained relatively consistent across 312 sampling periods. Sarbecovirus infection status was not associated with microbiome alpha diversity 313 metrics but was significantly related to diet composition along NMDS axis 2 (Supp. Table 4), with 314 increasing values associated with lower odds of infection (odds ratio = 0.48, p = 0.049). suggesting that 315 infected bats may occupy distinct dietary niches with slightly lower NMDS2 scores. The random effect 316 of seasonality explained substantial variation in Sarbecovirus infection probability (SD = 1.28 on the 317 log–odds scale), indicating strong seasonal heterogeneity independent of diet and microbiome 318 predictors. 319 According to the HMSC model, the residual association heatmap revealed strong, coherent structure 320 within the bacterial community and weak structure among arthropod prey. Bacterial taxa formed a 321 highly connected block of strong positive residual correlations, visible as the intense red region in the 322 upper–right portion of the heatmap (Figure 6). This indicates that many bacterial species co–occur more 323 often than expected from shared environmental responses alone, suggesting common unmeasured 324 drivers or facilitative interactions. In contrast, arthropod taxa exhibited only weak and diffuse 325 associations, forming a pale, largely unstructured block at the lower–left of the matrix, consistent with 326 limited residual co –occurrence among arthropods. A notable feature of the matrix is the presence of 327 several bacterial species that show little or no positive residual association with the rest of the bacterial 328 community, visible as pale vertical and horizontal bands interrupting the otherwise strongly correlated 329 bacterial cluster (Shigella flexneri, Escherichia fergusonii, Escherichia coli, Shigella dysenteriae and 330 Escherichia marmotae, Figure 6). These “weakly connected” bacterial taxa correspond precisely to 331 those that, in a separate analysis, were found to reach the highest abundances in infected samples (Figure 332 5). Together, these results suggest that infection is linked to the proliferation of a few dominant bacterial 333 species whose dynamics are largely independent of the structure of the wider microbiota or diet 334 composition. 335

336 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 27, 2026. ; https://doi.org/10.64898/2026.03.26.714368doi: bioRxiv preprint Overall, our study shows that Sarbecovirus infection in R. shameli is associated with a measurable and 337 consistent but moderate restructuring of the gut microbiome. Most importantly, (i) this signal was 338 characterized by increased interindividual dispersion and compositional shifts rather than by a marked 339 erosion of alpha diversity. In parallel, (ii) diet varied seasonally but showed limited global association 340 with infection status and weak coupling with microbiome structure, suggesting that trophic ecology is 341 unlikely to be the primary explanation for the infection –associated microbial signal. Finally, (iii) 342 infected bats were characterized by enrichment of bacterial taxa previously linked to inflammation or 343 epithelial stress states. 344 1. Sarbecovirus infection drives gut dysbiosis in R. shameli while alpha diversity remains stable. 345 The predominance of Pseudomonadota in both infected and non–infected individuals is consistent with 346 previous reports from insectivorous bats [27–28]. Contrary to our expectation that infection would be 347 associated with reduced bacterial diversity, alpha diversity metrics remained largely stable across 348 Sarbecovirus infection status, sex, and site, with only modest reductions in immature bats. This pattern 349 does not support a simple loss –of–diversity model of dysbiosis and instead suggests that infection –350 associated microbial responses in this system occur without major erosion of within–host diversity, as 351 documented in other bats [11, 20]. 352 Despite this alpha–diversity stability, Sarbecovirus infection was associated with modest but detectable 353 shifts in gut community composition and, importantly, greater interindividual variability. The absence 354 of alpha –diversity loss alongside increased beta –dispersion suggests infection –associated 355 destabilization of community structure rather than diversity erosion. In other words, infected bats did 356 not converge toward a single altered microbiome state; instead, their microbiomes became more 357 heterogeneous, consistent with Anna Karenina Principle expectations for stressed host –associated 358 communities. Infection was associated with compositional differences under Bray –Curtis and 359 unweighted UniFrac , and with increased dispersion under Bray –Curtis and weighted UniFrac, 360 indicating a recurrent infection –associated signal across complementary community metrics. These 361 differences were driven by changes in the relative abundance of dominant taxa, suggesting community-362 wide reorganization rather than compositional collapse. Similar increases in dispersion were observed 363 following experimentally induced inflammation in R. aegyptiacus [23], suggesting that immune or 364 epithelial stress responses may disrupt microbial assembly processes. This interpretation is strengthened 365 by the fact that the signal recurred across complementary beta –diversity approaches, even though 366 infection explained only a small fraction of total variance. In a naturally variable wildlife system, such 367 effect sizes are not unexpected. Field microbiomes integrate multiple sources of variation, including 368 host age, reproductive timing, diet, sociality, microhabitat, and stochastic exposure histories. In that 369 context, a modest but repeated infection signal, coupled with increased dispersion and the enrichment 370 of inflammation–associated taxa, is ecologically meaningful and consistent with a real restructuring of 371 microbial assembly processes rather than statistical noise. 372 Although causality cannot be established from this cross–sectional design, Sarbecovirus detection relied 373 on rectal-swab qPCR and confirmatory PCR/sequencing, which most plausibly reflects active or very 374 recent infection or shedding rather than long –past exposure. This temporal proximity between viral 375 detection and microbiome sampling increases the plausibility that the observed microbial signal reflects 376 infection–associated physiological perturbation, even if reverse or bidirectional relationships remain 377 possible. 378 379 2. Diet composition does not predict microbiome structure 380 The diet of R. shameli varied seasonally and was dominated by Coleoptera, Hemiptera, Lepidoptera, 381 and Diptera, in line with previous descriptions of insectivorous Rhinolophus diets in Cambodia [37]. 382 This trophic profile reflects the bat’s foraging strategy in dry deciduous and mixed evergreen forests of 383 northern Cambodia [33–34], where beetles and true bugs are abundant in the understory during both 384 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 27, 2026. ; https://doi.org/10.64898/2026.03.26.714368doi: bioRxiv preprint wet and dry seasons. The diversity of detected prey included representatives of families Cicadellidae, 385 Rhyparochromidae, and Scarabaeidae, suggesting that R. shameli exploits a wide range of arthropod 386 microhabitats, from foliage to soil detritus. However, Sarbecovirus infection did not produce a strong 387 global shift in overall diet composition, as shown by the non –significant PERMANOVA and the 388 absence of dispersion differences between infection groups. Although infected bats showed lower 389 values along the second dietary NMDS axis, this effect was modest and occurred against a broader 390

of seasonal dietary turnover. Infected individuals may still exhibit distinct foraging 391 preferences or altered prey consumption linked to seasonal availability or physiological state [50]. 392 More importantly, diet and microbiome structure were only weakly coupled. Mantel and Procrustes 393 analyses did not support meaningful covariation between dietary and bacterial community matrices, 394 and the HMSC model revealed weak residual structure among arthropod prey compared with the much 395 stronger structure observed among bacterial taxa. Taken together, these results suggest that trophic 396 ecology is unlikely to be the primary driver of the infection–associated microbiome signal. Rather, diet 397 appears to vary seasonally, while the microbiome signal associated with Sarbecovirus detection more 398 likely reflects infection–linked or host–physiological processes [22]. 399 3. Inflammation–associated bacterial taxa differentiate infection status 400 Differential abundance testing using ANCOM -BC2 identified 19 bacterial species differing 401 significantly between infection groups. Similar patterns were found in the HMSC model that included 402 the diet data, strengthening the link between bacterial association with Sarbeco–status, regardless of 403 diet items. The consistent enrichment of Shigella dysenteriae , S. flexneri , Escherichia coli , E. 404 fergusonii, and E. marmotae in infected bats aligns with inflammation –linked taxa reported in both 405 human and experimental bat studies [23–24]. In contrast, Enterobacter species, Enterococcus 406 casseliflavus, and Serratia ureilytica were depleted, indicating loss of potential mutualists involved in 407 mucosal homeostasis. This dual pattern of opportunistic enrichment and symbiont depletion points 408 toward an inflammatory shift in the gut ecosystem of infected individuals, and mirrors microbiome 409 alterations reported in human SARS -CoV-2 infections [18, 23], where Escherichia–Shigella 410 enrichment was linked to intestinal inflammation and epithelial barrier disruption. The parallel 411 enrichment of these inflammation–associated genera in R. shameli suggests that Sarbecovirus infection 412 could provoke subclinical gastrointestinal dysbiosis in bats, consistent with the known enteric tropism 413 of Sarbecoviruses [3]. However, such taxa should be interpreted as indicators of host physiological state 414 rather than direct drivers of pathology. Their enrichment may reflect immune –mediated changes 415 following infection rather than causal involvement in viral susceptibility. The absence of strong 416 correlations between infection and body condition or alpha diversity , as well as the absence of visual 417 symptoms during capture, further supports the idea that these microbial shifts occur under subclinical 418 infection and reflect tolerance mechanisms typical of bat hosts [11]. 419 Several limitations of this study should be acknowledged. First, the cross -sectional design precludes 420 inference on causal directionality between infection and microbiome dysbiosis. Second, viral RNA 421 detection by qPCR most likely reflects active or very recent infection/shedding, but it cannot distinguish 422 ongoing replication from very recent post –infection shedding . Third, although age effects were 423 considered, limited sample sizes constrained our ability to fully resolve age -by-infection interactions 424 previously reported in other bat systems. Finally, the lack of functional immune or metabolomic data 425 limits mechanistic interpretation of inflammation–associated microbial shifts. 426 427

428 While causality cannot be resolved, the combination of qPCR -based viral detection, gastrointestinal 429 viral tropism, stable alpha diversity, AKP –like increases in beta -dispersion, enrichment of 430 inflammation–associated taxa, and weak diet –microbiome coupling supports the interpretation that 431 Sarbecovirus detection is associated with destabilization of gut microbial community assembly in R. 432 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 27, 2026. ; https://doi.org/10.64898/2026.03.26.714368doi: bioRxiv preprint shameli rather than simple diversity loss. An alternative, non –exclusive interpretation is that pre –433 existing gut microbiome configurations influence host susceptibility or tolerance to Sarbecovirus 434 infection by modulating baseline immune responses, thereby affecting viral replication intensity and 435 detection probability. These results suggest that microbiome instability is more likely linked to host 436 physiological responses to infection than to trophic ecology, without implying strict causality. These 437 findings suggest that infection –associated microbiome instability is more closely linked to host 438 physiological responses than to trophic ecology, while still leaving open the possibility that pre-existing 439 microbiome configurations influence susceptibility or tolerance. Integrating such microbiome profiling 440 into wildlife viral surveillance may therefore improve understanding of host –virus coexistence and 441 tolerance mechanisms [51]. 442 443 Acknowledgments 444 We would like to acknowledge authorities from the Ministry of Agriculture, Forestry, and Fisheries for 445 their support in facilitating this work, and particularly the Department of Wildlife and Biodiversity 446 under the Forestry Administration for their support during field data collection, particularly San 447 Sovannary, Eam Sona and Chhun Vanna. We would like to warmly thank for their cooperation the local 448 authorities and the communities of the Thalaborivat district in Stung Treng province, Cambodia. We 449 also acknowledge the help of Cedric Marsboom (AVIA–GIS), Morgane Labadie (CIRAD), as well as 450 Anaïs Bompard from INRAE, Neil Furey, Tey Putita Ou, Vibol Hul and the entire virology unit of IPC. 451 452 Ethics approval 453 All procedures complied with relevant national and institutional guidelines for the care and use of 454 animals. Handling and sampling were performed by trained personnel in accordance with the Guidelines 455 [52] of the American Society of Mammologists for the use of wild mammals in research and education, 456 and with the statutory authorization of the Forestry Administration (Ministry of Agriculture, Forestry 457 and Fisheries, Cambodia). The Forestry Administration oversaw all field activities, as no animal ethics 458 committee exists in Cambodia. This study is reported in accordance with ARRIVE guidelines [53]. 459 Data availability 460 All data supporting the findings of this study are publicly available upon publication. This includes: (i) 461 sample–level metadata; (ii) a frequency -of-occurrence matrix of arthropod orders per sample derived 462 from DNA metabarcoding; (iii) the bacterial species abundance matrix generated from full–length 16S 463 rRNA sequencing; and (iv) the corresponding taxonomy file containing representative sequences and 464 tree. These datasets are deposited in an open -access repository 465 (https://figshare.com/s/6f68f9e937d1988e93d5), and accession links will be provided in the final 466 version of the manuscript. Additional code used for data processing and analysis will be made available 467 upon request. 468 469

470 1. Wong ACP, Li X, Lau SKP et al. Global epidemiology of bat coronaviruses. Viruses 2019;11:174. doi: 471 https://doi.org/10.3390/v11020174 472 2. Ruiz–Aravena M, McKee C, Gamble A et al. Ecology, evolution and spillover of oronaviruses from bats. Nat Rev 473 Microbiol 2022;20:299–314. doi: https://doi.org/10.1038/s41579–021–00652–2 474 3. Lau SKP, Wong ACP, Luk HKH et al. Differential tropism of SARS–CoV and SARS–CoV–2 in bat cells. Emerg 475 Infect Dis 2020;26:2961–5. doi: https://doi.org/10.3201/eid2612.202308 476 4. Cohen LE, Fagre AC, Chen B et al. Coronavirus sampling and surveillance in bats from 1996–2019: a systematic 477 review and meta–analysis. Nat Microbiol 2023;8:1176–86. doi: https://doi.org/10.1038/s41564–023–01375–1 478 5. Lan J, Ge J, Yu J et al. Structure of the SARS–CoV–2 spike receptor–binding domain bound to the ACE2 receptor. 479 Nature 2020;581:215–20. doi: https://doi.org/10.1038/s41586–020–2180–5 480 6. Syed A, Khan A, Gosai F et al. Gastrointestinal pathophysiology of SARS–CoV–2 – a literature review. J 481 Community Hosp Intern Med Perspect 2020;10:523–8. doi: https://doi.org/10.1080/20009666.2020.1811556 482 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 27, 2026. ; https://doi.org/10.64898/2026.03.26.714368doi: bioRxiv preprint 7. Schmulson M, Dávalos MF, Berumen J. Beware: gastrointestinal symptoms can be a manifestation of COVID–19. 483 Rev Gastroenterol Mex (Engl Ed) 2020;85:282–7. doi: https://doi.org/10.1016/j.rgmx.2020.04.001 484 8. Banerjee A, Kulcsar K, Misra V et al. Bats and coronaviruses. Viruses 2019;11:41. doi: 485 https://doi.org/10.3390/v11010041 486 9. Sánchez CA, Phelps KL, Frank HK et al. Advances in understanding bat infection dynamics across biological 487 scales. Proc Biol Sci 2024;291:20232823. doi: https://doi.org/10.1098/rspb.2023.2823 488 10. Caballero–Flores G, Pickard JM, Núñez G. Microbiota–mediated colonization resistance: mechanisms and 489 regulation. Nat Rev Microbiol 2023;21:347–60. doi: https://doi.org/10.1038/s41579–022–00833–7 490 11. Liu B, Chen X, Zhou L et al. The gut microbiota of bats confers tolerance to influenza virus (H1N1) infection in 491 mice. Transbound Emerg Dis 2022;69:e1469–87. doi: https://doi.org/10.1111/tbed.14478 492 12. Cappelle J, Furey N, Hoem T et al. Longitudinal monitoring in Cambodia suggests higher circulation of alpha and 493 betacoronaviruses in juvenile and immature bats of three species. Sci Rep 2021;11:24145. doi: 494 https://doi.org/10.1038/s41598–021–03169–z 495 13. Robinson DP, Klein SL. Pregnancy and pregnancy–associated hormones alter immune responses and disease 496 pathogenesis. Horm Behav 2012;62:263–71. doi: https://doi.org/10.1016/j.yhbeh.2012.02.023 497 14. Baker KS, Suu–Ire R, Barr J et al. Viral antibody dynamics in a chiropteran host. J Anim Ecol 2014;83:415–28. 498 doi: https://doi.org/10.1111/1365–2656.12153 499 15. Brook CE, Ranaivoson HC, Broder CC et al. Disentangling serology to elucidate henipa– and filovirus 500 transmission in Madagascar fruit bats. J Anim Ecol 2019;88:1001–16. doi: https://doi.org/10.1111/1365–501 2656.12985 502 16. Slack E, Hapfelmeier S, Stecher B et al. Innate and adaptive immunity cooperate flexibly to maintain host–503 microbiota mutualism. Science 2009;325:617–20. doi: https://doi.org/10.1126/science.1172747 504 17. Zaneveld JR, McMinds R, Vega Thurber R. Stress and stability: applying the Anna Karenina principle to animal 505 microbiomes. Nat Microbiol 2017;2:17121. doi: https://doi.org/10.1038/nmicrobiol.2017.121 506 18. Wu Y, Cheng X, Jiang G et al. Altered oral and gut microbiota and its association with SARS–CoV–2 viral load in 507 COVID–19 patients during hospitalization. npj Biofilms Microbiomes 2021;7:61. doi: 508 https://doi.org/10.1038/s41522–021–00232–5 509 19. Li J, Jing Q, Li J et al. Assessment of microbiota in the gut and upper respiratory tract associated with SARS–510 CoV–2 infection. Microbiome 2023;11:38. doi: https://doi.org/10.1186/s40168–022–01447–0 511 20. Wasimuddin, Brändel SD, Tschapka M et al. Astrovirus infections induce age–dependent dysbiosis in gut 512 microbiomes of bats. ISME J 2018;12:2883–93. doi: https://doi.org/10.1038/s41396–018–0239–1 513 21. Melville DW, Meyer M, Risely A et al. Hibecovirus (genus Betacoronavirus) infection linked to gut microbial 514 dysbiosis in bats. ISME Commun 2025;5: ycae154. doi: https://doi.org/10.1093/ismeco/ycae154 515 22. Nixon LO, Gillam EH. Glitter Guano: an effective marker–based method of determining gut transit time in bats. 516 Acta Chiropterol 2025;27:133–8. doi: https://doi.org/10.3161/15081109ACC2025.27.1.012 517 23. Berman TS, Weinberg M, Moreno KR et al. In sickness and in health: the dynamics of the fruit bat gut microbiota 518 under a bacterial antigen challenge and its association with the immune response. Front Immunol 519 2023;14:1152107. doi: https://doi.org/10.3389/fimmu.2023.1152107 520 24. Li J, Li L, Jiang H et al. Fecal bacteriome and mycobiome in bats with diverse diets in South China. Curr 521 Microbiol 2018;75:1352–61. doi: https://doi.org/10.1007/s00284–018–1530–0 522 25. Porras AM, Shi Q, Zhou H et al. Geographic differences in gut microbiota composition impact susceptibility to 523 enteric infection. Cell Rep 2021;36:109457. doi: https://doi.org/10.1016/j.celrep.2021.109457 524 26. Latinne A, Nga NTT, Long NV et al. One Health surveillance highlights circulation of viruses with zoonotic 525 potential in bats, pigs, and humans in Viet Nam. Viruses 2023;15:790. doi: https://doi.org/10.3390/v15030790 526 27. Ingala MR, Becker DJ, Bak Holm J et al. Habitat fragmentation is associated with dietary shifts and microbiota 527 variability in common vampire bats. Ecol Evol 2019;9:6508–23. doi: https://doi.org/10.1002/ece3.5228 528 28. Lobato–Bailón L, García–Ulloa M, Santos A et al. The fecal bacterial microbiome of the Kuhl’s pipistrelle bat 529 (Pipistrellus kuhlii) reflects landscape anthropogenic pressure. Anim Microbiome 2023;5:7. doi: 530 https://doi.org/10.1186/s42523–023–00229–9 531 29. Li J, Chu Y, Yao W et al. Differences in diet and gut microbiota between lactating and non–lactating Asian 532 particolored bats (Vespertilio sinensis): implication for a connection between diet and gut microbiota. Front 533 Microbiol 2021;12:735122. doi: https://doi.org/10.3389/fmicb.2021.735122 534 30. Clare EL, Oelbaum PJ. The diets of bats: think outside the guild. In: Russo D, Fenton B (eds). A Natural History 535 of Bat Foraging. London: Academic Press, 2024, 233–60. doi: https://doi.org/10.1016/B978–0–323–91820–536 6.00013–9 537 31. Falvo CA, Crowley DE, Benson E et al. Diet–induced changes in metabolism influence immune response and viral 538 shedding in Jamaican fruit bats. Proc R Soc B 2025;292:20242482. doi: https://doi.org/10.1098/rspb.2024.2482 539 32. Vanalli C, Falvo C, Crowley D et al. Diet composition affects bat viral shedding with potential consequences for 540 pathogen spillover. Proc R Soc B 2025;292:20250547. doi: https://doi.org/10.1098/rspb.2025.0547 541 33. Ith S, Soisook P, Bumrungsri S et al. A taxonomic review of Rhinolophus coelophyllus Peters 1867 and R. shameli 542 Tate 1943 (Chiroptera: Rhinolophidae) in continental Southeast Asia. Acta Chiropterol 2011;13:41–59. 543 34. Hassanin A, Tu VT, Curaudeau M et al. Inferring the ecological niche of bat viruses closely related to SARS–544 CoV–2 using phylogeographic analyses of Rhinolophus species. Sci Rep 2021;11:14276. doi: 545 https://doi.org/10.1038/s41598–021–93738–z 546 35. Ou TP, Guillebaud J, Baidaliuk A et al. Characterization and evolutionary history of novel SARS–CoV–2–related 547 viruses in bats from Cambodia. bioRxiv 2025 Apr 15. doi: https://doi.org/10.1101/2025.04.15.648942 548 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 27, 2026. ; https://doi.org/10.64898/2026.03.26.714368doi: bioRxiv preprint 36. Guillebaud J, Ou TP, Hul V et al. Study of coronavirus diversity in wildlife in northern Cambodia suggests 549 continuous circulation of SARS–CoV–2–related viruses in bats. Sci Rep 2025;15:12628. doi: 550 https://doi.org/10.1038/s41598–025–92475–x 551 37. Sin S, Chhorn S, Doeurk B et al. Diet preferences of insectivorous bats (Mammalia: Chiroptera) in Chambok, 552 Kampong Speu Province, Cambodia. Cambodian J Nat Hist 2020;2020(2):69–77. 553 38. Delaune D, Hul V, Karlsson EA et al. A novel SARS–CoV–2 related Coronavirus in bats from Cambodia. Nat 554 Commun 2021;12:6563. doi: https://doi.org/10.1038/s41467–021–26809–4 555 39. Francis CM. A Guide to the Mammals of Southeast Asia. Princeton, NJ: Princeton University Press, 2008, 392 p. 556 ISBN 9780691135512. 557 40. Chu DKW, Leung CYH, Gilbert M et al. Avian coronavirus in wild aquatic birds. J Virol 2011;85:12815–20. doi: 558 https://doi.org/10.1128/JVI.05838–11 559 41. Cuscó A, Catozzi C, Viñes J et al. Microbiota profiling with long amplicons using Nanopore sequencing: full–560 length 16S rRNA gene and the 16S–ITS–23S of the rrn operon. F1000Research 2019;7:1755. Doi: 561 https://doi.org/10.12688/f1000research.16817.2 562 42. Vescera C, Van Vyve C, Smits Q, Michaux JR. All–you–can–eat buffet: a spider–specialized bat species (Myotis 563 emarginatus) turns into a pest fly eater around cattle. PLoS One 2024;19:e0302028. doi: 564 https://doi.org/10.1371/journal.pone.0302028 565 43. Galan M, Pons JB, Tournayre O et al. Metabarcoding for the parallel identification of several hundred predators 566 and their prey: application to bat species diet analysis. Mol Ecol Resour 2018;18:474–89. doi: 567 https://doi.org/10.1111/1755–0998.12749 568 44. Gillet F, Tiouchichine ML, Galan M et al. A new method to identify the endangered Pyrenean desman (Galemys 569 pyrenaicus) and to study its diet, using next generation sequencing from faeces. Mamm Biol 2015;80:505–9. doi: 570 https://doi.org/10.1016/j.mambio.2015.08.002 571 45. Martin M. Cutadapt removes adapter sequences from high–throughput sequencing reads. EMBnet J 2011;17:10–2. 572 doi: https://doi.org/10.14806/ej.17.1.200 573 46. Li, H. (2021). New strategies to improve minimap2 alignment accuracy. Bioinformatics, 37:4572–4574. doi: 574 https://doi.org/10.1093/bioinformatics/btab705 575 47. Ovaskainen O, Tikhonov G, Norberg A et al. How to make more out of community data? A conceptual framework 576 and its implementation as models and software. Ecol Lett 2017;20:561–76. doi: https://doi.org/10.1111/ele.12757 577 48. Tikhonov G, Opedal ØH, Abrego N et al. Joint species distribution modelling with the R–package Hmsc. Methods 578 Ecol Evol 2020;11:442–7. doi: https://doi.org/10.1111/2041–210X.13345 579 49. Pol C, Belfield S, Martin R. Insects of Upland Crops in Cambodia. ACIAR Monograph No. 143. Canberra: 580 Australian Centre for International Agricultural Research, 2010, 132 p. 581 50. Tillman F E, Adelman J S. Searching while sick: How does disease affect foraging decisions and contact rates? 582 Functional Ecology 2023;37: 838–844. doi: https://doi.org/10.1111/1365–2435.14207 583 51. González R, Elena SF. The interplay between the host microbiome and pathogenic viral infections. mBio 584 2021;12:e02496–21. https://doi.org/10.1128/mBio.02496–21 585 52. Sikes RS, Animal Care and Use Committee of the American Society of Mammalogists. 2016 guidelines of the 586 American Society of Mammalogists for the use of wild mammals in research and education. J Mammal 587 2016;97:663–88. doi: https://doi.org/10.1093/jmammal/gyw078 588 53. Percie du Sert N, Hurst V, Ahluwalia A et al. The ARRIVE guidelines 2.0: updated guidelines for reporting animal 589 research. PLoS Biol 2020;18:e3000410. doi: https://doi.org/10.1371/journal.pbio.3000410 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 27, 2026. ; https://doi.org/10.64898/2026.03.26.714368doi: bioRxiv preprint Tables and Figures 614 615 Sampling session Viral infection status by Site 03.2023 05.2023 06.2023 12.2023 03.2024 05.2024 06.2024 08.2024 11.2024 Total Negative 22 22 22 21 11 18 21 22 15 174 Chab Pleurng 8 7 9 12 3 13 7 14 10 83 Chhngauk 12 8 4 4 6 2 6 7 2 51 Kar Ngaork 2 7 9 5 2 3 8 1 3 40 Sarbecovirus positive 20 16 7 43 Chab Pleurng 10 4 2 16 Chhngauk 8 9 2 19 Kar Ngaork 2 3 3 8 Total 22 42 38 21 11 25 21 22 15 217 Table 1. Sampling list of R. shameli individuals according to Sarbecovirus qPCR testing, sampling site and 616 session. 617 618 619 620 Figure 1. Upper left: R. shameli individual (Photo credit: Julien Cappelle), Bottom left: View of Phnom Kar 621 Ngoark (Photo credit: Julia Guillebaud). Right: Map of Cambodia (Source: Wikimedia commons) with sampling 622 sites. Pie charts represent viral status proportion variation between sites (red = Sarbeco–positive, black = 623 negative). 624 625 626 627 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 27, 2026. ; https://doi.org/10.64898/2026.03.26.714368doi: bioRxiv preprint 628 Figure 2.(A). Proportion of occurrence by arthropods order per Season and year in R. shameli separated by Sarbecovirus 629 infection status detected by PCR. No Sarbeco–pos were detected in the wet season of 2024. (B). NMDS ordination of R. 630 shameli dietary composition based on Jaccard distances (stress = 0.095). Each point represents one fecal sample, colored by 631 Sarbecovirus infection status. Arrows indicate environmental variables significantly correlated with ordination axes as 632 determined by envfit, with arrow length proportional to correlation strength (R² values). (C). Spearman rank correlations 633 between NMDS axes and relative frequencies of insect orders in Rhinolophus shameli diets. The x–axis shows correlation 634 coefficients (ρ), and the y–axis lists insect orders. Point color indicates correlation direction (blue = positive; red = negative), 635 and point size is proportional to the R² of the relationship. Black squares denote significant correlations (q < 0.05, 636 Benjamini–Hochberg correction). (D). Seasonal dynamics of Rhinolophus shameli dietary composition modeled by 637 Generalized Additive Models (GAMs). Smooth terms show fitted NMDS2 based on Sarbecovirus infection status (red = 638 positive, grey = negative) scores across the day of year (doy), with shaded areas representing 95% confidence intervals. 639 640 641 Figure 3. Relative abundance of most common bacterial phyla per R. shameli individual across seasonality, with 642 identification of Sarbecovirus infection status based on PCR detection. 643 Dry 2023 Wet 2023 Dry 2024 Wet 2024 0.00 0.25 0.50 0.75 1.00 Relative Abundance Phylum Acidobacteriota Actinomycetota Bacillota Bacteroidota Cyanobacteriota Fusobacteriota Mycoplasmatota Nitrospirota Phylum < 1% abund. Planctomycetota Pseudomonadota Individuals Non-infected Sarbeco-infected .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 27, 2026. ; https://doi.org/10.64898/2026.03.26.714368doi: bioRxiv preprint 644 645 Figure 4. (A) Principal Coordinates Analysis (PCoA) based on Bray –Curtis dissimilarities showing overall community 646 separation between infected (red) and non-infected (grey) individuals. (B). Multivariate dispersion (betadisper) showing Bray–647 Curtis distances of samples to group centroids (permutest, p = 0.05). 648 649 650 651 Enterobacter wuhouensis Enterococcus casseliflavus Enterobacter chuandaensis Enterobacter sichuanensis Enterobacter chengduensis Serratia ureilytica Cedecea lapagei Enterobacter bugandensis Enterobacter asburiae Lelliottia aquatilis Enterobacter kobei Yokenella regensburgei Kluyvera cryocrescens Leclercia adecarboxylata Escherichia fergusonii Escherichia marmotae Escherichia coli Shigella dysenteriae Shigella flexneri −1 0 1 Log Fold Change (LFC) Taxa (full dataset) Sarbeco-status Negative Positive Differentially Abundant Taxa (qPCR Sarbecovirus status)A 0 200 400 600 Cedecea lapagei Enterobacter as buriae Enterobacter bugandensis Enterobacter chengduensisEnterobacter chuandaensis Enterobacter kobei Enterobacter sichuanensisEnterobacter wuhouensisEnterococcus casselifl avus Esche richia coli Esche richia fergusonii Esche richia ma rmotae Kluy vera c ryocrescens Leclercia adecarb oxylata Lelliottia aquatilisSer ratia ureilytica Shigella dysente riae Shigella fl exne ri Yokenella regens burgei Mean relative abundance ± SE Sarbeco-status Negative Positive Differential abundance of bacterial species by Sarbecovirus statusB .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 27, 2026. ; https://doi.org/10.64898/2026.03.26.714368doi: bioRxiv preprint Figure 5. (A). Bias-corrected log fold changes (LFC) from ANCOM–BC2 at the species level comparing Sarbecovirus–652 positive vs. negative bats. Bars show LFC per taxon ordered by LFC magnitude. Multiple testing controlled with FDR (α = 653 0.05); only significant taxa (q < 0.05) after prevalence (≥10%) filtering are shown. (B). Mean ± standard error of relative 654 abundance of bacterial species differing between Sarbecovirus–positive (firebrick) and negative (grey) R. shameli 655 individuals. 656 657 658 Figure 6. Residual species associations from the HMSC model. The heatmap shows the residual correlation matrix 659 estimated by the HMSC model. Warmer colours (red) indicate positive residual associations, cooler colours (blue) indicate 660 negative associations, and white corresponds to near–zero correlations. 661 662 663 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 27, 2026. ; https://doi.org/10.64898/2026.03.26.714368doi: bioRxiv preprint