The recognition of microorganisms inhabiting the gut, urogenital system, skin, and lungs of humans has become widely accepted. Various studies provide evidence of microbial infiltration in tumor tissues, underscoring the potential role of the microbiome in cancer progression. Recent scientific inquiries challenge the traditional belief in the sterility of normal internal human tissues, presenting compelling findings about the existence of microbial communities within diverse internal organs. This perspective, though relatively unexplored, aligns with research in other species, such as mice. To delve deeper into the microbial presence in healthy human internal tissues, an analysis was conducted on 13,871 normal RNA-seq samples from 28 tissues obtained from the Genotype-Tissue Expression (GTEx) consortium. Acknowledging certain limitations within the GTEx analysis pipeline, it remains noteworthy that the GTEx consortium provides the most extensive and analytically robust dataset about RNA expression within healthy human tissues. Each sample's sequencing reads, not aligning with the human genome (assembly T2T-CHM13v2.0), underwent analysis using AGAMEMNON, an algorithm for aligning and quantifying microorganism abundance at various taxonomic levels. Reference database included 4034 bacterial, 489 archaea, and 11,259 viral genomes, along with 81 fungal transcriptomes. After cumulative sum scaling (CSS) normalization on species-level microbial abundances and retaining only species present in at least 10% of the samples for each tissue, gradient-boosting machine learning models were trained for each tissue. These models aimed to distinguish each tissue from all others based solely on their microbial profile. To validate the results, 38 healthy living tissue samples (from NCBI project PRJEB4337) were analyzed, as the GTEx samples were derived from post-mortem biopsies. The machine learning models exhibited robust performance for 11 out of 28 tissues (Muscle, Heart, Salivary Gland, Brain, Stomach, Colon, Small Intestine, Testis, Blood, Liver, and Bladder). Among these, 8 models demonstrated resilience to in silico contamination. The presence of resilient tissue-specific microbial signatures implies that microbial colonization is not a random occurrence. Testing the ML models on the external dataset of the 38 healthy living samples revealed high discriminatory performance for the Heart, Liver and Colon tissue, indicating the presence of a tissue-specific microbiome even in a living state for these tissues. Notably, the most crucial feature in the heart model was the fungus Sporisorium graminicola, for the colon model the gram-positive bacterium Flavonifractor plautii, and for the liver model was the gram-negative bacterium Bartonella machadoae. The presence of a distinct microbial signature within human internal organs, allowing differentiation of distinct organs through machine learning models, supports the assertion that even under healthy conditions, human internal organs host low-biomass microbial communities specific to each tissue. "BioCancer", “Bioradio 4”. University of Pennsylvania HEALTH SYSTEM (UPenn School of Medicine RedOnc), “Decision support systems to address Nosocomial Infections and Antimicrobial Resistance (MDR-CDSS).”. ‘Competitiveness, Entrepreneurship and Innovation’ [NSRF 2014 —2020], Greece. This is the full abstract presented at the American Physiology Summit 2024 meeting and is only available in HTML format. There are no additional versions or additional content available for this abstract. Physiology was not involved in the peer review process.
Read full abstract