Facts and Hopes in Colorectal Cancer Immunotherapy

Immune-modulating therapies have transformed the cancer treatment landscape. Immune-facilitated tumor responses are likely driven by two interacting disease features: (i) tumor-intrinsic foreignness derived from aberrant molecular alterations, and (ii) functionality of the tumor-extrinsic immune microenvironment (1–3). Highly immunogenic cancers, including non–small cell lung cancer (NSCLC), melanoma, and mismatch repair–deficient (MMRd) cancers, exhibit robust responses to immune checkpoint inhibition (ICI) of programmed cell death protein-1 (PD-1), programmed death ligand-1 (PD-L1), and cytotoxic T-lymphocyte–associated protein 4 (CTLA-4; refs. 4–7). However, this success is not seen in the vast majority of patients with mismatch repair–proficient (MMRp) colorectal cancers; immunotherapy is not approved for 95% of patients with colorectal cancer.

In this review, we analyze immunotherapy failures and promising signals to define an optimistic path for increasing immunotherapy benefit in colorectal cancer.

Immunotherapy efficacy requires recognition of molecular alterations that accompany cellular disorder. Tumor mutations can be translated into irregular peptides (neoantigens) which are presented for T-cell recognition via the MHC type I or type II (8–10). MMRd colorectal cancer is characterized by the loss of mismatch-repair protein(s) and the subsequent development of microsatellite instability, with accumulation of predominantly frameshift tumor mutations associated with high neoantigen burden. The landmark phase II clinical trial by Le and colleagues demonstrated that anti–PD-1 monotherapy provoked a 40% immune-related objective response rate (ORR) in heavily pretreated patients with MMRd cancers, including metastatic colorectal cancer (mCRC; ref. 11). The randomized, phase III KEYNOTE-177 trial demonstrated that first-line pembrolizumab (anti–PD-1) in MMRd mCRC led to a progression-free survival (PFS) benefit compared with first-line standard of care (16.5 months vs. 8.2 months; P = 0.0002; ref. 12). ORRs up to 65% were exhibited in a phase II study of nivolumab (anti–PD-1) and ipilimumab (anti–CTLA-4) in pretreated MMRd mCRC (13).

Early stage MMRd colorectal cancer tumors appear even more susceptible to ICIs. Complete response rates of 100% and 67% have been shown in phase II studies of stage II–III MMRd rectal or colon cancer, respectively (14–16). The explanation for this heightened efficacy is unknown, but may be related to favorable early microenvironmental features and/or a higher dependency of younger tumors on immune checkpoints for immune escape.

The outstanding success of ICI in MMRd colorectal cancer proves that sufficiently immunogenic colorectal cancer can be targeted by the immune system. However, for the 95% of patients with MMRp mCRC, ICI efficacy is poor.

MMRp mCRC tumor mutational burden (TMB) is on average half to 10-fold lower than tumors that exhibit high ICI response rates, such as melanoma, NSCLC and MMRd colorectal cancer (17, 18). No responses to anti–PD-1 monotherapy were witnessed in early studies of MMRp mCRC (11). An overall negative study of patients with MMRp mCRC randomized to combination durvalumab (anti–PD-L1) and tremelimumab (anti–CTLA-4) versus best supportive care showed that a subgroup of patients exhibited modestly increased survival in the ICI arm (6.6 months vs. 4.1 months; ref. 19). Notably, there was one objective response and no PFS difference in this small study.

One potential strategy to mitigate low MMRp mCRC immunogenicity is to select patients that exhibit tumor profiles akin to MMRd tumors, such as MMRd genomic signature patterns and/or especially high TMB. A retrospective analysis of 1,662 patients with advanced cancers, including colorectal cancer, determined that tumors in the top 20th TMB percentile per cancer type exhibited improved overall survival (OS) after ICI (17). Despite laudable efforts to define a predictive threshold to select ICI-appropriate cases, subsequent analyses have shown that this direct association between TMB and ICI-conferred OS disappears when only MMRp colorectal cancers are selected, with the rare exception of tumors harboring aberrant DNA polymerase δ and ε (POLD, POLE) that exhibit enormous mutational loads (1, 20, 21). TMB is a promising, but likely incomplete measurement of intrinsic tumor foreignness, and further controlled prospective studies are required to determine a threshold to predict ICI benefit in colorectal cancer. An increased focus on neoantigen qualities, such as peptide similarities to nonhuman infectious proteins and/or characteristics that increase the likelihood of HLA presentation and T-cell recognition, may better select disproportionately immunogenic tumors (22–24).

Immunotherapy success pan-cancer tends to disproportionately occur in tumors within a favorable “immune inflamed” microenvironment with infiltrating T cells and reduced stromal immunosuppressive signaling (2, 3). The colonic microenvironment is especially complex; functional digestion requires tolerance to commensal intestinal flora full of foreign antigens, swallowed material, and metabolic waste byproducts (25, 26). Higher microbial abundance and species diversity may enable especially pronounced immunotherapy responses in different cancer types (27, 28). The constellation of these external relationships integrated with intrinsic tumor aberrancy culminates in a diverse spectrum of overall colorectal cancer consensus molecular subtypes (CMS) from the functional, lymphocyte-dense MMRd signature (CMS1) to “T cell impaired” landscapes (CMS4; ref. 29). All colorectal cancer tumors and metastatic sites, however, may not exhibit the same degree of functional immunosuppression (30). Efforts to modify the local tumor-microenvironment interactions have been attempted to improve immunotherapy success in MMRp colorectal cancer.

The success of combination cytotoxic chemotherapy alongside ICI in multiple cancer types, including NSCLC and breast cancer, supports the growing hypothesis that chemotherapy may modulate the tumor environment and ICI-enabled disease control (31, 32). Cytotoxic chemotherapy, including oxaliplatin, may reduce immunosuppressive T-regulatory cells, induce cell surface death receptor density to facilitate immune-killing, and enhance neoantigen release in the microenvironment (33–35). This strategy has not been broadly successful in colorectal cancer (Fig. 1). First-line 5-FU, leucovorin, oxaliplatin, (FOLFOX) and bevacizumab with nivolumab provoked identical median PFS to chemotherapy alone, in a randomized phase II trial (36). However, the nivolumab arm exhibited higher response durability; 28% of patients in the chemo-ICI arm achieved 18-month PFS compared with 9% of patients with chemo, alone. Phase I/II single arm studies of first-line FOLFOX combined with immunotherapy, such as bevacizumab/ durvalumab/ oleclumab (COLUMBIA-1) or durvalumab/tremeliumumab (MEDITREME), have shown increases in response rates, but unclear PFS benefit compared with historical outcomes (37, 38). The phase II AtezoTRIBE study demonstrated that FOLFOXIRI/bevacizumab/atezolizumab modestly improves median PFS from 11.5 to 13.1 months (P = 0.018) compared with chemotherapy with no significant increase in response rate (38, 39).

Many studies examine whether radiation improves immunotherapy responses. Rarely, an abscopal effect is documented where radiation of a single tumor site provokes a response in a non-radiated metastasis, supporting a hypothesis that radiation-facilitated tumor damage may lead to systemic immune response (40–43). A phase II study of durvalumab (anti–PD-L1) and tremelimumab (anti–CTLA-4) plus radiotherapy in chemo-refractory patients with MMRp mCRC did not meet its overall primary endpoint, yet responses were seen in distant non-radiated lesions in 8% (2/24) patients, accompanied by increased CD8 T cells, suggesting potential systemic immunomodulation (44). A similar phase II study of radiotherapy concurrent with ipilimumab (anti–CTLA-4) and nivolumab (anti–PD-1) defined a disease control rate in 10 of 40 (25%) of patients with mCRC in the intention to treat cohort (45). In the future, optimization of radiation timing, fractionation, and duration relative to immunotherapy may be warranted (43).

Tyrosine kinase inhibitors (TKI) and/or VEGF inhibition have been hypothesized to reduce immunosuppressive signaling by increasing dendritic cell maturation and priming (anti-VEGF, bevacizumab; ref. 46), enhancing pro-stimulatory interferon gamma responses (multi-TKI; lenvatinib; ref. 47), and decreasing secretion of pro-tumor cytokines (BRAF inhibitor; vemurafenib; refs. 33, 48).

In colorectal cancer, studies of co-TKI/ICI therapies demonstrate variable results (Table 1). Single arm studies of lenvatinib with pembrolizumab (49), and regorafenib with nivolumab (50) defined modest ORRs of 22% (7/32) and 7% (5/70), respectively. A separate phase I mCRC study of regorafenib, nivolumab and ipilimumab demonstrated an ORR of 31% (9/29) with a disease control rate of 65% (51). Response rates to single agent regorafenib, and single-agent or combination ICI in MMRp mCRC, are less than 1%, suggesting that these combinations may have synergistic activity (50, 52). A phase III study is ongoing to evaluate lenvatinib/pembrolizumab in MMRp mCRC (NCT04776148).

Colorectal cancers often exhibit mutations in MAPK signaling mediators. In vitro, inhibition of pathway modulators, including MEK1/2, can reprogram effector CD8 T cells into a regenerative, stem cell–like memory phenotype that may assist ICI (53). In patients who previously received anti-EGFR therapy, cetuximab (anti-EGFR) rechallenge with avelumab resulted in a relatively low ORR of 8.5% in a single-arm phase II study (39). For patients with BRAF V600E mutated MMRp mCRC, a single-arm phase I/II trial of encorafenib (BRAF inhibitor), cetuximab, and nivolumab demonstrated an ORR of 45%, higher than the conventional ORR of 20% seen in the TKI-combo, alone (54, 55). These results led to a randomized phase II trial (SWOG 2107) of encorafenib/cetuximab with or without nivolumab (NCT04017650). Similar efficacy with an ORR of 24% was seen with trametinib (MEK inhibitor), dabrafenib (BRAF inhibitor) plus PDR001 (anti–PD-1) in a phase II study (29).

Unfortunately, clinical activity of other MAPK/ICI combinations is limited. In the IMBlaze370 phase III trial, no OS advantage occurred with atezolizumab (anti–PD-L1) plus cobimetinib (MEK inhibitor) versus atezolizumab monotherapy or regorafenib (56). A phase II study of durvalumab plus trametinib in a similar population also failed: only one patient responded to therapy (57). Overall, the measured successes in single-arm studies paired with failures in multi-arm trials demonstrate that combination ICI plus conventional therapies appear unlikely to dramatically shift the landscape of success for colorectal cancer. Novel combinations and/or predictive biomarkers for patient selection are needed.

A next generation of potent ICI agents may improve T-cell recognition even in the setting of low, MMRp colorectal cancer immunogenicity (Fig. 2). Botensilimab, an Fc-enhanced next generation CTLA-4 antibody, is designed to promote intratumoral regulatory T-cell depletion via Fc gamma receptor signaling, activation of natural killer (NK) cells and macrophages, and T-cell enhancement via CTLA-4 blockade. Preliminary results from the phase IA/B study of botensilimab and balstilimab (anti–PD-1) have demonstrated an ORR of 24% (10/41) in patients with pretreated MMRp mCRC (58).

Lymphocyte-activation gene 3 (LAG-3), an inhibitory immune checkpoint, is often co-expressed with immune checkpoint PD-1 on tumor-infiltrating lymphocytes; both contribute to T-cell exhaustion. LAG-3 inhibitors are approved in melanoma (59). In the dose confirmation phase of a multi-cohort, phase I study, favezelimab (anti–LAG-3) plus pembrolizumab or MK-4280A (favezelimab-pembrolizumab co-formulation), showed promising antitumor activity in MMRp colorectal cancer, compared with favezelimab alone, most notably in pts with PD-L1 CPS ≥ 1 tumors, with a 12-month survival rate of 50.6% compared with 29.5% in PD-L1 CPS low tumors (60). A randomized, phase III study of MK-4280A compared with standard of care in patients with PD-L1–positive colorectal cancer is underway (NCT05064059).

Immunorecognition of the modest MMRp mCRC neoantigen load may be facilitated by vaccine priming towards common mCRC alterations. Phase I evaluation of BN-CV301, a poxvirus-engineered vaccine against MUC1 and CEA gastrointestinal antigens, induced T-cell responses against target antigens and provoked one partial response in mCRC (61). Similarly, a TP53 synthetic peptide vaccine induced T-cell specific anti-antigen responses in 9 of 10 treated patients with MMRp mCRC; in over half of the patients anti-TP53 activity was durable for over 6 months (62). Addition of the seven antigen PolyPEPI1018 vaccine to maintenance chemotherapy in patients with MMRp mCRC may facilitate immune-related tumor responses (63). However, despite promising preliminary data, cancer vaccines remain slow to translate into conventional cancer treatment. In a phase II randomized, placebo study of patients with mCRC given the anti-MUC1 tecemotide vaccine after liver resection, OS was not improved with the vaccine (64). In separate trials, FOLFIRI plus administration of anti-CEA or antitumor antigen 5T4 vaccines demonstrated promising initial antibody responses, yet limited durability (65, 66).

An emerging strategy to redirect T cells involves the design of small molecules that create physical bridges between immune effector cells and tumor cell antigens. Bispecific antibodies are engineered to have separate binding sites directed against two different epitopes (67, 68). There are two major platforms of bispecific antibodies. IgG-like antibodies exhibit extended stability and solubility and possess Fc fragments which, depending on the design, may enable additional biological activity through antibody-dependent cell-mediated cytotoxicity and complement-dependent cytotoxicity (69). Non-IgG-like bispecific antibodies are smaller due to absence of the Fc component (70), such as blinatumomab, which is approved for use in acute lymphoblastic leukemia (67, 71).

The evolution of bispecific antibody clinical design in colorectal cancer is ongoing with considerable variability in drug efficacy and tolerability. In a phase I study of 39 patients given MEDI-565, no objective responses were observed, high rates of antidrug antibodies were witnessed, and several patients developed severe inflammation-related adverse events (72).

Bispecific antibody re-optimization to reduce off-target inflammation and improve drug potency in later generations appears promising. The CEA-CD3 T-cell bispecific antibody cibisatamab (RO6958688) generated potent T-cell engagement and activation against CEA-positive tumor cells in vitro (73). Clinical evaluation of cibisatamab alone or in combination with atezolizumab (anti–PD-L1) in 31 patients with advanced colorectal cancer determined high rates of drug-assisted tumor inflammation (74). Three of 14 (21.5%) evaluable patients with colorectal cancer given the combination of cibisatamab and atezolizumab exhibited a partial response, and tumor bulk reduction in excess of 10% was seen in an additional 36% of patients (5/14). Additional evaluation of this drug combination in MMRp mCRC after obinutuzumab pretreatment is ongoing (NCT03866239).

Additional bispecific antibodies in development in mCRC include the anti-EGFR and the CD28 T-cell receptor engager (REGN7075) in combination with cemiplimab (anti–PD-1; NCT04626635); the guanylyl cyclase C (GUCY2C) and the CD3 T-cell receptor engager (PF-07062119) alone and in combination with sasanlimab (anti–PD-1) or anti-VEGF (NCT04171141); and, the fibroblast activation protein-a (FAP) and CD137/4–1BB T-cell and NK cell receptor engager RO7122290, in combination with cibisatamab after obinutuzumab pretreatment (NCT04826003), among others (Fig. 2).

Another modality to enable colorectal cancer immunorecognition includes the autologous transplant of T cells modified with chimeric antigen receptors (CAR) specific to an intrinsic colorectal cancer antigen (75). CAR T cells have been more successful in liquid versus solid cancers (76–78). Data for CAR T cells in colorectal cancer is limited. In a phase I mCRC study, anti-CEA CAR T-cell transplantation was well tolerated and enabled disease control in 4 of 10 patients (2 with tumor shrinkage; ref. 79). However, considerable challenges remain, as similar phase I investigations of CEA-CAM5-specific (80) and TAG-72 specific (81) CAR T cells demonstrated limited CAR T clonal longevity and minimal antitumor responses. GCC19CART, designed to overcome limitations of conventional CAR T cells in solid tumors, pairs CAR T cells that target guanylate cyclase-C (GCC) with CD19 targeting CAR T cells to amplify proliferation and activation of the solid tumor CAR T component. A phase I, investigator-initiated clinical trial in China reported results from two dose escalation cohorts, 1 × 10⁶ or 2 × 10⁶ CAR T cells/kg, demonstrating an ORR of 15.4% (2/13) and 50% (4/8), respectively (82). This therapy continues to be evaluated (NCT05319314).

If therapies to improve intrinsic neoantigen recognition are insufficient, creative methods to: (i) induce increased native antigen expression and/or (ii) form new intrinsic antigens could transform MMRp mCRC into an MMRd molecular phenotype. Oncolytic viral-based platforms may be able to introduce new immunogenic tumor antigens into colorectal cancer to increase tumor immunogenicity (83, 84). Use of mutagenic agents, including temozolomide, may induce mutations and immune-related responses in a subset of patients with methylated methyl-guanine-methyl-transferase colorectal cancer in two phase II studies (85, 86).

Multiple studies across treatment modalities have shown that metastatic geography is an important feature for tumor immunoresponsiveness. In particular, across malignancy types, patients with liver metastases appear to exhibit worse immunotherapy outcomes (87–90). In the phase II study of patients with MMRp mCRC given combination regorafenib and nivolumab, 0 of 47 patients with liver metastases demonstrated a response, compared with 5 of 23 (22%) of patients with non-liver metastases (50). This was similarly observed in patients who received regorafenib with ipilimumab and nivolumab; 0 of 7 responses occurred in patients with liver metastases, compared with 9 of 22 (41%) of patients with non-liver metastases (47). Furthermore, in the combination botensilimab and balstilimab trial, an ORR of 42% (10/24) was seen patients without liver metastases, compared with 24% (10/41) in the overall population (26). These post hoc subgroup results provide optimism that future prospective studies enriched for patients without liver metastasis may show better ICI benefit. A randomized phase II study to compare botensilimab/balstilimab versus investigator's choice is underway in patients without liver metastasis (NCT05608044).

Explanations for these findings are under investigation. The liver is a tertiary immune organ with unique native antigen presenting cells which may create a tolerogenic microenvironment to prevent auto-immune reactions against filtered byproducts (87, 91–93). Notably, liver metastases confer a poor prognosis (94) and biological differences in the lower growth rates and reduced lethality of lung-predominant mCRC may confound observed trends.

Several broad patterns emerge from the history of immunotherapy in colorectal cancer that inform future efforts. MMRd colorectal cancer exhibit high TMB and sustained responses to ICI in early and metastatic settings. However, sustained antitumor responses are rarely observed in MMRp colorectal cancer across multiple trials of checkpoint inhibition, with or without the addition of chemotherapy, targeted therapies or radiotherapy. Tumor immunorecognition in MMRp colorectal cancer likely requires multi-modal activation and purposeful selection of complementary agents; several hopeful strategies have emerged.

T-cell redirecting bispecific antibodies can enable sustained T-cell engagement and heightened effector function within the colorectal cancer tumor microenvironment. Efforts to optimize T-cell redirecting therapies, and identify new surface targets and combination partners are especially promising. Next generation ICIs, such as the Fc-modified CTLA-4 inhibitor, botensilimab, and the LAG-3 inhibitor, favezelimab, have also shown preliminary signs of success, leading to randomized phase II and III studies

The mechanism of antigenic tolerance is not well defined, and heavy investment in immunobiology is needed to discover new targets. A shifting focus towards B-cell and Natural Killer therapies may help diversify the breadth of immune system activation to induce better immunological memory and decrease tolerogenic signals (95). In addition, close examination of MMRp mCRC immunotherapy responders may identify new biomarkers and facilitate a gradual transition to personalized immune-based regimens for especially immunogenic tumors. For all treatments, dosing and toxicity optimization will allow for more sustainable immune activation. Altogether, knowledge learned from prior immunotherapy failures and modest early successes encourages continued optimism for the future development of colorectal cancer immunotherapy.

G. Argilés reports personal fees from Gadeta BV and Amgen outside the submitted work. B. Rousseau reports grants and personal fees from Neophore during the conduct of the study; in addition, B. Rousseau has a patent for Methods and composition for cancer immunotherapy issued to US20230056846A1. N.H. Segal reports personal fees from Novartis, Puretech, Numab, AstraZeneca, GSK, ABL Bio, Revitope, Roche/Genentech, and Boehringer Ingelheim; grants from Roche/Genentech, Pfizer, Merck, BMS, AstraZeneca, Puretech, Immunocore, Regeneron, and Agenus; and other support from AstraZeneca and Regeneron outside the submitted work. No disclosures were reported by the other author.

MSK Cancer Center Support Grant/Core Grant (P30 CA008748).