Development of Rat Model (Rattus Norvegicus) Harboring Homogenous Profile of Human Microbiota Derived from Multiple Donors with Diverse Microbial Populations

The gut microbiota is a diverse system with important roles in host functions and well-being, and its dysbiosis is linked to metabolic disorders like obesity and diabetes...

studying the gut microbiota in human subjects is technically challenging, because there is no reliable, non-invasive method for collecting region-specific samples of the gut microbiota. Humanized models generated by transplanting human microbiota to germfree (GF) animals are useful for studying the changes in microbial population structure and functions following stress exposure, as well as the relationship between microbiota dysbiosis and changes in host functions [5,6], especially when the analysis requires tissue samples from the hosts [7,8]. GF models from commercial venders, however, are limited and GF rats are not commercially available in the United States. Establishing and maintaining a GF colony in-house is costly, due to the requirement of specialized facilities.
Additionally, humanized models, generated by mono-association with a single human donor, compound the variability that already exists due to inter-individual differences in gut microbiota, resulting in inconsistencies in the response of microbiota to experimental manipulation [9]. One way to overcome this challenge is to generate multiple humanized lines and use them as biological replicates; however, this greatly increases the cost and the number of animals needed to account properly for such variability.
The goal of this study was thus to develop a reliable and reproducible method for generating pseudo germ-free (PGF) rats that are suitable for the production of humanized rats harboring a stable, homogenous, gut microbiota derived from multiple human donors.
Others have described the use of antibiotics to deplete the native gut microbiota of rodents [9][10][11][12][13].The type of antibiotics used and method of delivery can affect the health of the animals [14] and the diversity of the remaining gut microbiota that can survive the treatment [11]. Although the microbiota that colonize the gut of GF animals after the transplant of human fecal microbiota showed similarity to the donor material, previous reports showed that the microbial population re-established in antibiotic-treated animals after fecal transplant has greater similarity with the native microbiota [12,13] highlighting the need to improve the procedures for generating PGF and humanized models. We report here the development of a new antibiotic cocktail, delivered in the drinking water.
This cocktail causes minimal systemic exposure, thereby avoiding potential adverse effects on host functions. It was optimized with the lowest concentration and type of antibiotics while capable of significantly decimating the rat native gut microbiota, thus generating PGF rats.
Humanization was accomplished using repeated inoculations of fecal microbiota from human donors, which has been successfully employed by others [9]; however, this study took advantage of coprophagia, the natural rodent behavior of consuming fecal materials, for the horizontal transfer and exchange of the transplanted human-derived microbiota between animals that do not share the same housing cage. This approach effectively increased the number of inoculations and further blended the microbiota across recipients, establishing a homogenous human microbiota population within the rat community. The homogeneity and stability of the transplanted microbiota was confirmed by metagenomic analysis targeting the hypervariable region of the 16S rRNA locus in the microbiota genomic DNA.
This procedure is capable of stably establishing human microbiota in antibiotic-generated PGF rats and eliminated donor-spe-

Animals and Husbandry
The study protocol was approved by the Wright-Patterson Air

Development of Antibiotic Cocktail
Initially, Enrofloxacin (500 mg/L), Neomycin (150 mg/L), Vancomycin (350 mg/L), Amphotericin B (1000 mg/L), and Ampicillin (150 mg/L) (all from Patterson Vet, Inc.) were chosen as candidates because of their low gastrointestinal absorption or rapid elimination thereby minimizing systemic exposure and residual effects. The individual and combined effectiveness of these agents were tested using plate cultures to determine their effects on microbial titer of the cecal contents from naïve rats as described above but while adding the antibiotic(s) to the blood-agar plates.
Amphotericin B and Ampicillin, which did not show significant antimicrobial activity when used alone, nor significantly improved the cocktail's effectiveness, were eliminated after the first round of testing. The concentrations of remaining antibiotics were adjusted empirically until a combination that effectively resulted in no detectable colony on agar plates was discovered.

Generating Pseudo Germ-Free (PGF) Rats
The animals were housed in Semi-Rigid GF Isolators (Charles River Laboratories) that were maintained in the PGF animal production room following the manufacturer's recommendations.
The GF status of the isolators was monitored using culture swabs and Pocket Swab Plus (Charm Sciences, Lawrence, MA). After two days of acclimation to the isolators, the rat native microbiota was depleted using the antibiotic cocktail in the drinking water, which was replaced daily for 14 days. Untreated rats, housed in a separate isolator and receiving regular food and drinking water, were used as control. Tests indicated that cecal content was a more reliable indicator than feces for the evaluation of the depletion of microbiota in the gut. Thus, the bacterial titer of the cecal contents was determined in two animals at each of Days 3, 6, 9 and 12 (Day 1 = the first day of antibiotics treatment), in order to monitor the depletion of the native rat microbiota. Initially, the antibiotic cocktail optimized using the plate culture method (65 mg/L Enrofloxacin, 1000 mg/L Neomycin, 500 mg/L Vancomycin) was used. The concentration of antibiotics were further optimized empirically until no detectable microbial colony (using the plate culture method) in the cecal content after three days of antibiotic treatment. Depersonalization of humanized rats began by forming two communities. Community 1 was comprised of half the animals from each of the following humanized lines: #74-associated, #80-associated, #88-associated, and 3-donor-mix-associated.

Human Fecal Microbiota
Community 2 was comprised of the remaining half of the #74-and #80-associated lines. The remaining half of the #88-and 3-donormix-associated animals were maintained as independent lines. Each community was depersonalized by coprophagia-mediated horizontal transfer of the gut microbiota within each community.
On every second day for 15 weeks, the bedding including all fecal pellets from all members of the respective community was pooled, mixed with fresh bedding, and redistributed back to the members' cages. This is to allow all members of the community had the opportunity to consume each other's feces, even though they did not share the same cage. Each community was housed in a separate isolator to prevent cross-contamination or the introduction of external microbes. Normal rats (with no antibiotic treatment or fecal transplant), housed in a separate isolator and receiving regular food and drinking water, were used as control. The acquisition of human microbiota through coprophagia alone was also tested by introducing newly generated PGF rats to the communities and providing them with a portion of the mixed bedding that  Note: Each box represents 1 week. The native gut microbiota of Sprague Dawley (SD) rats was depleted by antibiotics treatment in germ-free isolators to produce pseudo germ-free (PGF) rats. Humanization was accomplished with five (5) intragastric inoculations of human fecal material followed by 2 weeks of establishment for the colonization of the transplanted human microbiota. Some rats receiving fecal transplant were maintained as individual lines while others were assigned to specific communities for coprophagia-mediated depersonalization. Community bedding containing fecal pellets was mixed and redistributed to promote the horizontal transfer of microbiota. Fecal and cecal microbiota, as well as tissues samples were collected at indicated timepoints for 16s rRNA hypervariable regions sequencing analysis and the analysis of various markers for host functions.

Gut Microbiota
The population structure of the gut microbiota was analyzed  This kit contains two wide-ranging primer sets (V2-4-8 and V3-6, 7-9) for multiple hypervariable regions, allowing detection of a broad range of microbial species. DNA enrichment after PCR was confirmed by electrophoresis in a 2% agarose gel. PCR products was purified using Agencourt AMPure XP magnetic beads (Fisher) and quantified on the 4200 Tape Station using D1000 screen tape and reagents (Agilent Technologies, Santa Clara, CA were assigned based on 97% sequence identity to the Green Genes reference (v13.5) [16]. Beta-diversity was calculated using UniFrac [17] methodology, and three-dimensional principal coordinate analysis (PCoA) plots visualized with Emperor [18].

Development of Pseudo Germ-Free (PGF) Rats
The antibiotic cocktail developed using the plate culture method failed to eliminate all culturable microbial cells in the cecal content from the treated rats. After several rounds of optimization of the antibiotic concentrations, the finalized cocktail (200 mg/L Enrofloxacin, 800 mg/L Neomycin, and 850 mg/L Vancomycin) was able to eliminate all culturable cells in the cecal content after 3 days.
Comparing to two to three thousands of colonies normally obtained from the cecal content of control animals (diluted 10 6 -fold), this antibiotic cocktail is able to reduce the microbial load in the cecum (and perhaps in the entire intestinal tract) by at least 3-order of magnitude in 3 days. Based on this result, it was concluded that the treated animals likely reached a PGF status after the course of 14-day antibiotics treatment. However, since a significant portion of the gut microbiota is not culturable, this result may not provide a complete picture concerning the overall effectiveness of this antibiotic cocktail. We found that PCR products of the bacterial 16s rRNA gene hypervariable regions remained detectable in the fecal content from the animals that have received 12-day antibiotics treatment, provided the entire PCR product was used in gel electrophoresis. This result was not completely unexpected, since PCR is highly sensitive that it can detect a few molecules under optimal conditions. For instance, genomic DNA from non-culturable (or even non-proliferative) microbes would give positive results in PCR assays. Consistent with this, the result of metagenomic analysis showed that this antibiotic cocktail resulted in a 52.3% reduction in the number of OTUs detectable in the fecal content.

Microbiota-Humanization of PGF Rats
The fecal microbiota from six human donors and naïve rats (prior to antibiotic treatment) were analyzed using 16s rRNA hypervariable regions sequencing to determine their overall diversity and the relative abundancies of various bacterial taxa.  [17,20]. The human microbiota is notably different from that from naïve rats (Figure 2). In addition, other phyla including   Figure 4, with Donor #88 as a representative example).
Interestingly, antibiotics treatment of rats altered their microbial population, which is significantly different from that of the naïve rats or the human donor. However, the transplantation of human fecal material into the PGF rats resulted in a microbiota profile that is virtually indistinguishable from the that of the human donor ( Figure 4, #88-Associated Rats compared to Donor #88), thus confirming that this humanization procedure was highly effective in establishing the human microbiota in the PGF rats.   Note: The relative abundancies of detectable bacterial phyla in fecal samples were determined by 16s rRNA hypervariable regions sequencing analysis. Compared here are the average relative abundancies in the native rat microbiota collected at baseline, samples from PGF rats collected after 12-day antibiotic treatment, samples from #88-associated humanized line collected prior to depersonalization, and that from Donor #88.

Depersonalization of Humanized Rats
Elimination of the donor-specific characteristics of the gut microbiota in the humanized rats was accomplished through coprophagia, and its progress was monitored using 16s rRNA hypervariable regions sequencing analysis of the fecal samples collected at specific timepoints during the Depersonalization period. Nevertheless, Community 2 remained well separated from the baseline and control rat clusters, even after 15 weeks. These results further confirmed that human microbiota can be successfully established in the PGF rats that was generated using our newly developed antibiotic cocktail and maintained for at least 15 weeks.
Moreover, the donor-specific microbiota profile can be eliminated through the sharing of bedding material and fecal pellets. In each panel, cage control represents normal rats with no antibiotic's treatment or fecal transplant. Antibiotic/LB rats were treated with antibiotics but not human fecal material. The "3-Donor-associated" and "3-Donor-Line" are rats that were transplanted with the mix of materials from Donors #37, #60 and #65. "Donors" indicates fecal samples from individual donor or a mixture of the indicated human donors.
It was quite unexpected that the human microbiota was less stable in rat hosts when maintained as individual lines when compared to the two depersonalized communities. As shown in Figure   5d, only the 5-week samples from the #88-associated line remained separated from the naïve rat cluster. The samples collected at 10 and 15 weeks were closed to the naïve and control rat samples, indicating re-emerging of the native rat microbiota in these humanized animals. Similarly, the 3-donor mix-associated line also appeared to have reverted to the native rat microbiota by week 5. Analysis of markers related to HPA axis, immune, and intestine mucosal functions in the hosts showed that there were no significant changes in these parameters (see Supplemental Data). These results thus suggest that the human microbiota likely interacts with the rat intestinal tract in a similar way as the rat native microbiota, and that the procedure presented here for developing the DHR Model is unlikely to cause any significant change in the host functions.

Discussion
In this study, naïve SD rats were first converted to PGF rats using a highly effective antibiotic cocktail that can deplete all culturable microbial cells in the cecal content after 3 days of treatment, which represents an at least 3-order of magnitude reduction in the microbial load. The treatment was continued for another 11 days to ensure the animals reach a PGF status. Others have reported the use of 1 to 4 antibiotics in the drinking water or by oral gavage for 3 to 21 days in mice, and they achieved reductions of 1 to 5 orders of magnitude [12][13][14]. The authors of these studies concluded that the treated mice resembled GF mice anatomically and physiologically [2]. The PGF rats generated herein, which experienced a similar level of reduction in the bacterial load, thus are expected to be similar to GF rats. The humanized rats produced using the procedure described above harbored gut microbiota more closely resembled to that of the human donors than the naïve rats. A similar observation has been reported by others in mice, when a highly effective antibiotic treatment was employed [9] or when using GF animals [12]. However, a smaller reduction in microbial load produced a less diverse microbiota and greater similarity to the naïve microbiota than to that of the human donors [13]. Our procedure for creating PGF rats allowed the humanderived microbiota to successfully replace the native microbiota in the rat intestinal tract (see Figure 4). Although the gut microbiota of the humanized rats is similar to that of the human donors, the native rat microbiota is robust, as indicated by the partial recovery of the native rat microbiota in the PGF rats that did not receive fecal transplant (Antibiotic/LB in Figures 5b-d).The failure of PGF rats to become humanized through coprophagia alone also confirmed this notion.
The longest stability reported in humanized rodent models of gut microbiota to date was 12 weeks [9], but the mice in that study were each associated with a single donor and had distinct microbiota and metabolomic profiles. In this study, humanized rats generated from multiple donors with diverse microbiota were depersonalized to produce a high degree of homogeneity between the humanized rats, thereby eliminating donor-specific attributes in their gut microbiota. This homogenous profile of human microbiota was stable for 15 weeks. In addition, the humanderived microbiota of the depersonalized communities appeared to be more stable than that of individual humanized lines generated from the same human donors. One potential explanation is that the human-derived microbiota created in the community setting were better able to compete against the native rat microbiota than those in the individual lines. Alternatively, the communities may have benefited from coprophagia, which essentially re-inoculated each animal with the human-derived microbiota when they consumed feces contained within the redistributed bedding.
This study also showed that mixing the donor material prior to transplantation might hinder the humanization process. The individual humanized line generated from the mixture of 3 human donors (mix-associated) were more similar to the native rat microbiota than the human microbiota after 5 weeks (Figure 5b).
In addition, some members of Community 1 were mix-associated rats, and that community took longer to become homogenous and was more similar to the native rat microbiota than Community 2, which did not contain mix-associated rats (Figures 5b & 5c). Thus, the inoculum made by mixing human donor materials appears to compete poorly against the re-emerging, native rat microbiota compared to the inoculum made from a single human donor. In conclusion, we presented a procedure for generating PGF rats from commercially purchased animals, and the subsequent creation of communities of rats with a homogenous profile of humanderived microbiota that is stable for at least 15 weeks. With certain modifications, this procedure will likely be applicable for any natural or genetically modified strains of laboratory rodents, which in turn will greatly expand the models available for studying gut microbiota and its interaction with host functions.

Result
Supplemental Note: *Mean ± SE; 1 U of MPO activity = Amount of MPO needed to convert 1.0 μ mol of substrate to product per minute at 25°C; ** Statistically significant different from cage control (p < 0.05).  Table 2). This was confirmed by culture data that did not find any bacterial cells in any of the extra-intestinal tissues including mesenteric layer, mesenteric lymph nodes, spleen and liver (data not shown). There were some small changes in the cytokine levels, but none was statistically significant after the correction for multiple testing (Supplemental Data Table 3). For the hormones of the HPA axis, only sporadic changes in the level of adrenocorticotropic hormone (ACTH) and corticosterone occurred and were not likely a specific stress response to the experimental procedure or the colonization of the human microbiota (Supplemental Data   Table 4). In the #88-Associated Line, ACTH was significantly higher at 2 weeks and lower at 5 weeks than the cage control; and corticosterone levels were significantly higher than control at 10 weeks. The Depersonalized Community 1 had a significantly higher level of ACTH than the cage control at 15 weeks. The 3-donor mixassociated line had significantly higher corticosterone than the control at 15 weeks. Since these changes did not follow a specific pattern and did not correlate with other markers, the biological significance of these changes is not completely understood at this point.Taken together, these results suggested that the human microbiota probably interacts with the rat intestinal tract in a similar way as the native rat microbiota and is unlikely to cause any significant change in the host functions.