The study was performed under a protocol approved by the Tarleton State University Institutional Animal Care and Usage Committee (Animal Use Protocol 12-009-2016-A1). Four Balb/c females 8 weeks in age were utilized in this experiment. Females were housed together and raised on similar chow diets and similarly weaned. Mice 1 and 2 were siblings while mice 3 and 4 were siblings. All mice were euthanized with 150 microliters of sodium pentobarbital delivered intraperitoneally. Post injection, mice shed two to three samples of stool which were recovered utilizing sterile forceps and immediately frozen. Once deceased, 2.5 cm of the large distal colon from each mouse was removed. After, two small additional 0.5 cm samples of the large distal colon were excised from the specimen and immediately frozen. Colon tissue extractions were added to a sterile tissue grinder along with 5 mL of Dulbecco’s Modified Eagle Medium (DMEM; VWR, Radnor, PA). The sample was manually homogenized into a liquid solution.

Hydrogel (Corning, Corning, NY) was prepared using 8 mL of molecular grade water and 20 microliters of hydrogel to create a 0.25% solution. 150 microliters of the prepared solution were added to 6.5 mm transwell inserts (n = 8; Corning, Corning, NY) that were placed into a 24 well tissue culture plate (Corning, Corning, NY). In addition, 500 microliters of supplementary Dulbecco’s Modified Eagle Medium (DMEM) was added under each well insert. Once the culture plates were prepared, 250 microliters of the homogenized colon were added to the top of the hydrogel in the transwell inserts [22] [23] . Plates were checked for baseline pH by transferring a small drop of medium with a mechanical pipette onto litmus paper. The plates were then added to a single incubator, but to create an anoxic environment, plates were incubated in an anaerobic system (BD Diagnostics, Franklin Lakes, NY). Plates were incubated for 24 hours. The medium below each insert was again tested for pH again and the culture was transferred into sterile 2.5 mL storage tubes and frozen for future DNA extraction.

DNA was extracted from each sample using a modified protocol from Brady et al. [24] . After extraction, DNA was amplified utilizing prokaryote specific primers, 519F 5’-CAGCMGCCGCGGTAA-3’) and 785R (5’-TACNVGGGTATCTAATCC-3’), that target the V4 region of the 16S rRNA [25] [26] . PCR amplification was accomplished through denaturation at 95˚C for 3 minutes, followed by 35 cycles of 95˚C for 10 seconds, 55˚C for 30 seconds, and 72˚C for 30 seconds. DNA barcodes were added to samples with 10 cycles of the same PCR protocol. To prevent inhibition during PCR, samples underwent an additional cleanup with 20% Chelex 100. Sequences were size-selected with a Pippin Prep instrument (Sage Science, Beverly, MA) to a length of 300 - 600 base pairs. Sequencing was conducted on the Illumina MiSeq platform using 600 cycle paired end v3 sequencing kits at the Texas A&M University Genomics Core Facility. Raw sequences were processed through QIIME [27] and USEARCH [28] . Taxonomy was assigned using Greengenes 13.8 database [29] as a reference with UCLUST [28] , and OTU picking was conducted at 97% sequence similarity with the RDP [30] method in QIIME.

The microbes from four mouse distal colons were cultured in 12 cell culture lift inserts in a 24 well plate system. Additionally, a total of 9 fecal and 9 colon samples distributed across 4 mice were prepared for DNA extraction. Due limited spacing in the 96 well plate DNA extraction method, only one colon and fecal sample could be performed in triplicate, and it was randomly selected to be mouse 3.

Cumulative sum scaling was used to normalize the data and account for uneven sequencing depth between samples [31] . Biom files were constructed through QIIME and transferred into R [32] for further statistical analysis. Phyloseq [33] , ggplot2 [34] , and vegan [35] packages were utilized to evaluate alpha and beta diversity with seed set at 1400. Alpha diversity was assessed using the Shannon diversity index. Variation in alpha diversity for oxygen, pH, mouse, feces, and colon comparisons were first checked for normality using the Shapiro-Wilk test for normality [36] . The data was non-normal in distribution (Shapiro-Wilk test, w = 0.9506, p < 0.01); therefore, comparisons were made with non-parametric tests. All multivariate tests were corrected using false discovery rate (FDR) [31] . Comparisons of alpha diversity were assessed using Kruskal Wallis one-way analysis of variance (KW ANOVA) or Wilcoxon rank sums test (Wilcoxon test) while comparisons of beta diversity were assessed with unweighted unifrac distance metrics at 1000 permutations using permutational multivariate analysis of variance (PERMANOVA). Dunn’s test post-hoc analysis was done through the dunn.test package in R [37] . In addition, non-parametric t-tests were used for comparisons of mean abundance in individual bacterial strains between samples. A microbial network was constructed using the Co-occurrence Network Interferences (CoNet) application for Cytoscape [38] . Feces and colon data were removed before CoNet analysis. CoNet has been utilized in previous studies to investigate defined interactions between microbes [39] [40] . Spearman correlation coefficient with a cutoff ratio of 0.6 was utilized, and to focus the network, only microbes with sequence counts greater or equal to 20 were included. 1000 permutations were accomplished through a bootstrapping method with an FDR correction [40] .

pH readings of each plate were taken before and after incubation. As shown in Table 1, pH fluctuated from the original baseline of 8. In addition, mice maintained varying levels of pH due to differences in oxygen concentration (Table 1). Sample sizes are uneven due to losses during cleanup or DNA extractions.

After quality filtering, we had a total sample size of 111 samples and 3,133,666 sequences total (Supplemental Table S1). The sequence files were submitted to the National Center for Biotechnology Information (NCBI) database (Supplemental Table S2). The profile of the feces and the colon were characterized for microbial composition at the phylum (Figure 1(A)) and family (Figure 1(B)) levels. There was some variation between samples, even within the same mouse, but the core phylum composition of the colon and fecal samples was dominated by Firmicutes and Bacteroidetes (Figure 1(A)), with means of 47% and 49%, respectively, and standard deviations (SD) of 23%. In addition, the family S24-7 (order Bacteroidales) was highly abundant in all samples with a mean of 42% and a SD of 20% (Figure 1(B)). Analysis of beta diversity for each of the feces and colon samples revealed no difference in composition (Supplemental Table S3), and analysis of alpha (Shannon) diversity (Figure 1(C)) also revealed no difference (KW ANOVA P = 0.47). Therefore, samples were pooled together for comparison between feces and colon. Shannon diversity index was utilized for comparison of the bulk samples (Figure 1(D)). Results showed no difference

between the pooled feces and colon samples (Wilcoxon test, P = 0.44). In addition, beta diversity comparison of pooled samples showed no difference (data not shown, PERMANOVA, pseudo-F = 1.06, P = 0.37). Since it was determined that feces and colon samples are similar, all samples were pooled into one bulk sample, named “pooled microbiome”, for diversity comparisons with cultured prokaryotes.

In the cultured samples with 24-hour incubation, Firmicutes and Bacteroidetes were the dominant phyla (Figure 2(A)). Firmicutes had the highest average relative abundance, 70% (SD 28%), with Bacteroidetes averaging 18% (SD 16%). Mice 1 and 2 exhibited more species richness in the cultures than mice 3 and 4 at the phylum level (Figure 2(A)). Across all cultures, the impacts of oxygen levels were not evident at the phylum level (Figure 2(A)), at the genus level differences were observed in transient genera (data not shown), although these differences were not significant (Figure 2(B)). Shannon diversity index shows a difference between some of the cultures and the colon and fecal microbiome of the mice (KW ANOVA, Dunn’s test, Figure 2(B)). Post-hoc analysis shows that, compared to the microbiome, mouse cultures 1 and 2 were statistically similar to the pooled microbiome from colon and fecal isolates while mouse cultures 3 and 4 differed significantly (Supplemental Table S4, Figure 2(B)).

A community profile of cultural composition due to varying levels of oxygen and pH exposure was constructed (Figures 3(A)-(D)). The cultures were placed overnight into aerobic (20%) or anoxic (0%) conditions, and the dominant bacteria in the cultures produced pH shifts. The overnight culture of colon samples from mice 3 and 4 resulted in a final pH between 6 - 7 (Table 1) and the cultures are dominated by Firmicutes and Bacteroidetes with few transient phyla (Figure 3(A)). The overnight incubation of the cultures from mice 1 and 2 resulted in a higher final pH (9 - 10; Table 1) which resulted in more observed phyla (Figure 3(A)), and these cultures were more representative of the in vivo microbiome (Figure 2(B)). Comparison of Shannon diversity revealed that variation in pH led to significant differences in alpha diversity (KW ANOVA, Dunn’s test). Post-hoc

analysis revealed that plates reaching a pH of 6 and 7 were similar while all other comparisons differed (Figure 3(B)). Analysis of beta diversity also revealed differences in prokaryotic communities between plates of varying pH levels (Supplemental Table S5).

When samples cultured at identical oxygen concentration were pooled together, no difference in alpha diversity existed between the two oxygen levels among genera (Wilcoxon test, P = 0.34; Figure 3(C), Figure 3(D). However, shifts in the individual mice can be noted, especially with between the two oxygen concentrations. Additionally, distance-based linear modeling revealed that oxygen did not contribute significantly to community clustering (Supplemental Table S6).

Mice 1 and 2 were a sibling pair. Mice 3 and 4 were a sibling pair. A marked difference in prokaryote cultural composition was noted by familial relationship (Supplemental Table S7, Figure 4, Figure 5(A), Figure 5(B)). Mice 1 and 2 had greater diversity in their cultures than mice 3 and 4 (Figure 5(A)). Mouse 1 had much higher Lactobacillus, HA73 (Phylum Synergistetes), and Ruminofilibacter than mice 3 and 4 (Figure 5(B)) and mouse 2 had even more diversity with increases in the genus Clostridium (Figure 5(B)). Sibling relationship explains the changes in community composition as time elapsed in the incubators (Supplemental Table S6). Distance-based linear modeling indicated pH, individual mouse, and sibling effects all significantly contributed to microbial community variation when considered independently (P < 0.001), accounting for 20%, 18%, and 22% of the variation, respectively, while oxygen level did not impact the microbiota (Supplemental Table S6). However, when these variables were considered together, a most parsimonious model containing individual

mouse and sibling relationship accounted for 29.0% of the variation while pH did not contribute to explaining microbial variation if individual mouse and sibling relationship were already in the model. Additionally, Shannon diversity significantly varied between sibling groups (KW ANOVA, P < 0.01). Post-hoc analysis showed mouse 1 and mouse 2 were similar and varied from mouse 3 and mouse 4, which were also similar (Dunn’s test, Supplemental Table S4 and Figure 5(C)). Additionally, beta diversity varied according to familial relationship (PERMANOVA, P < 0.01).

The OTUs in the microbial network represent 88% of the relative sequence count for the cultured well plates (Figure 6). Many of the interactions were positive in

nature meaning co-presence in a shared-niche is the most abundant interaction type. Negative, mutually exclusive interactions are only between the microbial genus Enterococcus and an unclassified strain of Bacteroidales (Figure 5). The interaction between the 4 mutually exclusive OTUs account for 50% of all sequences. Enterococcus species were more well represented in the cultured samples than the colon and fecal samples (“pooled microbiome”), especially in cultures from mice 3 and 4 (Figure 7(A); Supplemental Table S8), which had the lower pH (Table 1, Figure 7(B)). The low pH of these cultures may be influenced by an increase in Lactobacillus (Figure 7(C)). The increase in Enterococcus and Lactobacillus in mice 3 and 4 resulted in a decline in genera from the phylum Proteobacteria (Figure 7(D), Supplemental Table S9). Although perhaps not proportional to the in vivo colonic microbiome, most representative

prokaryotes, including the hard to grow Archaea (Figure 7(E)) and Clostridium (Figure 7(F)) were isolated from these cultures. The addition of the chemical propidium monoazide (PMA) to the preparations indicated that the cultures were not just present, but also alive, as no significant differences were observed with or without PMA treatment at the level of family (Figure 8(A)). To conclude, the stability of the cultures was confirmed when a plate of eight transwells were frozen for one month at −80˚C, and again there were no statistically significant changes in the cultures at the level of family (Figure 8(B)).