Our study found significant differences in the microbiome between LS with or without CRN and non-Lynch controls and few between LS with CRC from LS without CRN. We found no global differences (alpha or beta diversity) between LS with CRN (either CRC or adenoma) and without CRN. There were taxon specific differences comparing individuals with LS and history of CRC and LS without any CRN, Streptococcus and Actinomyces. Notably, two Streptococcus species, Streptococcus bovis and Streptococcus gallolyticus have been frequently linked with colon cancer [26]. Additionally, Actinomyces has been associated with CRN in a systematic review [27]. However, we also found these same changes when we evaluated by history of surgery. Post-surgical anatomy may be an overriding factor confounding the association between history of colon cancer and microbiome composition as all individuals with a history of CRC underwent colorectal surgery.
While there were no global differences among LS comparing CRN to no history, we found both global and multiple taxon level differences between LS and non-LS controls. LS carriers had a higher alpha diversity compared to non-LS controls. Similarly, there was significant differences in community structure. Veillonella has been associated with inflammatory conditions including Crohn’s disease and hepatic encephalopathy and was enriched in LS. LS carriers can also have dysplastic crypts that may provoke chronic low-level inflammation despite being histologically undetectable. Lower abundances of bacteria regularly associated with being anti-inflammatory including Faecalibacterium and Romboustia were also depleted in LS [28]. These findings could be due to the significant differences in age between these groups. Veillonella was enriched in both LS without CRN and history of CRC compared to non-LS controls.
Because we observed more differences in microbiome composition comparing LS to non-LS controls than comparing LS groups to each other, it was unexpected that the machine learning models classifying 1) LS from non-LS controls and 2) LS-CRC from LS-without CRN performed equally well. The differences in microbiome composition between LS and non-LS controls might suggest that underlying microbiome composition of LS carriers is different potentially due to underlying genetic pathogenetic variants leading to changes in the colorectal epithelium or mucosal immunity. We observed fewer differences in microbiome composition among LS groups according to history of CRN, but the LS-CRC microbiome was different enough from LS- without CRN to allow a machine model to distinguish between these two groups. The observed differences between LS-CRC and LS-without CRN and the ability of our machine learning model to decipher the two groups might also be due to history of colon resection, which was associated with similar differences in microbiome composition. Conversely, we did not observe either global or taxon level differences between LS-adenoma and LS-without CRN and furthermore the model we built to classify these two groups did not perform well.
Other studies have similarly found that the microbiome compositions of subgroups of LS according to CRN history were like each other but different compared to the microbiome composition of non-LS controls. Mori et al. did not observe a difference between LS with CRC and LS with endometrial cancer (GC) but noted a statistically significant difference in the community structure between non-LS controls compared to LS. The authors hypothesized there might be an underlying fecal microbiota pattern associated with LS [16]. Similarly, Lu et al. found no difference in microbiome composition among LS carriers with and without cancer, but observed that LS carriers were enriched in B. fragilis and Parabacteroides distasonis, and Pseudomonadaceae family compared to non-LS controls [17]. Ferrarese et al. found stool microbiome from LS carriers were enriched in Bacteroidetes and Proteobacteria and depleted in in the Firmicutes and Ruminococcaceae compared to non-LS controls [18].
Fewer studies have noted differences in microbiome composition among patients with LS with adenomas and CRC compared to LS with no CRN. Gonzalez et al. found non-significant difference in several genera when they compared LS with cancer to LS without cancer, but sample size was very limited (n = 8).[29] In the largest study to date, Yan et al. found that Clostridiaceae was depleted and Desulfovibrio was enriched in LS with baseline adenomas compared to no adenomas at baseline. They also observed that history of surgery was the most significant contributor to differences in microbiome composition [15].
Only one other study explored whether different LS pathogenic variants were characterized by different gut microbial populations. Yan et al. found that MLH1 and MSH2 mutations carriers were depleted in Clostridiales and MLH1 were enriched in Blautia and Oscillospira [15]. We found that MLH1, MSH2 and MSH6 were depleted in Bifidobacterium while MLH1 and MS2 were enriched in Actinomyces and MSH2 was enriched in Streptococcus compared to PSM2. As noted above, we found similar differences in Streptococcus and Actinomyces when we evaluated by history of surgery. Post-surgical anatomy may also be confounding the association between LS pathogenic variant and microbiome composition as individuals with MLH1 and MSH2 were more likely to be diagnosed with cancer and require curative colorectal surgery compared to PSM2.
While correlation studies in humans have not yet provided evidence that microbiome contribute to cancer development in LS, LS animal models provide evidence the microbiome may drive carcinogenesis in LS. A study conducted in a MSH2 Lynch mouse model (APCMin/MSH2−/−) demonstrated that bacterial derived butyrate might drive hyperproliferation in MMR deficient cecal epithelial cells marked by deregulated beta-catenin activity.[30] By treating these mice with antibiotics or a diet low in carbohydrates, the study authors reduced both butyrate levels and the total number of polyps by 75%. This finding may explain the lack of any protective effect observed in LS carriers randomized to resistant starch despite increasing their levels of butyrate concentration in the CAPP2 randomized control trial in contrast to the protective effect observed in sporadic CRC [31]. Another MSH2 knockout mouse model exposed to conventional microbiome exhibited increased epithelial turnover rates, increased rate of spontaneous mutations in MSH2 deficient crypts and increased microsatellite instability compared to specific pathogen free (SPF mice) [32]. The authors hypothesized that bacterial presence in MMR deficient crypts drives epithelial turnover and subsequent mutations in DNA.
One strength of this study is that it is the largest study of LS to include non-LS controls for comparisons. Prior studies comparing LS patient to non-LS controls were based on small cohorts with less than 10 subjects in each groups making their findings subject to internal and external validity issues. Our similar findings of notable differences between LS and controls with a larger sample size strengthens and confirms this repeated finding.
There were several limitations to our study. The controls were significantly younger compared to patients with LS, which might confound the differences in microbiome composition we observed between these two groups. Age is an important contributing factor in microbiome variation. Furthermore, the study populations were female predominant and therefore our results might not be generalizable. Additionally, we did not collect stool prospectively and our investigation of differences in microbiome composition within LS carriers was based on prior diagnoses of colorectal adenomas and cancer. The stool microbiome is more significant and informative if collected prior to adenoma detection as our hypothesis is microbiome composition may lead to the risk of polyp and cancer formation. As this was a cross sectional study, stool collected might not accurately reflect the composition at the time adenoma or cancer was forming. Another limitation was that the sample size was still too small to do multigroup analyses and adjust for covariates. Larger studies with prospective stool collection are needed to untangle how microbiome protects against or contributes to carcinogenesis in LS.
In conclusion, we found significant global and specific differences in the microbiome composition when we compared individuals with LS to non-LS controls. This finding confirms observations from prior studies done with significantly smaller sample sizes. Further studies investigating how differences in the epithelial biology and immunology between LS carriers and non-LS controls might shape differences in the microbiome composition are warranted. We found no global differences and few specific taxa differences when we compared LS-CRC to LS-without CRN. While animal models suggest microbiome may contribute to colorectal carcinogenesis, no clinical studies in people have documented an association between CRC and microbiome in LS. A longitudinal collection of stool microbiome in LS to detect the contribution of stool microbiome composition to carcinogenesis in LS would be invaluable to investigating the role of microbiome composition in CRC development.