Familial adenomatous polyposis (FAP) is inherited as an autosomal dominant trait, characterized by numerous adenomas in the colon and rectum that progress to colorectal cancer by the 4th decade of life [12]. Most patients with FAP harbor a germline mutation in the adenomatous polyposis coli (APC) gene. Evidence is emerging for the role of microbiota in FAP based on pre-clinical and clinical metagenomic studies along with pre-clinical studies using gnotobiotic hosts [7]. A detailed exploration of the spatio-temporal changes in the microbiome in the adenomatous tissues versus the stool samples, however, has not been achieved. We performed a high-throughput sequencing and bioinformatics analysis to characterize the tissue and stool microbiota of the FAP patients and compared them with healthy controls. The intent of the current study was to investigate if differences exist in microbiota composition between polyps and stool samples and whether the longitudinal changes in the microbiome may indicate susceptibility to developing colon cancer later in life.

In previous studies, differences in community composition between cancerous tissues and surrounding areas have led to a bacterial driver-passenger model for CRC [13, 14] observed changes in rectal mucosal bacterial communities of adenoma patients as well as in healthy controls and suggested that rectal mucosal bacterial composition may reflect the presence of adenoma-specific bacterial communities. We describe for the first time, differences in microbiota composition between polyps and adjacent non-polyp mucosa and re-affirm that the bacterial populations in feces and mucosa are distinct and may, in fact, differ in how they are enriched and/or distributed over a period of time. Firmicutes, Bacteroidetes, Actinobacteria, and Proteobacteria were the dominant phyla in healthy controls, similar to previous studies on gut bacteria [15]. A large decrease in Firmicutes, Bacteroidetes, and Actinobacteria coincided with the relative expansion of Proteobacteria in the adenomatous or synchronous tissues of FAP patients. These changes in Proteobacteria phyla in tissues/polyps in years one to three were much higher than those recorded in the stool. Proteobacteria have been shown to be enriched in adenoma compared with non-adenoma tissue from the same patient [16] and compared with tissue from healthy volunteers [17]. They are reported as a major phylum in colonic biofilms from FAP patients compared to healthy individuals [7]. Several larger studies reported significantly higher carriage or abundance of Proteobacteria in CRC [18, 19]. In our cohort, as with the observed increased abundance of Fusobacteria, these changes were not observed in patient or control stool samples suggesting distinct evolution of tissue and polyp microenvironment. Our observation of increased abundance of Fusobacteria in polyp tissue over time compared with healthy tissue or stool – where it was barely detected, is consistent with an increase in the prevalence and/or abundance of Fusobacterium nucleatum reported in the colorectal cancer tissue and fecal samples compared to individuals with colorectal polyps [20] and, may have important implications in CRC development in FAP. Fusobacteria is an adherent and invasive Gram-negative anaerobic bacterium usually residing in the oral cavity and associated with periodontal disease [21]. It is a potential risk factor for CRC progression [22, 23] and a higher abundance of F. nucleatum in CRC is associated with shorter survival [24]. Mechanistically, Fusobacterium induces the Wnt signaling through multiple mechanisms. Specifically, F. nucleatum interacts with TLR4 inducing β-catenin phosphorylation by PAK-1. The Wnt/β catenin pathway is also activated through F. nucleatum-produced FadA adhesin binding to E-cadherin, resulting in up-regulation of Annexin A1 [23, 25]. Fusobacterium modulates CRC proliferation through Toll-like receptor four signaling to MYD88, leading to activation of the Nuclear Factor-κB (NF-κB) [26], resulting in increased TNF-α, IL-6, IL-8, and miR-135b. We also observed mucosa-associated E. coli-Shigella cluster to be more prevalent in P1–P3 samples, which has been shown to encode cyclomodulin, vital for mutational changes in colon crypt cells [27]. Thus, increases in Fusobacteria and E. coli_Shigella clusters, especially in year three cohorts, further delves into the significance of mucosal dysbiosis in the evolution of CRC in FAP patients.

As a member of the Clostridium leptum group, Faecalibacterium prausnitzii could represent the beneficial commensal bacteria, and previous studies are consistent with the anti-inflammatory properties of this bacterium [28]. We observed a decline in Faecalibacterium belonging to Firmicutes phyla in the patients' tissue and stool polyps in years one to three (P1–P3) [29]. Faecalibacterium negatively correlates with inflammatory bowel disease activity and is relatively overexpressed in healthy tissue compared with CRC-polyps [30]. Mechanistically, this may be related to chemoprotective butyrate production, which correlates with dietary fiber intake [31]. Butyrate appears to induce tumor apoptosis through the expression of E-Cadherin [32]. Faecalibacterium prausnitzii spp., which promote short-chain fatty acids production (SCFA), including butyrate, are decreased in patients with advanced colorectal adenoma compared with controls [33]. Interestingly, in a murine model, a decline in other species including Bacteroides uniformis and Bacteroides vulgatus in cohorts encompassing polyps, synchronous tissues, and stools, correlated with disease progression [34]. Bacteroides uniformis utilizes the O-glycans covalently attached to mammalian mucin and serves as a mucin-degrader that predominantly colonizes the mucosal surfaces, thereby interacting with the host [35]. A B. vulgatus strain was shown to protect against E. coli-induced colitis in IL-2−/− mice [38], while IL-10−/− mice mono-associated with pig isolates of B. vulgatus had significantly reduced colitis-associated colon tumor multiplicity compared with conventional IL-10−/− mice [36]. Thus, a decline in the levels of these species suggests that local microbiota disturbances may accompany disease progression.

When comparing the alpha diversity indices among groups using Student's t-test, we observed that both chao1 and Simpson indices declined in P3. Specifically, alpha diversity indices changed from low (P3) to high (S3) with less variability, suggesting that distinct changes in microbiome exist at these sites and that the development of adenomas/polyps may itself contribute towards microbiota imbalances. When we further delved into delineating the core microbiome at each site, we discovered ~ 1800 overlapping OTUs that were shared by the four groups in a Venn diagram regardless of whether it had high or low abundance. Yet, several unique OTUs were also present at each site. This combination of unique or overlapping core microbiomes may be integral to the development of CRC in FAP patients. The community structures of adenomatous/polyp tissues, when compared with stool samples, revealed hierarchical clustering upon PCoA analysis wherein, polyp and stool samples tended to cluster separately in FAP patients, particularly in year three (P3 vs. S3). These findings were further corroborated by ANOSIM data wherein, inter-group differences were significantly greater than intra-group variation in matched samples and by UPGMA clustering algorithm that clearly revealed segregation of bacterial communities in the tissue and stool samples. These site-specific alterations in the distribution of microbiota, whether causal or consequential, may dictate the kinetics of adenoma development as a prelude to CRC. In particular, we found that increased relative abundance of potential opportunistic pathogens such as Alpha/beta-Proteobacteria, E. coli/Shigella, Fusobacteria, etc., which contribute towards changes in intestinal homeostasis, might display robust inflammatory infiltration and directly or indirectly increase the risk of adenoma development.

We used machine learning to further provide insights into which bacterial populations are unique to patients with and without polyposis. The species that best discriminated between (stool from) subjects and controls were Archaea, Micrococcus luteus, and Eubacterium hallii. Euryarchaeota, the principal Archaea phylum in the intestinal microbiome, was one of the top 10 phyla in our combined OTU analysis. Euryarchaeota is highly diverse and includes methanogens, which in turn have been shown to be depleted in CRC [37] coincident with a progressive increase in halophilic spp. in stool from controls, adenoma then CRC. Furthermore, Archaea enrichment has been shown to relate to changes in the alpha-diversity observed in CRC [38]. Our observations suggest that perturbations in the Archaea subpopulation of the fecal microbiome may indeed modulate adenoma progression and may constitute useful biomarkers of syndromic adenomas. Another discriminant species in our ML analysis, Eubacterium hallii, distinguished stool from patients with and without polyposis (controls). Eubacterium hallii is recognized as a SCFA producing commensal and is decreased in diverse disease states, including inflammatory bowel disease and colorectal cancer [39].

Eubacterium hallii has been shown to detoxify carcinogenic heterocyclic aromatic amines present in processed meats [40]; its suppression, along with other SCFA producing species, has been implicated mechanistically in the relationship of high animal fat consumption in gut inflammatory and neoplastic processes [41]. Specifically, E. hallii has been proposed as a candidate next-generation probiotic [42].

The significance of Micrcoccus luteus is unclear; M. luteus is a well-described opportunistic pathogen usually in the context of immunocompromised hosts and has, to date, not been implicated in colorectal adenoma or cancer.

Thus, the realization that the microbiome modulates colorectal cancer risk introduces the possibility of altering the microbiome to change the risk of malignancy. Probiotics including Lactobacillus and Bifidobacterium have been shown to decrease adenoma formation in murine models of FAP [43, 44]. The mechanisms underlying the antitumor effects of probiotics include modulation of inflammatory pathways including NF-κB and downregulation of β-catenin; mechanisms that our observations suggest are active in the progression of adenomatous polyp evolution in our population. Specifically, for example, Lactobacillus supplementation has been shown to increase Roseburia in the ApcMin/ + mouse model reversing one of the characteristics of polyp evolution observed in our study [44]. The polyp-microbial interrelationships we observed in patients with FAP may be interpreted in the context of potential pre- and probiotic therapy, including a putative role for E. hallii.

Our study has several limitations. Given the rarity of the underlying diagnosis of pediatric pre-colectomy FAP, accrual of a robust cohort was difficult. Maintaining regular surveillance and sampling as well as identifying and recruiting a suitable control was challenging. Accordingly, we could not obtain all planned samples over the three-year study period in any of our patients. This does not detract, however, from the validity of observations between different sample groups during the same year. Recent genetic testing was not available in all subjects, this could account for some of the subjects having been tested several years ago, with less sensitive testing modalities, reported as APC mutation negative. Our choice of controls inherently presented a challenge insofar as cohabiting, unaffected, same-gender individuals would be preferred but was unfeasible in view of siblings, if any, in most instances being either younger and untested, or older and not cohabiting, we therefore consciously chose cohabitation, being the stronger determinant of similarity [45], and therefore necessarily the study design favored the inclusion of unaffected parents skewing the mean age of the control population. Ultimately, we recognize that any choice of control poses potential pitfalls as many factors [46, 47] influence the intestinal microbiome and therefore any realistic control cannot be perfectly matched. This may have a bearing on the observed differences between subject and control stool. Similarly with the challenge of control samples, we understand the small sample sizes might limit our interpretation of our results. However, we minimized the error in our study due to effect size, by (1) adopting robust comparison of between and within sample groups for PERMANOVA analyses [48]; (2) ensuring the samples satisfied the t-test assumptions [49]; (3) ensuring the LEfSe analyses are performed by ranking based on magnitude of variation and not statistical significance [11].

As implied above, the principal limitation of our study is the attrition in samples in years two and 3. We have therefore adopted a strategy to separately analyze year one where the bulk of the samples are, and then we examined the findings from longitudinal analyses of the small subset that have longitudinal samples. Our present work provided the first step into the understanding of the microbiome shifts due to FAP, and future work could be performed with more samples to provide further insights.

Our study further highlights the limitation of stool-sampling strategies in defining the polyp microenvironment in adenomatous polyposis and by extrapolation sporadic polyp. Although some interrelatedness between polyp, healthy mucosa, and stool microbiome was apparent, it is exceedingly challenging to correlate polyp microenvironment changes from stool samples. Future research may focus on whether the detection of certain bacterial concentrations within stool or biopsied polyps could serve as adjuncts to current screening modalities to help identify higher-risk patients.