Tumor development results from the interplay between cell-intrinsic factors -such as gene mutations, chromosome or gene expression alterations- and cell-extrinsic factors -namely microenvironmental factors linked to immune system activity. More recent studies indicate that some gut microbial species are involved in the onset and development of colorectal cancer (CRC), and the correlation between specific gut microbiome species and CRC have been exploited to define “microbiome signatures” distinguishing CRC patients and healthy controls. Moreover, ongoing studies have been exploring the potential of microbiome signatures for CRC staging.
In a recent paper by Gianmarco Piccinno et al., the authors, supervised by Nicola Segata -Group Leader of the department of experimental oncology of IEO and full professor at the university of Trento-, Curtis Huttenhower, Alessio Naccarati, and Eva Budinska, exploited metagenomic approaches to analyze, at high resolution, the microbiome species involved in the process of benign-to-malignant transformation, demonstrating that the presence of typically oral species in the gut are associated with CRC. They also showed that species-level analyses of the gut microbiome can distinguish cancer patients from healthy subjects quite accurately, as well as the different tumor stages. Indeed, they found species/level biomarkers in patients’ gut via stool metagenomic analyses and demonstrated that machine learning-based tools can be leveraged as disease-discovery tools.
Although the study was not set up to establish causal relationships that could enable in the future new therapeutic approaches, these results highlight that gut microbiome analyses can be leveraged as predictive tools, indicating that a gut microbiome-based non-invasive screening of gut dysbiosis could be soon used clinically for the early identification of CRC patients.
Prof Segata, how far we are from translating these findings to the clinical setting?
“From the methodological and technological viewpoint, we now have everything to offer a cost-effective metagenomic testing for colorectal cancer screening programs, and we are just performing the latest validation in a large cohort to hopefully confirm the accuracy of the predictions”.
By integrating existing -and new- datasets of metagenomic sequencing of stool samples from CRC patients (some of them were part of the ONCOBIOME initiative; see newsletter n. 063), with information on tumor stage and location (right vs left colon) and healthy controls, the authors assembled the largest existing database on CRC, allowing them to collect robust results. By using their previously developed computational tool -MetaPhlAn 4-, samples were characterized at high resolution, down to the single species of the microbiome.
Analyses of the overall gut microbiome richness (namely, the number of different species in the gut) showed that only the fraction of typically oral species present also in the gut −oral-to-gut introgression score− appeared different according to the different tumor stages and location. Indeed, while no significant correlation was found between richness of gut microbiome species and CRC stage, they observed differences of oral-to-gut introgression in CRC patients vs controls, in early vs late stage, in right vs left colon.
Some gut microbiome species correlate with CRC stage and location. Detailed analyses of the gut microbiome species revealed remarkable differences in control vs CRC microbiomes (confirming, in this large cohort, previous data). They found differences, in terms of species abundance, in the gut microbiome of patients with benign vs malignant lesions (specifically, ctrl/adenomas vs early/late carcinoma), highlighting that microbiomes start to differentiate already in the very early phases of the disease. Moreover, these analyses distinguished the different stages of tumor progression (namely, from stage I to stage IV). Some bacterial species were abundant at all tumor stages; for others, their abundance started to increase at stage I and kept growing towards stage IV. They also identified the specific species differentially abundant in the different tumor stages: 17 species were differentially abundant in late CRC, only 4 in early stage.
They also found few differentially abundant species in fecal samples from left vs right CRC, highlighting that the tumor location slightly affects gut microbiome composition.
Analyses of the microbial genes enriched in the gut of CRC patients as compared to healthy controls revealed the high presence of the cutC gene (key enzyme in the pathway responsible for the conversion of choline into trimethylamine); over 200 metabolic pathways associated with CRC; several enzymes; pathways associated with ammonia production. Previous studies have highlighted increased ammonia levels in CRC microenvironment, suggesting that the gut microbiota may be involved in ammonia regulation in the tumor microenvironment. Of note, previous studies showed that ammonia can reduce T cell activity (cell exhaustion).
Notably, they found several CRC-related gut microbiome species that were also correlated with cardiometabolic risk, Crohn disease and immune diseases.
Machine learning tools to predict CRC presence. Their study also confirmed that machine learning (ML) tools leveraging metagenomic gut microbiome data can be used as predictive tools, thus potentially exploitable for non-invasive CRC screening. Indeed, by testing three different previously developed ML tools, they showed that if trained on adequately large and diverse datasets, these tools are sufficiently reliable. Interestingly, they found that the predictive power of the ML tools tested largely depended on the presence, in the gut microbiome, of typically oral species, underlining the predictive relevance of the presence of oral species in the gut, for CRC.
CRC biomarker-species. By leveraging their huge dataset, they identified gut microbiome species serving as CRC biomarkers. Indeed, they found that some (bacterial) species were more abundant in CRC and some more abundant in healthy controls (no differences regarding eukaryotic species were scored), and a considerable fraction of these biomarker-species of CRC were typically oral, while no oral species were associated with controls, further highlighting the key impact of the presence of oral bacteria in the gut.