Once in the portal vascular system, CRC circulating tumor cells pass through the portal venules into the sinusoidal capillaries. There, the cells encounter liver-specific defense mechanisms, and many are phagocytosed by stellate macrophages (Kupffer cells (KCs)), a process potentiated by natural killer cells (NKs) [33,34]. However, cells that escape innate immune surveillance must stop in the sinusoids to extravasate and form a metastatic lesion. While this mechanical arrest contributes to the adhesion of circulating tumor cells to the sinusoidal vasculature, intravital microscopy has shown that CRC circulating cells are usually smaller than the vessel diameter [35,36]. This suggests that specific cell-adhesion interactions can facilitate tumor cell arrest.

The liver sinusoids are lined by liver sinusoidal endothelial cells (LSECs), which express the critical cell-adhesion protein, E-selectin. E-selectin is a cytokine-inducible protein implicated in CRC cell adhesion to liver sinusoidal endothelial cells [37]. Notably, the inhibition of endothelial E-selectin has been shown to reduce the number of CRLMs up to 97% relative to the placebo in syngeneic murine models [37,38]. Furthermore, Khatib et al. [39,40] showed that CRC circulating cells trigger the release of proinflammatory TNF-α and IL-1β from KCs, thereby upregulating the expression of E-selectin and other adhesion molecules on liver sinusoidal endothelial cells and enhancing liver metastases [39,40].

Once tumor cells have extravasated from the sinusoidal microvasculature into the space of Disse, hepatic stellate cells (HSCs) are activated by the pro-inflammatory cascade that was initiated in the microvascular phase. HSCs produce extracellular matrix (ECM) proteins, including collagen, laminin, fibrillin, and fibronectin, and provide a scaffold for tumor cells [41,42]. Additionally, KCs and neutrophils secrete matrix metalloproteinases (MMPs) and elastase, which degrade and remodel the ECM, facilitating tumor cell invasion [31,43]. Furthermore, cells in the TME have been implicated in the induction of an immunosuppressive niche, facilitating tumor cell growth. Huang et al. [44] demonstrated that HSCs co-cultured with T cells and CRC antigen-pulsed dendritic cells (DCs) significantly abrogated T-cell responsiveness and induced the expansion of immunosuppressive regulatory T cells (Tregs). In addition, portal injections of HSCs with tumor cells significantly increased CRLM compared to tumor cells alone [44].

CRLMs are predominantly supplied by hepatic arteries [45]. The vascularization of liver metastases can occur through vascular endothelial growth factor (VEGF)-mediated angiogenesis or by the “co-opting” of existing vasculature. Importantly, a phase III trial by Hurwitz et al. [46] showed that the addition of bevacizumab, an anti-VEGF agent, to standard chemotherapy improved the median survival of patients with mCRC; however, the majority of patients did not exhibit a response to therapy [46]. This suggests the importance of other mechanisms of tumor vascularization independent of VEGF in some CRCs [42,47]. Notably, Frentzas et al. [48] described the co-opting of vessels in CRLM, a process by which cancer cells invade along the existing liver microvasculature, replacing hepatocytes and occupying the space adjacent to sinusoidal vessels. This commonly occurs in tumors that display a replacement growth pattern (see Section 3.1.4. below below) and is associated with resistance to anti-VEGF therapy. Importantly, the inhibition of actin-related protein 2/3 (Arp2/3), which is associated with the co-opting of vessels, was found to reduce vascular density in combination bevacizumab significantly more than bevacizumab by itself. Combination therapy also potently inhibited the growth of CRLMs with reduction in lesion area by approximately 75% compared to controls [48].

The growth of vascularized micrometastases leads to tumor cell proliferation and the establishment of clinically detectable metastases. Three histologic growth patterns have been described by Eefsen et al. [25] in a series of 24 patients with resected CRLMs: desmoplastic, pushing, and replacement. In the desmoplastic pattern, a band of fibrotic tissue separates tumor cells from the surrounding liver parenchyma. In the pushing pattern, liver cells are pushed aside by the growing metastatic lesion and assume a flattened morphology with no fibrotic band separating the invasive tumor front from the hepatic parenchyma. The replacement pattern describes tumor cell infiltration into the existing liver parenchyma, such that it supplants the parenchyma instead of simply pushing it away. In 83% of patients, identical growth patterns were identified in all metastases [25]. This classification is notable because the dominant replacement growth pattern is associated with a worse prognosis [49]. These differences indicate that different interactions of tumor cells, TME cells, and cells of the hepatic parenchyma lead to varying CRLM phenotypes.

Tumor-associated macrophages (TAMs) have historically been dichotomized as “M1-like” macrophages with pro-inflammatory, immunostimulatory, and anti-tumorigenic properties or “M2-like” macrophages with immunosuppressive and pro-tumorigenic properties [62]. In reality, TAMs exist on a spectrum between these two states; however, this traditional classification provides a useful framework for discussion. M1 macrophages produce pro-inflammatory immunostimulatory cytokines, including IL-1β, IL-6, and TNF-α [62,63]. Although the M1 phenotype is generally thought to be anti-tumorigenic, this notion is context-specific. For example, M1 TAMs can promote chronic inflammation in induced colitis models, and this is a risk factor for the development of CRC [64]. In a colitis-associated CRC model, Wang et al. [65] demonstrated that the densities of M1 and M2 macrophages vary throughout the inflammation–carcinoma sequence, but overall macrophage count increased as tumors became metastatic. Exposure of CRC cell lines to conditioned media from M1 TAMs decreased viability and promoted pro-apoptotic morphologic changes, while treatment with M2-derived conditioned media increased cell proliferation and expression of the anti-apoptotic markers survivin and BMI-1 compared to control conditioned media [66]. Activated M2 TAMs also contribute to tumor angiogenesis through secretion of VEGF along with other TME cells [67,68,69].

M2 TAMs promote invasion and metastasis [70,71,72,73,74]. TAM secrete of MMP-9, a key proteolytic enzyme that contributes to ECM remodeling in multiple cancers, including CRC [72]. Concordantly, Afik et al. [74] demonstrated that CRC TAMs were enriched for molecular signatures associated with ECM remodeling in transcriptomic and proteomic analyses [74]. M2 TAMs also contribute to the production of TGF-β, which potentiates ECM remodeling, EMT, and metastasis [71]. EMT pertains to tumor cell loss of epithelial features (i.e., the loss of apical–basal polarity, dissolution of tight junctions, etc.) and assumption of a mesenchymal-like phenotype [75]. The mesenchymal phenotype is associated with increased invasiveness and migratory potential [75]. The loss of E-cadherin expression, a hallmark of the epithelial phenotype, is associated with a poor prognosis in CRC [76]. TAMs also promote EMT through IL-6-mediated activation of the JAK2/STAT3 pathway and subsequent inhibition of the miRNA tumor suppressor miR-506-3p and its target, FoxQ1, as well as through secretion of CCL22 and subsequent activation of the PI3K/AKT pathway [73,77].

M2 TAMs also facilitate immune evasion through secretion of immunosuppressive cytokines IL-10 and TGF-β, which facilitate suppression of T lymphocytes [65,78]. Herbeuval et al. [79] demonstrated that TAMs induced an immunosuppressive environment through the secretion of IL-6, leading to STAT3-mediated production of IL-10 by CRC cells. Thus, M2 TAMs in the CRC TME drive tumor progression, invasiveness, and a more aggressive tumor phenotype.

CD4+ helper T cells are key mediators of the adaptive immune system, secreting cytokines to modulate the activity of other immune cells in response to infection or cancer. The presence of tumor-infiltrating lymphocytes (TILs) is associated with improved survival in CRC. In a study of 342 resected CRC specimens, high CD4+ T-cell density was associated with improved relapse-free survival and disease-specific survival [86]. However, CD4+ T-cell subtypes in the TME have varying impacts on tumor behavior. T-helper 1 cells (T_H1) are traditionally thought to enhance CTL effector function and are instrumental for augmentation of the antitumor immune response [87]. Increased expression of T_H1-cluster genes in resected CRC specimens is associated with improved disease-free survival, while higher numbers of T-helper 17 cells (T_H17), which have an immunosuppressive function, are associated with poor survival [88]. This is consistent with prior preclinical studies, which suggest that T_H17-derived cytokines, such as IL-17, IL-21, and IL-22, promote CRC tumor growth [89].

DCs are professional antigen-presenting cells that are crucial for the initiation of the immune response to specific antigens through internalization of foreign antigens and subsequent presentation to T cells. To avoid immune surveillance, cancer cells may suppress DCs through multiple mechanisms, including the secretion of immunosuppressive TGF-β [95]. In CRC, myeloid DCs, the most common DC subtype associated with cell-mediated immunity and stimulation of naïve CD4+ T cells to the T_H1 phenotype, are found in increased frequency at the tumor invasive front and are associated with lymph node invasion [96]. In contrast, Gulubova et al. [97] noted that advanced-stage CRC tumors had a lower density of CD83+ DCs in the stroma and invasive margin [97].

The mechanism through which these DCs enhance invasiveness has not been well-elucidated. Orsini et al. [98] demonstrated that DCs from CRC patients had impaired antigen-presenting capacity and reduced co-stimulatory molecule expression compared to DCs from healthy controls. Furthermore, CRC patient-derived DCs secreted increased levels of immunosuppressive IL-10 and decreased levels of immunostimulatory IL-12 and TNF-α [98]. Concordantly, Nagorsen et al. [99] demonstrated that while S100+ DCs were found more frequently in limited disease and were associated with improved survival, their presence in the CRC TME was also positively associated with Treg infiltration, further highlighting the paradoxical pro-tumorigenic function of CRC-hijacked DCs. Indeed, Hsu et al. [100] demonstrated that DCs harvested from CRC patients highly expressed CXCL1, which enhanced tumor cell migration, cancer cell stemness, and EMT. Thus, DC immunostimulatory capacity may be suppressed by CRC cells and modulated into an invasive, immunosuppressive phenotype, perhaps reflecting a reversion back to an immature phenotype [101].

Tregs, characterized by CD25 and FoxP3 expression, are potent mediators of immunosuppression. Their presence in the TME is associated with increased metastasis and poor outcomes in numerous malignancies [106,107,108]. In their physiologic state, Tregs prevent autoimmunity and regulate the immune response by downregulating IL-2, releasing adenosine, and secreting immunosuppressive cytokines including TGF-β, IL-10, and IL-35 [109]. In CRC, the impact of Tregs in the TME is complex, and evidence suggests their function as pro- or anti-tumorigenic is context-specific. Ji et al. [110] demonstrated through mRNA profiling of 81 pretreatment biopsies of locally advanced rectal cancer patients that the Tregs were inversely correlated with prognosis. Conversely, other studies demonstrate that elevated FoxP3+ cell infiltration correlate with improved survival [86,111].

A potential explanation for the apparent contradictory effect of Tregs is likely due to their heterogeneity. Saito et al. [112] demonstrated that naïve Tregs (FoxP3^low/CD45RA+) in CRC specimens could be stratified into terminally differentiated immunosuppressive FoxP3^high/CD45RA-, and pro-inflammatory FoxP3^low/CD45RA- subgroups. Compared to other tumors, such as melanoma, some CRC specimens had significantly higher numbers of FoxP3^low/CD45RA- Tregs [112]. The authors therefore characterized CRC tumors as “type A” with a low expression of these inflammatory Tregs or “type B” with a high expression of inflammatory Tregs. Notably, type-B CRC demonstrated significant upregulation of genes involved with inflammation and immune response, including IL-12, TNF-β, and TGF-β [112]. Consistent with these findings, high FOXP3 expression in type-A CRC tumors was associated with a worse prognosis [112].

Neutrophils are the first responders to acute infection or inflammation [116]. They secrete lytic enzymes, phagocytose certain microbes, generate neutrophil extracellular traps, and secrete numerous cytokines that sustain the inflammatory response and recruit other immune cells as part of the innate immune response. Tumor-associated neutrophils (TANs) are associated with prognosis; as such, understanding the function of TANs is an emerging area of investigation [117]. Fridlender et al. [118] demonstrated that TGF-β production in the CRC TME polarizes TANs from an immunostimulatory “N1” phenotype to a pro-tumorigenic “N2” phenotype. TANs promote invasiveness through modulation of angiogenesis and resistance to VEGF inhibition. In a colitis-associated CRC mouse model, Itatani et al. [119] demonstrated that the induction of colitis increased granulocyte colony stimulating factor (G-CSF) levels in anti-VEGF therapy-resistant tumors, which was associated with increased Bv8/PROK2+ neutrophil infiltration. The dual inhibition of G-CSF or its receptor along with VEGF inhibition resulted in significantly suppressed tumor formation and angiogenesis [119].

TANs also contribute to immune evasion. In an inducible CRC murine model, Germann et al. [120] demonstrated that presence of T cells in adenomas was diminished and inversely correlated with neutrophil infiltration. These TANs were enriched for immunosuppressive gene expression and secreted MMP-9, which suppressed T-cell proliferation through the activation of TGF-β [120]. Overall, TANs may also facilitate the dissemination of tumor cells by augmenting degradation of the basement membrane, which enhances tumor mobility and promotes tumor cell extravasation [121].

CTLs are the primary effector cells of the adaptive immune response and are a crucial component of antitumor immunity. The presence of CTLs in patient-derived CRC tumor cores is associated with a reduced risk of recurrence, particularly in higher-stage tumors [131]. Concordantly, the CRC Immunoscore prognostic tool was developed to supplement the TNM staging paradigm and is based on the prevalence of infiltrating CD3+ cells and CTLs. Higher Immunoscore values positively correlate with improved survival and are corroborated by internal and external validation studies [132,133].

To promote immune evasion, CTL activity is suppressed through multiple mechanisms. O’Malley et al. [134] demonstrated that TNF-α–mediated inflammation in CT26 CRC cells induced PD-L1 expression in stromal cells, resulting in the inhibition of activated granzyme-secreting CD8+ T cells [134]. Tumor and stromal cells express PD-L1, which, upon binding to the PD-1 receptor on T cells, suppresses T-cell activation. In CRC, a higher expression of PD-L1 is associated with higher stage, higher grade, lymph node involvement, distant metastasis, and reduced overall survival [135,136,137]. The PD-1/PD-L1 axis is an attractive target to increase immune targeting. Indeed, PD-1 blockade with pembrolizumab was FDA-approved for MSI-high/mismatch repair-deficient subtype advanced tumors in the pivotal KEYNOTE 177 trial [4,138]. Other TME cells also curtail CTL function through TGF-β secretion, which inhibits CTL expression of the key cytolytic enzymes: perforin, granzyme, and Fas ligand [139]. Furthermore, Tregs recruited via the CCL5/CCR5 signaling pathway have a heightened ability to kill CTLs in CRC murine models [140].

The ECM comprises structural proteins, including collagen, proteoglycans, hyaluronic acid, and glycosaminoglycans, whose structure and function can be remodeled by cells in the TME [142]. In addition to being the physical scaffolding for tumor cells, the ECM dynamically contributes to cell–cell adhesion, paracrine signaling, tumor proliferation, immune evasion, and metastasis [143]. Remodeling of the ECM is a consistent feature of the TME in many indications including CRC. Yang et al. [144] demonstrated that density of the remodeling enzyme MMP-9 was markedly greater in CRC compared to normal mucosa and higher expression was associated with worse prognosis. MMP-2 has also been implicated in ECM degradation and correlates with lymphatic invasion and advanced stage [145,146]. MMP-2 and MMP-9 encode collagenases that target type IV collagen, which is found in the basement membrane; and the degradation of type IV collagen can enhance tumor cell motility and invasion [145,147,148]. The biomechanics of the ECM also contribute to CRC proliferation. Using atomic force microscopy, Brauchle et al. [149] demonstrated that collagen-rich regions of the CRC ECM were stiffer compared to normal tissue. Stiffness of CRC has been associated with metastasis and EMT [150]. Li et al. [143] observed that the density of type I collagen positively correlated with stage, while type IV collagen density was reduced in higher stage tumors. This latter finding is consistent with invasion through the basement membrane [151]. The inverse relationship between density of type IV collagen and CRC stage likely reflects increased ECM remodeling, partly through MMP-2 and -9 activity [147]. At distant sites, proteomic analyses of control vs. matched CRLM patient tissues suggest that, while the ECM of CRLM more closely resembles the primary tumor compared to liver parenchyma, it is enriched for several unique proteins (i.e., SPP1 and COMP) [152]. Yuzhalin et al. [153] found that the CRLM ECM had higher levels of citrullinating enzyme peptidylarginine deiminase 4 (PAD4) compared to benign liver tissue, primary CRC, or colonic mucosa. CRC cells grown in citrullinated type I collagen exhibited mesenchymal-to-epithelial transition (reversion to a proliferative epithelial phenotype of tumor cells) in vitro and enhancement of CRLM growth in vivo [153].

Cancer-associated fibroblasts (CAFs) contribute significantly to ECM maintenance, desmoplasia, angiogenesis, immunosuppression, invasion, and chemoresistance [154,155]. While a majority of CAFs arise from resident stromal fibroblasts, mesenchymal stem cells (MSCs) and mesothelial cells also contribute a significant proportion of the CAF progenitor population [154,156]. Bone-marrow-derived MSCs, which are pluripotent stem cells involved in tissue remodeling, may be recruited to tumor sites to aid in tumor growth and progression [157]. Once localized at the tumor site, MSCs are induced by tumor-derived factors to assume a CAF-like phenotype [158]. Meanwhile, mesothelial cells contribute to the CAF population in highly invasive tumors via mesothelial-to-mesenchymal transition, which has been implicated in CRC peritoneal carcinomatosis [159]. Overall, in CRC, two major subpopulations of CAFs have been identified through single-cell RNA sequencing: CAF-A (which express ECM remodeling-associated genes, including MMP2, DCN, and COL1A2) and CAF-B (myofibroblast-like that express ACTA2, TAGLN, and PDGFA) [160].

CAFs promote immune evasion [120,154,161,162]. In microsatellite stable CRC samples, Tauriello et al. demonstrated an inverse correlation between the immunostimulatory T_H1 to naïve T-cell ratio and the mean expression of CAF-specific genes [162]. Tumor invasive margins had high levels of stromal TGF-β, of which CAFs were the primary source. The elevated expression of TGF-β in the TME inhibited T_H1 T-cell function. Additionally, CAFs secrete CXCL8, which attracts monocytes to the CRC TME and promotes M2 polarization, further contributing to immunosuppression [163].

CAFs promote metastasis through ECM remodeling and EMT in multiple cancers, including CRC [164,165,166,167]. Their presence in the CRC intratumoral stroma is associated with lymphatic invasion [168]. Through the secretion of ECM remodeling enzymes, collagen, and other cytoskeletal proteins, CAFs promote desmoplasia, which can be identified in up to 78% of CRC tumor specimens [168]. TGF-β/Smad2 signaling is prominently activated in CAFs, and CRC cells themselves can stimulate TGF-β production in CAFs with resulting expression of α-SMA and differentiation to a myofibroblastic phenotype [169]. These upregulate the expression of invasion-related proteinases, including multiple MMPs [169]. Notably, TAMs may regulate CAF activity in the CRC ECM with respect to the expression of collagen types I and XIV [74]. The expression of collagen crosslinking enzyme lysyl oxidase–like 2 (LOXL2) by CAFs in CRC tumors is associated with a higher rate of recurrence and worse overall survival and disease-free survival [170]. Through LOXL2, CAFs also stimulate EMT through activation of the FAK pathway, with a resultant reduction in E-cadherin protein expression. CRC cells co-cultured with these activated fibroblasts demonstrated increased migration rates [167]. CAFs also promote angiogenesis through secretion of IL-6; Nagasaki et al. [171] showed that CRC cells potentiated IL-6 secretion from CAFs with resulting upregulation of VEGFA expression. Thus, CAFs contribute to CRC progression through immunosuppression, ECM remodeling, and the promotion of EMT.

Tumor vasculature prominently impacts CRC invasion and is an established therapeutic target [180,181]. Endothelial cells (ECs), along with pericytes, smooth muscle cells, and progenitor cells, comprise the vasculature [182]. While EC generally remain quiescent in healthy tissue, malignancy can induce vascular turnover and cell proliferation [183]. Local injury, hypoxia, and rapid malignant proliferation initiate paracrine signaling networks where factors such as VEGF, PDGF, CXCL8, and TNF-α secreted by multiple TME cells drive neovascularization to perfuse proliferating tumor tissue [184]. ECs are implicated in invasion and metastasis. Zhang et al. [185] demonstrated that expression of vascular cell-adhesion molecular 1 (VCAM1) expression in CRC was associated with more invasive features and poorer prognosis. Forced expression of VCAM1 in CRC cells promoted pseudopodia formation and increased trans-endothelial migration in vitro. Furthermore, after CRC cells have entered the systemic vasculature, circulating tumor-derived ECs have been recently identified that may reflect a unique EC state that contributes directly to tumor progression and metastasis [186].