Klebsiella oxytoca expands in cancer cachexia and acts as a gut pathobiont contributing to intestinal dysfunction

Cancer cachexia is a complex multi-organ syndrome characterized by body weight loss, weakness, muscle atrophy and fat depletion. With a prevalence of 1 million people in Europe and only limited therapeutic options, there is a high medical need for new approaches to treat cachexia. Our latest results highlighted microbial dysbiosis, characterized by a bloom in Enterobacteriaceae and altered gut barrier function in preclinical models of cancer cachexia. They also demonstrated the potential of targeting the gut microbial dysbiosis in this pathology. However, the exact mechanisms underlying the gut microbiota-host crosstalk in cancer cachexia remain elusive. In this set of studies, we identified Klebsiella oxytoca as one of the main Enterobacteriaceae species increased in cancer cachexia and we demonstrated that this bacteria acts as a gut pathobiont by altering gut barrier function in cachectic mice. Moreover, we propose a conceptual framework for the lower colonization resistance to K. oxytoca in cancer cachexia that involves altered host gut epithelial metabolism and host-derived nitrate boosting the growth of the gut pathobiont. This set of studies constitutes a strong progression in the field of gut microbiota in cancer cachexia, by dissecting the mechanism of emergence of one bacterium, K. oxytoca, and establishing its role as a gut pathobiont in this severe disease.


Klebsiella oxytoca is increased in tumor-bearing mice with cachexia independently of anorexia.
The C26 cachexia model consists of a subcutaneous injection of colon carcinoma cells 19 . Mice develop a relatively small tumor mass and display a decreased food intake and loss of body weight, due to both muscle atrophy and adipose tissue loss 18 . Feces from four cachectic mice were cultured on a coliform-selective medium and 15-16 isolates per mouse were identified based on DNA sequencing of their 16S rRNA encoding gene, with a majority of isolates classified as K. oxytoca (Table 1). This increase in K. oxytoca upon cancer cachexia was confirmed using qPCR in three independent mouse cohorts ( Fig. 1A-C).
Undernutrition and malnutrition have been shown to be associated with microbial dysbiosis 20 . To assess the role of a reduced caloric intake in the increased K. oxytoca levels found in cachectic mice, we included pair-fed animals in a similar in vivo experiment. Two additional groups of healthy mice were included. One was pair-fed to the CT group (PFtoCT) and the other to the C26 group (PFtoC26). The PFtoCT group serves as a control for the stress related to the pair-feeding procedure. The comparison of the PFtoCT and the PFtoC26 groups allows the strict evaluation of the caloric restriction consequences. As expected, C26 mice displayed increased cecal K. oxytoca levels compared to CT mice. However, mice pair-fed to cachectic mice (PFtoC26) exhibited levels of K. oxytoca similar to the ones present in mice pair-fed to the control mice (PFtoCT, Fig. 1D). Hence, we concluded that the reduced food intake alone does not drive the increase in K. oxytoca observed in cachectic mice.

Mechanisms of emergence of Klebsiella oxytoca in cachectic mice.
We have then investigated the mechanisms underlying the expansion of K. oxytoca in these mice. We know from the 1950s that the gut microbiota can confer "colonization resistance" against members of the family Enterobacteriaceae (phylum Proteobacteria) 21 . Therefore, we looked for bacteria which abundance would be reduced in cancer cachexia. At the phylum level, we observed that the expansion of Proteobacteria occurs at the expense of the Firmicutes ( Fig. 2A and Table S1). Such drastic decrease in the Firmicutes phylum (from 56% to 20%) was mainly explained by a decrease in the Clostridiales order (from 53% to 19%) and more specifically by a decreased abundance of the Ruminococcaceae and Lachnospiraceae families ( Fig. 2B and Table S1). Of note, these two families, alongside the Porphyromonadaceae, are the only ones that were strongly negatively associated with the Enterobacteriaceae levels in a multiple correlation analysis (Spearman rho of −0.81, −0.84 and −0.85, respectively, all q-value < 0.05, Fig. 2C). Similar information was retrieved from co-abundance network analysis (Fig. S1).
Interestingly, a bloom of Enterobacteriaceae at the expense of the Ruminococcaceae and Lachnospiraceae families has also been recently described by Byndloss and their colleagues after antibiotic treatment 22 . These authors demonstrated that a reduction in these butyrate-producing microbes by antibiotic treatment is driving a reduced PPAR-γ signaling, leading to reduced β-oxidation, glycolytic switch and increased expression of Nos2 (encoding the inducible nitric oxide synthase, iNOS). Finally, they showed that this host-derived nitrate, as a respiratory electron acceptor, can then foster the growth of Escherichia coli 22,23 . In accordance with their findings, we found that in cachectic mice, K. oxytoca expansion was associated in the cecal tissue with a reduced PPAR-γ signaling (as suggested by a reduced Pparg expression and increased iNOS expression, a target gene of PPAR-γ 22,24 ), reduced β-oxidation (as suggested by Cpt1a, encoding the carnitine palmitoyl-transferase 1a, which deficiency results in a decreased rate of β-oxydation 25 ) and higher glycolysis (as suggested by Hk2, encoding the hexokinase 2) (Fig. 3A-D). Hk2 catalyzes the first step of glycolysis in the intestinal tissue and is upregulated at the transcriptional level during glycolytic switch 26,27 . We also found out that cecal Nos2 expression was strongly correlated with cecal K. oxytoca levels (Fig. 3E). To further strengthen these findings, we asked the question whether nitrate could confer a growth advantage to K. oxytoca. Indeed, supplementation of a simplified growth medium with nitrate fostered the growth of K. oxytoca (Fig. 3F).
Byndloss and their colleagues further established that a reduction in the PPAR-γ signaling was necessary but not sufficient for increasing Enterobacteriaceae members. Indeed, a depletion in regulatory T cells (T regs ) was needed as a second trigger for such an increase to occur. Concordantly, in cachectic mice, we found a reduction in the cecal T regs pool (as measured by their surrogate marker, the transcription factor Foxp3) (Fig. 3G). Of note, these alteration in PPAR-γ signaling and T regs reduction could not be explained only by the reduced food intake as demonstrated by the pair-feeding experiment (Fig. S2). Pparg expression (as well as its target Nos2) was unsurprisingly decreased as caloric restriction was previously report to influence Pparg expression 28,29 . However, Cpt1a, Hk2 and Foxp3 expression were not affected.
Altogether, these findings strongly suggest that K. oxytoca expansion results from a cooperation between, on the one hand, the decreased abundance in the Ruminococcaceae, Lachnospiraceae and Porphyromonadaceae families accompanied by a reduced PPAR-γ signaling, and, on the other hand, the reduced cecal T regs pool.
Klebsiella oxytoca as a gut pathobiont. We next sought to determine the contribution of K. oxytoca to the intestinal alterations found in mice with cancer-related cachexia and, in fine, to cancer cachexia itself. Daily force-feeding can be a source of stress, especially for diseased mice. We therefore evaluated whether administration of K. oxytoca through the drinking water would be an appropriate method to increase in vivo K. oxytoca levels. In vitro experiments demonstrated that K. oxytoca's viability is not affected by −80 °C storage (up to 3 weeks) (Fig. S3A) and that the levels of the bacterium remain stable when administered in the drinking water of mice for 24 hours (Fig. S3B). Administration of K. oxytoca to healthy mice for 9 days resulted in a 3.5-log increase in fecal levels of K. oxytoca (Fig. 4A). These observations validate the administration mode of the bacterium to mice.
In healthy mice, K. oxytoca did not induce a translocation of coliform bacteria to the mesenteric lymph nodes (MLN) (Fig. 4B). In accordance with this result, K. oxytoca did not affect the mRNA expression levels of several key markers involved in the gut permeability (namely, Zonula Occludens-1, Occludin and Mucin-2, Fig. 4C). It also did not affect the mRNA expression levels of several key markers involved in epithelium renewal (namely, Ki67, the T-cell specific transcription factor 4, TCF4, and the Leucin-rich repeat-containing G-protein coupled receptor 5, Lgr5) and those reflecting mucosal immunity (namely, CD3g, Foxp3 and Tbet) ( Fig. 4D-E). A modest increase was found when examining the expression of antimicrobial peptides and associated markers (namely, the matrix metallopeptidase 7, the alpha-defensins and the regenerating islet derived protein 3 gamma, Reg3γ)  Fig. 4F). Altogether, these data allowed us to conclude that K. oxytoca does not affect gut barrier function in healthy mice.
In cachectic mice, we confirmed that K. oxytoca levels are increased. The administration of K. oxytoca in the drinking water induced an additional rise in its fecal levels (Fig. 5A). Analysis of the results of Figs 4A and 5A with an ANOVA model revealed a synergistic effect of exogenous administration of K. oxytoca and cachexia on fecal K. oxytoca levels (p-value interaction: 9.4 × 10 −8 ). We therefore concluded that cachectic mice display an impaired resistance to colonization with an exogenous source of K. oxytoca.
When cachectic mice were administered K. oxytoca, we found an increased bacterial translocation in the MLN of these mice (χ² test, p = 0.04, Fig. 5B). K. oxytoca and Enterobacter sp. were among the bacteria that translocated to the MLN (Table S2). Interestingly, within the group of cachectic mice receiving K. oxytoca, mice displaying translocation of coliform bacteria showed a higher fecal content in the bacterium compared to mice free of coliform bacteria in the MLN (Fig. 5C). In accordance with these results, the expression of several key markers of the  Table S1, n = 5 for sham-injected mice (CT) and n = 8 for C26-injected mice (C26). gut barrier function was affected by K. oxytoca, such as Occludin, Mucin-2, Ki67 and TCF4 ( Fig. 5D-F). Similarly to what was observed in healthy mice, the expression of antimicrobial peptides and associated markers were increased by the administration of K. oxytoca (Fig. 5G).
As we found out that K. oxytoca can worsen the gut barrier dysfunction in cachectic mice, we wondered whether K. oxytoca could also impact cancer progression and other cachectic features. However, neither tumor growth nor any of the cachectic features evaluated in these mice (namely body weight change and muscle atrophy markers) were significantly impacted by K. oxytoca supplementation ( Table 2).

Discussion
So far, most of the research concerning the role of the gut microbiota in the management of metabolic disorders has been performed in the context of overnutrition 30,31 . The gut microbiota might also play a role at the other end of the spectrum, in metabolic and immune alterations associated with undernutrition 20 . In this context, we initiated an original axis of research ten years ago focusing on the causal role and therapeutic potential of the gut microbiota in the metabolic alterations associated with cancer progression, usually defined as cancer cachexia. We found out that preclinical models of cancer cachexia were characterized by intestinal dysbiosis, with one of the main traits being an increase in Enterobacteriaceae 15,18 . Following up on this finding, we report here that the main Enterobacteriaceae species increased in cancer cachexia is K. oxytoca which behaves as a gut pathobiont. We also highlight in the present study by which mechanisms such a pathobiont can emerge in the context of cancer cachexia.
The Enterobacteriaceae family encompasses gram-negative facultative enteropathogens and pathobionts. Enterobacteriaceae reside in the gut at low levels and are localized in close proximity to the mucosal epithelium due to their relative higher tolerance to oxygen diffused from the epithelium 32 . Bloom of Enterobacteriaceae has been described in various pathological contexts, such as inflammatory bowel diseases, obesity, colorectal cancer, celiac disease and antibiotic treatment 32,33 . Several factors can contribute to the emergence of these bacteria, such as for instance, an increased nitrate production, higher levels of oxygen in the gut, release of phospholipids and sialic acid, depletion of microbiota-derived inhibitory products and increased nutrient availability 21,32 .
Host-derived nitrate can result from the reaction of superoxide radicals with nitric oxide, produced by the iNOS. Such nitrate can boost the growth of E. coli through its use as a terminal electron acceptor for the anaerobic respiration 23 . Beside this function, nitrate is a source of ammonium for biosynthesis and can generate proton-motive force for energy through its respiration 34 . Byndloss and colleagues highlighted last year a complex framework for the lower colonization resistance to Enterobacteriaceae after antibiotic treatment 22 . In brief, antibiotic treatment depletes butyrate-producing bacteria, thereby changing host cell metabolism (characterized among others by reduced PPAR-γ signaling and metabolic switch from β-oxidation to anaerobic glycolysis) to elevate iNOS synthesis and reduce host oxygen consumption. Thereby, antibiotics raise the concentration of host-derived respiratory electron acceptors such as nitrate, which can drive the expansion of Enterobacteriaceae 21,22 .
Building up on this conceptual framework for colonization resistance, we reasoned that the emergence of Enterobacteriaceae in cancer cachexia could be driven by similar factors (Fig. 6). In accordance with this hypothesis, we found out that three butyrate-producing families (Ruminococcaceae, Lachnospiraceae and  Porphyromonadaceae 35 ) are strongly and negatively correlated with the bloom of Enterobacteriaceae. These microbial changes are associated with alterations of gut epithelial metabolism and immunity, namely markers reflecting a reduction in PPAR-γ signaling, an increase in iNOS activity, a reduced β-oxidation and an increased glycolysis. Finally, we demonstrated that nitrate confers a growth advantage to K. oxytoca.
From our data, we concluded that although anorexia seems to play a role in the modification of the cecal PPAR-γ signaling, it does not drive alone the expansion of K. oxytoca found in cachectic mice. One explanation might involve the cecal T regs pool. Foxp3 expression, a marker of the T regs pool, was decreased in cachectic mice but not affected in mice pair-fed to cachectic mice versus mice pair-fed to control mice. Byndloss and colleagues showed that a shrinkage in T regs pool, concomitantly with a reduction in butyrate-producing members and PPAR-γ signaling, is needed to recapitulate the bloom of Enterobacteriaceae 22 . Therefore, the absence of effect of anorexia on the T regs pool might explain the lack of expansion of K. oxytoca in pair-fed mice, a hypothesis that would need to be experimentally confirmed.
The term 'pathobionts' has been suggested to describe resident microbes with pathogenic potential. Pathobionts are innocuous to the host under normal conditions, distinct from traditional pathogens which may cause disease even in healthy hosts 36 . According to this definition, our data clearly show that K. oxytoca behaves as a gut pathobiont in cancer cachexia. Indeed, K. oxytoca worsened the gut barrier dysfunction in cachectic mice whereas no effects on the gut barrier function were observed when administered to healthy mice. To our knowledge, so far, K. oxytoca has been described as a gut pathobiont in only one other pathology, namely  Table 2. Body weight gain, tumor weight and expression of atrophy markers in the gastrocnemius of mice receiving a sham injection and receiving the vehicle in their drinking bottle (CT), mice transplanted with C26 cells and receiving the vehicle in their drinking bottle (C26) and mice transplanted with C26 cells and receiving Klebsiella oxytoca in their drinking bottle (C26-Kox). Data are presented as mean ± SEM. *p < 0.05; ***p < 0.001. Figure 6. Based on our findings and previous literature in the field, we propose a conceptual framework for the emergence of K. oxytoxa in cachectic mice. In such framework, reduction in three butyrate-producing microbial families (Ruminococcaceae, Lachnospiraceae and Porphyromonadaceae) fosters a reduced PPAR-γ signaling, which leads to an increased iNOS expression. This reduced PPAR-γ signaling promotes a glycolytic switch (higher glycolysis and reduced β-oxidation), that together with a reduced T regs abundance, allows the emergence of K. oxytoca. Host-derived nitrate production also contributes to the growth of the bacterium. antibiotic-associated hemorrhagic colitis 37 . In this context, a toxin produced by K. oxytoca was isolated and identified as tilivalline. The authors showed that tilivalline induced apoptosis in cultured human cells in vitro and disrupted epithelial barrier function in vivo 38 . However, the K. oxytoca strain administered to the mice did not present any sign of cytotoxin production such as tilivalline in the tested conditions (Fig. S4) suggesting that, in cancer cachexia, the mechanism explaining its pathobiont character is independent from tilivalline. Future research will need to determine the nature of the molecular mediator(s) driving the pathobiont ability of this K. oxytoca strain. Strikingly, K. oxytoca administration hampers gut barrier function in cachectic mice but did not worsen any of the other cachectic features that we assessed. From this observation, it is tempting to speculate that the bloom of Enterobacteriaceae does not contribute significantly to the systemic alterations found in cancer cachexia. However, we showed that supplementation with a synbiotic approach reduced the expansion of Enterobacteriaceae, reduced cancer cell proliferation in the liver, partially protected cachectic mice from muscle atrophy and increased survival 15 . One explanation could be that a combination of microbial changes (namely a full reversal of the intestinal dysbiosis) is needed to observe a benefit for synbiotics in cachexia. In addition, we can also not exclude that exogenous administration of K. oxytoca does not fully mimic the endogenous K. oxytoca expansion which is associated with a defect in the intestinal immune response such as for example the reduced T reg cell pool.
The translational potential of our findings extends beyond cancer cachexia itself. In clinical practice, cancer cachexia is often worsened by anticancer therapy 8 , which exerts drastic antimicrobial effects and can drive pathogen overgrowth 39 . Antibiotics are often administrated during anticancer therapy to prevent or treat life-threatening infections 39 . Antibiotics are also part of the conditioning regimen for hematopoietic stem cell transplantation (often used as a treatment for hematological malignancies), where the gut microbiota appears to play a role, especially in the development of the main complications, the graft-versus-host disease and infections 40 . In a study involving 80 recipients of allo-hematopoietic stem cell transplantation, the post-engraftment microbiota of subjects who died differed significantly from the microbiota of patients who survived, harboring a greater abundance of Enterobacteriaceae and a lower abundance of Lachnospiraceae 41 . Our data suggested that colonization resistance to Enterobacteriaceae (and Proteobacteria in general) is decreased by the presence of the cancer alone. In clinical practice, such mechanism could therefore act conjointly with the antibiotics and anticancer treatments to foster the emergence of life-threatening pathogens.
Altogether, our findings established a role for K. oxytoca as a gut pathobiont in cancer cachexia and exposed new mechanisms of emergence of this bacterium in cancer cachexia, thereby further expanding the conceptual framework underlying the concept of colonization resistance to Enterobacteriaceae. We believe this set of studies constitutes a strong progression in the field of gut microbiota in cancer cachexia, by dissecting the mechanism of emergence of one bacterium and investigating its role in this severe disease. Clinical studies investigating the gut microbiota composition in cancer patients with and without cachexia are currently ongoing and should reveal key information about the translational potential of our findings.
Mouse experiments. Male CD2F1 mice (7 weeks old, Charles River Laboratories, Italy) were kept in specific pathogen free conditions and housed in individually ventilated cages with a 12 h light/dark cycle and fed an irradiated chow diet (AO4-10, 2.9 kcal/g, Safe, France). After one week acclimatization, either a saline solution or C26 cells (1 × 10 6 cells in 0.1 ml saline) were injected subcutaneously. All C26-injected mice displayed a tumor mass observable at day 7. Food intake and body weight were recorded. Eight mice were randomly assigned in each group based on their body weight on the day of cell injection.
The pair-feeding experiment was composed of 4 groups of mice: CT group (sham-injected and fed ad libitum), C26 group (receiving an injection of C26 cancer cells and fed ad libitum), PFtoCT group (sham-injected and fed the mean amount consumed by the CT mice) and PFtoC26 group (sham-injected and fed the mean amount consumed by the C26 mice). Pair-fed mice received daily in 2 equal portions the amount of food consumed by the group they were matched to, with one week delay.
When appropriate, mice received K. oxytoca or the vehicle in their drinking bottles at a concentration of 10 9 CFU/ml, from day 1 after cell injection until the end of the experiment. Solutions were replaced every evening.
Nine to ten days after cancer cell injection, fresh feces were collected, mice were fasted from 7AM to 1PM (except for one experiment) and tissue samples and cecal content were harvested following anesthesia (isoflurane gas, Abbot, Belgium). Tissues were weighed and frozen in liquid nitrogen. All samples were stored at −80 °C until further analyses. DNA of these bacteria was extracted using the QI Amp Fast DNA stool Minikit (Qiagen, Germany) and amplified using 8 F and 1391 R primers (primer sequences in Table S2). The PCR products were purified and subjected to sequencing using the services of DNAVision (Belgium) or Eurofins Genomics (Germany). The sequences were compared with those available in the RDP and EZ BioCloud databases.
Growth of K. oxytoca in a medium with or without nitrate. In anaerobic conditions, 100 ml of a No-Carbon E medium (NCE) supplemented with trace elements 23 and with or without sodium nitrate (40 mM) was inoculated with 1 ml of saturated culture of K. oxytoca and placed at 37 °C. 1 ml of this solution was sampled at each time point to measure the OD at 600 nm.
Growth of K. oxytoca for in vivo administration. K. oxytoca was grown in tryptic soy broth. Bacterial culture was centrifuged, supernatant removed and pelleted bacteria stored at −80 °C. Bacterial numeration was performed after thawing by plating on tryptic soy agar (Sigma-Aldrich, USA), and the bacteria were suspended in the drinking water (vehicle) at an adequate concentration. Identity of the bacterium was checked for each batch using PCR with primers targeting K. oxytoca and visual inspection of the PCR product on agarose gel (Fig. S5). Picture was acquired with a common digital camera. No further processing were applied, beyond a croping of the borders of the image.
Isolation of coliform bacteria from the mesenteric lymph nodes (MLN). Mesenteric lymph nodes were harvested and homogenized in 1 ml sterile PBS using a dounce. 200 µl of the solution were plated on MacConkey agar plates and incubated in aerobic and anaerobic conditions at 37 °C. DNA was directly amplified from an aqueous bacterial suspension using 8 F and 1391 R primers (primer sequences in Table S2). The PCR products were purified and subjected to sequencing using the services of Eurofins Genomics (Germany). The sequences were compared with those available in the RDP and EZ BioCloud databases.
Evaluation of cytotoxin production. Evaluation of cytotoxin production was performed following the principle described by Darby and colleagues 43 . Briefly, HeLa cells (10 000 cells/well, 96-well plate) were incubated in triplicate in presence of various dilutions of bacterial supernatant. After 48 h, cell viability was assessed using a MTT test and data were expressed as percentage of the signal detected for cells incubated with the same dilution of the bacterial medium. The experiment was repeated with three independent bacterial isolates. The MTT assay is based on the principle that metabolically active cells will cleave yellow thiazolyl blue tetrazolium bromide (MTT, 0.5 mg/ml) to form purple formazan crystals. The formazan absorbance was measured at 570 nm, from which a background value, measured at 650 nm, was subtracted.
Tissue mRNA analyses. The isolation of RNA, preparation of complementary cDNA and real-time polymerase chain reaction were performed as previously described 15 (primer sequences in Table S2).
Gut microbiota analyses. Genomic DNA was extracted from the cecal content or feces using the QIAamp DNA Stool Mini Kit, including a bead-beating step. Absolute quantification of the K. oxytoca species levels were performed using qPCR (primer sequences in Table S2). The samples were PCR-enriched for the V5-V6 region of the 16S rRNA gene and then underwent a library tailing PCR (primer sequences in Table S2). The amplicons were purified, quantified and sequenced using an Illumina MiSeq to produce 2 × 300 bp sequencing products. Initial quality-filtering of the reads was conducted with the Illumina Software, yielding an average of 164 603 pass-filter reads per sample. Quality scores were visualized and reads were trimmed to 220 bp (R1) and 200 bp (R2). The reads were merged with the merge-Illumina-pairs application 44 . For all samples, a subset of 25 000 reads was randomly selected using Mothur 1.32.1 45 . The UPARSE pipeline implemented in USEARCH v10.0.240 46 was used to further process the sequences. Putative chimaeras were removed using the cluster_otus command. Non-chimeric sequences were subjected to taxonomic classification using the RDP MultiClassifier 1.1 from the Ribosomal Database Project 47 . The phylotypes were computed as percent proportions based on the total number of sequences in each sample. Significantly affected phyla and families were identified by Welch's t-test. The p-value was adjusted to control for the false discovery rate (FDR) for multiple tests according to the Benjamini and Hochberg procedure 48 . The 16S rDNA sequencing dataset generated and analyzed for this study can be found in the SRA database (project ID: SRP144622, https://www.ncbi.nlm.nih.gov/sra/SRP144622).

Statistical analyses.
Outliers were removed using the Grub's test. The statistical significance of differences between groups was assessed using Student's t-test when comparing two groups or one-way ANOVA followed by Tukey's multiple comparison tests when comparing several groups. Two-way ANOVA followed by Bonferroni post-tests was used to assess the significance of two independent variables for one dependent variable. Multiple correlation analyses based on Spearman correlation were done in R, with adjustment of the p-value according to the Benjamini and Hochberg procedure. The heatmap was drawn using the heatmap.2 function in the gplots package (v2.17.0) as well as the RColorBrewer package (v1.1-2). Co-abundance network was built using CoNet and Cytoscape (v3.6.1) 49,50 . Statistical analyses were performed using GraphPad Prism 5.0 and R 51 . P < 0.05 was considered statistically significant.
Data Availability Statement. The 16S rDNA sequencing dataset generated and analyzed for this study can be found in the SRA database (project ID: SRP144622, https://www.ncbi.nlm.nih.gov/sra/SRP144622).