Three-way relationships between gut microbiota, helminth assemblages and bacterial infections in wild rodent populations

Background Despite its central role in host fitness, the gut microbiota may differ greatly between individuals. This variability is often mediated by environmental or host factors such as diet, genetics, and infections. Recently, a particular attention has been given to the interactions between gut bacteriota and helminths, as these latter could affect host susceptibility to other infections. Further studies are still required to better understand the three-way interactions between gut bacteriota, helminths and other parasites, especially because previous findings have been very variable, even for comparable host-parasite systems. Methods In our study, we used the V4 region of the 16S rRNA gene to assess the variability of gut bacteriota diversity and composition in wild populations of a small mammal, the bank vole Myodes glareolus. Four sites were sampled at a regional geographical scale (100 km) along a North-South transect in Eastern France. We applied analyses of community and microbial ecology to evaluate the interactions between the gut bacteriota, the gastro-intestinal helminths and the pathogenic bacteria detected in the spleen. Results We identified important variations of the gut bacteriota composition and diversity among bank voles. They were mainly explained by sampling localities and reflected the North/South sampling transect. In addition, we detected two main enterotypes, that might correspond to contrasted diets. We found geographic variations of the Firmicutes/Bacteroidetes ratio, that correlated positively with body mass index. We found positive correlations between the specific richness of the gut bacteriota and of the helminth community, as well as between the composition of these two communities, even when accounting for the influence of geographical distance. The helminths Aonchotheca murissylvatici, Heligmosomum mixtum and the bacteria Bartonella sp were the main taxa associated with the whole gut bacteriota composition. Besides, changes in relative abundance of particular gut bacteriota taxa were specifically associated with other helminths (Mastophorus muris, Catenotaenia henttoneni, Paranoplocephala omphalodes and Trichuris arvicolae) or pathogenic bacteria. Especially, infections with Neoehrlichia mikurensis, Orientia sp, Rickettsia sp and P. omphalodes were associated with lower relative abundance of the family Erysipelotrichaceae (Firmicutes), while coinfections with higher number of bacterial infections were associated with lower relative abundance of a Bacteroidales family (Bacteroidetes). Conclusions These results emphasize complex interlinkages between gut bacteriota and infections in wild animal populations. They remain difficult to generalize due to the strong impact of environment on these interactions, even at regional geographical scales. Abiotic features, as well as small mammal community composition and within host parasite coinfections, should now be considered to better understand the spatial variations observed in the relationships between gut bacteriota, gastro-intestinal helminths and bacterial infections.

individual. Age groups (juveniles and adults) were defined according to body mass and sexual 177 maturity. This latter was inferred using testes length and position, and seminal vesicle 178 development for males, or uterus size for females. Body condition was estimated using the 179 body mass index (BMI = weight/length 2 ). The digestive tract and the spleen were removed 180 and stored respectively in 96% ethanol and RNA later solution (-20°C). (stomach, small intestine, large intestine and caecum), and classified by morphotype then 258 stored in 95% ethanol for further accurate identification. The latter was based on 259 unambiguous morphological criteria using conventional microscopy and generalist 260 identification keys or specific literature when available (Anderson et al., 2009 ;Khalil et al., 261 1994  We estimated the beta diversity, i.e., the dissimilarity between host individuals in their gut 280 bacteriota using Bray-Curtis distances. We considered the relative abundance of OTUs 281 (family). 282 283 Influence of host and environmental factors on gut bacteriota diversity and composition 284 We tested the influence of individual characteristics (age class, gender, BMI) and localities, 285 independently on the F/B log-ratio and on the alpha diversity using generalized linear models 286 (GLM). We considered a negative binomial error distribution for the F/B ratio and the specific 287 richness, and a gaussian distribution for the Shannon index. Best model selection was 288 performed considering models with all possible combinations of factors and the DREDGE 289 function of the MUMIN package. The best model was selected using the Akaike information 290 criterion corrected for small sample size AICc, (J. B. Johnson & Omland, 2004). We assessed 291 the effect of each factor in the best model with the ΔAICc index. When the factor locality was 292 significant, Tukey's post-hoc tests were applied to evaluate pairwise differences between 293 localities, using the MULTCOMP package (Hothorn et al., 2008). Residuals were checked to 294 graphically to ensure that all assumptions regarding normality, independence and the 295 homogeneity of variance were satisfied. 296 We evaluated the influence of geographic distance on the dissimilarities in gut bacteriota by 297 performing Mantel tests and using Pearson correlation (10,000 permutations). These tests 298 have less statistical power to address questions related to the variation in community 299 composition data among sites. Therefore, we also analysed the factors shaping the 300 dissimilarities in gut microbiota composition using several functions of the VEGAN package 301 We estimated the alpha diversity of the gastro-intestinal helminhth and pathogenic bacteria 319 community using the richness index (presence/absence data). We used GLMS and model 320 selection process described above to analyse whether the alpha diversity of each intra-host 321 community (gut bacteriota and pathogenic communities) was influenced by the alpha 322 diversity of the two other ones. 323 We estimated the beta diversity of the gastro-intestinal helminth and pathogenic bacteria 324 community using the Jaccard index (presence/absence data). The relationships between intra-325 host community dissimilarities were investigated using three approaches. i) We applied 326 partial Mantel tests using MULTI.MANTEL (phytools package Revell, 2012) to analyse the 327 correlation between two matrices of dissimilarities (corresponding to two different 328 communities), while controlling for the effect of a third dissimilarity matrix (third 329 community). ii) We used db-RDA to analyse more deeply the relationships between the gut 330 bacteriota and the pathogenic (bacteria and helminths) communities. We included the alpha 331 diversity indices (richness specific) and infectious status as presence / absence) of pathogens 332 with prevalence greater than 10% in at least one locality as explanatory variables in these 333 analyses. We selected the best model using the ORDIR2STEP method. iii) We used DESEQ2 to 334 determine the gut bacteria taxa whose relative abundances changed with significant 335 explanatory variables. 336 337 localities. For technical reasons (e.g., poor sample preservation, missing data), we could study 342 the three intra-host communities for 124 rodents only. 343 344 Characterization of the gut bacteriota: taxa and enterotypes 345 Once the quality control steps were applied, the gut bacteriota dataset included 161 bank 346 voles. We detected 10 phyla and 61 families of bacteria. At the phylum level, we found six 347 predominant taxa that represented 99% of the gut bacteria relative abundance ( Figure S2   We found that the Firmicutes / Bateroidetes ratio varied significantly between localities 368 ( Figure 1C). Overall, northern localities exhibited lower F/B ratio than southern ones, with all 369 pairwise comparisons being significant except Chatillon versus Chaux-des-Crotenay. 370 Individual characteristics did not influence this ratio (Table S1). 371 The sampling locality had a significant global effect on the alpha diversity of the gut 372 bacteriota (GLMs. Specific richness: F = 8.49, P < 10 -3 ; Figure 2A; Shannon index: F = 4.74, 373 P = 3 x 10 -3 ; Figure 2B; Table S2A). The locality Cormaranche exhibited a higher specific 374 richness than all other localities (Tukey post hoc test. Mont-sous-Vaudrey: Z = 5.13, Padj < 10 -375 3 , Chaux-des-Crotenay: Z = 4.57, Padj < 10 -3 and Chatillon: Z = 3.62, Padj = 1.7 x 10 -3 ) but a 376 lower level of diversity than Mont-sous-Vaudrey when considering taxa relative abundance 377 (Tukey post hoc test. Shannon index: Z = -3.64, Padj = 1.5 x 10 -3 , Figure 2B). Body condition 378 (BMI) was also found to have a significant effect, but only when considering specific richness 379 (t = 2.91; P = 4 x 10 -3 ) -with higher values of BMI associated with increasing species 380 richness. All these results are detailed in Table S2A   variance, Figure 2C). However, this result has to be taken cautiously as significant differences 401 of data dispersion were detected between localities (betadisper. P = 1 x 10 -3 ). The locality 402 Cormaranche showed a lower dispersion compared to all other localities (Tukey multiple 403 comparisons, Table S3B). 404 We detected significant differences in the relative abundance of specific taxa using DeSeq2 405 (Table 2; Table S3C). The main changes (Log2 fold values higher than 20) were detected 406 between the northern (Mont-sous-Vaudrey) and southern localities. The northern population 407 was involved in 75% of all significant pairwise differences (Log2 fold change in composition 408 > 10). The gut bacteriota of these bank voles includes less Clostridiales (one unknown family; 409  Differences in the composition of the gut bacteriota between males and females bank voles 424 were driven by the phylum Firmicutes, with males exhibiting higher relative abundance of 425 this taxa than females (Table S3C). 426

427
Relationships between the diversity of the three intra-host communities 428 We found a significant relationship between the specific richness of the gut bacteriota and the 429 richness of the helminth community. A more diverse gut bacteriota was associated with a 430 greater number of helminth species infecting bank voles (GLM. F = 14.09, P < 10 -3 ; Figure  431 3A; Table S2B). We also found a positive relationship between the specific richness of the 432 pathogenic bacteria and the richness of the gastro-intestinal helminth community (GLM. F= 433 6.99, P = 9 x 10 -3 ; Figure 3A; Table S2B). Lastly, we found a significant effect of the specific 434 richness of both the gut bacteriota and of pathogenic bacteria on the richness of the gastro-435 intestinal helminth community (GLM. Gut bacteriota. t = 3.50, P < 10 -3 ; pathogenic bacteria t 436 = 2.38, P = 0.019; Figure 3A, Table S2B). 437 438 439   Table S3A.   (Table S3D). 475

476
The specific richness of the gastro-intestinal helminth community, as well as infections with 477 A. murissylvatici and H. mixtum, were only slightly associated with different relative 478 abundance of particular gut bacteria taxa (DeSeq2. Log2 fold changes did not exceed 3.5). 479 These changes concerned four main families. Rhizobiaceae and Spirochaetaceae showed 480 negative associations with gastro-helminth specific richness and A. murissylvatici. Mollicutes Considering pathogenic bacteria, we found that higher levels of specific richness were 499 associated with lower relative abundance of an undetermined Bacteroidales family 500 (Bacteroidetes), and that Neoehrlichia mikurensis, Orientia tsutsugamushi and Rickettsia sp 501 infections were associated with strong decreases in relative abundance of Erysipelotrichaceae 502 (Firmicutes) ( Table 3; Table S3E). Other associations between bacterial infections and 503 changes in relative abundance of specific gut bacteriota taxa were detected, but with little size 504 effect (DeSeq2. log2 fold changes lower than 5). 505 506 Discussion 507 Understanding the complex interlinkages between host microbiota, host-pathogen interactions 508 and health in wild animal populations has become a key topic in disease ecology. Such 509 understanding is instrumental for deciphering population dynamics, and designing strategies 510 for zoonotic risk management or biodiversity conservation. Here, we use a combination of 511 metabarcoding and community ecology approaches to (i) describe the gut microbiota of wild 512 rodent populations and their variations at a regional geographical scale, and (ii) explore the 513 three-way relationships between the gut bacteriota and communities of gastro-intestinal 514 helminths and pathogenic bacteria. 515 516

Spatial variations of gut bacteriota and their potential causes 517
The gut microbiota of bank voles has been mainly examined in the context of exposure to 518 radioactive pollutants (e.g., Lavrinienko (2018), but see Knowles et al. (2019)). In this study, 519 we focused on localities sampled at a regional scale (100 km) along a North-South gradient in host factors such as gender and age played little role. Interestingly, we found that all 523 individuals were clustered within two groups, which could be distinct enterotypes if 524 longitudinal surveys confirmed this clustering through time (Arumugam et al., 2011). There are now many evidence that the environment in which hosts evolve (through abiotic 545 and biotic factors) is likely to shape variations in gut bacteriota composition between 546 localities sampled and studies. Previous works have already shown that the structure of rodent 547 gut microbiota varied between localities at large spatial scales due to biogeographic or genetic 548 factors (Linnenbrink et al., 2013). Geographic variability has also been found at smaller 549 spatial scales (e.g., few km Goertz et al., 2019). Here, our results provide significant evidence 550 for spatial structure of gut bacteriota between bank vole populations that are between 50 and 551 130 km away, with no clear barrier to dispersal or gene flow (Dubois et al., 2018). 552 We observed gradual changes between bank voles from the northern and southern populations 553 in terms of gut bacteriota richness, evenness, composition and particularly 554 Firmicutes/Bacteroidetes ratio. Although the links between the diversity and functional 555 capacity of the gut bacteriota still need to be better understood (Worsley et al., 2021), it is 556 largely assumed that changes in diversity are associated with shifts in metabolism (Reese & 557 Dunn, 2018). Bank voles from southern populations exhibit higher specific richness and lower 558 evenness of the gut bacteriota, as well as lower dispersion of gut bacteriota composition. They 559 have higher levels of body condition and F/B ratio, which are indicative of an optimisation of 560 calorie intake and absorption, weight gain and fat storage (see refs in Wolf et al., 2021). 561 Altogether, these results could suggest strong constraints on gut bacteriota function to 562 maximise energy extraction. The northern populations show the opposite patterns. Lower 563 BMI and lower levels of F/B ratio might reflect energy production and conversion, amino acid 564 transport and metabolism. Diversity patterns (higher evenness and lower specific richness of 565 the gut bacteriota) could suggest lower stochasticity and/or directional selection. Further 566 studies are required to investigate the eco-evolutionary processes driving these changes in gut 567 bacteriota. 568 Lastly, these differences in gut microbiota composition between the northern and southern 569 populations might also reflect physiological variations related to physiology, health and 570

Three-way relationships between intra-host communities 584
At the community level, we have not found strong evidence for three-way relationships 585 between the gut microbiota (diversity or composition), the gastro-intestinal helminth and The strong associations between gastro-intestinal helminths and gut bacteriota may be 641 interpreted under two perspectives. First, the strong positive associations between the 642 diversity of helminth community and gut bacteriota might corroborate the hypothesis. 643 Besides, experimental evidence showed that helminths have the capacity to maintain higher 644 gut microbiota diversity and may represent gut homoeostasis (Kreisinger et al., 2015). Indeed, 645 low-intensity, chronic helminth infections are commonly linked to high microbial diversity 646 and predominance of bacteria typically associated with gut health (Peachey et al., 2017). 647 Nevertheless, this interpretation has to be taken cautiously as the diversity of both 648 communities was strongly influenced by the localities of sampling. The environment might 649 therefore shape similarly gut bacteriota and helminth community diversity. 650 Second, significant associations between helminth community and gut bacteriota composition 651 -which remain significant even when potential geographic confounding effects were 652 removed -may be linked to the fact that both communities reside in the same environmental could expect therefore potentially strong interactions and reciprocal influence between them, 656 Unfortunately, the causal processes behind these gut microbiota and helminths interactions 657 are complex, multifaceted and difficult to assess. This intricacy is amplified by the fact that 658 experimental studies mostly focus on single helminth infections while interactions 659 between/within community are the rule within host organisms. The field of microbiota 660 research would thus benefit from taking into account the whole composition of gastro 661 intestinal helminth community rather than single helminth infections only. 662 In this study, we also highlight a large number of species-specific associations between 663 helminths infections and members of the gut bacteriota. High-intensity, acute helminth 664 infections may correlate with changes in hosts gut microbiota, through direct and indirect 665 interactions (e.g., immune or other processes such as malnutrition; Peachey et al., 2017). 666 Nevertheless, the patterns of shifts in gut bacteriota associated with helminth infections 667 remain hardly predictable so far. As such, research works addressing this issue with 668 laboratory or wild animals have provided variable, and sometimes even contradictory 669 conclusions. Most surprising is that these inconsistent patterns are also found when focusing 670 on single host-helminth models. A potential explanation is that these infection-associated