Emerging multifaceted roles of the microbiome in cancer susceptibility

Abstract

Identifying the factors that contribute to individual susceptibility to cancer is essential for both prevention and treatment. The advancement of biotechnologies, particularly next-generation sequencing, has accelerated the discovery of genetic variants linked to cancer susceptibility. While hundreds of cancer-susceptibility genes have been identified, they only explain a small fraction of the overall cancer risk, a phenomenon known as "missing heritability". Despite progress, even considering factors such as epistasis, epigenetics, and gene-environment interactions, the missing heritability remains unresolved. Recent research has revealed that an individual's microbiome composition plays a significant role in cancer susceptibility through several mechanisms, such as modulating immune cell activity and influencing the presence or removal of environmental carcinogens. In this review, we examine the multifaceted roles of the microbiome in cancer risk and explore gene-microbiome and environment-microbiome interactions that may contribute to cancer susceptibility. Additionally, we highlight the importance of experimental models, such as collaborative cross mice, and advanced analytical tools, like artificial intelligence, in identifying microbial factors associated with cancer risk. Understanding these microbial determinants can open new avenues for interventions aimed at reducing cancer risk and guide the development of more effective cancer treatments.

Key Words: Cancer susceptibility; Genetic variants; Genome-wide association study; Missing heritability; Microbiome; Microbiome-wide association study

Core Tip: Understanding individual susceptibility to cancer is critical for effective prevention and treatment strategies. While genetic studies have identified numerous cancer-susceptibility genes, much of the heritable risk remains unexplained. Emerging evidence indicates the microbiome as a key contributor to this "missing heritability". Microbiomes influence cancer risk through immune modulation, metabolic activity, and interaction with environmental exposures. This review discusses the complex gene-microbiome-environment interplay in cancer susceptibility and emphasizes the value of integrative models and tools, such as collaborative cross mice and artificial intelligence, for uncovering microbial determinants of cancer risk.

INTRODUCTION

Cancer susceptibility, a term used to describe an individual’s increased risk of developing certain cancers, is a complex genetic trait. The genome-wide association study (GWAS) approach combined with the advent of biotechnologies has successfully identified genetic variants contributing to cancer risk[1-11]. However, it is strikingly found that these genetic variants identified so far account for only a small fraction of overall cancer risk, implying that much of the genetic risk remains to be missing, known as “missing heritability”[1,2,8,12,13]. Therefore, continuing to explore the genetic basis of cancer susceptibility and to better understand the molecular mechanisms by which genetic variants influence cancer risk is still one of the holy grails of cancer research.

The missing heritability of cancer is a notable unsolved problem for genetic association studies. Since few cancers result in mutations in a single gene, cancer susceptibility is usually highly polygenic, i.e., influenced by a multitude of common genes[14-17]. The effect sizes of each common gene often fell below significance thresholds that can be identified by GWAS, which may account for most missing heritability[1]. A combination of the methods to sequence genomes and the advanced computational tools allows us to detect subtle differences in the genetic background of individuals and to discover rare variations that mitigate cancer risk[18,19], but the gaps are still not filled. Several other mechanisms have also been proposed for missing heritability, such as epistasis and epigenetics[12,13,20-22]. Also, the gene-environment (GxE) interactions are known to contribute to cancer susceptibility[23]. Moreover, recent studies provide evidence that analysis of GxE interactions holds promise for identifying additional genetic and environmental determinants. Therefore, GxE interactions can provide valuable insights into missing heritability[24]. However, taking all counts, the proportion of missing heritability still has not been satisfactorily explained. Other strategies have been proposed to identify the missing heritability[12]. Recently, researchers suggested that the human microbiome must be considered in order to improve our estimates of the heritability of phenotypes[25,26].

There is a growing understanding and appreciation for the functional significance of the human microbiome in health and disease such as cancer[27-32]. A sharply increasing number of studies have revealed the roles of microbiomes in cancer development and progression, diagnosis, prognosis, and treatment. Some of these aspects have been extensively reviewed[27-31]. While sufficient evidence shows that the microbiome composition differs greatly between individuals[33-36], a critical question remains open as to whether inter-individual variation in the microbial community contributes to cancer susceptibility. This review will first overview the links between microbiome and cancer susceptibility. Subsequently, we will discuss the gene-microbiome and environment-microbiome interactions, where the role of the microbiome is a mediator of genetics and environmental exposures in cancer susceptibility (Figure 1). Finally, we summarized the importance of mouse models, such as the collaborative cross (CC) mice and gnotobiotic mouse models, and powerful new tools, such as artificial intelligence, for discovering the microbes associated with cancer susceptibility. Understanding these microbial determinants of cancer susceptibility can lead to a new avenue of interventions and prevention, which ultimately minimizes cancer risk and sheds light on new strategies to implement appropriate cancer treatment.

Figure 1 Contribution of host genetics, microbiome, environmental factors, and their bidirectional interactions to cancer susceptibility. The host gene-environment, host gene-microbiome, and microbiome-environment bidirectional interactions play critical roles in cancer susceptibility.

VARIATION IN HUMAN MICROBIOME COMPOSITION AND DIVERSITY

Our body harbors 10-100 trillion microbial cells, including bacteria, viruses, and fungi, the majority of which live in the gut[37-39]. All these microbial cells are defined as human microbiota, which develops to reach a climax stage during the first 3 years of life. In healthy adults, the commensal microbial communities typically maintain a stable composition, although the relative abundance of each microbe oscillates over time. As evidenced by an increasing number of epidemiological and experimental studies, disrupting the infant microbiota is significantly associated with immune and metabolic disorders and other diseases, such as cancer, in later life, indicating that the early human gut microbiota plays a critical role in modulating disease susceptibility, including cancer susceptibility (see below).

The genetic material of these microbial cells is collectively defined as the human microbiome, which is also referred to as the human “second genome”[40,41]. We humans have 20000-25000 genes in each of our cells, which are inherited and largely static, whereas the human microbiome contains approximately 500 times more, which is acquired and dynamic through life[37,42]. In 2007, the National Institutes of Health initiated The Human Microbiome Project to characterize the human microbiome and to understand how it contributes to normal physiology and predisposition to disease[42]. The findings from this initiative and numerous other studies have shown that the human microbiome exhibits huge both intra- and inter-individual variability in composition and function[35,36,42]. In contrast to the human genome, which is about 99.9% identical among individual humans, they are completely different in their microbiome[37]. The 50% of all genes in the human microbiome are individual-specific[37,39,42]. Therefore, the human microbiome is a great source of genetic diversity, which can be a modifier of disease, such as cancer.

It is well known that many factors can influence gut microbiota, including host genetics and environmental factors[43-45]. The 5%-45% of inter-individual variation in microbiome composition and diversity can be explained by genetics[33]. Hundreds of genetic variants have been discovered to be significantly associated with the abundance of specific gut microbes by large GWASs[33]. In addition to genetic influences, environmental factors including diet, pharmacological use, and lifestyle have been shown to shape microbial composition, which is associated with human cancer[43]. These collective findings indicate that the microbiome might be the missing link among genes, environments, and cancer. The following sections discuss how bidirectional interactions between genetics and microbiota, as well as between environmental factors and microbiota, impact cancer susceptibility.

Importance of the microbiome in cancer susceptibility

Accumulating evidence has shown that the human microbiome contributes to all stages of cancer development by modulating the hallmarks of cancer (Figure 2). For example, some pathogens directly target genes that regulate cell growth; others cause chronic inflammation that can eventually lead to cancer development. Some infections weaken the people’s immune system, reducing the body’s ability to combat malignancy. Different mechanisms have been reviewed for microbiota to cause cancer[27,30,32]. The tremendous interindividual variation in microbiome diversity and function, coupled with the critical roles of the human microbiome in cancer, suggests that the human microbiome may also be an important factor driving individual cancer susceptibility.

Figure 2 The microbiome and the hallmarks of cancer. The microbiome modulates the majority of the hallmarks of cancer.

The first evidence for the contribution of specific bacterial species to cancer susceptibility is the link between Helicobacter pylori (H. pylori) and gastric cancer[27,46]. In addition to the oncogenic bacterium H. pylori, many other cancer-related pathogens have been identified, including seven oncogenic viruses: Hepatitis virus B and C, human papillomavirus, human T-cell lymphoma virus 1, Epstein-Barr virus, Merkel cell polyomavirus, and Kaposi’s sarcoma virus (or HHV8), and three oncogenic parasites: Schistosoma haematobium, Opisthorchis viverrini, and Clonorchis sinensis[27,46]. Some studies have reported that an estimated 12%-20% of cancers are caused by viruses or bacteria[46-48]. It is believed that this figure will significantly increase since advanced tools and methods allow for the detection of more microbes controlling cancer susceptibility. For example, microbiome-wide association studies (MWAS), together with next-generation sequencing technologies, will provide unprecedented views into the association of the human microbiome with cancer risk. MWAS has successfully identified microbial taxa or functions associated with human cancer[49]. These findings reveal that the higher abundance of some bacteria decreases while others increase cancer risk, consistent with the concept that our body harbors both “good” and “bad” bacteria. Identification of microbial determinants of cancer susceptibility will provide new avenues for cancer prevention. However, we still face many challenges in human studies. The most challenge is that the human microbiome is sensitive to many environmental factors that are almost impossible to control.

HOST GENETICS-MICROBIOME INTERACTIONS IN CANCER SUSCEPTIBILITY

Increasing attention has focused on bidirectional interactions between host genetics and microbiome. On the one hand, host genetics determines the microbiome composition and diversity (Figure 1), which have been documented in both human and mouse studies[33,50-54]. Conversely, the microbiome exerts substantial influence on the host genome (Figure 1). Growing evidence suggests that bidirectional interactions between host genetics and microbiome are critical determinants of disease susceptibility, including cancer[55,56].

Although recent breakthroughs in cancer genetics have identified hundreds of genetic variations responsible for the increased cancer susceptibility, the missing heritability of cancer remains a challenge. Strong links between genetics and microbiome and between microbiome and cancer lead us to hypothesize the effect of some genetic variants on cancer susceptibility through the microbiome, which is defined as an indirect effect. Based on this hypothesis, we could divide pathogenic variants into three categories: (1) Genetic variants, which pose a direct effect; (2) Genetic variants that pose both direct and indirect effects; and (3) Genetic variants that pose an indirect effect on cancer susceptibility (Figure 3). The GWAS could hardly detect Category 3 and some Category 2 variants, which can possibly be identified by combining GWAS of the microbiome and MWAS of cancer with mediation analysis (details see below).

Figure 3 Direct and indirect genetic effect on cancer susceptibility. The microbiome can serve as a mediator of genetic susceptibility.

Mounting evidence has shown that gut microbiota modulates host gene expression[57-59]. Several mechanisms for microbial control of host gene regulation have been discovered, including disruption of the epigenetic landscape, interference with host signaling pathways, chromatin remodeling, altered RNA splicing, and modulations by miRNAs[60-63]. In addition, gut microbiota can produce genotoxic metabolites that cause DNA damage, subsequently leading to genomic instability, chromosomal aberrations, and gene mutations[60-63]. Three bacterial genotoxins have been identified: Cytolethal distending toxin, typhoid toxin, and colibactin[64-67]. The most well-studied genotoxic bacteria are Escherichia coli, which produce colibactin[68,69]; enterotoxigenic Bacteroides fragilis, which produce a toxin[70,71]; and Campylobacter jejuni strains, which produce the cytolethal distending toxin[68]. A recent study has identified more bacterial strains that show DNA-damaging activity[72]. It is possible that more genotoxic bacteria will be discovered in the future. The discovery of genotoxic bacteria and other microbial determinants of cancer susceptibility is critical for cancer prevention, diagnosis, and treatment.

ENVIRONMENT-MICROBIOME INTERACTIONS IN CANCER SUSCEPTIBILITY

Scientific research has begun to elucidate the diverse mechanisms through which the human microbiome interacts with environmental exposures[73-75]. The gut microbiota plays a central role in modulating the toxicity of environmental pollutants, while, as discussed earlier, environmental exposures are key determinants of the composition and function of human microbial communities (Figure 4). Compelling evidence has indicated that the human microbiome can modulate exposure to environmental pollutants, which modifies disease susceptibility, such as cancer, to environmental exposures[76]. Microbial metabolism of environmental pollutants might affect the host favorably or unfavorably through metabolic byproducts or modulation of compound toxicity. Many studies have shown that the gut microbiome directly participates in the metabolic transformation of environmental chemicals, including metals, polycyclic aromatic hydrocarbons, nitrosamines, aromatic amines, and organochlorine pesticides[76,77]. Research also shows that gut microbiome indirectly involves metabolisms of environmental pollutants through deconjugation of host-generated metabolites, modulation of epithelial permeability to facilitate the transport or excretion of chemicals, and regulation of the expression or activity of key endogenous metabolic pathways[76,77]. While the roles of microbiomes in the metabolism of environmental pollutants are well established, substantial gaps remain in our understanding of the full range of metabolic pathways for environmental pollutants within individual microbiomes. Moreover, how substantial differences in microbiome composition among different individuals influence the metabolism and disposition of environmental pollutants has received little attention, which might play an important contribution to individual cancer susceptibility.

Figure 4 Bidirectional interactions between the microbiome and environmental pollutants and their contribution to cancer susceptibility. The interactions between the microbiome and environmental pollutants modulate cancer susceptibility.

On the other hand, environmental pollutants can, directly and indirectly, alter the microbiome. Many environmental pollutants have been identified to disrupt the microbiome composition and diversity[78,79]. Animal studies have demonstrated that exposure to a variety of pesticides, metals, artificial sweeteners, and drugs can induce microbiome changes[78,79]. Alterations in the microbiome by environmental pollutants can be characterized using multi-omics approaches. It should be noted that the effects of environmental pollutants on the microbiome composition and diversity can be modulated by host genetics. However, the mechanisms by which environmental pollutants induce gut microbiota dysbiosis and the subsequent impact on cancer susceptibility remain to be elucidated in future studies.

MOUSE MODEL SYSTEMS FOR INVESTIGATING THE ROLE OF THE MICROBIOME IN CANCER SUSCEPTIBILITY

Environmental exposures are difficult to quantify and control for human populations, which severely hinders the identification of microbial determinants of cancer susceptibility in humans. In contrast, mice offer many advantages for the study of the role of microbiome in cancer susceptibility. These advantages include well-designed populations tailored to specific research questions, standardized husbandry conditions, comprehensive phenotypic analyses, and the ability to manipulate the ecosystem to investigate the role of the microbiome in tumor progression and metastasis. Moreover, mouse models have been instrumental in developing our knowledge of the microbiome in vivo and of the aberrations in microbiome that are causally associated with disease. Human and mouse gut microbiota exhibit approximately 90% similarity at the phylum level and 89% at the genus level[80-83]. Therefore, mouse models will be pivotal in laying the foundation for microbial oncogenesis by providing information on the crosstalk between cancer and the microbial community.

The CC mouse population model is an ideal system for studying the role of microbiomes in cancer susceptibility

The CC is a large panel of novel inbred mouse strains derived from eight genetically diverse founders (A/J, C57BL/6J, 129S1/SvImJ, NOD/LtJ, NZO/HlLtJ, CAST/EiJ, PWK/PhJ, and WSB/EiJ) through a targeted funneling breeding strategy designed to randomize the genetic composition of each inbred line. The CC captures nearly 90% of the known genetic variation found in laboratory mice, making it a powerful tool to model the genetic diversity of human populations[84]. This resource enables high-resolution genetic mapping and the identification of individual genes underlying complex traits and diseases. A key advantage of the CC is that each strain’s genome is fully characterized and reproducible, allowing consistent phenotypic measurements across genetically distinct lines. Studies have documented extensive phenotypic variation in CC mice, including spontaneous tumor development[50,85-94]. Moreover, the gut microbiome composition also varies significantly among strains[52], and has been linked to traits such as a sleep phenotype, memory, anxiety-like behavior, and Azoxymethane-induced toxicity and colon tumor development[50,88,95]. One recent study highlighted the critical role of host genetics-microbiome interactions in determining susceptibility to colorectal cancer (CRC)[50], pointing to novel avenues for personalized prevention and treatment of CRC. Collectively, these findings suggest that CC mice closely mimic the phenotypic diversity seen in humans. So, CC population model is an ideal system for identifying microbial contributors to individual variability and for dissecting GxE-microbiome interactions in cancer susceptibility. When combined with other models, such as knockout and gnotobiotic mice, the CC enables detailed investigations into the causal mechanisms of disease at an unprecedented resolution.

Gnotobiotic mouse model

A major driver of advances in microbiome research has been the use of gnotobiotic (or germ-free) mice, which are devoid of all microorganisms and maintained under sterile conditions[96,97]. Gnotobiotic mice effectively allow us to transfer selective bacterial species or whole fecal microbiota[98-100]. Especially, establishing a humanized gnotobiotic mouse model through the fecal microbiota transplantation (FMT) of humans into germ-free mice offers an innovative and powerful approach for recapitulating the human microbial system[81,101-103]. Microbial transplantation into germ-free mice can be used to explore the complex relationship between the specific microbe or microbiome and cancer etiology and progression, to characterize the impact of specific microbe or microbiome on the toxicity of environmental pollutants, and to evaluate potential microbiome-based prevention and therapeutics in validated in vivo cancer model paradigms[98,104-109].

ANALYTICAL TOOLS FOR IDENTIFYING MICROBIAL DETERMINANTS OF CANCER SUSCEPTIBILITY

Microbiome-related studies on cancer development and treatment have notably increased recently[110-112], which provide both the challenges and opportunities for developing and deploying advanced analytical tools to explore the host-microbiome determinants or their intermediate role towards phenotypic outcomes, including cancer susceptibility. These analytical tools majorly fall into two categories: (1) Classical machine learning (ML) and statistical models; and (2) Artificial intelligence techniques, all of which play important roles in microbiome studies towards diagnostic and predictive biomarker detection and many other applications (Figure 5).

Figure 5  Analytical methods for microbiome-wide association study.

Classical ML techniques and statistical models in microbiome studies

Classical ML and statistical models continue to play an essential role in microbiome data analysis, among which logistic regression (LR), linear discriminant analysis (LDA), support vector machine (SVM), naïve Bayes classifiers, and random forest (RF) are among the most popular techniques that have been extensively used.

LR: LR is a statistical model for binary outcome (Y) predictions based on one or more independent variables (X). LR and its variations have been used to predict cancer prognosis and drug response from the tumor microbiome[113]; to improve CRC diagnosis using gut microbiome data[114]; to identify gut microbiome associations with immune checkpoint inhibitor response in advanced melanoma[115]; and to identify non-invasive diagnostic biomarkers in colorectal carcinogenesis using the metabolomics and/or metagenomics profiles from fecal samples[116].

LDA: LDA is a generalization of Fisher's linear discriminant, aiming to maximize the separability of two or more classes of objects or events through an optimized linear combination of features. LDA and its variations have been used to evaluate the microbial effects between tumor and normal groups[117]; to support high-dimensional metagenomic biomarker discovery and explanation between two or more biological conditions (or classes)[118]; and to assess and characterize differences in the gut microbiota among CRC patients according to the location of the tumor (i.e., left-sided CRC and right-sided CRC)[118].

SVM: SVM is a ML technology for both classification and regression analysis. It aims to discover a hyperplane with the maximal functional margin to the nearest training data points (i.e., support vectors) of any class. It provides a robust solution with both linear and non-linear capabilities through the kernel trick. SVM and its variations have been used to construct a phylogenetic approach for classifying oral microbiota, and to predict the clinical outcomes of oncology patients on immunotherapies using baseline gut microbiome features[119].

Naïve Bayes classifiers: Naïve Bayes classifiers are a collection of classification algorithms based on Bayes’ Theorem with strong (naïve) independence assumptions between the features. Naïve Bayes classifiers are highly scalable and have been used rapidly to assign rRNA sequences into the new bacterial taxonomy[120]; and to investigate the impact of training sets on taxonomy classification of high-throughput bacterial 16s rRNA gene sequences, where a selected region of target sequences helps improve the performance of naïve Bayes classifier[121].

RF: RF is an ensemble learning method consisting of multiple decision trees, mainly for classification and regression tasks, where the final decision of RF is the consensus of its decision trees. RF and its variations have been used to identify the association between gut microbes and CRC, to reveal host microbial determinants of clinical response to FMT therapy in type 2 diabetes patients, and to increase the prediction performance of CRC disease status based on metagenomic shotgun sequencing data.

In addition, other statistical models, including the mediation model and Cox proportional-hazards model, have also been widely used in microbiome studies to explore the intermediate role of the microbiome, together with other factors (e.g., genetics), towards phenotypic outcomes; or to construct microbiome signatures for time-to-event prediction (e.g., prognosis)[122].

Artificial intelligence techniques in microbiome studies

Motivated by recent neuroscience findings[123,124], one area of ML has attempted to develop computational modules that emulate properties of the neocortex for constructing information representations. Although state-of-the-art ML architectures have evolved, the concept of the hierarchical structure for information processing has remained the central theme. Over the past decade, deep learning has gained momentum due to its demonstrated capability to enhance performance across diverse automation tasks and its promising potential for future research. Among different artificial intelligence techniques, convolutional neural networks (CNNs), predictive sparse decomposition (PSD), and transformers have been widely used in microbiome studies.

CNN: CNN is a class of artificial neural networks consisting of multiple layers of neurons with trainable weights and biases toward the outcome (e.g., classes/Labels of input data). Although CNNs have been mostly applied to imaging data, they have also been used to accurately classify host phenotypes from metagenomic data[125,126].

PSD: PSD is a class of unsupervised sparse learning technology with a hierarchical learning framework designed to capture higher-level dependencies of input variables, thereby improving the ability of the system to identify underlying regularities in the data. Similar to CNN, PSD has also been widely applied to imaging data. Until recently, it has been used to integrate tumor microbiome biomarkers with cellular morphometric biomarkers and gene biomarkers for improved multi-modal risk stratification in breast cancer patients[127].

Transformer: Transformer is a deep learning model originally designed to process sequential input data, such as natural language processing[128]. Transformers have now also been widely used in the fields of computer vision[129] and audio processing[130]. A recent study introduced transformer models to the field of microbiome analysis for the accurate identification of bacteriophages from metagenomic data[131], which opens a new opportunity for microbiome studies with large language models or even pretrained systems on microbiome data (microbGPT), similar to generative pre-trained transformers[132].

In summary, these analytical tools, empowered by classical ML and statistical models or more advanced artificial intelligence techniques, provide new avenues to microbiome data processing and knowledge mining with promising potential to further advance the discovery of microbial determinants of cancer susceptibility.

CONCLUSION

Mounting evidence highlights the microbiome as a vital and dynamic factor in cancer susceptibility. Research is increasingly revealing that the microbiome is not merely a passive passenger but an active modulator of host physiology, immunity, and even the efficacy of anticancer therapies. From influencing genetic stability and inflammation to modulating metabolic pathways and immune surveillance, the microbiome emerges as a critical determinant in both cancer initiation and progression. Despite these advances, much remains to be discovered about the full spectrum of microbes that individually or synergistically affect cancer risk, as well as the mechanisms behind their effects. Progress in this area will require sophisticated analytical tools, including artificial intelligence, to unravel these complexities. Notably, the context-specific roles of the microbiome, ranging from protective to pathogenic, highlight the complexity of its interactions with host and environmental factors. Certain microbial taxa have been implicated in carcinogenesis through genotoxic metabolite production or pro-inflammatory mechanisms, while others exhibit tumor-suppressive effects by enhancing immune responses or maintaining epithelial barrier integrity. This duality underscores the need for precision approaches in microbiome research, tailored to individual host genetics, lifestyle, and tumor microenvironments. Importantly, future research must aim to move beyond association studies and unravel causal mechanisms through integrative multi-omics, longitudinal cohort designs, and functional validation in preclinical models. A major challenge remains in distinguishing correlation from causation and understanding how specific microbial consortia dynamically influence oncogenic processes over time. Longitudinal studies of the fecal microbiome in at-risk populations could help identify microbial markers of increased cancer susceptibility, potentially leading to new early detection tools. Since the microbiome can also modify the effects of environmental pollutants, identifying microbes capable of neutralizing harmful compounds may pave the way for novel preventive or therapeutic strategies. Ultimately, integrating microbiome science into the broader oncology landscape could transform our understanding of cancer etiology and herald a new era of microbiome-informed diagnostics and therapeutics. To fully harness this potential, interdisciplinary collaboration across microbiology, immunology, oncology, and bioinformatics will be essential. In conclusion, the microbiome stands as a pivotal frontier in the quest to understand and mitigate cancer susceptibility. More rigorous research and clinical trials are needed to translate these findings into effective strategies for cancer prevention, treatment, and management.