Impact of a Model Used to Simulate Chronic Socio-Environmental Stressors Encountered during Spaceflight on Murine Intestinal Microbiota

During deep-space travels, crewmembers face various physical and psychosocial stressors that could alter gut microbiota composition. Since it is well known that intestinal dysbiosis is involved in the onset or exacerbation of several disorders, the aim of this study was to evaluate changes in intestinal microbiota in a murine model used to mimic chronic psychosocial stressors encountered during a long-term space mission. We demonstrate that 3 weeks of exposure to this model (called CUMS for Chronic Unpredictable Mild Stress) induce significant change in intracaecal β-diversity characterized by an important increase of the Firmicutes/Bacteroidetes ratio. These alterations are associated with a decrease of Porphyromonadaceae, particularly of the genus Barnesiella, a major member of gut microbiota in mice and humans where it is described as having protective properties. These results raise the question of the impact of stress-induced decrease of beneficial taxa, support recent data deduced from in-flight experimentations and other ground-based models, and emphasize the critical need for further studies exploring the impact of spaceflight on intestinal microbiota in order to propose strategies to countermeasure spaceflight-associated dysbiosis and its consequences on health.


Introduction
Gut microbiota (GM) form a complex microbial ecosystem whose balance and homeostasis are essential to the well-being of the host. Its composition is affected by numerous intrinsic and extrinsic factors such as antibiotics or diet [1,2]. Recent works have shown that host stress, particularly Table 1. Comparison of socio-environmental stressors encountered during space missions with those delivered using the CUMS model, and limitations of this model (adapted from [26]). Figure 1a)

Socio-Environmental Stressors Encountered during Spaceflights Limitations of CUMS Model
Mice confined in a small cage during 1 or 2 h. Confinement throughout the mission.
Mice confined during short periods (ethical point of view) /astronauts confined for several months in the ISS.

Days of CUMS Exposure Do Not Induce a Major Stress Response
Male mice were divided in two groups: ten mice submitted to 21 days of CUMS and ten controls placed in another room of the animal facility. Animals presenting injuries, such as bites that could induce inflammation, were discarded resulting in ten CUMS mice and seven controls at the end of the experiment. To evaluate stress, mice were weighted at the end of the experimental procedure and the amount of corticosterone in peripheral blood was quantified by ELISA. Figure 1b,c show that these two parameters were similar in both groups of mice. We also determined thymus weight since it is well known that stress induces its involution. This organ weight was normalized to body weight ( Figure 1d). Again, no statistically significant difference could be noted between the two groups of mice.

Intestinal Microbiome β-Diversity Is Significantly Modified by CUMS
To evaluate the effects of CUMS exposure on GM, we quantified by qPCR the number of 16S rRNA encoding gene copies per mg of intracaecal content. Figure 2a reveals that bacterial load was not significantly affected by CUMS exposure (CUMS: 1.12 × 10 8 ± 1.34 × 10 7 vs. controls: 1.39 × 10 8 ± 1.67 × 10 7 , p = 0.19) (Figure 2a). We also preformed pyrosequencing experiments. They generated an average of 9024 reads per sample (ranging from 4276 to 24,502) with a mean length of 527 bp (ranging from 517 to 533 bp). Individual rarefaction curves (Supplementary Materials Figure S1) showed that the mean numbers of observed operational taxonomic units (OTUs), 140 taxa (ranging from 61 to 210 OTUs), reached in all samples a plateau of approximately 5000 sequence reads. The read coverage was therefore sufficient to capture most of the bacterial diversity of each intracaecal microbiome. The within-sample diversity (α-diversity) indicated no significant difference between CUMS and control mice (Figure 2b). This suggests that CUMS mice had no change in microbial richness and evenness. However, in terms of β-diversity, Principal Component Analysis (PCA) showed distinct clustering between samples from control and CUMS mice indicating a significant change in microbiome composition (Figure 2c, PERMANOVA p = 0.029).

Impact on Caecal Microbiome Composition
A more in-depth taxonomic analysis of bacterial types revealed several changes in microbiome composition, and variations appeared at different phylogenetic levels. Nine divisions were identified by pyrosequencing. In all samples, the majority of caecal bacteria (ranging from 92 to 98% of total 16S) belonged either to the Firmicutes (ranging from 49.3 to 94.4%) or to the Bacteroidetes phylum (ranging from 2.5 to 46.8%), with a small proportion (2-8% of the identified sequences) of bacteria from seven others phyla: Actinobacteria, Candidatus Melainobacteria, Candidatus Saccharibacteria (TM7), Cyanobacteria, Proteobacteria, Tenericutes and Verrucomicrobia (Supplementary Materials Database S1). Moreover, 16 classes, 26 orders, 53 families, and 123 genera were identified.
CUMS led to an increase of the Firmicutes phyla (p = 0.0041) and a decrease of the Bacteroidetes taxa (p = 0.0062) compared to control mice ( Figure 3a). These alterations induced a significant rise of the Firmicutes/Bacteroidetes ratio from 2.28 ± 0.38 in controls to 11.75 ± 3.43 in CUMS mice ( Figure 3b, p = 0.00072). The gain of Firmicutes in CUMS mice was not clearly associated to the expansion of distinct genera, except for the Clostridiales members Anaerotruncus, Coprococcus and Sporobacter (Figure 3c), but seemed rather to be due to a general moderate rise of several taxa within the phylum. Concerning the diminution of Bacteroidetes, it is clearly linked to a significant decrease of Porphyromonadaceae (p = 0.022) and Flavobacteriaceae (p = 0.073) with the corresponding impacted genera being Barnesiella, Prevotella, Coprobacter, Porphyromonas, Pricia, Parabacteroides, Dysgonomonas and the vanishing of Nonlabens and Maribacter (Figure 3c, Supplementary Materials Database S1). We also noticed the lowering of another Bacteroidetes (Candidatus Armantifilum and Odoribacter) and of members of the genus Akkermansia.
At the species level, of the 389 taxa assigned, 275 species were found in control mice and 337 species in CUMS mice, corresponding to 223 species recovered in both groups ( Figure 3d). Among them, only 27 were shared by all animals (core microbiome).

Discussion
It is increasingly evident that chronic psychosocial stresses influence intestinal homeostasis. Such alterations in microbiome composition can lead to local or central dysregulations that could be involved in the onset or exacerbation of chronic disorders such as IBD or psychiatric disorders [2,13,27,28]. During spaceflight, astronauts are subjected to various chronic physical and psychosocial stressors which could lead to dysbiosis, in a context of limited medical procedures and facilities. It has already been shown, using ground-based murine models, that weight modulation induces disruption of intestinal microbiota [3,29]. In this study, we used the CUMS model to mimic chronic socioenvironmental stresses encountered during space travels and explore their impact on intestinal microbiota. Indeed, we previously showed that this model replicates some spaceflight-induced immunological changes observed in astronauts [26]. Furthermore, it is recognized as a reliable and effective rodent model of depression [9,13,15,28,[30][31][32].
Our results revealed that after 3 weeks of CUMS exposure, a duration chosen to simulate a six-month flight at the human scale [33], there was no significant change in murine caecal bacterial load. Additionally, no statistically significant modification of the α-diversity was observed in CUMS mice by comparison to controls, indicating that the within-community diversity was not altered by this model of chronic stress. Although these results are in agreement with other studies using variants of the rodent CUMS model [9,28], they are discrepant when compared to other works describing a decrease of α-diversity [11,15,31,32]. Such differences could be explained by variation in the CUMS models (species, strains, age, gender, feeding conditions, type of stressors, duration of exposure to individual stress), the origin of the samples (fecal or intraluminal), or protocols (DNA extraction method, PCR parameters) [3].
However, significant change in intracaecal global β-diversity was observed after CUMS treatment. Indeed, an important increase of the Firmicutes/Bacteroidetes ratio was observed in CUMS mice, which is consistent with other reports using variants of the rodent CUMS model [9,15,28,31]. Within the Bacteroidetes phylum, we observed a decrease of Porphyromonadaceae that has already been noted with other chronic stress such as restraint stress [34] and multifactorial model of early-life adversity [35]. Within this family, the greatest impact of CUMS was observed on the relative abundance of Barnesiella sp., a genus composed of Barnesiella intestinihominis and Barnesiella viscericola, belonging to the core microbiome of the murine and human gut. These species are described as having beneficial effects, such as protecting against colitis [36], enhancing the efficacy of antitumor treatments [37] and conferring resistance to intestinal colonization by pathogenic microorganisms [38]. These data raise the question of the impact of the decrease of this major member of GM in CUMS mice.
On the other hand, the increase of Firmicutes in CUMS mice cannot be statistically correlated with the increase of specific OTUs. This lack of correlation could be due to high interindividual variability in GM illustrated by the small number of species shared by all animals, stressed or not, suggesting the existence of only a reduced core microbiome. Such variability could also explain the lack of statistical significance at low taxa level and the fact that the impact of CUMS was manifest only at the phylum level. It is noteworthy that CUMS is associated with the appearance of several new taxa (114, Figure 3d), mainly belonging to Firmicutes, among them various OTUs of Lactobacillus with a great interindividual variability. Some protective taxa appeared (Lactobacillus johnsonii) while other decreased (Lactobacillus murinus), potentially offsetting each other. Interestingly, we observed opposite results when using a 3G-hypergravity model with a lowering of L. johnsonii and a rise of L. murinus [3]. Moreover, 3G-hypergravity was associated with increased bacterial load and α-diversity, as well as with a significant impact on the relative abundance of 50 intestinal species, whereas 2G-hypergravity seemed to modulate only moderately the GM composition. As described for the 2G-hypergravity model, the moderate alteration of GM observed with the CUMS model could be due to a lower activation of the HPA axis as no elevation of corticosterone level was noted in mice sera. This hypothesis is supported by higher serum corticosterone concentrations noticed in mice exposed to 3G during 21 days [39], as well as during the first two weeks of exposition to the chronic mild stress model (CMS) which is more intense than CUMS because of water and food deprivation periods [40]. So, as previously reported for the TCRβ repertoire [41,42], chronic socio-environmental stressors seem to have less impact on intestinal microbiota than gravity changes.
The results of the present study demonstrate that 3-weeks of exposure to chronic unpredictable psychosocial and environmental stressors alter mice GM, although to a lower extent than gravity changes. One limitation of this study is the small sample size that could lead to miss some modifications of GM because of intraindividual variability precluding their statistical detection. However, alteration of GM must receive attention and should be monitored in crewmembers, especially since it has been recently shown that a fecal transfer of GM from CUMS to healthy mice induces despair-like behaviors associated with alterations in serotonin pathway [32]. Furthermore, these data provide additional arguments to the countermeasure protocol proposed by experts against spaceflight-associated perturbations to the immune system [22]. Their recommendations include physical and psychological exercises for stress management, pre-or probiotics supplementation and dietary approaches, that could also permit to limit dysbiosis and its consequences on health. Finally, note that the results of this study go beyond astronaut health protection because the CUMS model can also be used to study the impact of everyday life stresses and it is well established that stress can contribute to the development or aggravation of several pathologies [2,43].

Exposure to Chronic Unpredictable Mild Psychosocial and Environmental Stressors (CUMS Model)
Isolated animals (one mouse per cage) were subjected during 21 days to different unpredictable mild psychosocial and environmental stressors, according to Pardon et al. (2000) [44]. The CUMS procedure presented in Figure 1a was scheduled over a 1-week period and repeated throughout the 3 weeks of experimentation. Stress periods were always separated by stress-free intervals of at least 2 h to avoid any habituation process. The control group was left undisturbed in another room of the animal facility, five mice per standard cage (37.5 cm × 21.5 cm × 18 cm). Animals presenting injuries (such as bites that could induce inflammation) were discarded resulting in 7 control mice and 10 CUMS mice.

Sample Collection
At the end of the experiment, CUMS and control mice were anesthetized using isoflurane, weighed and then put to death by cervical dislocation. All samples were immediately processed to avoid degradation and/or contamination. The intestine was dissected in by excising the entire caecum. Samples were opened longitudinally and their contents were removed by two successive washes in DEPC (1% )-treated PBS. Intra-luminal contents were immediately frozen in liquid nitrogen and stored at −80 • C until DNA isolation.

Corticosterone Quantification
Corticosterone was quantified in serum samples without any extraction procedure using the Corticosterone Enzyme Immunoassay kit (ArborAssays, Ann Arbor, MI, USA). Samples were analyzed in duplicate. Absorbance at 405 nm was measured and concentrations, calculated from a standard curve established using calibrators, were expressed as ng/mL.

Intracaecal Microbiota Sequencing
Barcoded primers Bact-515F (5 -GTGCCAGCMGCNGCGC-3 ) and Bact-1061R (5 -CRRCACGAGCTGACGAC-3 ) described by Klindworth et al. (2013) [46] were used for the initial amplification of the V4-V6 region of the 16S rRNA gene as previously described [3]. PCR reactions contained 2.5 U of Taq DNA Polymerase (Invitrogen, Cergy Pontoise, France), 5 µL of 5X buffer, 75 nmol MgCl 2 , 1 µL of 10 mM dNTPs, 1 µL of each primer (50 µM) and 50 ng of DNA. Three PCR reactions were run for each sample as follows: 95 • C for 5 min, followed by 40 cycles at 95 • C for 45 s, 60 • C for 45 s, 72 • C for 45 s and a final extension at 72 • C for 5 min. PCR reactions from the same sample were pooled, purified using the QIAquick PCR purification kit (Qiagen, Courtaboeuf, France) and quantified using a Qubit 2.0 Fluorometer (Life Technologies, Carlsbad, CA, USA) using the dsDNA HS Assay Kit (Life Technologies). To ensure equal representation of each sample in the sequencing run, each barcoded sample was standardized by calculating equimolar amounts (100 ng/sample) using the SequalPrep Normalization Plate Kit (Invitrogen) prior to pooling. Pooled samples of the 16S rRNA gene multiplexed amplicons were sequenced on a Roche 454 Genome Sequencer FLX Titanium instrument using the GS FLX Titanium XLR70 sequencing reagents and protocols (Beckman Coulter Genomics, Danvers, MA, USA).

Amplicon Sequencing Data Analysis
Analysis of amplicon sequencing data was carried out using the MEGAN pipeline [47]. After demultiplexing, combined raw sequencing data plus metadata were filtered to exclude low-quality reads.
Next, data were denoised and clustered using the MIRA 4 software (http://mira-assembler.sourceforge.net). Sequences with ≥98% similarity were binned and assigned to the same OTU to approximate species-level phylotypes. Representative sequences of each OTU, derived from clusters or singletons, were assigned at different taxonomic level by using the Ribosomal Database Project II Classifier [48]. To avoid a potential bias linked to variation of sequence coverage between samples, the data were normalized to 100,000 sequences per samples. Rarefaction curves were constructed to evaluate sequencing depth. Relative abundances of each OTU were compared according to the different experimental conditions. Bacterial richness and diversity across samples were estimated by calculating the following indexes as previously described [3]: Shannon index, Evenness index, OTU's number, Simpson's index of diversity, and Simpson's reciprocal index. PCA was conducted to appreciate overall distance between microbial communities, using relative abundance and taxa-to-taxa distance estimates. Obtained 16S rRNA gene sequences have been deposited into NCBI's Sequence Read Archive database (https://www.ncbi.nlm.nih.gov/sra) under accession number SRP153311.

Intracaecal Bacterial Load Quantification
The amount of total bacteria was assessed by amplifying 0.5 ng of DNA extracted from each fecal sample with pan-bacterial primers targeting the 16S rRNA gene as previously described [3]. Briefly, PCR assays were performed using the MESA FAST qPCR MasterMix for SYBRAssay as recommended by the manufacturer (Eurogentec, Seraing, Belgium). DNA extracted from the Barnesiella intestinihominis DSM 21032 T strain using the QIAamp DNA Mini Kit (Qiagen) was used to establish the standard curves. All assays were performed in triplicate. The following thermocycling conditions were applied with the MyiQ™2 real-time PCR system (Bio-Rad Laboratories, Hercules, CA, USA): initial denaturation at 95 • C for 5 min followed by 40 cycles of 95 • C for 15 s and 60 • C for 1 min. Melting curves were obtained immediately after the amplification under the following conditions: 70 cycles of 10 s with an increment of 0.5 • C/cycle starting at 60 • C.

Statistical Analysis
Comparison of body weights, corticosterone concentrations, normalized thymus weights, bacterial loads quantified by qPCR, relative abundances, and phylogenetic diversity indexes were performed using the Mann-Whitney U test with a significance level α of 0.05. p-values comprised between 0.05 and 0.10 indicate trend. The p-values were adjusted for multiple hypotheses testing using the False Discovery Rate method [49] for all the results within each taxonomy level. The PERMANOVA analysis (99 permutations) was conducted on dissimilarity indices produced by the Bray-Curtis method [50]. The β-diversity PCA was produced using Marti Anderson's procedure for the analysis of multivariate homogeneity of group dispersions [51]. All the analysis were performed using R version 3.5.0 (https://www.R-project.org/).  Figure S1. Rarefaction curves of bacterial 16S rRNA gene sequences obtained from (A) control and (B) CUMS mice. These curves were used to evaluate if further sequencing would likely detect additional taxa. Datasets S1: Statistical analysis of intracaecal microbiomes comparing relative abundance using the nonparametric Mann-Whitney U test at the phylum level (Dataset S1_a), at the class level (Dataset S1_b), at the order level (Dataset S1_c), at the family level (Dataset S1_d), at the genus level genera (Dataset S1_e) and at the species level (Dataset S1_f).

Acknowledgments:
We thank Amandine Simeon for technical assistance and Tevrasamy Marday for helping with the CUMS procedure.

Conflicts of Interest:
The authors declare no conflict of interest.