Converting cold to hot: Strategies to sensitize microsatellite-stable colorectal cancer to immunotherapy

Abstract

Colorectal cancer (CRC) is a leading cause of cancer-related mortality worldwide. While immune checkpoint inhibitors (ICIs) have transformed outcomes for patients with microsatellite-instability-high or mismatch-repair-deficient CRC across disease stages, the vast majority of tumors are microsatellite-stable (MSS) or proficient mismatch repair and remain largely refractory to current immunotherapies. These “cold” tumors are characterized by low tumor mutational burden and neoantigen load, impaired antigen presentation, exclusion or dysfunction of cytotoxic T cells, and a profoundly immunosuppressive tumor microenvironment shaped by oncogenic signaling pathways, stromal components, and the gut microbiota. This review summarizes the immunobiology of MSS CRC and dissects key mechanisms of primary and acquired resistance to ICIs, including Wnt/β-catenin, mitogen-activated protein kinase and transforming growth factor-β signaling, myeloid-derived suppressor cells, regulatory T cells, angiogenic and metabolic reprogramming, and organ-specific immunity in liver and peritoneal metastases. We then discuss emerging strategies to convert MSS CRC from cold to hot disease, focusing on rational ICI-based combinations (chemotherapy, anti-vascular endothelial growth factor/vascular endothelial growth factor receptor, targeted agents, radiotherapy, dual checkpoint blockade), cancer vaccines, oncolytic viruses, adoptive cell therapies, microbiota-directed interventions, and loco-regional immunotherapy approaches. Finally, we highlight evolving biomarkers and trial designs that may enable precision immunotherapy for MSS CRC and outline future directions for translational research and clinical practice.

Key Words: Adoptive cell therapy; Anti-angiogenic therapy; Biomarkers; Immune checkpoint inhibitors; Immune exclusion; Microsatellite-stable colorectal cancer; Tumor microenvironment; Vaccine

Core Tip: Most colorectal cancers are microsatellite-stable (MSS)/proficient mismatch repair and remain largely refractory to single-agent immune checkpoint inhibitors. This review dissects major resistance determinants low neoantigen burden, impaired antigen presentation, immune exclusion by cancer-associated fibroblast-rich stroma and abnormal vasculature, and myeloid-dominant suppression, particularly in liver metastases. We synthesize current evidence for rational “cold-to-hot” strategies, spanning chemo/radiotherapy priming, anti-angiogenic and anti-epidermal growth factor receptor regimens, mitogen-activated protein kinase/Wnt/transforming growth factor-β/CXCR4-signal transducer and activator of transcription 3 pathway targeting, and emerging vaccines, oncolytic viruses, cell therapies, and microbiome modulation. A biomarker-guided roadmap is proposed to personalize combination immunotherapy in MSS colorectal cancer.

INTRODUCTION

Colorectal cancer (CRC) is a major global health burden, ranking as the third most commonly diagnosed malignancy and the second leading cause of cancer-related death worldwide, with nearly 2 million new cases and about 900000 deaths reported in 2020-2022[1]. Despite advances in screening, minimally invasive surgery, and multimodal systemic therapy, many patients present with or subsequently develop metastatic disease, most commonly in the liver, followed by lung and peritoneum, and unresectable metastases remain the principal cause of death[2]. Modern chemotherapy combined with anti-vascular endothelial growth factor (VEGF) or anti-epidermal growth factor receptor (EGFR) agents has improved outcomes, but durable long-term survival is still achieved in only a minority of patients, underscoring the need for treatments that can establish lasting systemic immune control rather than transient cytoreduction[3].

Immune checkpoint inhibitors (ICIs) targeting programmed cell death protein-1 (PD-1), programmed death ligand-1 (PD-L1), and cytotoxic T-lymphocyte-associated protein-4 (CTLA-4) have transformed the management of several solid tumors[4]. In CRC, however, their success is largely confined to tumors with microsatellite instability-high (MSI-H) or deficient mismatch repair (dMMR). MSI-H/dMMR cancers are characterized by high tumor mutational burden (TMB), abundant frameshift-derived neoantigens, and a pre-existing inflamed tumor microenvironment (TME) with dense cluster of differentiation (CD) 8⁺ T-cell infiltration, which together confer pronounced sensitivity to PD-1 blockade. The phase III KEYNOTE-177 trial established pembrolizumab as a first-line standard of care in MSI-H/dMMR metastatic CRC (mCRC), while the CheckMate 142 study demonstrated high response rates and durable remissions with nivolumab, alone or combined with low-dose ipilimumab, in previously treated and treatment-naive MSI-H/dMMR mCRC[5,6]. These data represent a paradigm of biomarker-driven immunotherapy in gastrointestinal oncology.

By contrast, the much larger group of patients with microsatellite-stable (MSS) or mismatch repair-proficient (pMMR) CRC derives little benefit from current ICIs. Epidemiologic data indicate that dMMR/MSI-H tumors account for only 10%-15% of all CRCs and an even smaller fraction of metastatic cases, meaning that approximately 85%-95% of mCRC tumors are MSS/pMMR. These tumors usually have low TMB and sparse neoantigen landscapes, frequently exhibit impaired antigen processing and presentation, and are embedded in an immunosuppressive or immune-excluded TME enriched with tumor-associated fibroblasts, myeloid-derived suppressor cells (MDSCs), and regulatory T cells (Tregs). Consistent with this biology, PD-1/PD-L1 or CTLA-4 blockade alone yields objective response rates (ORRs) rarely exceeding 5% in unselected MSS populations and has not produced clear survival gains[7]. Even combinations with chemotherapy, targeted therapy, or anti-angiogenic agents have delivered only modest, context-dependent benefits. MSS CRC is therefore regarded as a prototypical “cold” tumor, characterized by ineffective T-cell priming, poor effector T-cell trafficking or retention, and activation of oncogenic and stromal pathways such as Wnt/β-catenin and transforming growth factor-β (TGF-β) that reinforce immune exclusion. Organ-specific immunologic niches, particularly within the liver, further dampen systemic antitumor immunity and predict poor ICI response in mCRC[8].

Closing this efficacy gap between MSI-H/dMMR and MSS/pMMR disease has become a central challenge in CRC immuno-oncology. Conceptually, the key objective is to convert MSS/pMMR tumors from an immune-cold, ICI-refractory state into an immune-hot, responsive phenotype by enhancing tumor antigenicity, restoring antigen presentation, overcoming stromal and myeloid barriers, and reprogramming the TME to support effective cytotoxic T-cell responses. A broad spectrum of strategies is under investigation, including rational ICI-based combinations, activation of innate immune pathways, oncolytic viruses (OVs) and vaccines, adoptive cell therapies, and interventions targeting the gut microbiota and tumor metabolism[8]. In this review, we focus on the immunobiology of MSS CRC, summarize clinical experience with ICIs in this setting, and discuss emerging approaches aimed at rendering MSS disease amenable to immunotherapeutic control.

Beyond reporting IMblaze370 as “negative”, the trial is mechanistically informative and highlights why mitogen-activated protein kinase (MAPK)/MAPK kinase (MEK) blockade did not reliably sensitize MSS CRC to PD-L1 inhibition[9]. First, MEK inhibition is immunologically pleiotropic and context dependent: While tumor-intrinsic MAPK suppression has been proposed to increase antigen presentation and favor T-cell infiltration, systemic and continuous MEK blockade can concurrently dampen T-cell priming/activation [e.g., attenuating T-cell receptor (TCR) signaling, interleukin (IL)-2 dependent expansion, and effector cytokine output] as well as antigen-presenting cell function, potentially offsetting any tumor-intrinsic “sensitization” when administered in a chronic schedule. Second, the dominant resistance architecture in MSS CRC myeloid-driven immunosuppression, TGF-β/cancer-associated fibroblasts (CAFs)-mediated stromal exclusion, and abnormal vasculature limiting trafficking is unlikely to be dismantled by MEK + PD-L1 alone, such that modest immunogenic modulation may not translate into durable tumor control. Third, the late-line population and an active control arm can further narrow the therapeutic margin, making small biological gains clinically unapparent. Collectively, IMblaze370 argues that MEK inhibitors, if used as immune primers in MSS CRC, likely require schedule optimization (short priming or pulsed dosing) and/or multi-node combinations that simultaneously restore priming and trafficking rather than relying on PD-L1 blockade alone.

IMMUNOBIOLOGY OF MSS CRC

MSS or pMMR tumors account for roughly 85% of all CRC and more than 90% of metastatic cases, and they are overwhelmingly responsible for the “immune-cold” clinical phenotype and poor responsiveness to ICIs[10]. Unlike dMMR/MSI-H tumors, which show dense lymphocytic infiltrates and high neoantigen burden, MSS CRC is typically characterized by low TMB, few neoantigens, and a profoundly immunosuppressive TME with features of immune exclusion or immune desert[10]. These biological differences form the mechanistic basis for the limited efficacy of PD-1/PD-L1 blockade in unselected MSS CRC and underpin the current interest in “heating up” these tumors with rational combinations.

Molecular subtypes and immune phenotypes of CRC

Transcriptomic profiling led to the consensus molecular subtype (CMS) classification of CRC into four major groups (CMS1-4)[11]. CMS1 (“MSI-immune”) comprises about 14% of CRC and is enriched for MSI-H tumors, hypermutation, strong T helper (Th) 1/CD8⁺ T-cell infiltration and high expression of immune checkpoints. CMS2 (“canonical”) is the most common (approximately 35%), dominated by chromosomal instability and Wnt/β-catenin activation. CMS3 (“metabolic”) is characterized by metabolic reprogramming (KRAS mutations, altered glycolysis and lipid metabolism), whereas CMS4 (“mesenchymal”) shows a prominent stromal and angiogenic signature with abundant CAFs, TGF-β signaling and the worst prognosis[12].

From an immune standpoint, CMS1 and CMS4 are “immune-rich” but qualitatively distinct: CMS1 tumors display adaptive immune activation with cytotoxic and memory T-cell infiltrates, while CMS4 is heavily infiltrated by myeloid cells, CAFs and immunosuppressive factors (e.g., TGF-β, VEGF, IL-6). CMS2 and CMS3, in contrast, are immunologically “cold”, with low expression of immune-related genes, sparse lymphoid and myeloid infiltration and reduced major histocompatibility complex (MHC) expression[13]. Because MSI-H tumors cluster predominantly within CMS1, the large majority of MSS/pMMR tumors belong to CMS2-4. CMS2/3 typically correspond to immune-desert or immune-excluded phenotypes, while CMS4 combines intense stromal remodeling with an immunosuppressive, “inflamed-but-ineffective” immune infiltrate[10].

At the level of spatial immune architecture, MSS CRC frequently falls into the “immune-excluded” or “immune-desert” categories defined across solid tumors: Immune-inflamed tumors harbor CD8⁺ T cells within tumor nests; immune-excluded tumors show T cells trapped in the peritumoral stroma; immune-desert tumors lack meaningful effector infiltration altogether[14]. MSS CRC is enriched for the latter two patterns, which strongly correlates with primary resistance to ICIs.

TMB, neoantigens, and antigen presentation

MSI-H CRC typically exhibits a TMB > 10 mutations/Mb with frequent frameshift indels in coding microsatellites, creating abundant immunogenic neoepitopes that drive brisk lymphocytic infiltration and sensitivity to PD-1 blockade[10]. In contrast, most MSS tumors display a low to intermediate TMB (often approximately 2-5 mutations/Mb) and fewer predicted neoantigens, limiting the probability that T cells will recognize tumor cells as foreign[10]. Rare hypermutated MSS subsets driven by POLE/POLD1 mutations form an exception and may behave more like MSI-H tumors immunologically, but they represent only a small fraction of CRC.

Defects in antigen processing and presentation further blunt immunogenicity in MSS CRC. Multiple studies have reported downregulation or loss of MHC class I molecules on CRC cells, often through loss of heterozygosity or mutations affecting human leukocyte antigen (HLA) heavy chains, β2-microglobulin (B2M), or components of the antigen-processing machinery (e.g., TAP1/TAP2, NLRC5)[13]. Even in tumors without structural HLA alterations, oncogenic signaling and epigenetic repression can suppress HLA class I expression, reducing CD8⁺ T-cell recognition. Furthermore, interferon (IFN) signaling defects [e.g., Janus kinase (JAK) 1/2 loss-of-function] and chronic exposure to TGF-β or IL-10 can impair antigen presentation and dampen the transcriptional response to inflammatory cytokines[15].

Collectively, low TMB/neoantigen load and impaired antigen presentation converge to create a “poorly visible” tumor to the immune system, helping explain the minimal activity of single-agent ICIs in unselected MSS CRC.

The TME of MSS CRC

The TME of MSS CRC is a complex ecosystem in which tumor cells, immune cells, CAFs, endothelial cells, extracellular matrix (ECM) components and the gut microbiota co-evolve[10]. Compared with MSI-H tumors, MSS lesions show lower densities of Th1, CD8⁺ cytotoxic T cells and follicular helper T cells, but enrichment of immunosuppressive populations such as Tregs, tumor-associated macrophages (TAMs) and MDSCs, alongside a dense desmoplastic stroma[10]. These features underpin the predominant immune-excluded or immune-desert phenotypes.

Effector T-cell exclusion and dysfunction: In MSS CRC, CD8⁺ T cells, when present, are frequently confined to the invasive margin or peritumoral stroma rather than infiltrating the tumor core[10]. Spatial transcriptomic and pathology-based analyses have linked nuclear β-catenin activation and stromal exclusion programs to reduced intratumoral CD8⁺ T-cell density[16]. Even when T cells reach tumor nests, they often exhibit an exhausted phenotype characterized by high expression of PD-1, T cell immunoglobulin and mucin domain-containing protein 3 (TIM-3), lymphocyte activation gene-3 (LAG-3), and T cell immunoglobulin and ITIM domain (TIGIT), diminished cytokine production, and impaired cytotoxicity[16]. Mechanistically, this dysfunction reflects a convergence of chronic antigen stimulation, metabolic stress (hypoxia, acidosis, and nutrient competition within the TME), and soluble immunosuppressive signaling that jointly constrains priming, trafficking, and effector function (see the integrated IL-10/TGF-β/VEGF node in “shared soluble immunosuppressive nodes: The IL-10/TGF-β/VEGF axis”)[10]. Collectively, these spatial and functional defects in effector T cells are major determinants of the immune-cold state and represent central obstacles to ICI efficacy.

Immunosuppressive myeloid and lymphoid populations: MSS CRC is enriched in MDSCs and M2-polarized TAMs, which suppress T-cell proliferation and cytotoxicity via arginine depletion (ARG1), nitrosative/oxidative stress, inhibitory ligand expression, and immunoregulatory programs that reinforce tolerance[10]. Spatial single-cell studies further implicate niche-specific interactions such as HLA-G⁺ cancer cells engaging SPP1⁺ (osteopontin-expressing) macrophages at the invasive front in shaping a desmoplastic, immunosuppressive microenvironment associated with poor prognosis and resistance to immunotherapy.

Tregs also accumulate in MSS CRC and dampen antitumor immunity through CTLA-4 mediated suppression, competition for key metabolites, and adaptation to hypoxic, glycolysis-dominated conditions (e.g., enhanced glucose transporter type 1-dependent glycolytic flux), which sustains their survival and suppressive capacity while impairing CD8⁺ T-cell function. Together, these myeloid and lymphoid suppressor populations form a myelo-lymphoid barrier that can neutralize nascent cytotoxic responses induced by chemotherapy, radiotherapy, or targeted agents. Importantly, many of these suppressive effects are functionally integrated through the shared soluble hub described in “shared soluble immunosuppressive nodes: The IL-10/TGF-β/VEGF axis”, rather than operating as isolated cytokine pathways.

Stromal barriers, vasculature, and ECM: CAFs are dominant stromal components in MSS CRC and are particularly abundant in CMS4 tumors. They remodel the ECM by depositing collagen, fibronectin, and proteoglycans, increasing tissue stiffness and generating a dense desmoplastic reaction that physically impedes T-cell infiltration. CAFs also shape immune exclusion through chemokine and growth-factor programs (e.g., CXCL12 and HGF) that promote tumor cell survival and recruit suppressive myeloid populations, while stromal TGF-β signaling can consolidate exclusionary matrix states (see “shared soluble immunosuppressive nodes: The IL-10/TGF-β/VEGF axis”)[17,18].

In parallel, abnormal, leaky, and poorly perfused tumor vasculature produces hypoxia and acidosis and contributes to endothelial dysfunction (“endothelial anergy”), collectively impairing effector T-cell trafficking and amplifying immune escape[19]. VEGF-driven vascular pathology is best interpreted as part of an integrated “trafficking failure” module within the IL-10/TGF-β/VEGF axis (“shared soluble immunosuppressive nodes: The IL-10/TGF-β/VEGF axis”), providing a mechanistic rationale for vascular/immune normalization strategies in combination regimens.

Thus, the CAF-rich stroma, aberrant vasculature, and stiff ECM constitute a coupled physical and biochemical barrier underlying the immune-excluded phenotype characteristic of many MSS CRCs.

Shared soluble immunosuppressive nodes: The IL-10/TGF-β/VEGF axis: Across the barriers described above, several “soluble hubs” recur and function as integrative nodes rather than isolated factors. IL-10, TGF-β, and VEGF collectively couple tumor-intrinsic programs to microenvironmental and systemic immune failure by converging on three actionable bottlenecks: Priming, trafficking, and effector function. First, these mediators impair priming by promoting tolerogenic antigen-presenting cells; in CRC, dendritic-cell maturation and antigen presentation can be inhibited in a coordinated manner by TGF-β, IL-10, and VEGF, biasing the immune contexture toward tolerance rather than productive tumor-reactive T-cell activation[20]. Second, TGF-β-driven fibroblast activation and matrix remodeling establish a physical and biochemical exclusion barrier that limits T-cell entry and sustains immune-evasive phenotypes, a theme repeatedly observed in CRC stromal biology[21]. Third, VEGF-mediated vascular dysfunction reinforces trafficking defects through hypoxia, endothelial anergy, and aberrant vessel architecture; the concept of vascular and immune normalization provides a mechanistic rationale for combining anti-angiogenic strategies with immunotherapy to restore perfusion and immune infiltration[22].

Importantly, this axis is not purely descriptive: It is clinically and experimentally actionable. For example, IL-10 blockade has been shown to potentiate anti-tumor immune function in human CRC liver metastases, highlighting IL-10 as a tractable suppressive node in metastatic immune failure[23]. Likewise, TGF-β and VEGF pathways are repeatedly implicated as dominant drivers of immune suppression and resistance programs across solid tumors, reinforcing the need for multi-node strategies rather than single-factor “lists”[24].

In the remainder of this review, we therefore reference IL-10/TGF-β/VEGF primarily as an integrated “node” that explains recurrent patterns of immune escape across barriers, and we focus barrier-specific sections on the cell/ecosystem contexts in which this node is wired (myeloid, fibroblast, endothelial, and metastatic niches).

Oncogenic and signaling pathways driving immune evasion

Oncogenic signaling in CRC not only drives proliferation and metastasis but also actively shapes the immune landscape. Wnt/β-catenin activation especially prominent in CMS2 correlates with reduced CD8⁺ T-cell infiltration and poor response to immunotherapy. Mechanistically, β-catenin signaling suppresses CCL4/CCL5 production, limiting the recruitment of Batf3⁺ dendritic cells (DCs) required for priming tumor-specific CD8⁺ T cells, and can promote Treg stability and Th17 polarization[25].

KRAS and BRAF mutations, present in approximately 40%-50% and approximately 8%-10% of CRCs respectively, hyperactivate the RAS/RAF/MEK/extracellular signal-regulated kinase cascade[26]. This signaling axis upregulates PD-L1 expression, induces pro-inflammatory yet immunosuppressive cytokines (e.g., IL-6, IL-8) and enhances recruitment of MDSCs and TAMs, collectively promoting immune evasion[26]. Similarly, the phosphatidylinositol 3-kinase (PI3K)/protein kinase B (AKT)/mammalian target of rapamycin (mTOR) pathway frequently activated via PIK3CA mutations or PTEN loss drives tumor growth and is tightly linked to PD-L1 expression and metabolic competition with T cells, thereby reinforcing an exhausted T-cell phenotype[15].

TGF-β signaling, especially in stromal cells, is a hallmark of MSS/CMS4 CRC and is strongly associated with T-cell exclusion, desmoplasia and resistance to PD-1/PD-L1 blockade[27]. JAK/signal transducer and activator of transcription (STAT) pathways also play dual roles: IL-6/STAT3 activation in tumor and myeloid cells supports immunosuppressive polarization, whereas defects in IFN-γ/JAK1/2 signaling can abrogate antigen presentation and PD-L1 induction, contributing to ICI resistance[28]. These oncogenic and cytokine pathways therefore represent critical nodes at which targeted therapies might be combined with ICIs to reprogram the MSS TME.

Gut microbiota and metabolomic influences on antitumor immunity

The gut microbiota is an integral component of the CRC TME. Patients with CRC exhibit characteristic dysbiosis, including enrichment of Fusobacterium nucleatum, enterotoxigenic Bacteroides fragilis and certain Escherichia coli strains, alongside depletion of beneficial genera such as Bacteroides and Lactobacillus[29,30]. These bacteria can directly damage DNA (e.g., colibactin), activate pro-inflammatory pathways (nuclear factor kappa-B, STAT3, Wnt/β-catenin) and secrete adhesins/toxins (FadA, BFT) that foster tumorigenesis and immunosuppression[31]. Fusobacterium nucleatum, for example, engages TIGIT on T and natural killer (NK) cells via Fap2 and activates a β-catenin/CCL28 axis that enriches FOXP3⁺ Tregs, thereby dampening antitumor immunity and promoting chemotherapy resistance[32].

Microbiota-derived metabolites further modulate the MSS CRC immune milieu. Short-chain fatty acids (SCFAs) such as acetate, propionate and butyrate, secondary bile acids, and tryptophan metabolites (indoles, kynurenine) can act locally in the colon and systemically[33]. SCFAs have context-dependent effects: They can enhance epithelial barrier integrity and CD8⁺ T-cell function but also promote Treg expansion or, in some settings, drive metabolic adaptations that confer resistance to ICIs for example, butyrate-induced CPT1A-dependent fatty acid oxidation has been linked to resistance to anti-PD-1 therapy in CRC models[34].

Metabolic reprogramming of tumor and stromal cells in MSS CRC high glycolytic flux, lactate accumulation, altered lipid metabolism and enhanced tryptophan catabolism creates an immunosuppressive biochemical landscape. Lactate accumulation acidifies the TME, impairs CTL and NK-cell function, supports MDSC recruitment (e.g., via HCAR1 signaling) and stimulates CAFs to secrete IL-6, collectively promoting immune escape and resistance to ICI[35]. The tryptophan kynurenine pathway, driven by IDO1/TDO and stabilized by factors such as USP14, depletes tryptophan and accumulates kynurenine, which inhibits effector T and NK cells while expanding Tregs and MDSCs; targeting IDO1 or its regulators sensitizes CRC models to PD-1 blockade[36-38].

Altogether, dysbiotic microbiota and metabolite-driven immunometabolic rewiring are key contributors to the immune-cold, therapy-resistant state of MSS CRC. They also offer attractive, and potentially reversible, levers through diet, probiotics, antibiotics, metabolite mimetics or enzyme inhibitors to modulate the TME and enhance responses to immunotherapy in this large patient population (Table 1). These multi-layered determinants of the immune-cold phenotype in MSS CRC are summarized in Figure 1.

Figure 1 Immunobiology of “cold” microsatellite-stable/mismatch repair-proficient colorectal cancer and multilevel barriers to immune checkpoint blockade. Concentric layers summarize how tumor cell-intrinsic programs, the local tumor microenvironment (TME), and systemic gut-liver crosstalk converges to generate an immune-excluded, immune checkpoint inhibitor (ICI)-refractory phenotype in microsatellite-stable (MSS) colorectal cancer (CRC). Tumor cell-intrinsic mechanisms include low tumor mutational burden and limited neoantigenicity, impaired antigen processing/presentation (e.g., β2-microglobulin or Janus kinase/signal transducer and activator of transcription pathway alterations), and oncogenic signaling (Wnt/β-catenin and KRAS/BRAF/mitogen-activated protein kinase) that reduces cluster of differentiation 8⁺ T-cell infiltration and promotes programmed death ligand-1 upregulation and immune evasion; tumor- and stroma-derived transforming growth factor-β (TGF-β) further reinforces immune suppression and exclusion. In the TME, dense cancer-associated fibroblast-rich desmoplastic stroma increases tissue stiffness and forms a physical barrier, while aberrant vasculature and hypoxia impair T-cell trafficking. Immunosuppressive populations including myeloid-derived suppressor cells and M2-like tumor-associated macrophages producing interleukin-10, TGF-β, and arginase-1, along with regulatory T cells blunt effector function. Metabolic stress (glycolysis-driven lactate accumulation and acidic potential of hydrogen) further suppresses T-cell activity. At the systemic/gut-liver axis level, dysbiotic microbiota (e.g., Fusobacterium and enterotoxigenic Bacteroides fragilis) and microbiota-derived metabolites (bile acids, short-chain fatty acids, and tryptophan-kynurenine pathway products) reshape systemic immunity and therapeutic responsiveness; the tolerogenic liver niche promotes immune privilege of liver metastases. Collectively, these factors yield a composite immune-cold state characterized by scarce functional T cells, dominant immunosuppressive myeloid/stromal elements, and microbiome-linked resistance, underpinning primary non-response to ICIs in MSS CRC. TMB: Tumor mutational burden; TME: Tumor microenvironment; B2M: Β2-microglobulin; JAK/STAT: Janus kinase/signal transducer and activator of transcription; MAPK: Mitogen-activated protein kinase; PD-L1: Programmed death-ligand 1; CD: Cluster of differentiation; TGF-β: Transforming growth factor-β; ARG1: Arginase-1; MDSC: Myeloid-derived suppressor cell; IL: Interleukin; CAF: Cancer-associated fibroblast; TAM: Tumor-associated macrophage; SCFA: Short-chain fatty acid; ETBF: Enterotoxigenic Bacteroides fragilis.

Table 1 Major immunobiological barriers to immune checkpoint inhibition in microsatellite-stable/proficient mismatch repair colorectal cancer and actionable therapeutic opportunities.

Immune barrier/feature	Mechanistic basis	Representative readouts/contexts	Actionable strategies
Low immunogenicity and weak antigen presentation	Low mutation load (MSS/Low TMB); impaired APC priming; HLA/B2M loss; IFN-JAK/STAT defects	Low TMB; low IFN-γ/CXCL9-CXCL10; MHC-I downregulation	Chemo/RT priming; STING/TLR agonists; epigenetic priming; vaccines/neoantigen approaches[13-15,25]
T cell exclusion by stromal/TGF-β programs	CAF-derived TGF-β; dense ECM; reduced trafficking/retention immuneexcluded phenotype	CMS4/stromal signature; high TGF-β; desmoplasia	TGF-β blockade/traps; CAF/ECM modulation; CXCR4 blockade; vessel normalization[27,28]
Oncogenic signaling-driven immune escape	KRAS/BRAF/MAPK or WNT/β-catenin suppress antigen presentation/chemokines and promote myeloid recruitment	RAS/BRAF mutations; MAPK/WNT activation signatures	Targeted therapy to reprogram TME (e.g., BRAF/MEK/EGFR); rational combinations with ICI[15,26]
Myeloid-dominant suppression (TAM/MDSC)	VEGF/CSF1/IL-8 axis; MDSC expansion; TAM M2 polarization; arginase/ROS-mediated T-cell suppression	High MDSC/TAM signatures; high NLR; VEGF-high tumors	Anti-VEGF/VEGFR TKIs; CSF1R/CXCR1/CXCR2 inhibition; CD47/SIRPα blockade (investigational)[10,16,19]
Checkpoint redundancy and T cell exhaustion	Co-expression of PD-1/PD-L1 with CTLA-4/LAG-3/TIGIT/TIM-3 dysfunctional CD8 pool	Exhaustion signatures; multiple checkpoints on TILs	Dual/next-gen checkpoint blockade; costimulatory agonists; intratumoral delivery[10,14]
Immunometabolic suppression	Hypoxia, lactate; adenosine (CD39/CD73); tryptophan catabolism (IDO/TDO); nutrient competition	High CD73; hypoxia markers; kynurenine/adenosine signatures	A2A/CD73 axis inhibitors; metabolic modulation; normalization of hypoxia/vasculature[35,36]
Microbiome-driven immune modulation	Dysbiosis (e.g., Fusobacterium) shapes myeloid programs and alters T cell function via microbial metabolites	Stool metagenomics; Fusobacterium abundance; bile acid/SCFA profiles	Diet/probiotic/FMT strategies; antibiotic stewardship; microbiome-informed stratification[33,34]
Metastatic niche effects (especially liver)	Hepatic tolerogenic myeloid cells; sequestration/deletion of activated T cells systemic suppression	Presence of liver metastases; low intrahepatic CD8 density	Liver-directed RT/SIRT/TACE + ICI; consider liver metastasis status in trial design[45,66,70]

TGF-β: Transforming growth factor-β; MDSC: Myeloid-derived suppressor cell; TAM: Tumor-associated macrophage; MSS: Microsatellite stable; TMB: Tumor mutational burden; APC: Antigen-presenting cell; B2M: Β2-microglobulin; HLA: Human leukocyte antigen; JAK/STAT: Janus kinase/signal transducer and activator of transcription; IFN: Interferon; CAF: Cancer-associated fibroblast; ECM: Extracellular matrix; MAPK: Mitogen-activated protein kinase; VEGF: Vascular endothelial growth factor; IL: Interleukin; CSF: Colony stimulating factor; PD-1: Programmed cell death protein 1; PD-L1: Programmed death-ligand 1; ROS: Reactive oxygen species; CTLA-4: Cytotoxic T-lymphocyte-associated protein-4; LAG-3: Lymphocyte activation gene-3; TIGIT: T cell immunoglobulin and ITIM domain; CD: Cluster of differentiation; TIM-3: T cell immunoglobulin and mucin domain-containing protein 3; MHC: Major histocompatibility complex; CMS: Consensus molecular subtype; NLR: Neutrophil-to-lymphocyte ratio; TIL: Tumor-infiltrating lymphocyte; SCFA: Short-chain fatty acid; RT: Radiotherapy; STING: Stimulator of interferon genes; TLR: Toll like receptor; TME: Tumor microenvironment; MEK: Mitogen-activated protein kinase kinase; EGFR: Epidermal growth factor receptor; ICI: Immune checkpoint inhibitor; VEGFR: Vascular endothelial growth factor receptor; TKI: Tyrosine kinase inhibitor; SIRPα: Signal regulatory protein alpha; FMT: Fecal microbiota transplantation; SIRT: Selective internal radiation therapy; TACE: Transarterial chemoembolization.

CLINICAL EXPERIENCE WITH ICIS IN MSS CRC

Limited efficacy of ICI monotherapy in MSS/pMMR disease

Early clinical experience with PD-1 blockades in unselected mCRC provided the clearest demonstration of the immunologic divide between MSI-H/dMMR and MSS/pMMR tumors. In the pivotal phase II trial by Le et al[39], pembrolizumab achieved an immune-related ORR of 40% in dMMR mCRC, whereas no objective responses were observed among 18 patients with pMMR MSS mCRC; disease control was only 11%, with a median progression-free survival (PFS) of 2.2 months and median overall survival (OS) of 5.0 months in the pMMR cohort[37]. Subsequent analyses and real-world data have consistently confirmed that PD-1 monotherapy has essentially no meaningful activity in MSS/pMMR CRC, in stark contrast to its durable benefit in MSI-H/dMMR disease[38].

Large phase II and III trials of PD-1/PD-L1 antibodies have primarily focused on MSI-H/dMMR CRC, where pembrolizumab (KEYNOTE-164/177) and nivolumab, alone or combined with ipilimumab (CheckMate-142), have produced ORRs of approximately 30%-60% and long median PFS/OS[39]. In these studies, patients with MSS tumors were either excluded or, when included in early cohorts, showed negligible responses, reinforcing the notion that MSI status is a dominant predictive biomarker.

In MSS/pMMR-dominant mCRC, single-agent PD-L1 blockade has repeatedly shown minimal clinical activity. In the phase III IMblaze370 trial (92% confirmed MSS), neither atezolizumab monotherapy nor atezolizumab plus the MEK inhibitor cobimetinib improved OS vs regorafenib (median OS: 7.10 months vs 8.87 months vs 8.51 months), and objective responses remained in the low single digits (ORR: 2% vs 3% vs 2%, respectively)[40]. Importantly, this negative result is biologically plausible and highlights a key translational pitfall: MEK blockade can be immunologically bidirectional. While MAPK inhibition may enhance tumor-cell antigen presentation and transiently favor intratumoral immune remodeling in selected contexts, continuous systemic MEK inhibition can simultaneously blunt T-cell priming/expansion and effector differentiation, thereby limiting the magnitude and durability of anti-tumor immunity even when PD-L1 is blocked. In addition, the deeply layered resistance architecture typical of MSS CRC (myeloid-driven suppression, stromal/TGF-β linked exclusion, and VEGF-associated vascular dysfunction) is unlikely to be dismantled by a simple MEK⁺ PD-L1 doublet in unselected late-line populations[9,41]. Collectively, these data argue that future MEK-based immunotherapy strategies should prioritize schedule optimization (short priming or intermittent/pulsed dosing) and multi-node combinations that concurrently restore priming, trafficking, and effector function rather than relying on a two-drug doublet[40,42].

Collectively, these data indicate that ICI monotherapy offers little benefit to the vast majority of patients with MSS/pMMR CRC. The low mutational burden, paucity of neoantigens, and “cold” or immune-excluded TME characteristic of MSS tumors provide a biologic explanation for this primary resistance and have motivated intensive efforts to develop rational combination strategies[43].

Organ-specific immunity: Liver and peritoneal metastases

Liver metastases as an “immune-privileged” site: Beyond global MSS biology, accumulating evidence indicates that metastatic topology particularly hepatic involvement critically shapes responses to checkpoint blockade. The liver is a uniquely tolerogenic organ continuously exposed to gut-derived antigens and enriched in specialized antigen-presenting and scavenger populations, including Kupffer cells, liver sinusoidal endothelial cells, and tolerogenic DCs. These hepatic circuits promote deletion, dysfunction, or regulatory skewing of activated CD8⁺ T cells and support expansion of regulatory programs through coordinated suppressive signaling (conceptually captured by the integrated IL-10/TGF-β/VEGF suppressive hub, “shared soluble immunosuppressive nodes: The IL-10/TGF-β/VEGF axis”), thereby enforcing systemic immune restraint[44]. In preclinical liver-metastasis models, cross-presentation of tumor antigens by hepatic myeloid cells can induce apoptosis or functional impairment of tumor-specific cytotoxic T lymphocytes within the liver and, importantly, blunt distal antitumor immunity even at extrahepatic sites[44]. Together, these findings support the concept of liver metastases as an “immune sink” that can reprogram systemic antitumor immunity rather than merely adding tumor burden.

Clinically, the adverse impact of liver metastases on ICI outcomes in MSS CRC is increasingly well documented. In a retrospective cohort of 95 heavily pretreated MSS mCRC patients treated with PD-1/PD-L1 targeting regimens, Wang et al[45] reported an overall ORR of 8.4%; notably, ORR was 19.5% in patients without liver metastases but 0% in those with hepatic involvement, with median PFS of 4.0 months vs 1.5 months, and multivariable analysis identifying liver metastases as the strongest negative factor for progression under immunotherapy (hazard ratio approximately 7.0)[45]. A similar site-dependent pattern has been reported across regorafenib-based combinations (including REGONIVO and follow-up cohorts), where responses were enriched in lung-predominant disease yet declined to single digits or absent in patients with liver metastases[45]. While these datasets are heterogeneous in design and geography, their convergence reinforces metastatic site as a clinically meaningful modifier of immunotherapy benefit in MSS CRC.

Collectively, these observations argue that CRC liver metastases constitute an immune-privileged sink that sequesters and inactivates effector T cells and can limit systemic benefit from PD-1/PD-L1 blockade. Practically, this supports incorporating metastatic pattern into patient selection and trial stratification for ICI-containing regimens, and it provides a rationale for liver-directed or myeloid/vascular/stromal-targeted strategies aimed at restoring priming and trafficking within hepatic lesions to unlock systemic responses in this subgroup.

Peritoneal metastases and unique immune constraints: Peritoneal metastases from CRC constitute a distinct metastatic niche with unique anatomic and immunologic constraints. Clinically, peritoneal carcinomatosis is associated with poor prognosis and poses delivery challenges for systemic therapies, as disease often presents as diffuse, multifocal nodules on peritoneal surfaces with heterogeneous vascular access[46]. Biologically, the peritoneal cavity is a fluid-rich and frequently hypoxic, inflammatory, and fibrotic environment enriched in TAMs, mesothelial cells, and myofibroblast-like stromal cells. Together, these cellular programs reinforce immune suppression through coordinated soluble signaling and barrier formation conceptually captured by the integrated IL-10/TGF-β/VEGF suppressive hub described in “shared soluble immunosuppressive nodes: The IL-10/TGF-β/VEGF axis” thereby limiting effective priming, trafficking, and effector function within peritoneal deposits.

Importantly, emerging translational evidence suggests that peritoneal metastases are not uniformly “immune deserts”. Sundström et al[47] profiled immune infiltrates in CRC peritoneal metastases (CRC-PM) and observed immune populations in all lesions; infiltrating CD8⁺ T cells often expressed exhaustion markers such as PD-1 yet retained the capacity to mount Th1-type cytokine responses upon ex vivo stimulation, implying a potentially reversible dysfunction that could be amenable to checkpoint-based reinvigoration in selected contexts. Comparative transcriptomic and immune deconvolution analyses further support that immune composition differs across metastatic sites (liver, lung, peritoneum), and that peritoneal deposits may exhibit a site-specific balance of effector and regulatory subsets that could variably shape ICI responsiveness.

From a therapeutic standpoint, cytoreductive surgery plus hyperthermic intraperitoneal chemotherapy (CRS-HIPEC) remains a mainstay for selected patients with CRC-PMs. Beyond cytoreduction, HIPEC may provide an immunologic “priming” component: Experimental and early clinical observations suggest that heated intraperitoneal chemotherapy can promote immunogenic cell death and increase CD8⁺ T-cell infiltration, supporting the rationale for combining intensified local therapy with systemic immunomodulation[48]. Consistent with this concept, preclinical work integrating HIPEC-like hyperthermia with immune stimulation indicates that locally intensified treatment within the peritoneal cavity may help relax physical and immunologic barriers and expand a window for checkpoint blockade[49]. However, prospective clinical data for ICIs specifically in CRC-PMs remain limited, and defining optimal sequencing (peri-HIPEC vs post-HIPEC), combination partners (e.g., anti-angiogenic or stromal/myeloid-directed agents), and site-aware endpoints represents a key research priority.

Molecular correlates of primary and acquired resistance

The near-universal primary resistance of MSS/pMMR CRC to ICI monotherapy reflects coordinated constraints spanning tumor-intrinsic immunogenicity and microenvironmental permissiveness. Rather than a list of correlates, available multi-omic and translational data can be organized into three actionable bottlenecks that determine ICI efficacy: Priming competence, trafficking/infiltration, and effector persistence.

At the priming level, MSS CRC typically exhibits low antigenicity (low TMB and paucity of clonal neoantigens) and tumor-intrinsic programs (e.g., Wnt/β-catenin) that limit productive de novo T-cell priming; exclusionary stromal programs further restrict access to tumor nests[49]. In addition, suppressive soluble signaling captured as an integrated IL-10/TGF-β/VEGF hub (“shared soluble immunosuppressive nodes: The IL-10/TGF-β/VEGF axis”) can bias antigen presentation toward tolerance and blunt immune initiation[50].

Prospective biomarker analyses from combination trials provide more granular guidance. In TASNIVO (anti-PD-1 plus anti-angiogenic therapy), responders were enriched for POLE mutations or other features consistent with higher mutational load and an “inflamed” signature (IFN-γ signaling, cytotoxic markers, CXCL9/CXCL10), whereas non-responders showed stromal-rich, TGF-β high profiles with elevated myeloid/Treg signatures and multiple inhibitory checkpoints (LAG-3, TIM-3, TIGIT), consistent with dominant exclusion and dysfunction[51]. These findings support a practical interpretation: MSS CRC benefit is most likely when regimens co-target priming and trafficking constraints rather than intensifying PD-L1 blockade alone.

When responses occur under combination regimens, acquired resistance may emerge through erosion of immune recognition. Canonical routes include defects in IFN-γ signaling (e.g., JAK1/2 alterations) and antigen presentation (e.g., B2M loss), which can disrupt inducible MHC-I and impair CD8⁺ T-cell recognition[52]. However, CRC-specific context matters: Analyses in MSI-H CRC suggest B2M and JAK1/2 alterations do not uniformly confer resistance, underscoring the need for disease-tailored validation of putative escape mechanisms[53]. PD-L1 and HLA class I expression are also heterogeneous and dynamic, enabling region-specific escape under drug pressure and amplifying resistance in low-immunogenic tumors[54].

Collectively, these correlates indicate that MSS CRC resistance is typically a composite phenotype limited antigenicity, stromal-myeloid exclusion, and fragile effector function, often amplified by organ-specific niches thereby motivating cold-to-hot strategies that are bottleneck-matched, schedule-aware, and multi-node by design, which we detail in the next section.

STRATEGIES TO CONVERT MSS CRC FROM “COLD” TO “HOT”

Accordingly, we propose a bottleneck-matched framework in which composite readiness scores inform both the selection and sequencing of priming-, trafficking-, and effector-restoring combinations, rather than applying uniform regimens to unselected MSS CRC populations. MSS CRC is typically immunologically “cold”, characterized by limited T-cell infiltration and poor responsiveness to ICIs. To overcome these constraints, a broad range of “cold-to-hot” combination strategies is being explored with the goal of increasing tumor immunogenicity, facilitating immune cell recruitment and penetration, and ultimately restoring effective antitumor effector function. In this section, we synthesize the principal approaches, including ICIs combined with chemotherapy, anti-angiogenic agents, targeted therapies, radiotherapy, vaccines, OVs, adoptive cell therapies, microbiome modulation, metabolic reprogramming, and loco-regional delivery platforms. Figure 2 provides a conceptual overview of these major therapeutic levers and illustrates how they can be integrated with PD-1/PD-L1 blockade.

Figure 2 Multi-scale immunobiological barriers to immune checkpoint inhibitor response in microsatellite-stable/proficient mismatch repair colorectal cancer. This schematic summarizes how microsatellite-stable/proficient mismatch repair (MSS/pMMR) colorectal cancer (CRC) resists immune checkpoint inhibitors (ICIs) across the major phases of antitumor immunity priming, trafficking, and effector function. Following tumor cell death, released antigenic peptides are captured by dendritic cells (DCs) and presented in draining lymph nodes to initiate T-cell priming. In MSS/pMMR CRC, tumor-intrinsic resistance (low tumor mutational burden/neoantigen scarcity, oncogenic Wnt/β-catenin signaling, and impaired antigen processing/presentation) limits productive priming and T-cell generation. Even when primed, T cells frequently fail to reach or penetrate tumor nests due to a stroma-vasculature exclusion module, characterized by cancer-associated fibroblast/extracellular matrix physical barriers, dysfunctional/endothelial anergy that reduces adhesion and transmigration, and hypoxia/hypoxia-inducible factor-driven vascular and metabolic constraints. Within tumors, a myelo-lymphoid suppression module (e.g., myeloid-derived suppressor cells/tumor-associated macrophages/regulatory T cell) further suppresses immunity through arginase-1/reactive oxygen species and upregulation of inhibitory checkpoint ligands, fostering cluster of differentiation (CD) 8⁺ T-cell exhaustion marked by programmed cell death protein 1, cytotoxic T-lymphocyte-associated protein-4, and T cell immunoglobulin and mucin domain-containing protein 3. These layers converge into a soluble suppressive hub enriched in interleukin-10, transforming growth factor-β, and vascular endothelial growth factor, which globally dampens priming, trafficking, and effector cytotoxicity. At the organ level, hepatic involvement can function as a liver immune sink: Liver sinusoidal endothelial cells, Kupffer cells, and tolerogenic DCs promote CD8⁺ T-cell deletion/dysfunction and systemic immunosuppression, reducing the likelihood of durable ICI benefit in MSS/pMMR metastatic CRC. IL: Interleukin; TGF-β: Transforming growth factor-β; VEGF: Vascular endothelial growth factor; TMB: Tumor mutational burden; LN: Lymph node; DC: Dendritic cell; CAF: Cancer-associated fibroblast; ECM: Extracellular matrix; CD: Cluster of differentiation; LSECs: Liver sinusoidal endothelial cells; PD-1: Programmed cell death protein 1; PD-L1: Programmed death-ligand 1; ROS: Reactive oxygen species; ARG1: Arginase-1; CTLA-4: Cytotoxic T-lymphocyte-associated protein-4; TIM-3: T cell immunoglobulin and mucin domain-containing protein 3.

Chemotherapy plus immune checkpoint blockade

Rationale: Immunogenic cell death and antigen release: Several cytotoxic agents used in CRC, particularly oxaliplatin-based regimens, can trigger immunogenic cell death, turning the tumor into an in situ vaccine source. Dying cancer cells expose and release damage-associated molecular patterns (DAMPs) such as calreticulin, high mobility group box 1 protein (HMGB1), adenosine triphosphate (ATP) and heat-shock proteins, which activate DCs, enhance cross-presentation of tumor antigens, and prime tumor-specific CD8⁺ T cells with the potential to generate immunologic memory[55,56]. Recent preclinical work in CRC models has confirmed that oxaliplatin-induced immunogenic cell death enhances the efficacy of PD-1/PD-L1 blockade or other immunotherapies and that immunogenic cell death-related signatures correlate with T-cell infiltration and outcome[55]. Contemporary reviews further emphasize that conventional chemotherapy can increase MHC expression, augment tumor-infiltrating lymphocytes (TILs) and reshape the cytokine milieu, thereby lowering the threshold for effective immune checkpoint inhibition[56]. Together, these data support the concept that cytotoxic chemotherapy can act as an immune adjuvant by providing both antigenic cargo and danger signals, offering a clear mechanistic rationale for combining FOLFOX- or CAPOX-based regimens with ICIs to help convert immunologically “cold” MSS CRC into more inflamed, T-cell-infiltrated tumors[55,56].

Key trials of chemo-ICI combinations in MSS mCRC: Early clinical experience with chemo-ICI combinations in unselected MSS or pMMR mCRC has been sobering. In the phase Ib KEYNOTE-651 trial, pembrolizumab plus mFOLFOX7 or FOLFIRI in first- or second-line MSS/pMMR mCRC was feasible and showed manageable toxicity, but ORRs and PFS were not clearly superior to historical chemotherapy benchmarks, and no new efficacy signal emerged in biomarker-unselected patients[57,58]. Likewise, in the biomarker-driven umbrella trial MODUL, adding atezolizumab to fluoropyrimidine-bevacizumab maintenance after induction therapy did not improve progression-free or OS in the BRAF wild-type cohort, underscoring that simply layering PD-L1 blockade onto standard chemotherapy and anti-VEGF therapy is insufficient for most MSS tumors[59].

More favorable activity has been observed in biomarker-enriched subsets, although robustness across cohorts remains to be established. The FFCD 1703-POCHI phase II trial selected previously untreated pMMR/MSS mCRC with high immune infiltrate by Immunoscore and treated them with CAPOX plus bevacizumab and pembrolizumab. This TILs-high population achieved a disease control rate of 96%, an ORR of 75% with a notable proportion of complete responses, and a 12-month progression-free rate close to 70%, suggesting that chemo-ICI combinations can induce deep and durable responses when pre-existing immune activation is present[60]. These results align with immune-score data indicating that only about 10%-15% of MSS/pMMR CRCs are highly infiltrated by CD3⁺/CD8⁺ lymphocytes and may behave more like MSI-H tumors in terms of immunotherapy sensitivity[59,60]. Ongoing trials, including immune-score-enriched studies based on NCT04262687, are now prospectively testing chemo plus anti-VEGF with or without PD-1 blockade in MSS CRC with predefined immune biomarkers[59,60]. Overall, current evidence suggests that chemotherapy can facilitate “cold-to-hot” conversion through immunogenic cell death, but meaningful clinical benefit from chemo-ICI combinations in MSS mCRC is largely confined to biologically selected subgroups with either pre-existing immune infiltration or additional favorable features that lower immunologic resistance.

Anti-angiogenic therapy and vascular normalization

VEGF/VEGF receptor signaling and immune exclusion: Tumor angiogenesis not only supports growth but also promotes an immune-excluded, suppressive microenvironment. VEGF-A, the master pro-angiogenic factor in CRC, has potent immunomodulatory effects: It recruits immunosuppressive cells (Tregs, MDSCs) and impairs DC maturation[61,62]. The abnormal, leaky vasculature driven by VEGF results in hypoxia and high interstitial pressure, which together hinder effector T cell trafficking into the tumor parenchyma. Additionally, VEGF directly upregulates inhibitory checkpoints on T cells (e.g., via chronic exposure leading to T cell exhaustion) and expands Treg populations in the tumor[62]. In essence, VEGF/VEGF receptor signaling creates a feed-forward loop of immune evasion: It builds physical and biochemical barriers that exclude cytotoxic lymphocytes while nurturing suppressor cells. Indeed, higher VEGF levels in CRC have been correlated with reduced T cell infiltration and poorer ICI efficacy[61].

By blocking VEGF or its receptors, anti-angiogenic therapy can normalize the tumor vasculature and reprogram the tumor immune microenvironment. The concept of vascular normalization (championed by Huang et al[63]) posits that optimally dosed anti-VEGF treatment prunes aberrant vessels while stabilizing remaining vasculature, leading to improved perfusion and oxygenation. This in turn facilitates T lymphocyte entry and reduces hypoxia-driven immunosuppression[63]. Preclinical studies show that low-dose anti-VEGF receptor 2 antibodies can indeed shift the microenvironment from immune-suppressive to immune-supportive[63]. Moreover, VEGF blockade relieves the direct immunosuppressive signaling on immune cells; for example, a selective CD73 inhibitor (which blocks the VEGF adenosine axis) was shown to suppress PD-L1 expression on TAMs[64]. Collectively, these data indicate that anti-angiogenic therapies have an immunotherapeutic bonus: They open the tumor gates for immune cells and diminish local immune suppression. This provides a strong rationale to combine VEGF/VEGF receptor inhibitors with immunotherapy, to convert an immune-deserted CRC into one permissive for T cell attack.

Clinical data of bevacizumab or tyrosine kinase inhibitors combined with ICIs: Anti-angiogenic therapy provides a strong mechanistic rationale to partner with ICIs in MSS CRC by partially restoring vascular function and immune trafficking, and by modulating myeloid and stromal programs. However, early-phase activity signals have proven heterogeneous and, in some instances, geographically non-reproducible, warranting cautious interpretation and biomarker-driven refinement. An initial phase Ib study conducted in Japan (REGONIVO/EPOC1603) reported objective responses in a heavily pretreated CRC cohort that was almost entirely MSS/pMMR (ORR 36% in CRC; median PFS 7.9 months), generating substantial interest in regorafenib PD-1 combinations as a potential “cold-to-hot” lever[65].

Subsequent studies in Western populations, however, reported more modest efficacy, underscoring that the signal is not uniformly reproducible across geographic cohorts or unselected late-line settings. In a multicentre phase 2 study across 13 United States sites (NCT04126733), regorafenib plus nivolumab yielded an ORR of 7% (5/70) with a median PFS of 1.8 months, and all responders lacked liver metastases at baseline[66]. A separate phase I/Ib study similarly concluded that regorafenib plus nivolumab had limited anticancer activity in pMMR CRC (ORR approximately 10% among evaluable patients)[67]. Collectively, these discrepant outcomes suggest that apparent benefit may be confined to selected clinical/biological contexts (e.g., metastatic pattern, baseline immune sensitivity) and may be influenced by dosing intensity and schedule. Accordingly, regorafenib PD-1 combinations should be framed as hypothesis-generating rather than practice-changing, and ongoing efforts should prioritize prospective enrichment/stratification (including liver-metastasis status) and rational multi-node regimens that concurrently address trafficking, myeloid suppression, and stromal exclusion.

Other anti-angiogenic combos are showing preliminary activity in early-phase cohorts. Fruquintinib, a selective VEGF receptor-1/2/3 tyrosine kinase inhibitor (TKI) approved in refractory CRC, has been combined with the PD-1 antibody sintilimab in a phase II trial. Among 75 MSS CRC patients (all progressing after standard lines), fruquintinib + sintilimab achieved an ORR of 12.5% and disease control rate of 76%[68]. While modest, these responses were encouraging in a chemo-refractory setting. Median PFS was 4.1 months and median OS 15.3 months, comparing favorably to historical data for regorafenib or trifluridine/tipiracil. Importantly, echoing the regorafenib experience, liver metastasis was a negative predictor patients without liver mets had significantly longer PFS (7.6 months vs 3.2 months)[68]. The combination’s toxicity was manageable (only approximately 19% grade 3-4 events, no unexpected immune toxicities)[68]. This suggests that in patients with lower tumor burden or non-hepatic disease, anti-angiogenic TKIs can sufficiently alter the microenvironment for ICIs to work.

Several trials in Asia and globally are ongoing to refine these strategies. For example, single-arm studies of regorafenib + nivolumab + ipilimumab (triple therapy) have reported some complete responses in MSS CRC, though with increased toxicity[69,70]. Exploratory biomarkers from these trials indicate that responders tend to show evidence of immune activation on treatment (e.g., increased CD8 T cells or IFN-γ-signature) and often had immune “excluded” tumors at baseline that became inflamed after therapy[66]. Additionally, circulating factors like TGF-β and angiopoietin-2 are being evaluated as predictors of response to VEGF/ICI combos. In summary, anti-angiogenic therapies help unlock the tumor for immune attack by normalizing vessels, reducing VEGF-mediated immunosuppression, and perhaps direct modulation of immune cells. When coupled with ICIs, they have induced meaningful responses in some MSS CRC patients. Ongoing research is focused on identifying who these patients are (e.g., those without liver mets, with pre-existing immune excluded phenotype) and on optimizing drug dosing to maximize immune benefits (e.g., “vascular normalizing” doses)[63]. This combination approach has already become a standard in other cancers (e.g., VEGF/ICI in kidney cancer and hepatocellular carcinoma), and it is now one of an active investigational direction to turn MSS CRC from cold to hot.

Targeted therapies reshaping the immune microenvironment

EGFR, human epidermal growth factor receptor 2, BRAF, and RAS-targeted agents: Molecularly targeted therapies against oncogenic drivers in CRC can have complex, bidirectional effects on the tumor immune microenvironment. On one hand, inhibiting certain oncogenic pathways can enhance immune recognition; on the other, some targeted agents may unintentionally impair immune function, so understanding context is key.

EGFR inhibition: Approximately 50% of CRCs are RAS wild-type and can be treated with anti-EGFR antibodies (cetuximab or panitumumab). Beyond direct anti-tumor effects, EGFR blockade can modulate the immune milieu. Cetuximab induces antibody-dependent cellular cytotoxicity (ADCC) by engaging Fcγ receptors on NK cells, leading to immunologic killing of tumor cells and release of tumor antigens. Even more intriguing, recent research in an inflammatory cancer model showed that EGFR inhibition can actively convert an immunosuppressive TME into an immunostimulatory one[71]. In that study, an anti-EGFR antibody reduced immunosuppressive chemokines (via downregulating EGR1 transcription factor activity) and led to increased infiltration of cytotoxic CD8⁺ T cells with concomitant reductions in Tregs and M2 macrophages[71]. The remodeled microenvironment improved responsiveness to anti-PD-L1 therapy[71]. Translating this to CRC: EGFR signaling in tumors is known to upregulate PD-L1 and other escape mechanisms, so blocking EGFR may lower those signals[72]. Indeed, panitumumab has been noted to improve immune cell infiltration in some CRC biopsies (potentially part of its mechanism of action in addition to ADCC). On the flip side, in EGFR-mutant lung cancers (an immunologically cold tumor type), oncogenic EGFR fosters an environment with few T cells and high levels of inhibitory cytokines; EGFR-TKI treatment in that context can upregulate MHC and make tumors slightly more visible to the immune system[72,73]. In summary, targeting EGFR in CRC seems to generally help anti-tumor immunity by direct cell killing that releases antigen and by altering cytokine profiles to favor T cell recruitment. These findings provide a rationale for combining cetuximab with ICIs[74]. (Notably, a small trial of cetuximab plus avelumab in MSS CRC reported hints of activity in cetuximab-refractory patients, supporting further exploration[75]).

Human epidermal growth factor receptor 2 (HER2)-targeted therapy: A subset (approximately 3%-5%) of RAS/BRAF wild-type CRCs harbor HER2 amplification. Anti-HER2 therapies (e.g., trastuzumab, pertuzumab, or ADCs like trastuzumab deruxtecan) have shown efficacy in these patients. Like EGFR monoclonal antibodies, HER2 antibodies engage immune effector cells via ADCC[76,77]. In breast cancer, the presence of TILs correlates with better response to trastuzumab, suggesting immune involvement. In CRC, preclinical data indicate that dual HER2 blockade can increase tumor T cell infiltration and upregulate Th1 cytokines. There is an ongoing trial (HERACLES-III) adding an anti-PD-1 to trastuzumab-based therapy in HER2⁺ CRC to see if immune responses can be amplified. One challenge is that HER2 is expressed on some normal epithelial cells, so off-tumor immune activation must be monitored. Overall, though, HER2 targeting is thought to synergize with immunotherapy by providing an initial immune attack (through ADCC and partial tumor destruction), which ICIs can then boost[78,79].

BRAF/MEK targeting: BRAF V600E mutant CRC (approximately 8% of mCRC) often has aggressive biology; These tumors can be either MSI-H (immune hot) or MSS (immune cold). Oncogenic BRAF signaling in MSS CRC is associated with an immune-desert phenotype BRAF activation drives VEGF production, suppresses IFN response genes, and can increase PD-L1 expression on tumor and myeloid cells[80]. In melanoma (where BRAF inhibitors are routine), it’s well-documented that blocking mutant BRAF causes a transient increase in T cell infiltration and melanocyte differentiation antigens, but also upregulates PD-L1 hence combining BRAF/MEK inhibitors with ICIs has proven beneficial in melanoma patients[81,82]. In CRC, the approved regimen for BRAF-mutant MSS tumors is combined BRAF (encorafenib) + EGFR inhibition. This likely has immunologic effects: EGFR blockade as noted can enhance immunity, and BRAF inhibition may increase presentation of tumor antigens. Preclinical work supports that BRAF or MEK inhibitors can restore anti-tumor immunity by reversing MAPK-driven immune evasion. For instance, selective BRAF inhibitors were shown to induce marked CD8⁺ T cell infiltration in mouse models and upregulate MHC-I on tumor cells[80,83]. There is a completed trial (COMMIT) of encorafenib + cetuximab ± nivolumab; while results are not yet published in full, exploratory analyses may reveal if the triplet improved immune activation. Another study (INCEPTION) is testing a BRAF vaccine plus PD-1 therapy, leveraging mutant BRAF as a neoantigen in CRC a different twist on targeting the oncogene to get T cells. In summary, inhibiting the MAPK pathway (BRAF/MEK) can make tumors more visible to the immune system, but may also induce counter-regulatory mechanisms like PD-L1 upregulation, so pairing with ICIs is logical.

RAS targeting: Roughly 45% of CRCs have KRAS mutations. Direct KRAS G12C inhibitors (sotorasib, adagrasib) were recently developed and have modest activity in CRC (approximately 10% ORR). How RAS inhibition affects the TME is under intense study. KRAS-mutant tumors tend to secrete factors [e.g., granulocyte-macrophage colony stimulating factor (GM-CSF), IL-8] that recruit suppressive myeloid cells; they also often have an upregulated β-catenin pathway leading to exclusion of DCs. Inhibition of KRAS signaling could thus alleviate some of these immunosuppressive loops. Early evidence in lung cancer models suggests KRAS-G12C inhibition can increase cytotoxic T cell infiltration and synergize with anti-PD-1 (some trials are adding ICIs after a “lead-in” of KRAS inhibitor therapy). In CRC, concrete clinical data are scant so far, but ongoing trials are combining KRAS G12C inhibitors with anti-PD-1 to see if responses deepen once the MAPK pathway is shut off. On the cautionary side, effective RAS blockade might also reduce tumor inflammation too much if the oncogene was driving production of T cell attractants; this balance is not yet fully understood. Nonetheless, targeting RAS is another strategy that could reshape the immune microenvironment from pro-tumor to anti-tumor.

In summary, targeted agents against EGFR, HER2, BRAF, and RAS can have dual immunologic effects: By killing tumor cells or blocking oncogenic signaling, they may release antigens and dismantle tumor-driven immune evasion, effectively “heating up” the tumor. EGFR/HER2 antibodies clearly engage the immune system via ADCC and have been associated with increased TILs[71]. BRAF/RAS pathway inhibitors can upregulate antigen presentation and T cell recruitment[80,83]. The flipside is that tumors often adapt (e.g., via checkpoint upregulation), so these targeted therapies are best deployed alongside immunotherapies. Clinical trials combining EGFR or BRAF/MEK inhibitors with ICIs in MSS CRC are ongoing. If successful, these combinations could transform subsets of molecularly-defined MSS CRC (e.g. RAS wild type, BRAF mutant) into immunotherapy-responsive disease.

MEK, PI3K, and other pathway inhibitors: Beyond the well-known driver mutations, MSS CRCs commonly harbor dysregulated signaling pathways (Wnt/β-catenin, PI3K-AKT-mTOR, JAK/STAT, etc.) that contribute to immune escape. Preclinical studies suggest that inhibiting certain oncogenic or metabolic pathways can “lift the brakes” on anti-tumor immunity. Notably, MEK inhibitors and PI3K inhibitors have been explored as immune modulators.

MEK inhibition: MEK inhibition in MSS CRC highlights a recurrent translational pitfall: Tumor-intrinsic immune “sensitization” does not automatically translate into clinical benefit if systemic immune competence and dominant exclusion programs are not simultaneously addressed. Constitutive MAPK signaling (e.g., downstream of RAS/BRAF alterations) can promote an immune-excluded milieu, in part through pro-angiogenic and inflammatory mediators that impair trafficking and favor myeloid suppression. In CRC mouse models, MEK inhibition produced a transient remodeling of the TME reducing immunosuppressive myeloid populations (including MDSCs), increasing class I MHC expression on tumor cells, and elevating intratumoral CD8⁺ T-cell numbers[59,71]. However, this immunologic “window” may be short-lived and counterbalanced by adaptive resistance mechanisms. Clinically, the cobimetinib plus atezolizumab doublet was tested in refractory MSS CRC in IMblaze370, yet outcomes were not improved vs regorafenib, with similarly low response rates (approximately 2%)[59]. A plausible biological explanation is the bidirectional immunology of MEK blockade: While it can favor antigen presentation within tumor cells, continuous systemic MEK inhibition may also blunt T-cell priming/expansion and effector function, thereby limiting the magnitude and durability of anti-tumor immunity even when PD-L1 is blocked[84,85]. In addition, the dominant resistance architecture in MSS CRC myeloid-driven suppression, stromal/TGF-β-linked exclusion, and VEGF-associated vascular dysfunction is unlikely to be dismantled by MEK⁺ PD-L1 alone in unselected late-line populations. Therefore, future directions should prioritize schedule optimization (short priming or intermittent/pulsed dosing) and multi-node combinations that concurrently restore priming and trafficking, rather than relying on a simple MEK⁺ PD-L1 doublet[86-88].

PI3K pathway inhibition: The PI3K-AKT pathway is another axis frequently activated in CRC (e.g., by PIK3CA mutations or loss of PTEN). Beyond tumor-intrinsic effects, PI3K signaling in immune cells (especially PI3K-γ in myeloid cells) is a major contributor to immunosuppression. In TAMs and MDSCs, PI3K-γ drives a suppressive, “M2” polarization. A novel approach has been to use PI3K-γ selective inhibitors to reprogram these myeloid cells. Eganelisib (IPI-549) is one such oral inhibitor; in a phase I study across solid tumors, IPI-549 + nivolumab showed favorable safety and some evidence of immune activation (e.g., increased M1 macrophages)[89]. While objective responses were rare in MSS CRC patients, there were signals of longer stable disease. Another PI3K inhibitor, copanlisib (pan-class I PI3K inhibitor), was combined with avelumab in CRC patients in a trial, but results were not notable. It’s thought that more specific targeting (like PI3K-γ or δ isoforms) is needed to avoid suppressing effector T cells (which rely on PI3K-δ for function). Interestingly, PI3K and MEK pathways converge on many immune-related genes; dual inhibition of MEK and PI3K together is highly immunogenic in preclinical models but at the cost of significant toxicity.

Other pathways: Wnt/β-catenin signaling is commonly upregulated in CRC and is known to cause exclusion of DCs from the tumor, leading to T cell ignorance. Though direct Wnt inhibitors are not yet clinical, experimental agents or tankyrase inhibitors (indirect Wnt blockers) have shown to restore DC infiltration and priming of T cells in mouse models potentially “heating up” the tumor. STAT3 is another factor often active in CRC (downstream of IL-6); STAT3 in tumors and myeloid cells fosters an immunosuppressive environment. Small-molecule STAT3 inhibitors or even STAT3 small interfering RNA nanoparticles are being explored to boost anti-tumor immunity. CXCL12-CXCR4 axis: Some CMS4 subtype CRCs have a fibroblast-rich stroma secreting CXCL12, which keeps T cells out; a CXCR4 inhibitor (plerixafor) in a preclinical CRC model improved T cell penetration and synergized with PD-1 blockade. These represent emerging strategies where blocking a single immunosuppressive pathway might not be enough, but in combination with ICIs could tip the balance.

In essence, inhibiting oncogenic and stromal pathways can reshape the immune contexture of MSS CRC. MEK and PI3K inhibitors provided proof-of-concept that tumors can become more permissive to immune attack when these pathways are dampened[71,89]. The disappointing monotherapy or combo trial results so far underscore that tumors adapt quickly so multi-pronged approaches or careful patient selection are needed. Future trials are incorporating these inhibitors with other immune therapies (vaccines, myeloid cell modulators) to generate a more potent and sustained “hot” TME. The overall strategy remains compelling: Disable tumor-intrinsic signals of immune resistance to allow immune clearance. As our understanding of resistance mechanisms grows, we expect smarter combinations (e.g., transient MAPK blockade, specific PI3K-γ inhibition, etc.) will yield greater success in turning cold CRCs hot.

Radiotherapy and locoregional treatments as immune adjuvants

Radiotherapy-induced immunogenic cell death and abscopal effect: Radiation therapy (RT) has long been recognized to cause tumor cell death, but only in the past decade did we appreciate its role in stimulating anti-tumor immunity. High-energy radiation can induce immunogenic cell death of cancer cells, much like certain chemotherapies do[56]. Irradiated tumor cells release DAMPs (calreticulin exposure, HMGB1, ATP, etc.), which recruit and activate DCs that then prime T cells against tumor antigens[56,90]. This process effectively turns the irradiated tumor into an in-situ vaccine. In addition, RT causes local inflammation increasing chemokines that attract T cells and NK cells into the tumor area. The net result can be conversion of a cold tumor site into a hot, inflamed one, especially when radiation is delivered in hypofractionated regimens (moderate doses per fraction that favor immunogenic cell death over purely apoptotic death).

A frequently cited illustration of RT-enabled systemic immunity is the abscopal effect; however, in CRC it appears uncommon and remains largely anecdotal outside carefully selected settings. Though rare, abscopal responses have been documented in various cancers (melanoma, lymphoma) and more recently in anecdotal cases of CRC when combined with ICIs. For example, one case report described a metastatic MSS CRC patient who had progression on anti-PD-1 therapy, but after receiving localized radiotherapy to a liver lesion, the patient experienced regression of distant metastases and a pathological complete response an abscopal effect attributed to RT unleashing tumor antigens for the primed immune system[91,92]. In a small phase II study (NICHE-P trial), researchers are evaluating PD-1 blockade with or without stereotactic body radiotherapy (SBRT) to a single metastasis in MSS CRC; preliminary results suggest that the RT + ICI arm had a higher rate of disease control outside the radiation field (consistent with abscopal immune responses)[93]. Similarly, retrospective analyses indicate that CRC patients who received palliative RT to one lesion while on immunotherapy had prolonged stable disease in some cases compared to those on immunotherapy alone.

It must be noted that the abscopal effect in CRC is infrequent, likely due to the strongly immunosuppressive baseline environment. Nonetheless, RT clearly can augment antigen release and T cell priming[56]. Preclinical studies support combining RT with ICIs: In murine CRC models, adding CTLA-4 or PD-1 blockade to localized RT led to eradication of both irradiated and unirradiated tumors in a proportion of mice, whereas either modality alone could not[94]. Mechanistically, RT upregulates MHC-I and tumor antigen presentation on cancer cells, and can increase PD-L1 expression (a potential adaptive resistance that justifies PD-L1 blockade after RT)[94]. It also can induce phenotypic changes like converting some “cold” tumors into an “inflamed” subtype by increasing intratumoral IFN-γ and CXCL10, as seen in some mouse CRC models.

Clinically, radiation is being harnessed as an immune adjuvant in multiple ways: (1) In oligo mCRC, SBRT to a few lesions combined with ICIs is hypothesized to induce systemic immunity that might control micro-metastatic disease; (2) In the neoadjuvant setting (e.g., rectal cancer), adding ICIs to chemoradiation is being explored to increase pathological response rates via immune-mediated tumor killing (early trials like NICHE-2 showed that short-course RT plus dual PD-1/CTLA-4 blockade yielded major pathologic responses even in some pMMR rectal cancers); and (3) Using RT to temporarily overcome resistance e.g., a patient progressing on immunotherapy could receive a radiation boost to a growing lesion, potentially releasing new antigens and cytokines to re-engage T cells.

In summary, radiotherapy can convert a tumor into an endogenous vaccine site and reduce immune exclusion[94]. When strategically combined with ICIs, it holds the potential to produce systemic anti-tumor immunity (the coveted abscopal effect). Ongoing clinical trials in MSS CRC are testing RT + ICI combinations, and while challenges remain (optimal dose, timing, and patient selection), RT is increasingly viewed as an immunotherapy partner rather than just local therapy.

Combining radiotherapy, transarterial chemoembolization, selective internal radiation therapy, or HIPEC with ICIs: Building on the above, various locoregional therapies are being integrated with immunotherapy in MSS CRC, aiming to induce or enhance immune responses in situ.

Radiation + ICI in metastatic sites: In CRC liver metastases, investigators are experimenting with combining SBRT or radioembolization [selective internal radiation therapy (SIRT) with Y-90 microspheres] with checkpoint inhibitors. The liver is an immunotolerant organ that often dampens systemic immunity, but localized radiation can break this tolerance. Small series have reported that MSS CRC patients treated with Y-90 radioembolization to liver metastases followed by anti-PD-1 had prolonged disease control in a few cases compared to historical controls[93]. The hypothesis is that radioembolization causes tumor necrosis and antigen release within the liver, which with ICI on board allows T cells to overcome the hepatic immunosuppressive environment. Similarly, for lung or lymph node metastases, SBRT can inflame the TME and promote T cell infiltration that might then attack other lesions under ICI effect. Clinical trials like METASTORM are randomizing patients to SBRT + PD-1 vs PD-1 alone to quantify any improvement in response rate in MSS mCRC.

Transarterial chemoembolization (TACE): TACE is mostly used in liver-dominant cancers (like hepatocellular carcinoma), but occasionally in CRC liver metastases. TACE delivers high-dose chemotherapy directly into the tumor artery with embolization that causes ischemic necrosis. The resultant tumor cell kill could be highly immunogenic, and indeed in hepatocellular carcinoma, TACE has been shown to increase PD-L1 expression and TILs (hence combinations of TACE + anti-PD-1 are being tested in hepatocellular carcinoma). By analogy, combining TACE with ICIs in CRC liver metastasis is hypothesized to yield synergistic benefit, though published data are scarce. One phase I trial (No. NCT04384899) is exploring TACE with hepatic artery infusion of nivolumab for CRC metastases; results are pending. The concept is compelling: TACE might turn a “cold” liver metastasis into a “hot” site by massive tumor antigen release, and ICIs then prevent peripheral tolerance to those antigens. Time will tell if this approach produces robust abscopal effects or prolonged control.

HIPEC and intraperitoneal (IP) immunotherapy: MSS CRC with peritoneal metastasis has very poor prognosis and limited systemic therapy efficacy. Cytoreductive surgery with HIPEC is sometimes offered. Interestingly, HIPEC itself can induce immunogenic changes the heat and chemo can increase heat shock proteins and other DAMP signals, and analyses of HIPEC-treated peritoneal tumors show increased infiltration of immune cells post-treatment[95]. Building on this, researchers are considering adding ICIs or other immunotherapies either immediately after HIPEC or as maintenance. For example, a Chinese phase II trial is evaluating HIPEC combined with intravenous toripalimab (PD-1 antibody) for CRC-PMs. Another angle is IP immunotherapy: Directly injecting immune agents into the peritoneal cavity to concentrate the drug at tumor deposits. Preclinical work showed that IP injection of a Toll like receptor (TLR) 9 agonist alongside anti-PD-1 led to cures of CRC peritoneal tumors in mice[96]. This has led to a trial of IP nivolumab plus IP CMP-001 (a TLR9 agonist) for patients with peritoneal carcinomatosis the idea being to stimulate local innate immunity (via TLR9) and systemic T cell response (via PD-1 blockade). Similarly, chimeric antigen receptor (CAR)-T cell therapy is being trialed with IP delivery for peritoneal tumors (discussed in “CAR-T and CAR-NK cells”). Early reports suggest it is feasible and may reduce off-target toxicity compared to IV infusion, since the cells stay mostly in the peritoneal cavity attacking tumors.

SIRT: As noted, SIRT with yttrium-90 microspheres causes high-dose radiation confined to liver tumors. SIRT can lead to intense tumor necrosis, which might be immunogenic. A trial called EMLALA-1 (presented in abstract form) combined SIRT with durvalumab/tremelimumab (PD-L1 and CTLA-4 blockade) in MSS mCRC patients. While the primary endpoint (safety) was met, efficacy results were modest but a few patients did have prolonged stability, correlating with an increase in T cell clonality in blood post-SIRT, hinting at immune activation. To improve on this, future studies will refine timing (e.g., giving ICI before and after SIRT to capture the wave of antigen release) and possibly add immune stimulants (like GM-CSF injections) to further enhance DC uptake of tumor debris (Table 2).

Table 2 Selected clinical evidence for immunotherapy combination strategies in microsatellite-stable/proficient mismatch repair metastatic colorectal cancer.

Combination strategy/regimen	Clinical setting (population)	Key outcomes	Predictors/caveats	Ref.
Pembrolizumab + mFOLFOX7 or FOLFIRI (KEYNOTE-651)	1 L/2 L advanced CRC (predominantly pMMR/MSS)	Safety acceptable; no clear signal beyond historical chemo in unselected MSS cohorts	Single-arm; chemotherapy confounds ORR; highlights need for biomarker enrichment	Gallois et al[57]; Kim et al[58]
Maintenance FP + bevacizumab ± atezolizumab (MODUL)	Post-induction maintenance in mCRC (unselected; largely MSS)	No significant PFS (HR = 0.92) or OS (HR = 0.94) improvement vs control	Negative phase II; subgroup signals require validation	Wang et al[59]
CAPOX + bevacizumab + pembrolizumab (FFCD1703-POCHI)	1 L pMMR/MSS mCRC with high TIL/Immunoscore	ORR = 73% (17% CR); DCR 100%; 12-month PFS 52%; 24-month OS 80%	Biomarker-enriched (high Immunoscore); confirm in larger/controlled studies	Yamaguchi et al[60]
Regorafenib + nivolumab (REGONIVO, phase Ib)	Refractory MSS mCRC (Japan)	ORR approximately 33%-36% in CRC cohort; median PFS approximately 7.9 months	Early-phase; later Western studies show lower activity; liver metastases may blunt benefit	Fukuoka et al[65]
Regorafenib + nivolumab (phase II)	Refractory MSS mCRC (multicenter United States)	ORR 7% (5/70 PR); median PFS 1.8 months; OS 11.9 months	All responders had no liver metastases	Fakih et al[66]
Fruquintinib + sintilimab	Refractory MSS mCRC (China phase II)	ORR 12.5%; DCR 76.4%; median PFS 4.1 months; OS 15.3 months	Liver mets: PFS 3.2 months vs 7.6 months; NLR/albumin/ECOG predictive	Zhang et al[68]
Regorafenib + ipilimumab + nivolumab	Refractory MSS mCRC (single-center phase I)	ORR 27.6%; median PFS 4.0 months; OS 20 months; non-liver mets ORR 36.4%	Responses largely confined to nonliver metastatic disease; dose-related skin/immune AEs	Xiao et al[69]; Fakih et al[70]
Atezolizumab + cobimetinib (IMblaze370)	Refractory mCRC (predominantly MSS; phase III)	No OS benefit vs regorafenib; ORR approximately 2%	Suggests MEK inhibition alone is insufficient; scheduling/partner choice critical	Eng et al[40]
Radiotherapy + ICI (various early-phase)	MSS mCRC; often liver-dominant disease	Clinical benefit inconsistent; abscopal responses uncommon	Dose/fractionation, target lesions, and systemic priming likely determinants	Yang et al[91]; Lee et al[92]; Nelson et al[93]; Hang et al[94]
Locoregional therapy + ICI (HIPEC/HAIC/TACE/SIRT; early-phase)	Liver or peritoneal metastases	Primarily early-phase/ongoing; rationale is in situ antigen release and myeloid reprogramming	Safety/sequence critical; endpoints include immune correlatives	Nelson et al[93]; Nevo et al[95]; Jiang et al[96]

CAPOX: Capecitabine + oxaliplatin; FP: Fluoropyrimidine; ICI: Immune checkpoint inhibitor; HIPEC: Hyperthermic intraperitoneal chemotherapy; HAIC: Hepatic arterial infusion chemotherapy; TACE: Transarterial chemoembolization; SIRT: Selective internal radiation therapy; mCRC: Metastatic colorectal cancer; MSS: Microsatellite-stable; pMMR: Mismatch repair-proficient; CRC: Colorectal cancer; TIL: Tumor-infiltrating lymphocyte; OS: Overall survival; PFS: Progression-free survival; CR: Complete response; DCR: Disease control rate; ORR: Objective response rate; HR: Hazard ratio; PR: Partial response; NLR: Neutrophil-to-lymphocyte ratio; ECOG: Eastern Cooperative Oncology Group; AE: Adverse event; MEK: Mitogen-activated protein kinase kinase.

Cancer vaccines and OVs

Neoantigen and tumor-associated antigen vaccines: Therapeutic vaccines are designed to prime or amplify tumor-reactive T cells, a particularly appealing concept in MSS CRC, where baseline T-cell infiltration and spontaneous antitumor immunity are often limited. Two main strategies are under active investigation: Shared tumor-associated antigen (TAA) vaccines and personalized neoantigen vaccines.

TAA vaccines [e.g., carcinoembryonic antigen (CEA), mucin 1 (MUC1)] are attractive because they target broadly expressed antigens, but clinical efficacy has historically been modest, likely reflecting immune tolerance and dominant immunosuppressive mechanisms in established MSS CRC. Recent combination attempts illustrate the central challenge: In a randomized phase II study adding a CEA adenoviral vaccine plus PD-L1 blockade (avelumab) to first-line mFOLFOX6/bevacizumab, the regimen was immunologically active (multifunctional T-cell responses and antigen spreading), yet did not improve PFS vs standard therapy[97]. In the minimal residual disease/adjuvant setting, a randomized phase II metastasectomy study using poxviral vectors/DC-based vaccination against CEA/MUC1 reported feasibility and encouraging survival signals vs contemporary controls, but recurrence-free survival between vaccine strategies was similar, underscoring that platform optimization and rational combinations remain essential[98].

Personalized neoantigen vaccines avoid central tolerance and can broaden the T-cell repertoire against truly tumor-specific targets. Although evidence in MSS CRC is still early and largely small-scale, a clinical study of individualized neoantigen vaccination in advanced MSS CRC demonstrated feasibility and safety, with neoantigen-specific immune responses in approximately 2/3 of patients, and longer PFS among immune responders[99]. These data support continued development most plausibly in low tumor burden/minimal residual disease settings and in combinations that enhance intratumoral trafficking and sustain effector function (e.g., checkpoint blockade, VEGF/TGF-β axis modulation, radiotherapy/chemotherapy priming).

OVs

OVs can provide direct tumor lysis and an “in situ vaccination” effect via immunogenic cell death, antigen release, and type I IFN signaling features that could help “heat up” MSS CRC. Clinically, however, systemic delivery barriers, antiviral immunity, and the suppressive CRC microenvironment have limited efficacy to date. In mCRC, a randomized phase II trial adding pelareorep (reovirus) to FOLFOX6/bevacizumab increased ORR but resulted in inferior PFS, likely confounded by reduced intensity/duration of standard agents, illustrating that OV integration into chemo backbones requires careful scheduling and tolerability management[100]. For vaccinia-based OV therapy, a phase 1b trial of intravenous Pexa-Vec (JX-594) in heavily pretreated mCRC showed an acceptable safety profile and disease stabilization in a majority, supporting feasibility but highlighting the need for stronger combination partners (e.g., ICIs, stromal/vascular modulators)[101]. Overall, current data position OVs primarily as immune primers rather than stand-alone therapies in MSS CRC.

Adoptive cell therapy in CRC: TIL and engineered T Cells

TILs and engineered TCR-engineered T cells: Adoptive cell therapy (ACT) aims to supply CRC patients with large numbers of tumor-reactive T cells generated ex vivo, thereby bypassing the weak endogenous T-cell priming that typifies pMMR/MSS disease. Two clinically advanced ACT strategies in CRC are: (1) Expansion of autologous TILs; and (2) Genetic redirection of peripheral T cells with tumor-reactive TCR (TCR-T)[102].

TIL therapy: The main bottleneck for MSS CRC is the scarcity and functional impairment of naturally occurring tumor-reactive TILs, particularly in immune-excluded lesions and liver metastases. Recent progress has come from neoantigen-reactive/selected TIL products, which enrich for mutation-reactive clones before rapid expansion. In a phase 2 Nature Medicine study across refractory metastatic gastrointestinal cancers, neoantigen-specific TIL therapy demonstrated feasibility and objective responses, with improved activity when integrated with anti-PD-1 blockade, supporting the concept that optimized TIL selection plus checkpoint support can partially overcome T-cell dysfunction in “cold” gastrointestinal tumors[103]. Proof-of-principle in MSS CRC was established by the classic KRAS G12D case: Infusion of ex vivo expanded, KRAS G12D-reactive CD8⁺ T cells mediated regression of mCRC, but subsequent progression was linked to immune escape via loss of the restricting HLA allele highlighting both potency and antigen-presentation vulnerability of ACT[104].

TCR-T: TCR-T therapy can “manufacture” tumor specificity by introducing a defined TCR recognizing a neoantigen (or shared hotspot mutation) presented by HLA, enabling treatment even when baseline TILs are sparse. Clinically, a key advance for MSS CRC is the emergence of personalized neoantigen-reactive TCR-T: In a Nature Medicine phase 2 interim report in heavily pretreated metastatic pMMR CRC, transfer of autologous peripheral T cells transduced with individualized neoantigen-reactive TCRs produced RECIST responses in a subset of patients and showed measurable persistence of engineered cells (including long persistence in a responder), demonstrating clinical feasibility in a setting historically resistant to immunotherapy[105].

Parallel efforts are building “semi-off-the-shelf” TCR-T products against recurrent driver mutations (e.g., KRAS) or shared neoantigens (e.g., TP53 R175H), but these approaches remain constrained by HLA restriction, variable antigen presentation, and the risk of tumor immune escape through HLA/antigen downregulation[106].

Implications for MSS CRC: Collectively, contemporary data suggest ACT is most likely to succeed in MSS CRC when: (1) Tumor-reactive clones are enriched (neoantigen-selected TIL or neoantigen-defined TCR-T); (2) Tumor burden and immune suppression are managed (lymphodepletion ± cytokine support, plus rational combinations such as anti-PD-1); and (3) Escape mechanisms (antigen/HLA loss, heterogeneous expression) are anticipated with multi-target strategies and correlative monitoring[103].

CAR-T and CAR-NK cells: In CRC, CAR-based therapies are being explored to avoid HLA restriction by targeting surface antigens, but efficacy is limited by trafficking barriers, suppressive myeloid/TGF-β programs, and on-target/off-tumor toxicity for shared epithelial antigens. Early clinical experience in CRC includes CEA-directed CAR-T trials demonstrating feasibility yet underscoring safety constraints, while CLDN18.2 CAR-T (CT041) has shown notable activity across CLDN18.2⁺ gastrointestinal cancers and may be relevant to selected CRC subsets with confirmed antigen expression[107]. GUCY2C CAR-T has generated an efficacy signal in early-phase reports in mCRC, but (to date) key efficacy figures are primarily available as conference outputs and should be presented as preliminary[108].

Microbiota-directed interventions

Modulation of the gut microbiome to enhance ICI response: The gut microbiome is a modifiable determinant of systemic immune tone and ICI efficacy. Landmark clinical metagenomic studies across tumor types established that baseline microbial ecology and antibiotic-associated dysbiosis can correlate with primary resistance to PD-1/PD-L1 blockade, with functional support from “avatar” fecal transfer experiments demonstrating partial causality[109].

In CRC, the microbiome-immunity link is especially relevant because tumors arise within (and are exposed to) dense luminal and mucosa-associated microbial communities. CRC-specific evidence is emerging from both mechanistic and clinical datasets. Notably, Fusobacterium nucleatum promotes immune evasion by engaging inhibitory receptors on T/NK cells (e.g., TIGIT; CEACAM1) and can drive immunotherapy resistance through microbiota-metabolite-innate signaling crosstalk[32]. Clinically, baseline microbiome features have also shown predictive value in MSS/pMMR locally advanced rectal cancer treated with PD-1 blockade plus chemoradiotherapy, supporting microbiome-informed stratification in combination regimens[110].

From a practical standpoint, microbiome-disrupting co-medications are plausible modifiers of ICI outcomes; antibiotic exposure has been repeatedly linked to poorer ICI benefit in large clinical series, while acid suppression (e.g., proton pump inhibitors) is also associated with inferior outcomes in meta-analyses factors worth controlling for in CRC immunotherapy trials and real-world studies[109].

Fecal microbiota transplantation, live biotherapeutics (probiotics), and diet: Microbiome manipulation is being tested to convert resistant phenotypes into ICI-sensitive states. Proof-of-concept fecal microbiota transplantation (FMT) trials in PD-1-refractory melanoma demonstrated that responder-derived microbiota can re-sensitize a subset of patients, catalyzing broader solid-tumor programs[111]. More recently, a cell host and microbe clinical trial reported that FMT combined with anti-PD-1 could improve efficacy in anti-PD-1 refractory advanced solid cancers, providing contemporary clinical support for this strategy and informing ongoing gastrointestinal/CRC trial design[112].

Compared with whole-community transfer, “live biotherapeutics” aim for standardized consortia, but caution is warranted: Observational data in ICI-treated patients suggest over-the-counter probiotics may correlate with worse outcomes, potentially via reduced community diversity arguing for mechanism-guided, trial-tested products rather than empiric supplementation[113]. Diet is a lower-risk lever; higher dietary fiber intake has been associated with favorable microbiome configurations and improved ICI outcomes in translational cohorts, supporting dietary optimization as an adjunct in CRC immunotherapy protocols (ideally evaluated prospectively with stool multi-omics and predefined endpoints)[113].

Emerging small molecules and metabolic modulators

Non-immune-targeted small molecules can still “recondition” antitumor immunity by reversing metabolic checkpoints imposed by tumors and suppressive myeloid/stromal cells. In CRC, three pathways are most consistently implicated in T-cell dysfunction and ICI resistance: Tryptophan catabolism [IDO1/TDO2-kynurenine-aryl hydrocarbon receptor (AhR)], ARG1, and purinergic signaling (CD39/CD73 adenosine A2AR), with lactate/acidification acting as a broader metabolic brake[36].

IDO1/TDO2-kynurenine-AhR axis: IDO1 is frequently upregulated in CRC and correlates with reduced T-cell infiltration and adverse outcomes, consistent with a nutrient-deprivation/toxic-metabolite model (tryptophan depletion plus kynurenine accumulation)[36]. Kynurenine activates AhR signaling, promoting regulatory programs and dampening effect or T-cell function, thereby creating a permissive immune-evasive niche[114]. Clinically, enthusiasm for IDO1 inhibition was tempered after epacadostat failed to improve outcomes with pembrolizumab in melanoma (ECHO-301/KEYNOTE-252), shifting the field toward better target coverage (e.g., dual IDO1/TDO approaches), biomarker-driven selection, and rational combinations rather than empiric add-on use[115].

Arginine depletion via ARG1 (myeloid metabolic checkpoint): ARG1 produced by MDSCs/M2-like macrophages depletes extracellular arginine, impairing T-cell proliferation and cytotoxicity. Early-phase clinical studies of the arginase inhibitor INCB001158 (alone or combined with anti-PD-1) demonstrate acceptable safety and on-target pharmacodynamic effects, but clinical activity has been modest, suggesting arginase blockade may be most impactful when paired with broader myeloid-reprogramming or trafficking strategies[116].

CD39/CD73-adenosine-A2AR pathway: Extracellular ATP is converted to immunosuppressive adenosine by CD39/CD73, and adenosine signaling through A2AR potently suppresses T and NK cell function while reinforcing suppressive myeloid states[117]. Early clinical development of A2AR antagonists (e.g., AZD4635) and CD73-targeting agents supports feasibility and pathway inhibition, but CRC benefit likely requires combination with ICIs and careful patient stratification (e.g., CD73-high tumors, adenosine-signature–positive microenvironments)[118].

Lactate/acidification and glycolytic rewiring: CRC glycolysis generates lactate and an acidic microenvironment that suppresses effector T/NK activity and can upregulate inhibitory programs (including PD-1 on Tregs) in highly glycolytic tumors[36]. In CRC models, lactate dehydrogenase-A inhibition can enhance anti-PD-1 efficacy, supporting lactate control as an immunotherapy adjunct[119]. Targeting lactate transport (e.g., monocarboxylate transporter blockade) is also emerging as a strategy to reverse lactic-acid–driven immunosuppression and potentiate immunotherapy[120].

Microbiome-linked metabolites and liver-dominant disease (brief note): Metabolites shaped by the gut-liver axis, including bile acids, can tune hepatic immune surveillance and may contribute to the relative immunotherapy refractoriness of liver-dominant disease. Recent translational evidence in liver mCRC links altered microbiota bile-acid metabolism with the metastatic immune niche, nominating bile-acid-targeted strategies as a future adjunct in selected patients[121].

Loco-regional and interventional immunotherapy approaches

Loco-regional immunotherapy aims to overcome anatomic “sanctuary” barriers (e.g., peritoneum and liver) by delivering immune-active agents directly to the disease compartment, thereby increasing local exposure, intensifying immune priming, and potentially reducing systemic toxicity. In MSS CRC where spontaneous T-cell infiltration is often limited these approaches are increasingly viewed as immune “ignition” platforms to be paired with systemic ICIs and/or other microenvironment modulators[122].

IP strategies for peritoneal metastases

CRC-PM are clinically difficult to control and can be biologically distinct from primary tumors. Pressurized IP aerosol chemotherapy (PIPAC) enables repeated, high-concentration IP exposure. A multicenter phase II study in unresectable CRC-PM demonstrated feasibility/safety signals but limited radiologic responses, supporting PIPAC mainly as a platform for rational combinations rather than a stand-alone strategy[123].

Recent reverse-translational work in CRC-PM treated with PIPAC-oxaliplatin indicates that loco-regional therapy can induce immune features associated with ICI sensitivity, including reduced immunosuppressive programs (e.g., hypoxia/IL-10/TGF-β signatures) and increased lymphocyte infiltration with tertiary lymphoid structure (TLS) formation suggesting a mechanistic basis for “PIPAC priming ICI amplification[124]”.

A first-in-human phase I trial combining PIPAC-oxaliplatin with systemic nivolumab (PIANO; gastric peritoneal metastases) further supports the combination concept, reporting tolerability and enhanced IP T-cell infiltration with reductions in M2 macrophage signals on translational profiling highly relevant for gastrointestinal peritoneal disease program design (including CRC-PM)[122].

Regional ACT

Regional delivery can also be used to improve effector-cell contact with peritoneal deposits. Notably, a 2025 clinical report describes IP infusion of NK group 2D CAR-NK cells in advanced CRC with evidence of immune activation (as reflected in the study’s endpoints/title), supporting feasibility of compartmental adoptive immunotherapy in CRC[125]. In parallel, ongoing early-phase trials are evaluating tumor-associated glycoprotein-72 CAR-T approaches in epithelial malignancies, reflecting renewed interest in peritoneal-directed engineered-cell strategies where systemic trafficking is a bottleneck.

Liver-directed immunotherapy and interventional combinations

CRC liver metastases are shaped by a tolerogenic hepatic milieu and are frequently associated with systemic immune suppression. A multicenter phase I study delivered the oncolytic vaccinia virus TG6002 via the intrahepatic artery (with oral 5-fluorocytosine) in liver-dominant mCRC, establishing clinical feasibility for hepatic-arterial delivery of an immunogenic biologic[126].

However, “local injury as an in-situ vaccine” may be insufficient on its own in liver-dominant disease. In the EORTC-1560 (ILOC) phase II trial, durvalumab + tremelimumab plus partial local ablative therapy (radiofrequency ablation/SBRT) did not induce responses in untreated liver lesions, underscoring the need for stronger, multi-axis combination design (e.g., myeloid/vascular/metabolic rewiring) when targeting hepatic immune resistance[127].

Precision local delivery

To maximize tumor-restricted activity and minimize systemic exposure, stimulus-responsive delivery systems are being developed. A 2024 study demonstrated potential of hydrogen-activatable nanoparticles enabling spatially confined PD-L1 degradation with enhanced radioimmunotherapy efficacy in vivo, offering a proof-of-principle for locally activated checkpoint modulation that could be adapted to percutaneous, intra-arterial, or implantable delivery formats in mCRC[128].

Overall, loco-regional immunotherapy in MSS CRC is best positioned as a priming/conditioning layer that reshapes immune contexture in hard-to-treat compartments (peritoneum and liver), to be coupled with systemic ICIs and biomarker-guided escalation (e.g., TLS induction, myeloid reprogramming, vascular normalization) in prospective trials[124].

BIOMARKERS AND PRECISION IMMUNOTHERAPY IN MSS CRC

The disappointing performance of ICIs in unselected MSS CRC has shifted the focus from “whether” to use immunotherapy to “in whom” and “how” to use it. A growing body of work suggests that responsiveness in MSS/pMMR disease is confined to biologically distinctive subsets defined by genomic, immunologic, and microenvironmental features rather than by anatomy alone. Recent reviews have started to organize these biomarkers into four broad domains tumor genomics, TME, host factors, and dynamic blood-based markers but their clinical implementation remains nascent, especially for MSS CRC[115]. A biomarker-guided, site-aware roadmap linking sampling, multi-omic profiling, stratification, and tailored combination modules is proposed in Figure 3.

Figure 3 Bottleneck-guided “cold-to-hot” combination design for microsatellite-stable/proficient mismatch repair colorectal cancer. Schematic framework that maps rational combination immunotherapy to the dominant rate-limiting step (“bottleneck”) along the cancer-immunity cycle in microsatellite-stable (MSS)/proficient mismatch repair colorectal cancer (CRC): Priming, trafficking, and effector function. For each bottleneck, “compact therapy modules” are shown that are intended to restore productive antitumor immunity rather than simply add a second agent to programmed death ligand-1 (PD-L1) blockade. Priming modules aim to increase immunogenic antigen availability and innate sensing (chemo/radiotherapy-mediated priming, oncolytic viruses, cancer vaccines, stimulator of interferon genes/Toll like receptor agonists, and strategies that enhance antigen presentation). Trafficking modules target stromal/vascular exclusion and immune cell access [anti-vascular endothelial growth factor (VEGF)-mediated vascular normalization, transforming growth factor-β (TGF-β)/cancer-associated fibroblast (CAF) blockade, extracellular matrix remodeling, disruption of the CXCL12-CXCR4 retention axis, and endothelial activation to support adhesion/extravasation]. Effector modules focus on reversing intratumoral dysfunction and exhaustion [programmed cell death protein-1 (PD-1)/PD-L1 blockade, cytotoxic T-lymphocyte-associated protein-4 blockade, co-stimulatory agonism such as 4-1BB/OX40, myeloid reprogramming, and metabolic rescue]. These interventions are coordinated against a shared soluble suppressive hub (interleukin-10/TGF-β/VEGF) that can concurrently dampen priming, trafficking, and effector cytotoxicity. The right panel highlights liver metastasis as an immune sink, where a suppressive hepatic microenvironment (e.g., Kupffer cell-driven tolerogenicity) can systemically blunt immune checkpoint inhibitor efficacy; “liver-directed adjuncts” (local radiotherapy, ablation, optional hepatic arterial infusion/transarterial chemoembolization approaches, and myeloid-targeted strategies) are positioned as context-specific intensifiers to overcome hepatic immune resistance. The bottom timeline emphasizes schedule optimization short immune priming, intermittent/pulsed dosing (illustrated by pulsed mitogen-activated protein kinase kinase priming to avoid continuous immune suppression), and window-of-opportunity biopsies to mechanistically validate target engagement and refine sequencing. Overall, the model underscores that multi-node combinations (addressing more than one bottleneck) are more likely to outperform single doublets in unselected, late-line MSS CRC, with example “combo recipes” illustrated (anti-VEGF + PD-1 with priming; TGF-β/CAF + PD-1 with vaccine/oncolytic priming; myeloid reprogramming + PD-1 with radiotherapy). IL: Interleukin; TGF-β: Transforming growth factor-β; VEGF: Vascular endothelial growth factor; RT: Radiotherapy; STING: Stimulator of interferon genes; TLR: Toll like receptor; MEK: Mitogen-activated protein kinase kinase; ECM: Extracellular matrix; CAF: Cancer-associated fibroblast; PD-1: Programmed cell death protein 1; PD-L1: Programmed death-ligand 1; CTLA-4: Cytotoxic T-lymphocyte-associated protein-4; MSS: Microsatellite stable; CRC: Colorectal cancer; HAI: Hepatic arterial infusion; TACE: Transarterial chemoembolization; OV: Oncolytic viruses.

Established and emerging biomarkers

MMR/MSI status, TMB and PD-L1 expression: dMMR/MSI-H remains the only fully validated biomarker for checkpoint blockade in CRC, underpinning regulatory approvals of PD-1 inhibitors in metastatic and, more recently, locally advanced and adjuvant settings[129]. By contrast, the vast majority of patients are MSS/pMMR and derive minimal benefit from current immune monotherapies, highlighting the limitations of using MSI/dMMR alone to guide treatment in this population[130].

TMB has been explored as a pan-cancer biomarker, and a threshold of ≥ 10 mutations/Mb formed the basis for tissue-agnostic pembrolizumab approval. However, CRC has emerged as an outlier: Several analyses indicate that MSS-TMB-high mCRC shows heterogeneous and often modest responses to PD-1 blockade, with some series reporting low ORRs despite high TMB[60]. These data suggest that TMB in CRC should be interpreted alongside mutation context (e.g., POLE/POLD1) and immune microenvironment rather than as a stand-alone threshold.

PD-L1 immunohistochemistry is widely used in lung and gastric cancer but has limited predictive power in CRC. Multiple series show that PD-L1 expression is often low, spatially heterogeneous, and only weakly associated with benefit from checkpoint inhibitors once MSI status is accounted for[115]. As a result, major guidelines do not recommend PD-L1 alone to select MSS CRC patients for ICI outside of clinical trials.

Genomic predictors: POLE/POLD1 and immune-evasive alterations: A small but clinically important subset of CRCs carries proofreading-deficient mutations in the exonuclease domain of POLE or POLD1, generating an “ultra-mutated” phenotype with extremely high neoantigen loads. Pooled analyses across tumor types and CRC-specific cohorts show striking response rates and durable disease control with checkpoint inhibitors in these patients, even when tumors are formally MSS[131]. Consequently, many experts now consider POLE/POLD1 status as a high-priority biomarker when contemplating immunotherapy for MSS disease.

Conversely, alterations that disrupt antigen presentation or T-cell recruitment may predict primary or acquired resistance. Tumor-intrinsic Wnt/β-catenin signaling is enriched in “immune-cold” CRCs and correlates with exclusion of CD8⁺ T cells from the tumor bed, in part through reduced chemokine production and impaired DC recruitment[132]. Defects in B2M and other components of the MHC class I pathway, as well as JAK/STAT signaling alterations, have been implicated in immune escape in MSI-H CRC and other tumor types, although their predictive value in MSS CRC remains less clear and may be context-dependent[131].

Other gene-level features including specific RAS/RAF alterations, STK11 and KEAP1 co-mutations, and transcriptional subtypes characterized by TGF-β or stromal activation are being investigated as modulators of ICI sensitivity or resistance, but evidence is still emerging and often extrapolated from non-CRC settings[133].

Immune infiltration metrics and immunoscore: TILs, particularly CD3⁺ and CD8⁺ T cells in the tumor center and invasive margin, have strong and independent prognostic value in early-stage CRC. The Immunoscore^® assay, which quantifies these populations digitally, has been validated in large international cohorts of stage II-III colon cancer and outperforms tumor node metastasis staging for relapse risk stratification[134]. Recent analyses from randomized adjuvant trials suggest that Immunoscore may also predict benefit from oxaliplatin-based chemotherapy and is now being explored as a stratification factor for immunotherapy trials[135].

In metastatic disease, high baseline CD8⁺ T-cell density and “inflamed” gene signatures appear enriched among the rare MSS CRC responders to combination ICI regimens (e.g., regorafenib plus nivolumab or CTLA-4/PD-1 doublets), whereas non-responders often display myeloid-dominant, TGF-β rich or angiogenic transcriptional profiles[51]. Although not yet standardized for routine practice, these findings support the concept that a composite “immune score” incorporating TILs, cytokine signatures, and myeloid markers may be more informative than any single parameter in MSS CRC[115] (Table 3).

Table 3 Candidate biomarkers for patient selection and monitoring in microsatellite-stable/proficient mismatch repair colorectal cancer.

Biomarker	Assay/readout	Rationale in MSS CRC	Clinical application and limitations	Ref.
MSIH/dMMR	IHC (MLH1/MSH2/MSH6/PMS2), PCR, NGS	High neoantigen load and inflamed TME	Established selection for PD-1 (± CTLA-4); rare in metastatic CRC	Gandini et al[130]; Ambrosini et al[131]
TMB-high; POLE/POLD1 ultra-mutation	NGS panel; TMB cutoffs	May confer immunogenicity despite MSS	Identifies small ICI-sensitive subset; platform/cutoff heterogeneity	Xue et al[132]; Domingo et al[135]
PD-L1 expression	IHC (TPS/CPS; immune cell staining)	Surrogate of immune activation, but weak predictor in CRC	Limited standalone utility; may contribute in combination contexts	Nakamura et al[137]
Immunoscore/TIL density (CD3/CD8)	Standardized IHC quantification (center + invasive margin)	Captures pre-existing antitumor immunity; enriches for ICI-responsive biology	Potential for patient selection (e.g., POCHI); requires harmonization and cutoffs	Yamaguchi et al[60]; Esmail et al[134]
Inflamed vs excluded gene signatures (IFN-γ; TGF-β/stromal)	RNA-seq/nano string signatures	Defines immune-hot vs immune-excluded states and matches mechanisms	Trial stratification; resource-intensive; signatures not yet standardized	Mortezaee and Majidpoor[27]; Li et al[28]; Esmail et al[134]
Liver metastasis status	Imaging; metastatic pattern	Liver promotes systemic immune tolerance and attenuates ICI benefit	Negative predictor in multiple ICI + TKI datasets; guides patient selection/Locoregional consolidation	Fakih et al[66]; Fakih et al[70]; Taïeb et al[138]
ctDNA kinetics	Plasma NGS (VAF dynamics; MRD)	Early molecular response/resistance signal	Monitoring and adaptive strategies; assay and threshold variability	Hou et al[140]
Host inflammatory/nutrition indices	NLR, albumin, ECOG	Correlates with myeloid skewing and immune fitness	Adjunct prognostic/predictive markers (not standalone)	Zhang et al[68]
Gut microbiome features	Metagenomics; targeted qPCR (e.g., Fusobacterium)	Microbiota shapes systemic immunity and therapy response/toxicity	Exploratory; actionable interventions (diet/FMT) under evaluation	Luo et al[33]; Luu et al[34]; Ramachandran et al[139]

MSI-H: Microsatellite instability-high; dMMR: Deficient mismatch repair; TMB: Tumor mutational burden; PD-L1: Programmed death-ligand 1; TIL: Tumor-infiltrating lymphocyte; CD: Cluster of differentiation; IFN: Interferon; TGF-β: Transforming growth factor-β; ctDNA: Circulating tumor DNA; IHC: Immunohistochemistry; NGS: Next-generation sequencing; PCR: Polymerase chain reaction; qPCR: Quantitative polymerase chain reaction; CPS: Combined positive score; TPS: Tumor proportion score; RNA-seq: RNA-sequencing; MRD: Minimal residual disease; VAF: Variant allele frequency; NLR: Neutrophil-to-lymphocyte ratio; ECOG: Eastern Cooperative Oncology Group; TME: Tumor microenvironment; MSS: Microsatellite-stable; CRC: Colorectal cancer; ICI: Immune checkpoint inhibitor; PD-1: Programmed cell death protein 1; CTLA-4: Cytotoxic T-lymphocyte-associated protein-4; TKI: Tyrosine kinase inhibitor; FMT: Fecal microbiota transplantation.

Composite and dynamic biomarkers

Given the multifactorial nature of immune resistance in MSS CRC, single-axis biomarkers are unlikely to adequately capture who will benefit from immunotherapy. Multi-omics approaches that integrate genomics, transcriptomics, proteomics, and spatial immunophenotyping are beginning to define composite signatures of response.

A landmark multi-omic study of MSS/pMMR mCRC treated with combination immunotherapy (regorafenib plus nivolumab and related regimens) identified distinct responder and non-responder clusters characterized by differences in T-cell inflammation, IFN-γ signaling, antigen presentation, and angiogenic and myeloid gene programs[51]. Responders frequently exhibited features such as higher T-cell-inflamed gene expression scores, intact HLA class I machinery, favorable microbiome signatures, and absence of extensive liver metastases, whereas non-responders had immune-excluded, VEGF-driven or TGF-β dominated microenvironments. Composite biomarker scores may outperform single markers in selected contexts, but their translational value depends on how they are operationalized, prospectively locked, and validated within biomarker-driven trial designs[8,136].

Dynamic blood-based biomarkers offer a complementary dimension. Circulating tumor DNA (ctDNA) has emerged as a highly sensitive tool for detecting minimal residual disease and predicting recurrence after curative-intent surgery in stage II-III CRC, with multiple prospective trials showing that post-operative ctDNA positivity is strongly associated with relapse and can guide adjuvant chemotherapy decisions[137]. Although most of these data pertain to chemotherapy, early evidence suggests that on-treatment ctDNA dynamics may also reflect sensitivity or resistance to checkpoint blockade. In MSI-H mCRC, a pre-planned analysis of the SAMCO-PRODIGE 54 trial demonstrated that a marked ctDNA decline within the first month of anti-PD-L1 therapy predicted improved progression-free and OS, whereas persistently detectable ctDNA identified patients at high risk of early progression[138]. Case reports in POLE-mutant MSS CRC similarly show tight concordance between ctDNA clearance and radiologic response to PD-1 inhibitors[139].

Beyond ctDNA, circulating immune cell signatures such as baseline neutrophil-to-lymphocyte ratio, expansion of peripheral effector/memory T-cell clones, or changes in MDSCs have been associated with ICI outcomes across tumor types and are beginning to be explored in CRC cohorts[140]. Extracellular vesicles and exosomes carrying PD-L1, immunomodulatory microRNAs, or metabolic enzymes represent another layer of potential dynamic biomarkers, although current data are mostly preclinical or small-scale[133].

Taken together, these developments suggest that future biomarker strategies in MSS CRC will likely rely on integrated, longitudinal models combining tumor tissue (genomic and spatial immune) information with serial liquid biopsies and peripheral immune profiling rather than static, single-time-point assays[141].

Patient selection and trial enrichment strategies

The extremely low response rates to ICIs in unselected MSS mCRC argue strongly for biomarker-enriched trial designs. Several principles can be distilled from current evidence.

First, prioritize highly immunogenic genomic subsets. Patients with POLE/POLD1-mutated, ultra-mutated, or otherwise TMB-high tumors after careful exclusion of sequencing artifacts and confirmation of strong neoantigenic signatures should be preferentially enrolled into ICI-containing arms, given their consistently high response rates across studies and case series[131].

Second, incorporate tumor immune contexture and metastatic pattern into eligibility and stratification. Enrichment for an “inflamed” microenvironment (e.g., high Immunoscore or CD8⁺ T-cell density, T-cell-inflamed gene signatures) and relative absence of extensive liver metastases may increase the likelihood of benefit from chemo-ICI, VEGF-TKI-ICI, or dual-checkpoint strategies, whereas heavily liver-dominant or immune-excluded phenotypes might be better triaged to trials combining immunotherapy with locoregional approaches or agents that remodel the myeloid/angiogenic niche[8].

Third, embed adaptive and basket/umbrella elements. Basket trials that enroll patients with POLE/POLD1 mutations or high T-cell-inflamed signatures across gastrointestinal malignancies, and umbrella trials that test different ICI-based combinations in molecularly defined MSS CRC subgroups (e.g., RAS-mutant/VEGF-high vs TGF-β-dominant vs DNA repair-deficient), can accelerate signal detection and allow for early stopping of futile arms[115]. Incorporating early on-treatment ctDNA and immune-profiling readouts into trial endpoints would enable adaptive randomization based on dynamic biomarkers rather than solely on baseline histology[138].

Finally, co-development of companion diagnostics is essential. As multi-omic signatures and composite scores progress from discovery to validation, their analytical performance, reproducibility across platforms, and clinical utility must be prospectively tested. Several consortia are now standardizing Immunoscore, TMB assays, and ctDNA platforms, providing a framework for future MSS CRC studies to integrate these tools in a way that is both scientifically robust and translatable to routine practice[142].

In summary, the move toward precision immunotherapy in MSS CRC hinges on shifting from single, static biomarkers to integrated, dynamic models that capture tumor genomics, immune contexture, metastatic pattern, and treatment-induced changes over time. Rational trial enrichment based on these principles will be crucial to convert small “exceptional responder” anecdotes into reproducible benefits for well-defined patient subgroups.

Operationalizing and validating composite scores for MSS CRC immunotherapy

Although conceptually appealing, composite scores are only clinically useful when they are defined as a locked, assay-specific algorithm with prespecified cutoffs and demonstrated reproducibility. A pragmatic operational roadmap for MSS CRC is to design composite scores around the three mechanistically actionable bottlenecks highlighted in this review priming competence, trafficking/infiltration, and effector persistence so that each component reflects a distinct resistance module rather than redundant markers.

Score construction (discovery to lock-down): Candidate features should be selected a priori from orthogonal biological layers that map to the bottlenecks, such as tumor antigenicity (e.g., TMB/neoantigen surrogates), T-cell inflamed activity (e.g., IFN-γ linked expression programs), and exclusion/suppression modules (e.g., stromal/TGF-β and myeloid programs), optionally integrating site-specific modifiers such as liver metastasis status. The model should be trained in a discovery cohort using penalized regression or other parsimonious approaches, and performance should be assessed via internal cross-validation and then external validation across independent cohorts to mitigate optimism bias and geographic/center effects. The final formula, assay platform (next-generation sequencing vs RNA panel vs immunohistochemistry), and decision threshold(s) must be locked before prospective testing[136].

Analytical validation (assay readiness): Prior to clinical use, the score requires analytical validation, including specimen requirements, pre-analytic stability, inter-laboratory reproducibility, and batch-effect controls. Importantly, composite scores should specify whether they are intended as binary enrichment tools (high vs low) or continuous predictors (treatment-by-score interaction), because this choice dictates statistical testing and trial implementation[143].

Prospective validation and trial integration: A realistic path is a stepwise program: (1) Prospective-retrospective validation in archived samples from completed trials with prespecified hypotheses; and (2) Prospective confirmation in biomarker-stratified or biomarker-enriched studies. From a trial-design perspective, composite scores can be incorporated through: (1) Stratified randomization with prespecified interaction testing; (2) Enrichment designs that restrict enrollment to score-high patients when biological rationale and prior data support differential benefit; or (3) Adaptive enrichment designs that allow preplanned modification of enrollment based on interim evidence while controlling type I error. Master-protocol approaches (umbrella/platform) can further evaluate score-guided combinations efficiently across modules (e.g., priming + trafficking vs trafficking + effector), which is particularly relevant for MSS CRC where resistance is multi-layered[143].

Actionability and clinical workflow: To maximize translational relevance, a composite score should be paired with a prespecified action rule that maps dominant bottlenecks to combination choices (e.g., exclusion-high, add stromal/vascular normalization; myeloid-high, add myeloid reprogramming; inflamed-low, add priming agents). Such “score-to-therapy” mapping converts composite biomarkers from descriptive correlates into trial-ready decision tools.

CHALLENGES, KNOWLEDGE GAPS, AND FUTURE DIRECTIONS

Biological unknowns and translational hurdles

Despite intensive effort, most MSS CRCs remain refractory to ICIs and we still do not fully understand what a successful “cold-to-hot” conversion actually looks like in humans. Current combination trials often rely on coarse endpoints (ORR, PFS) and limited pre-/on-treatment biopsies, which makes it difficult to deconvolute causality: Did a regimen truly induce de novo T-cell priming and intratumoral trafficking, or did it simply transiently upregulate PD-L1 or expand pre-existing, non-productive lymphocytes? Recent reviews emphasize that multiple resistance mechanisms (defective antigen presentation, myeloid suppression, stromal barriers, metabolic constraints) usually co-exist in MSS CRC, so “turning hot” likely requires coordinated changes across several compartments, not just an increase in CD8⁺ density[8].

To precisely validate “cold-to-hot” transitions, we need prospective studies with: (1) Serial, spatially resolved biopsies or resection specimens (e.g., neoadjuvant window-of-opportunity designs); (2) Integrated multi-omics (genome, transcriptome, spatial proteomics, TCR/B cell receptor sequencing); and (3) Matched liquid-biopsy and peripheral immune profiling. Some neoadjuvant trials in MSS rectal and colon cancer are beginning to incorporate such intensive correlative work, but sample sizes remain small and assays heterogeneous[86]. Establishing standardized definitions (e.g., a composite “conversion score” incorporating T-cell infiltration, TCR clonality, IFN signatures, and myeloid/stromal remodeling) is an unmet need if we want to compare strategies systematically[144].

A second major knowledge gap is the longitudinal interplay among gut microbiota, liver immune tolerance, and tumor evolution. Multiple lines of evidence indicate that the gut-liver axis shapes CRC progression and immunotherapy responses: Dysbiosis can promote CRC initiation and liver metastasis; Microbial metabolites influence intrahepatic immune cell polarization; And liver metastases, in turn, act as systemic “sinks” for cytotoxic T cells[145]. Yet most data are cross-sectional. We lack time-resolved studies tracking how microbiome composition, liver immune cell states, and tumor subclones co-evolve under chemotherapy, anti-VEGF therapy, and ICIs in MSS CRC[146].

Similarly, the evolutionary dynamics of MSS CRC under immune pressure are incompletely defined. We know that under effective ICI in MSI-H CRC, subclones with B2M or JAK/STAT pathway defects can be selected, leading to acquired resistance; whether analogous processes occur in the minority of MSS responders to combination immunotherapy is poorly characterized[147]. Systematic sampling of multiple metastatic sites over time (including liver and peritoneum) will be required to understand how regional immune niches and treatment sequences shape clonal architecture and immune escape.

Optimizing combination strategies and sequencing

Most current strategies in MSS CRC involve layering ICIs on top of chemotherapy, anti-angiogenic agents, TKIs, radiotherapy or local treatments, yet we have little evidence-based guidance on the optimal order, dose, and timing. Retrospective and early prospective data suggest that some regimens may require priming or “induction” phases (e.g., low-dose VEGF/VEGFR blockade to normalize vasculature and reduce myeloid suppression) before PD-1/PD-L1 is introduced, whereas others may benefit from truly concurrent administration to exploit immunogenic cell death[87].

For example, regorafenib or fruquintinib plus PD-1/PD-L1 inhibitors show modest but real activity in refractory MSS mCRC, with meta-analytic ORRs around 10%-15% and better outcomes in patients without liver metastases[148]. Yet the ideal TKI dose (full vs “vascular-normalizing” reduced dose), whether to start TKI first then add PD-1, and how long to continue each drug remain unsettled, and may differ between agents (e.g., regorafenib vs fruquintinib vs lenvatinib)[149]. Likewise, combinations that initially looked promising such as atezolizumab plus cobimetinib, or lenvatinib plus pembrolizumab in unselected mCRC ultimately failed to improve survival in phase III, underlining that pharmacodynamic synergy in preclinical models does not guarantee clinically meaningful benefit at tolerable doses[150].

Radiotherapy and local interventions (SIRT, TACE, CRS-HIPEC) add further complexity. Their immunologic impact is highly schedule-dependent: Different fractionation schemes, dose rates, or timing relative to ICI can either prime robust T-cell responses or instead exacerbate lymphopenia and T-cell exhaustion[150]. Prospective “schedule-optimization” trials that explicitly randomize sequencing (e.g., induction chemo/TKI, ICI vs ICI lead-in, chemo, or SBRT before vs after PD-1) are still rare in MSS CRC.

A parallel concern is the tendency toward “over-stacking” multiple targeted agents and ICIs, which may push patients into overlapping toxicities (hepatic, gastrointestinal, immune-related) without proportionate efficacy gains. Recent reviews caution that more complex regimens triplets combining PD-1, CTLA-4, anti-VEGF, and a TKI, for example should be justified by strong biologic rationale and biomarker enrichment rather than empirical escalation[151]. Rational trial design will require early pharmacodynamic readouts (e.g., on-treatment biopsies, ctDNA, immune cell phenotyping) to identify non-beneficial arms quickly and avoid unnecessary exposure.

Toxicities, cost, and quality-of-life considerations

As combination immunotherapy moves into earlier lines of MSS CRC treatment, immune-related adverse events (irAEs) and financial toxicity become central issues. While checkpoint inhibitors are generally better tolerated than multi-agent chemotherapy, adding them to existing regimens can substantially increase risks of endocrinopathies, colitis, pneumonitis, hepatotoxicity, and rare but severe events such as myocarditis[147]. VEGF-TKI-ICI combinations also accentuate hypertension, hand-foot syndrome, bleeding and proteinuria, requiring meticulous dose-modification and multidisciplinary management[148].

Real-world studies across tumor types indicate that irAEs can both impair and, in some cases, correlate with improved outcomes, complicating clinical decision-making around re-challenge or treatment discontinuation[152]. For CRC specifically, long-term health-related quality-of-life (HRQoL) analyses with ICIs are limited, but early data suggest that while survivors often experience stable or improved global QoL compared with cytotoxic therapy, chronic toxicities (fatigue, arthralgia, endocrine dysfunction) can persist and require survivorship care planning[152].

Economically, adding high-cost biologics and ICIs to already expensive mCRC regimens poses serious questions of value. Global cost-of-illness analyses show that mCRC accounts for a disproportionate share of direct medical expenditures, largely driven by extended use of targeted agents and hospitalizations for complications[153]. Population-level analyses of ICI use indicate meaningful survival gains but also highlight substantial drug and toxicity management costs, with financial toxicity particularly pronounced in lower-income health systems[154]. For MSS CRC where response rates are low outside biomarker-selected subsets the risk of exposing broad populations to high-cost, low-yield therapies is non-trivial. Robust cost-effectiveness models that incorporate predictive biomarkers, irAE incidence, and real-world HRQoL will be essential to inform reimbursement and guideline decisions[153].

Outlook: Toward personalized immunotherapy for MSS CRC

Future progress in MSS CRC immunotherapy is likely to be driven by precision integration rather than any single new agent. In practice, this means combining tumor-intrinsic profiling (genomics, transcriptomics, spatial immune architecture) with host- and site-specific variables (microbiome, liver vs peritoneum niche) and real-time disease kinetics (serial ctDNA). Recent spatially resolved studies in CRC highlight that where immune cells sit e.g., immune-active tumor-stroma boundaries or permissive macrophage T cell niches can track with checkpoint sensitivity, supporting the concept of “immune ecotypes” within MSS disease that can be computationally learned from multiplex pathology and multi-omics and then used for treatment allocation. Meanwhile, ctDNA has matured into a pragmatic dynamic biomarker for risk stratification and on-treatment adaptation, making response-guided de-escalation/escalation a realistic design principle for MSS trials. On the therapeutic side, “targeted therapy + ICI” backbones continue to consolidate, including regimens such as regorafenib + PD-1 blockade and fruquintinib + PD-1 blockade, and triplets such as regorafenib-ipilimumab-nivolumab, with multiple datasets reinforcing that organ context (especially liver involvement) is a major determinant of benefit. Finally, biologically defined “MSS exceptions” are becoming actionable: POLE/POLD1 proofreading-deficient mCRC can show ICI outcomes that rival or exceed classical MSI-H/dMMR cohorts, underscoring that MSS should be treated as a spectrum rather than a uniform exclusion label[155].

Personalization will also increasingly incorporate microbiome and immunometabolic state as manipulable co-variables, not just correlates. Clinical FMT data in refractory solid tumors provide a plausible template for CRC-focused protocols that aim to convert non-permissive microbiota states into “responder-like” configurations, while parallel programs test metabolic checkpoint relief (e.g., adenosine/tryptophan axes) as enablers of T-cell function in otherwise noninflamed tumors. Procedure-enabled immunotherapy is likely to expand for selected MSS CRC subsets: Personalized neoantigen-reactive TCR-engineered ACT has shown feasibility and early regression signals in metastatic gastrointestinal cancers, and regional delivery concepts are gaining clinical traction (e.g., intrahepatic arterial administration of an oncolytic vaccinia platform for liver-dominant mCRC), aligning with the broader strategy of locally igniting immunity in anatomically immunosuppressive niches and then sustaining it systemically. The operational endpoint for the next 5-10 years is an adaptive, biomarker-co-developed paradigm (umbrella/basket or enrichment designs) that assigns MSS patients to rational combinations or cellular/Loco-regional intensification based on immune ecotype plus ctDNA-guided kinetics, while prospectively accounting for toxicity, feasibility, and cost[112].

CONCLUSION

MSS CRC remains the immunologic “blind spot” of checkpoint blockade because resistance is typically multi-layered limited antigenicity and/or antigen-presentation competence, stromal-vascular exclusion that restricts trafficking, and durable myeloid-dominant suppression that can be amplified by organ-specific sanctuaries (notably liver and peritoneum) and host modifiers such as the microbiome. The field has therefore shifted from ICI monotherapy to multidimensional “cold-to-hot” combinations; however, the next translational step is to convert this catalog into testable, biomarker-driven strategies. We propose that durable benefit in MSS CRC will generally require simultaneous relief of at least two bottlenecks priming competence, trafficking/infiltration, and effector persistence rather than intensifying PD-L1 blockade alone, a hypothesis that can be prospectively evaluated by paired-tissue pharmacodynamics (CD8⁺ influx, IFN-γ programs), systemic immune readouts, and early ctDNA kinetics. We further hypothesize that liver metastases function as a systemic immune “sink” that imposes a dominant site-specific penalty, such that metastasis-aware trial stratification and liver-niche targeted adjuncts should be required to unlock extrahepatic benefit in this subgroup. Finally, because static single biomarkers are insufficient in MSS CRC, prospectively locked composite scores coupled to early on-treatment dynamics should guide adaptive escalation/de-escalation and sequencing, enabling enrichment for biologically plausible responders while containing toxicity and cost. Overall, progress will depend less on identifying additional combinations and more on designing bottleneck-matched, metastasis-aware, and schedule-informed regimens tested in prospective, biomarker-enriched trials with clinically actionable decision rules, with the goal of making durable immunotherapy benefit in MSS CRC a predictable outcome for defined patient subsets rather than an exceptional event.