The gut vascular barrier is a complex system that acts as both a physical and chemical defense, preventing harmful intestinal microbes and antigens from entering the host circulation [50]. Penetration of the gut vascular barrier is a critical step in CRC invasion and metastasis and is mediated, in part, by the gut microbiome. Wu and colleagues [39] found the metalloproteinase enterotoxin of ETBF cleaves the extracellular domain of E-cadherin, a critical zonula adherens protein. This process increases gut barrier permeability and activates the $\beta$-catenin oncognic pathway, leading to increased cellular proliferation [39]. Corroborating these findings, Grivennikov et al. [51] found that pre-malignant colorectal adenomas are characterized by increased epithelial permeability due to the decreased expression of multiple barrier components, including mucin 2 (Muc2); junctional proteins JAM-A and JAM-B; and claudin. Microbial products are thus able to penetrate defective tumor-associated tight junctions and activate myeloid cells, leading to upregulated IL-23, which enhances tumor growth through IL-17 signaling. Translocation of gut microbes through a leaky gut vascular barrier sets the stage for the hepatic pre-metastatic niche, discussed further below.

The liver is the most common site of CRC metastasis, with colorectal liver metastases (CLM) developing in up to 50% of CRC patients at some point during their disease course [52]. Recent evidence suggests that a hepatic pre-metastatic niche (PMN) is formed prior to the development of overt CLM, and implicate the gut and tumoral microbiomes in its formation.

PMN are organ specific microenvironments that are primed for the implantation and outgrowth of disseminated tumor cells prior to their arrival [53]. PMNs are modulated by soluble secreted factors and extracellular vesicles from the primary tumor and are characterized by vascular permeability, extracellular matrix remodeling, angiogenesis, and an immunosuppressive microenvironment rich in regulatory T cells (Treg), myeloid derived suppressor cells (MDSCs) and fibroblasts [54].

Recent work by Bertocchi et al. [55] elucidated the role of the gut and tumoral microbiomes in hepatic PMN formation. The authors showed that gut vascular barrier impairment triggered by E. coli at the primary tumor site, as indicated by plasmalemma vesicle-associated protein-1 (PV-1) expression, correlated with bacterial levels in paired CLM samples. They further demonstrated that targeted antibiotic treatment with neomycin reduced the recruitment of innate immune cells to the liver and abrogated CLM formation [55]. Dysbiosis secondary to a high-fat diet has also been linked to liver metastasis with increased expression of multiple PMN markers, including matrix metallopeptidase 2 (MMP2), matrix metallopeptidase 9 (MMP9), fibronectin, and C-X-C chemokine ligand 12 (CXCL12), in non-tumoral hepatic tissue. This pro-tumoral cytokine signature was abrogated by antibiotic treatment, suggesting that its formation is microbiome-dependent [47]. The role of bacteria in PMN establishment was further supported in a recent study by Galeano Niño et al. [56], using spatial transcriptomics to show that tumor-associated microbial communities are not distributed randomly, but comprise organized, immunosuppressive microniches that promote tumor progression.

Epithelial-mesenchymal transition (EMT) is a highly complex process by which malignant epithelial cells lose their normal polarity and assume a mesenchymal stem cell phenotype, enabling invasion, migration, and metastasis. Studies across multiple tumor types suggest that gram negative bacteria interface with the EMT program to promote cancer dissemination.

Zhao and colleagues found that lipopolysaccharide (LPS) in the gram-negative bacterial cell wall can induce EMT in the liver. Benign biliary epithelial cells cultured in the presence of LPS exhibited upregulation of mesenchymal markers including S100A and $\alpha$-smooth muscle actin as well altered polarity and mesenchymal morphology in a process dependent on TGF-$\beta$1/Smad2/3 signaling [57]. Kim et al. [58] confirmed a causative link between LPS and EMT and found that simvastatin treatment could inhibit mesenchymal transformation by reducing toll-like receptor 4 (TLR4) expression and NF-$\kappa$$\beta$ signaling. In breast cancer cell lines, exposure to ETBF toxin led to $\beta$-catenin-dependent upregulation of EMT-related transcription factors, including those of the TWIST and SNAI family, as well as cancer stem cell markers, including OCT4 and NANOG [59]. In CRC, consensus molecular subtype 4 (CMS4) tumors are characterized by the overexpression of EMT-associated transcription factors including SLUG, TWIST, and ZEB1/2, and are associated with high rates of metastasis and a poor prognosis [60]. However, in a CRC mouse model driven by colonic expression of Zeb2, microbiota depletion with antibiotics was able to completely prevent cancer development, further supporting a link between the microbiome and EMT [61].

The potential for CRC cells to migrate and metastasize is also increased in the presence of F. nucleatum. Chen et al. [62] found that F. nucleatum infection leads to upregulation of the non-coding antisense RNA KRT7-AS via NF-$\kappa$$\beta$ signaling. This led to increased KRT7 expression, which was associated with increased transwell migration capacity in a CRC cell line, a higher rate of metastasis in vivo, and higher rates of nodal disease in human patients [62].

Fusobacterium nucleatum plays yet another role in CRC progression at the juncture of circulating tumor cell endothelial adhesion and extravasation. Zhang et al. [63] serendipitously found that the human CRC cell line HCT116 exhibited markedly increased adherence to vascular endothelium in the presence of F. nucleatum as compared to E. coli or PBS. F. nucleatum infection also enhanced cell migration relative to E. coli and Akkermansia muciniphilia through the upregulation of ICAM1 via ALPK1-mediated NF-$\kappa$$\beta$ activation. The authors confirmed the activation of this signaling pathway in vivo using murine tail vein injection to model lung metastases and in a human CRC tissue microarray [63].

The process by which extravasated tumor cells survive and proliferate in the target organ is exceedingly complex and involves interplay between tumor cells, the target organ microenvironment, and the host immune system. Studies in breast cancer have demonstrated that tumor cells may disseminate early in the disease course and remain in a dormant/quiescent state for years before metastatic outgrowth is triggered [64, 65]. In fact, stress from surgery to remove the primary tumor may, in some cases, precipitate metastatic outgrowth in patients with no clinical evidence of metastatic disease [66, 67].

Cytotoxic necrotizing factor 1 (CNF1), a bacterial toxin produced by E. coli, plays a role in the quiescence of disseminated CRC cells. CNF1 blocks cytokinesis, elicits endoreplication and polyploidization, and drives cells into a reversible dormant state [68]. Dormant tumor cells exhibit intrinsic resistance to chemotherapy agents that are conventionally designed to target and eradicate rapidly proliferating tumor cells [69, 70].

The multifarious triggers that cause tumor cells to exit dormancy and proliferate have not been fully elucidated, but it is known that a favorable tumor immune microenvironment (TIME) is required [71]. The gut and tumor-associated microbiomes interact with both the innate and adaptive immune systems to produce an inflammatory milieu that favors metastatic outgrowth.

Tumor-associated macrophages (TAMs) and neutrophils (TANs) play important roles in the promotion of metastatic outgrowth. TAMs exist in a state of flux between the M1 and M2 polarization states. M1 macrophages promote cytotoxicity and effectuate a tumor-suppressive microenvironment while M2 macrophages are associated with immunosuppression and the expression of cytokines that promote tumor survival and proliferation [72]. Both E. coli and F. nucleatum have been shown to shift this balance in favor of the M2 state. Li and colleagues [73] found that E. coli gavage stimulated the secretion of cathepsin K from MC38 cells implanted in the cecal mesentery of antibiotic treated mice. Binding of cathepsin K to TLR4 produced M2 polarization of TAMs and was associated with more numerous liver metastases [73]. F. nucleatum infection similarly facilitates metastasis through the downregulation of miR-1322, leading to increased CCL20 expression and M2 macrophage differentiation [74].

Activated TANs also support metastatic outgrowth through the expression of multiple cytokines including MMP-9, VEGF, CXCL4, and CCL5, and through the formation of weblike structures known as neutrophil extracellular traps (NETs). Bacteria residing in the metastatic niche activate NETosis through pathogen-associated molecular patterns (PAMPs), including LPS, fMLP, and Nigercin [75, 76, 77]. Cleavage of the extracellular matrix protein laminin by NET-associated proteases then activates dormant tumor cells through integrin signaling [75, 78]. NETs can further promote metastasis by shielding circulating tumor cells, stimulating angiogenesis, and promoting the formation of tumor thrombi [75].

Finally, tumor-associated microbes also modulate the adaptive immune response to promote metastatic outgrowth. Sakamoto et al. [79] found that F. nucleatum levels were associated with significantly lower cytotoxic (CD8${}^{+}$) T cell density in a sample of 181 CLM specimens, suggesting that anti-tumor inflammation is blunted in the presence of this bacteria. Oral gavage with F. nucleatum also resulted in decreased NK, CD3${}^{+}$, CD4${}^{+}$, and CD8${}^{+}$ T cells, and a significant increase in Tregs in a mouse model of CLM [80].

Systemic chemotherapy is known to alter gut microbial composition, diversity, and function [81]. However, the gut and tumoral microbiomes can also mediate cytotoxic therapy metabolism and response. Multiple chemotherapeutic agents, including cyclophosphamide, oxaliplatin, irinotecan, and 5-FU are known to interact with the intestinal microbiome [82, 83, 84, 85]. With regard to CRC, Yu et al. [30] found that colon cancer cell lines attained resistance to both 5-FU and oxaliplatin when cocultured with F. nucleatum. Chloroquine assays revealed that this process is dependent on autophagy driven by TLR4 and MYD88 signaling [30]. F. nucleatum-induced TLR4/NF-$\kappa$$\beta$ signaling has also been shown to incite 5-FU chemoresistance through upregulation of baculoviral IAP repeat C3 (BIRC3) [86]. Interestingly, a specific gut microbial metabolite, urothilin A, has been associated with increased 5-FU sensitivity and reduced murine xenograft tumor growth in both CRC and pancreatic adenocarcinoma, supporting the importance of a balanced microbiome in cancer patients [87, 88].

Immune checkpoint blockade (ICB) has revolutionized the treatment of multiple solid tumors. Several important studies have linked the gut microbiome to ICB response and toxicity in melanoma and non-small cell lung cancer. In these investigations, increased relative abundance of specific bacterial taxa, including Akkermansia muciniphila, Bacteroidaceae, Ruminococcaceae, and Bifidobacterium, were associated with improved ICB response [89, 90, 91, 92].

In CRC, the role of ICB has been largely limited to patients with microsatellite unstable tumors [93]. These cancers, characterized by a high mutational burden and increased expression of immune checkpoints, exhibit increased immunogenicity and response to both CTLA-4 and PD-1 inhibition [94, 95, 96]. The impact of the microbiome in this setting and the potential for microbe-based sensitization of microsatellite stable tumors to ICB has not yet been established; however, early preclinical studies provide promising results. Destefano and colleagues found that BRAF${}^{V\mathit{}\mathrm{600}\mathit{}E}$ mutant ETBF-induced colon tumors were uniquely sensitive to PD-L1 inhibition compared to BRAF wild type tumors, suggesting a unique interaction between a known driver mutation, microbiome-mediated carcinogenesis, and ICB response [97]. Jiang et al. [31] demonstrated that succinic acid from F. nucleatum confers resistance to anti-PD-1 therapy that could be abrogated by fecal microbiota transfer from F. nucleatum-low responders or the antibiotic metronidazole.

Peritoneal metastases (PM) occur in 5 to 10% of CRC patients and are associated with worse clinical outcomes than any other metastatic site [98]. The peritoneal microenvironment and the role of the microbiome in CRC carcinomatosis have not been well studied or defined. In ovarian cancer, another solid tumor with a strong connection to dysbiosis, local gram-negative peritoneal colonies and decreased microbiome diversity have been associated with peritoneal spread [99, 100]. In PM from appendiceal cancer, enteric bacteria, including those of the Proteobacteria, Actinobacteria, Firmicutes, and Bacteroidetes phyla, have been identified in both PM and associated mucinous ascites, implicating these taxa in the pathobiology of this disease site. Ongoing studies will reveal more about the unique TIME of CRC PM and the potential role of gut and tumoral microbes in its development.

Strong correlations between the gut microbiome and oncologic outcomes have led to the study of microbiome modulation as a therapeutic strategy. In their aforementioned study of paired CRC and CLM, Bullman and colleagues found that treatment with the antibiotic metronidazole led to reduced Fusobacterium levels and decreased tumor cell proliferation [32]. Two forthcoming trials are now evaluating the efficacy of neoadjuvant metronidazole in patients with CRC (NCT04264676, NCT05748145).

Dietary modification also presents a low-risk means to modulate the gut microbiome as well as CRC outcomes. CRC patients who consume a high-fat diet are known to have significantly shorter 5-year disease-free survival [101]. On the other hand, work by Spencer et al. [102] showed that a high-fiber diet produced taxonomic and structural changes in the gut microbiome that correlated with anti-PD-1 response in conventionally housed mice and in human patients who reported sufficient dietary fiber intake. Numerous trials are now evaluating the effect of diet-based microbiome modulation on CRC outcomes. At our institution, the Beans to Enrich the Gut Microbiome vs Obesity’s Negative Effects (BEGONE) trial is evaluating the longitudinal effect of dietary fiber on the gut microbiome and risk of CRC recurrence.

Fecal microbiota transplant (FMT) is the most direct means of altering the gut microbiome and is now widely accepted as an effective treatment for refractory Clostridium difficile colitis [103]. FMT has also been used to successfully treat inflammatory bowel disease and is under investigation in the treatment of other autoimmune disorders [104, 105]. In oncology, FMT has been used to overcome melanoma resistance to ICB in multiple preclinical studies and at least two clinical trials [106, 107]. Multiple ongoing trials are investigating the role of FMT in modulating toxicity and improving response to ICB in other solid tumors, including CRC.

While still in the preliminary stages of investigation, study of both additive (FMT) and subtractive (antibiotics and novel therapeutics) modulation of the gut and tumoral microbiome in CRC carries enormous clinical potential. As suggested by the preclinical studies covered in this review, these efforts may provide new avenues through which to improve CRC detection and increase the efficacy of existing standard treatments. As our understanding of the microbiome’s role in gut vascular permeability and modulation of the host immune response deepens, we anticipate a new class of microbiome-based interventions that may even stymie CRC dissemination and metastatic outgrowth.