Fecal Microbiota Restoration Modulates the Microbiome in Inflammation-Driven Colorectal Cancer

Simple Summary Inflammation of the colon (colitis) can increase the risk of developing colorectal cancers. Gut microbiota may play a role in the development of this cancer. One potential way to prevent this cancer is to manipulate the community of bacteria in the colon. In a study using mice, we found that transferring the bacteria from the bedding of healthy mice to those with colitis increased the presence of Akkermansia, which may help reduce inflammation in the colon. Meanwhile, untreated mice had an increase in other types of bacteria, such as Anaeroplasma and Alistipes, which may contribute to inflammation. These findings suggest that certain types of bacteria may help reduce inflammation in the colon and prevent cancer. Abstract Chronic inflammation of the colon (colitis) is a known risk factor for inflammatory-driven colorectal cancers (id-CRCs), and intestinal microbiota has been implicated in the etiology of id-CRCs. Manipulation of the microbiome is a clinically viable therapeutic approach to limiting id-CRCs. To understand the microbiome changes that occur over time in id-CRCs, we used a mouse model of id-CRCs with the treatment of azoxymethane (AOM) and dextran sodium sulfate (DSS) and measured the microbiome over time. We included cohorts where the microbiome was restored using cage bedding swapping and where the microbiome was depleted using antibiotics to compare to untreated animals. We identified consistent increases in Akkermansia in mice receiving horizontal microbiome transfer (HMT) via cage bedding swapping, while the control cohort had consistent longitudinal increases in Anaeroplasma and Alistipes. Additionally, fecal lipocalin-2 (Lcn-2), a marker of intestinal inflammation, was elevated in unrestored animals compared to restored and antibiotic-treated counterparts following HMT. These observations suggest a potential role for Akkermansia, Anaeroplasma, and Alistipes in regulating colonic inflammation in id-CRCs.


Importance
Preclinical models of inflammation-driven colorectal cancer (id-CRC) have shown that the microbiome regulates intestinal inflammation. However, most of these studies have relied on short periods of longitudinal microbiome sampling and a single mouse strain to conclude bacterial genera influencing id-CRC pathology. Here we utilized a mouse model of id-CRC. We show that horizontal microbiome transfer (HMT) through fecal restoration with healthy animals can longitudinally alter the microbiome of mice from Balb/c and C57BL/6

Mice and Animal Husbandry
Three to four-week-old female C57BL/6 and BALB/c mice purchased from Jackson Laboratory were used in this study. On arrival at the animal facility, mice were randomly assigned to restored or unrestored groups (n = 5/group) and received an ear punch. Bedding on days 1-3 was pooled between experimental groups to normalize the baseline microbiome between restored and unrestored groups. Following baseline normalization, mice underwent AOM/DSS treatment, and stool samples were longitudinally collected 3x/week for 18 weeks. In week 18, mice were sacrificed and dissected. Mouse colons were fixed in 10% neutral buffered formalin and processed for histological analysis. All mice were housed in specific pathogen-free conditions in fully autoclaved cages for the duration of the experiment. The Institutional Animal Care and Use Committee approved all animal studies.

AOM/DSS Tumor Induction
Depending on the experimental design, female 3-4-week-old BALB/c or C57BL/6 mice were injected intraperitoneally with 10 mg azoxymethane (Sigma) per kg mouse weight. After five days, mice were treated with two cycles (1 week/cycle) of 2% dextran sodium sulfate (MP Biomedicals M.W. = 36,000-50,000) for five days through their drinking water, followed by a supply of regular water for the duration of the experiment. Mice were sacrificed in week 18 following AOM induction. Depending on the experimental setup, the antibiotics cocktail of 0.05 g/L of vancomycin (Boynton Health), 0.025 g/L of metronidazole (Boynton Health), and 0.2 g/L of streptomycin (Sigma) was administered in the drinking water of mice following the second cycle of DSS administration. During necropsy, colons were dissected and flushed with PBS to visualize tumors in the colon. Following visualization of tumors, tumor tissues were fixed in 10% neutral buffered formalin and sectioned.

Fecal Restoration and Stool Sample Collection
Mouse stool samples were longitudinally collected 3 times weekly for 18 weeks. Stool sample collection was conducted as follows: Mice were separated into individual autoclaved SPF cages lacking food and bedding and left for 1 h. On average, 3-5 fecal pellets were collected per mouse. Depending on group assignment, mice were removed from individual cages and returned to their respective cages. Stool samples were pooled by experimental groups and stored at −80 • C until processing for DNA isolation. During weeks 8, 9, and 10, the bedding from normal mice without AOM/DSS was distributed to cages for fecal restoration by HMT.

DNA Isolation and 16S rRNA Amplicon Sequencing and Analysis
Mouse fecal pellets of approximately 0.1 g were homogenized and processed using the DNA PowerSoil kit (Qiagen) with the QIACube automated platform following the manufacturer's inhibitor removal technology (IRT) protocol. The V5-V6 hypervariable regions were amplified and sequenced using the BSF784/1064R primer set [29]. Sterile water and no template controls were carried through amplification and sequencing and did not produce amplicons. Samples underwent paired-end sequencing on a single run of a MiSeq platform (Illumina, Inc., San Diego, CA, USA) at a length of 301 nucleotides (nt). All sequence processing was performed using Mothur software version 1.35.1 [30]. Raw fastq files were trimmed to 150 nt sequences to remove low-quality regions and were paired-end merged using the fastq-join function [31]. Reads were further trimmed to remove reads with mean quality scores < 35 over a 50 nt window, homopolymers > 8 nt, ambiguous bases, and more that 2 nt mismatches to primer sequences. High-quality sequences were aligned to the SILVA database (version 132) [32] and subjected to a 2% pre-clustering step to remove likely errors [33]. Chimeric sequences were identified and removed using the UCHIME package [34]. Operational Taxonomic Units (OTUs) were assigned at 99% similarity using the complete-linkage clustering algorithm, and taxonomy was assigned using version 16 of the Ribosomal Database Project [35]. The raw fastq files are available at https://www.ncbi.nlm.nih.gov/sra/PRJNA915052.

Histopathological Analysis
Mouse colons were fixed overnight in 10% neutral buffered formalin at the experimental endpoint. Following fixation, solutions were changed to 70% ethanol before submission to the clinical and translational sciences institute at the University of Minnesota. Tissues were paraffin-embedded, and two sections per mouse were stained with hematoxylin and eosin. Board-certified pathologists reviewed tissue sections for dysplasia and acute and chronic inflammation.
3.6. Quantification of Fecal Lipocalin-2 (Lcn-2) by ELISA Initially, 0.2 g of frozen fecal samples was reconstituted in PBS with 0.1% Tween 20 (100 mg/mL). Subsequently, 3 mm glass beads were added to the mixture and vortexed 3 times in 30 s bursts to obtain a homogenous fecal suspension. The samples were then centrifuged for 10 min at 12,000 rpm at 4 • C. Clear supernatants were collected and stored at −20 • C prior to analysis. Lcn-2 levels were approximated using a Duoset murine Lcn-2 ELISA kit (R&D Systems, Minneapolis, MN) following the manufacturer's protocol. Four time points before and after fecal restoration (W2, W7, W10, W15) were selected for fecal Lcn-2 analysis. Samples for each experimental condition were run in triplicate. Calibration standard curves (500 pg/mL-7.81 pg/mL) were used to extrapolate Lcn-2 concentrations and averaged across experimental time points and conditions. Reagent blank 1.0% BSA in PBS was used as a negative control.

Statistics
Alpha diversity of microbial communities was assessed using Shannon and Chao1 indices. Bray-Curtis dissimilarity matrices [36] were calculated and used for ordination through principal coordinates analysis [37]. These matrices were also used to determine differences in beta diversity by analysis of similarity (ANOSIM) [38]. Bonferroni corrections for multiple comparisons were performed for ANOSIM analyses. The SplinectomeR R package [39] and its permuspliner functions were used to assess significant differences in the longitudinal abundances of bacterial genera between restored and unrestored groups using standard 999 permutations. One-way unpaired equal variance Student's T-tests were conducted to assess statistically significant differences in concentrations of fecal Lcn-2. All statistics were evaluated at α = 0.05 unless otherwise corrected for multiple comparisons.

Microbial Community Composition and Diversity between Restored and Unrestored Groups Are Longitudinally Different
We longitudinally compared the microbial community changes following AOM/DSS administration between unrestored (no HMT) and restored (HMT from normal Balb/c fecal bedding swap once weekly during weeks 8, 9, and 10) groups. Mouse fecal pellets were longitudinally collected three times weekly for 17 weeks from Balb/c mice in unrestored n = 5 and restored n = 5 groups ( Figure 1A). The overall mean percentage of the relative abundance of bacterial genera in restored and unrestored groups was relatively stable and similar ( Figure 1B). Longitudinally, the microbial communities in mice with restored microbiota had significantly lower Shannon index values than unrestored mice (p = 0.035; Figure 1C); however, no significant differences were observed in Chao1 (p = 0.084; Figure 1D). abundance of bacterial genera in restored and unrestored groups was relatively stable and similar ( Figure 1B). Longitudinally, the microbial communities in mice with restored microbiota had significantly lower Shannon index values than unrestored mice (p = 0.035; Figure 1C); however, no significant differences were observed in Chao1 (p = 0.084; Figure  1D).  Next, we sought to investigate the difference in community composition between restored and unrestored groups using Bray-Curtis dissimilarity. Samples in both treatment groups were considered as "early" (weeks 0-7; before microbiota restoration) and "late" (weeks 8-18; at the time of and following HMT). Principal coordinates analysis (PCoA) showed significant early and late changes in the microbiome of Balb/c mice treated with AOM/DSS ( Figure 1E). The unrestored group had significant early and late changes (ANOSIM R = 0.499, p < 0.001) within the group ( Figure 1E). Similarly, we also observed significant early and late changes in the restored group (ANOSIM R = 0.261, p < 0.001). Microbiota composition also differed significantly between experimental groups among the later "late" samples, but not samples collected prior to week 8 (R = 0.530 and 0.081, p < 0.001 and p = 0.01; Bonferroni-corrected α = 0.008; Figure 1E) Spearman correlation was used to identify bacterial genera associated with the late differences in the microbial community between unrestored and restored groups ( Figure 1E). Akkermansia, Clostridium_clade XVIII, and members of the order Clostridiales that could not be further classified were compositionally correlated with the restored group ( Figure 1E). This was an interesting observation as Akkermansia and Clostridium_XVIII have been implicated as having probiotic properties and are associated with reduced intestinal inflammation through production of short-chain fatty acids such as butyrate [40,41].
In contrast, microbial communities in the restored group were correlated with greater relative abundances of the bacterial genera Alistipes, Anaeroplasma, and Ruminococcaceae, as shown in Figure 1E. We were unable to classify Rumminococcaceae further in our 16S rRNA analysis. Interestingly, Ruminococcus gnavus is associated with Crohn's disease, likely through the ability of R. gnavus to synthesize and secrete glucorhamnan polysaccharides, which can lead to TNFα secretion by dendritic cells [42]. Additionally, the increased abundance of potentially pathogenic Alistipes is consistent with recent investigations of AOM/DSS-induced cancer [43]. These data suggest that HMT altered the microbial composition in the restored group compared to the unrestored group, potentially shifting toward a less inflammatory community composition.

Longitudinal Differences in the Microbiome between Restored and Unrestored Groups Is Significant
Next, we sought to investigate longitudinal changes in bacterial genera between restored and unrestored groups using the SplinectomeR permuspliner function, which can be used to assess longitudinal microbiome data [39]. Significant longitudinal differences were observed between groups among the genera Alistipes (p = 0.032), Akkermansia (p = 0.001), Anaeroplasma (p = 0.032), Ruminococaceae (further unclassified) (p = 0.001), Clostridiales (further unclassified) (p = 0.001), and Clostridium_XVIII (p = 0.001) (Supplemental Figure S1A-F). Significant longitudinal differences in the Shannon index (p = 0.035) between groups were also observed (Supplemental Figure S1G). Notably, the longitudinal significant difference between unrestored and restored is a somewhat conflicting result as previous work has shown that alpha diversity has decreased longitudinally in AOM/DSS models of intestinal inflammation.
The longitudinal abundance of Akkermansia was significantly different between unrestored and restored groups following fecal restoration (Supplemental Figure S1B). For Akkermansia, the mean relative abundance was significantly greater in the restored group compared to the unrestored in weeks 9-17. This was also reflected in our Bray-Curtis dissimilarity and Spearman correlation analysis in Figure 1E. Many bacterial genera identified from Spearman correlations were also longitudinally altered between unrestored and restored groups in SplinectomeR permuspliner analysis.

Intestinal Inflammation Marker Lipocalin-2 Is Decreased in Restored Balb/c Mice
To assess differences in intestinal inflammation before and after HMT in AOM/DSStreated animals, we utilized a semi-quantitative sandwich ELISA (R&D systems, Minneapolis, MN) to probe fecal concentrations of lipocalin-2 (Lcn-2) [44,45]. Fecal Lcn-2 concentrations are a non-invasive marker of intestinal inflammation. Two time points were used before HMT (weeks 2 and 7), and two time points post-HMT (weeks 10 and 15) were selected to quantify fecal concentrations of Lcn-2 in restored and unrestored groups. In weeks 2 and 7, we did not observe significant differences in fecal Lcn-2 concentrations (week 2, 136.1 ± 8.0 pg/mL and 123.7 ± 9.8 pg/mL, p = 0.082, n = 3) (week 7, 171.1 ± 21.7 pg/mL and 154.2 ± 17.2 pg/mL, p = 0.154, n = 3) between unrestored and restored animals ( Figure 2A). Post-HMT, we did observe significant decreases in fecal Lcn-2 concentrations (week 10, 206.1 ± 20.9 pg/mL and 139.4 ± 17.8 pg/mL, p = 0.007, n = 3) (Week 15, 335.4 ± 16.2 pg/mL and 252.4 ± 27.3 pg/mL, p = 0.005, n = 3) in unrestored and restored animals ( Figure 2A). These data suggest that HMT could potentially reduce Lcn-2 concentrations and could have transiently reduced intestinal inflammation in restored animals compared to unrestored counterparts.  Despite significant differences in fecal Lcn-2 concentrations, we did not observe any pathological differences at the experiment endpoint when looking at the hematoxylin and eosin staining of fixed colon tissues ( Figure 2B). Both restored and unrestored animals at the experiment endpoint presented histology ranging from high-grade dysplasia to high-grade intramucosal adenocarcinoma. Additionally, restored and unrestored animals presented with acute and chronic inflammation ranging from focally mild to focally moderate and focally marked. These observations also align with attempted colonoscopy of unrestored and restored animals in the early stage. However, a colonoscopy was impossible at late stages due to bloody stool and inflammation within the colon. Although there were no significant histological differences at the experiment endpoint, we observed significant differences post-HMT in fecal Lcn-2 concentrations. This observation led us to consider that the protective effects of HMT were likely transient, and more frequent treatment with HMT could impact pathology between groups.

Bacterial Community Compositions Are Similar between Balb/c and C57BL/6 in Late Restored Groups
We previously observed altered longitudinal fecal microbiomes of restored Balb/c mice compared to unrestored mice. Considering the variable fluctuations in microbiome analysis, we questioned whether our observations were specific to Balb/c. We used 3-4-week-old Balb/c and C57BL/6 mice in the AOM/DSS model to test this. Additionally, we included a group of 3-4-week-old C57BL/6 mice treated with an antibiotics cocktail of vancomycin (0.05 g/L), metronidazole (0.025 g/L), and streptomycin (0.2 g/L) to deplete mouse microbiomes as a control to ensure our longitudinal data acquisition and analysis pipelines were working appropriately. Interestingly, antibiotics-treated mice observed lower-grade dysplasia compared to unrestored and restored counterparts, which was in line with previous studies [20]. Fecal restoration via HMT was again performed during weeks 8, 9, and 10.
Longitudinally stable mean relative abundances of genera between Group 6A (Balb/c), Group 6B (C57BL/6 unrestored), Group 7 (C57BL/6 +Antibiotics), and Group 8 (C57BL/6 restored) were observed ( Figure 3A,B). These data from C57BL/6 were consistent with our previous observation in Balb/c mice. As expected, we observed depleted alpha diversity in Shannon and Chao1 indices for mice receiving antibiotics compared to restored and unrestored groups ( Figure 3C and Supplemental Figure S2). Similar alpha diversity was observed between the restored and unrestored groups, which was not significant when compared using ANOSIM. Interestingly, we did not notice decreasing alpha diversity in restored C57BL/6 mice compared to unrestored mice, as was observed in Balb/c mice.
Next, we used Bray-Curtis dissimilarity and PCoA to compare restored and unrestored groups' early and late microbial compositions. The early compositional signature between Balb/c and C57BL/6 unrestored animals differed significantly (ANOSIM R = 0.833, p < 0.001; Figure 3D). However, unrestored mice showed more similarity in late microbial compositions between Balb/c and C57BL/6 mice (ANOSIM R = −0.101, p = 0.723; Figure 3D). These observations suggested that even though C57BL/6 and Balb/c unrestored mice had significantly different initial microbial compositions after AOM/DSS treatment, similar compositions were observed between strains of mice when left untreated.  Furthermore, we also observed significant differences between late restored C57BL/6 and unrestored C57BL/6 mice (R = 0.897, p < 0.001; Figure 3D). Spearman correlation was used to investigate bacterial genera associated with differences in ordination position. Interestingly, we noticed similar taxa (i.e., Bacteroides, Clostridium_XVIII, and Akkermansia) were correlated with late compositions of the restored group ( Figure 3D), similar to our prior observations in Balb/c mice. Additionally, unrestored C57BL/6 communities were associated with a greater relative abundance of Anaeroplasma, Alistipes, Lachnospiraceae, Bacteroidales (not further classified), and Ruminococcaceae (not further classified). Notably, the genera Ruminococcaceae, Alistipes, and Anaeroplasma were found in greater abundance in late samples among untreated Balb/c and C57BL/6 mice. Similarly, Akkermansia and Clostridium_XVIII were found at greater relative abundances in the restored microbiota of both mouse genotypes.

Inflammatory and Pathological Signatures between Balb/c Mice and C57BL/6 Mice Are Different
We again used ELISA to probe fecal concentrations of Lcn-2 pre-HMT (weeks 2 and 7) and post-HMT (weeks 10 and 15) in restored, unrestored, and antibiotics-treated animals to observe alterations in intestinal inflammation between groups. We did not observe Furthermore, we also observed significant differences between late restored C57BL/6 and unrestored C57BL/6 mice (R = 0.897, p < 0.001; Figure 3D). Spearman correlation was used to investigate bacterial genera associated with differences in ordination position. Interestingly, we noticed similar taxa (i.e., Bacteroides, Clostridium_XVIII, and Akkermansia) were correlated with late compositions of the restored group ( Figure 3D), similar to our prior observations in Balb/c mice. Additionally, unrestored C57BL/6 communities were associated with a greater relative abundance of Anaeroplasma, Alistipes, Lachnospiraceae, Bacteroidales (not further classified), and Ruminococcaceae (not further classified). Notably, the genera Ruminococcaceae, Alistipes, and Anaeroplasma were found in greater abundance in late samples among untreated Balb/c and C57BL/6 mice. Similarly, Akkermansia and Clostridium_XVIII were found at greater relative abundances in the restored microbiota of both mouse genotypes.
Even though we observed a statistically significant decrease in fecal Lcn-2 concentrations in week 10 following HMT, this did not influence the pathological observations at the experimental endpoint between restored and unrestored animals ( Figure 4B). Mice in restored and unrestored groups at the experiment endpoint presented with high-grade dysplasia to high-grade intramucosal adenocarcinoma. Furthermore, both restored and unrestored animals presented with acute inflammation, ranging from minimal to focally moderate, and chronic inflammation, ranging from mild to focally moderate. These observations suggested that any beneficial effect from HMT, as determined by concentrations of fecal Lcn-2, was likely transient and did not influence overall pathological outcomes between unrestored and restored groups. However, mice receiving antibiotics were absent of dysplasia and showed only minimal acute inflammation and mild chronic inflammation. This was consistent with fecal Lcn-2 concentrations as antibiotics-treated animals had significantly reduced levels of Lcn-2 compared to restored and unrestored animals throughout all sampled time points. These data further support a role or niche of the gut microbiota in regulating intestinal inflammation as mice treated with antibiotics presented with less advanced-stage disease than unrestored and restored groups.

Alistipes, Akkermansia, and Anaeroplasma Are Longitudinally Altered in C57BL/6 Restored and C57BL/6 Unrestored Mice
Next, we investigated longitudinal changes in bacterial genera between C57BL/6 restored and unrestored groups using the SplinectomeR permuspliner function [39]. We observed significant longitudinal differences in Alistipes (p= 0.001), Akkermansia (p = 0.001), Bacteroides (p = 0.001), Erysipelotrichaeae (p = 0.001), and Anaeroplasma (p = 0.001) between unrestored and restored groups ( Figure 5 and Supplemental Figure S3). These bacterial genera were also identified in our Bray-Curtis dissimilarity analysis and were correlated using Spearman correlation. Additionally, we observed longitudinally significant differences in Bacteroides between unrestored C57BL/6 mice and restored C57BL/6 mice. However, this was not observed in Balb/c mice. Compared to healthy controls, Bacteroides are commonly observed in patients with irritable bowel disease [46,47]. When comparing Balb/c and C57BL/6 from all experiments, we consistently saw longitudinal increases in Akkermansia in the restored group compared to the unrestored, along with consistent observations of an increased abundance of Alitipes and Anaeroplasma in unrestored compared to restored Balb/c and C57BL/6 mice. 4.6. Alistipes, Akkermansia, and Anaeroplasma Are Longitudinally Altered in C57BL/6 Restored and C57BL/6 Unrestored Mice Next, we investigated longitudinal changes in bacterial genera between C57BL/6 restored and unrestored groups using the SplinectomeR permuspliner function [39]. We observed significant longitudinal differences in Alistipes (p= 0.001), Akkermansia (p = 0.001), Bacteroides (p = 0.001), Erysipelotrichaeae (p = 0.001), and Anaeroplasma (p = 0.001) between unrestored and restored groups ( Figure 5 and Supplemental Figure S3). These bacterial genera were also identified in our Bray-Curtis dissimilarity analysis and were correlated using Spearman correlation. Additionally, we observed longitudinally significant differences in Bacteroides between unrestored C57BL/6 mice and restored C57BL/6 mice. However, this was not observed in Balb/c mice. Compared to healthy controls, Bacteroides are commonly observed in patients with irritable bowel disease [46,47]. When comparing Balb/c and C57BL/6 from all experiments, we consistently saw longitudinal increases in Akkermansia in the restored group compared to the unrestored, along with consistent observations of an increased abundance of Alitipes and Anaeroplasma in unrestored compared to restored Balb/c and C57BL/6 mice.

Discussion
Interactions between the microbiota that contribute to host intestinal inflammation are increasingly being investigated. Associations have been reported between bacterial genera and disease phenotypes in inflammatory bowel diseases and CRC [7,48,49]. Many variables render causative associations between the microbiome and disease pathology challenging [14][15][16]. Our study aimed to use an established mouse model of id-CRC [17,18] to longitudinally track shifts in the microbiome of AOM/DSS-treated mice and assess how fecal restoration influences longitudinal microbial signature and disease pathology. Our study demonstrated that longitudinal compositional changes over 17 weeks are similar between Balb/c and C57BL/6 mice following AOM/DSS treatments (Figures 1 and 3). Additionally, we demonstrated that horizontal microbiome transfer (HMT) restoration from untreated mice through bedding swaps led to consistent longitudinal increases in Akkermansia in Balb/c and C57BL/6 mice compared to their respective unrestored controls, along with consistent observations of an increased abundance of Anaeroplasma and Alistipes in unrestored Balb/c and C57BL/6 mice.
Clinical investigations have reported that the relative abundance of Akkermansia is decreased in stool samples of patients with active ulcerative colitis compared to patients with quiescent ulcerative colitis and healthy controls [50]. We reason that observing consistent longitudinal increases in Akkermansia in restored mice compared to unrestored mice may represent a more quiescent phenotype following fecal restoration. Additionally, it has been reported that oral administration of A. muciniphila strain BAA-835 significantly ameliorated intestinal inflammation following DSS-induced inflammation and was dependent on NLRP3 expression [51]. Others have shown that A. muciniphila strain BAA-835 can reduce inflammatory cytokine expression and reduce mucosal inflammation in DSS models of inflammation [39,40]. Knowing that Akkermansia can reduce inflammatory cytokine expression could explain why we observed decreased fecal Lcn-2 concentrations in week 10 following HMT in both genotypes of mice (Figures 2 and 4). Nevertheless, these temporal increases in Akkermansia in restored groups did not yield a difference in pathological outcomes between restored and unrestored groups, so this effect was likely transient. More frequent administration of HMT could potentially yield alterations in pathological outcomes; however, without further investigation, the overall impact of HMT on disease pathology remains unknown.
Additionally, mice treated with antibiotics observed less dysplasia and inflammation than restored and unrestored animals. Our observations corroborate previous investigations and suggest that microbiota regulates intestinal inflammation, which can be seen in mice lacking a microbiome exhibiting less aggressive disease [20]. Furthermore, our Lcn-2 data support different pathological outcomes in antibiotics-treated mice as, throughout all sampled time points observed, there were significantly lower concentrations of Lcn-2 compared to unrestored and restored animals ( Figure 4). We also observed significantly lower concentrations of Lcn-2 in C57BL/6 mice compared to Balb/c mice across treatment conditions. This supports our observations from hematoxylin and eosin staining, where Balb/c mice tended to present more frequently with intramucosal adenocarcinoma and focally marked inflammation than C57BL/6 counterparts. In general, C57BL/6 mice presented with less aggressive disease than Balb/c mice. Our studies suggest that these observations should be considered in future investigations of id-CRCs, as choosing an appropriate model can influence pathological outcomes. Nevertheless, these observations suggest that fecal restoration could contribute to an altered colonic inflammatory response through longitudinal increases in Akkermansia.
We also observed significant longitudinal increases in Alistipes and Anaeroplasma in the unrestored compared to the restored group in Balb/c and C57BL/6 mice (Figures 1 and 3 and Supplemental Figures S1 and S3). The observation of an increased abundance of potentially pathogenic Alistipes is consistent with recent investigations of AOM/DSSinduced CRCs [43]. Additionally, more mechanistic insight has been gained by studies showing that Alistipes abundance is sufficient to induce colitis in IL10 -/mice and can lead to tumorigenesis [52]. Alistipes has also been identified as a predominant genus of CRC patients from metagenome-wide association studies and is more abundant in carcinomas compared to adenomas and healthy controls [53]. Our observation of an increased relative abundance of Alistipes in unrestored groups compared to restored groups could reflect an increased abundance of Alistipes contributing to id-CRCs. Furthermore, Alistipes abundance has been implicated in the response to immune checkpoint inhibitors (ICIs) in preclinical models [54]. Iida et al. showed that antibiotic-treated animals bearing MC38 tumors observed resistance to anti-IL10 antibody treatment [55]. Additionally, Alistipes and Ruminoccocus were positively correlated with TNF-dependent anti-tumor immune activation in response to immunotherapy [55]. Routy et al. also implicated Alistipes indistinctus in response to ICIs and other commensal gut microbes in reversing resistance to ICIs [56]. Routy et al. administered Alistipes indistinctus to tumor-bearing animals and reversed resistance to ICIs [56]. Further studies are needed to fully illuminate how genera can modulate immune responses in the context of ICI resistance to enhance response rates in different malignancies so this knowledge can be translated into clinics.
We also observed consistent increases in the relative abundance of Anaeroplasma in unrestored compared to restored groups for both murine backgrounds. It has been reported that Anaeroplasma is enriched in mice receiving AOM/DSS compared to healthy controls, [21] and our work demonstrates a similar profile. However, conflicting results have also been reported. For instance, Anaeroplasma was shown to be decreased in the stool of mice with a common CRC driver gene APC mutation compared to the wild type [57]. However, different inflammatory niches between models could explain this observed difference in the abundance of Anaeroplasma. Further, it has been reported that Anaeroplasma strongly correlates with intestinal IgA and TGF-B secretion and regulates intestinal inflammation [58]. This is also conflicting as IgA increases in patients with IBD [59], while TGF-B is a known immunosuppressive cytokine [60].
Lastly, we also observed significant longitudinal increases in Ruminococcaceae in Balb/c restored mice compared to unrestored Balb/c but not C57BL/6 mice. Interestingly, Ruminococcus gnavus is associated with Crohn's disease, likely through the ability of R. gnavus to synthesize and secrete glucorhamnan polysaccharides, which can lead to TNFα secretion by dendritic cells [42]; it is also enriched in CRC patients compared to normal controls [61]. We did not see this recapitulated in Balb/c and C57BL/6 mice. However, we reason this as evidence that fecal restoration longitudinally shifted the microbiome and modulated fecal Lcn-2 concentrations between restored and unrestored mice between genotypes.

Conclusions
This study investigated how HMT through cage bedding swapping with untreated animals could influence the microbiome and pathological outcomes of Balb/c and C57BL/6 mice treated with AOM/DSS to mimic id-CRC. We observed consistent and significant increases in Akkermansia in restored groups compared to unrestored counterparts in both genotypes. At the same time, we observed consistent increases in Anaeroplasma and Alistipes in unrestored animals compared to restored counterparts across genotypes. These observations were correlated with reduced fecal Lcn-2 concentrations post-HMT in restored mice compared to unrestored counterparts. We also observed differences in dysplasia and inflammatory signatures between Balb/c and C57BL/6 mice, which should be considered in future investigations.
In comparison, fecal restoration through HMT did not yield pathologically different outcomes between restored and unrestored animals. Temporal differences in fecal Lcn-2 concentrations were observed post-HMT. Future studies should seek to illuminate the direct microbiota-host mechanisms of Akkermansia, Anaeroplasma, and Alistipes for their roles in regulating intestinal inflammation and CRC development in different model systems.