Mismatch Repair–Proficient Colorectal Cancer Can Evade Immune Surveillance through an Intrinsic Suppressive Program

Introduction

Colorectal cancers can be classified based on the functional status of the DNA mismatch repair (MMR) system into two main subtypes: MMR-proficient (MMRp) and MMR-deficient (MMRd) tumors (1). In MMRd colorectal cancers, the inactivation of the MMR machinery is most commonly due to genetic or epigenetic alterations causing loss of function of the MLH1, MSH2, MSH6, and PMS2 genes. This leads to genetic hypermutability and the accumulation of insertion–deletion mutations at microsatellite loci, a phenomenon known as microsatellite instability (MSI; 2). In colorectal cancer, MSI is therefore considered a hallmark feature of MMR deficiency. Conversely, MMRp tumors retain an intact MMR system and are characterized by microsatellite stability (MSS). MMRp/MSS tumors dominate the colorectal cancer landscape, representing approximately 85% of primary colorectal cancer cases. When considering the metastatic setting, their prevalence is even higher, accounting for nearly 95% of patients with colorectal cancer (3). For clarity and consistency with experimental readouts throughout this study, tumors are hereafter referred to according to their microsatellite status as MSI or MSS.

The advent of cancer immunotherapy, in particular the use of checkpoint inhibitors targeting programmed cell death protein 1 (PD-1) and cytotoxic T-lymphocyte–associated protein 4 (CTLA-4), has shown impressive clinical results in a subset of patients with solid tumors (4). Patients affected by metastatic MSI colorectal cancer often respond to therapies based on immune checkpoint blockade (ICB), whereas those with MSS tumors do not (5).

MSI tumors usually display high tumor mutational burden (TMB), a feature linked to rapid progression but also to favorable prognosis in colorectal cancer (6). Because of their high mutational burden, MSI tumors express a high number of neoantigens, novel protein sequences resulting from tumor-specific mutations that, when expressed on MHC class I (MHC-I) or II (MHC-II) molecules, may be recognized by patient T cells. According to the prevalent view, high TMB leading to elevated number of neoantigens increases tumor’s visibility to the T cell–based immune system (7). Levels of TMB are correlated with therapeutic response to ICB and, as a result, the US Food and Drug Administration (FDA) has granted tumor-agnostic approval for treatment with anti–PD-1 pembrolizumab for adult and pediatric patients with unresectable or metastatic solid tumors exhibiting a TMB higher than 10 mut/Mb (8, 9). Therapies based on checkpoint inhibitors are emerging as the standard-of-care treatment for MSI colorectal cancers (10–15).

With the exception of specific cases (e.g., POLE-mutated tumors), MSS tumors are instead characterized by a much lower number of mutations, and the prevalent view functionally links this feature to ineffective immunogenicity and lack of response to immune checkpoint inhibitors (ICI; 3). However, recent insights suggest a more nuanced picture. For instance, upon administering immunotherapy, MSS tumors show markers of immune activation, such as surge in CD8⁺ T cells or heightened IFNγ score, albeit with minimal (or absent) tumor regression (11). Moreover, with the goal of increasing the immunogenicity of MSS tumors, innovative clinical strategies have been implemented. For example, the alkylating agent temozolomide has been used to induce a “hypermutant” status in MSS patients, as a priming treatment to boost the number of tumor’s neoantigens, prior to anti–PD-1 treatment (16–18). Yet, despite an increase in the TMB, only a subset of patients benefited from ICB, whereas the majority of the patients remained refractory to checkpoint inhibition. These results suggest that factors beyond neoantigen load may be involved in therapeutic resistance of MSS tumors to cancer immunotherapies (19). It is conceivable that during their development, colorectal cancer could have acquired active mechanisms to evade immune surveillance.

Elucidating the mechanisms behind the poor response of MSS tumors is the key to develop new therapeutic strategies. However, the complexity of the tumor–immune system interaction, consisting of many different intrinsic and extrinsic effectors, makes it difficult to functionally pinpoint the underlying mechanism(s) of resistance (20). To tackle this issue, we developed a model system whereby both MSI and MSS tumor cells express equal levels of a given antigen which is recognized by engineered T cells. Using this model, we tested the hypothesis that the lack of T cell–mediated tumor recognition that characterizes MSS colorectal cancer might be due to an intrinsic resistance to T-cell attack and not only due to the low mutational burden.

Results

To test the potential impact of intrinsic characteristics of MSS tumors on immune system recognition, we developed a tightly controlled experimental platform (Fig. 1A). To ensure unbiased assessment, it was necessary to keep the tumor recognition conditions identical across all samples. To achieve this, peripheral blood mononuclear cells (PBMC) were isolated from a healthy donor, followed by sorting and transduction of the CD8⁺ T cells fraction with a T-cell receptor (TCR) of known specificity. This preparation assures consistent and reproducible recognition across all experiments. In parallel, patient-derived tumor organoids (PDO) from both MSI and MSS colorectal cancers were established and screened for expression of a specific MHC-I molecule, HLA-A2. This step ensured that an identical antigen of choice could be uniformly presented to the previously engineered T cells, allowing for a controlled comparison between MSI and MSS tumor recognition.

We successfully established and cultured human tumor organoids from surgical specimens obtained from a cohort of 32 patients with either MSI (n = 6) or MSS (n = 26) colorectal cancer tumors (Table 1). Immunohistochemistry (IHC) assessments confirmed the MMRd/MSI or MMRp/MSS profile of these organoids, based on the detection of MLH1, MSH2, MSH6, and PMS2 expression, as shown in Supplementary Fig. S1. Tumor organoids screening for MHC expression revealed variable levels of surface expression of MHC-I and MHC-II molecules (Supplementary Fig. S2A and S2B). The evaluation was conducted both with and without prestimulation using IFNγ. The presence of the primary HLA allele of interest, HLA-A2, was tested and detected in 12 different PDOs (Supplementary Fig. S2C). Furthermore, we checked programmed cell death ligand 1 (PD-L1) expression levels upon IFNγ stimulation (Supplementary Fig. S3) to determine whether the addition of ICIs in subsequent experiments would be required.

In parallel, we isolated CD8⁺ T cells from healthy donor PBMCs and retrotransduced them with the 1D3-TCR (MART1_25–35-specific) antigen (Supplementary Fig. S4A; 21). After antibiotic selection and expansion, we tested antigen-specific activation capabilities by evaluating the number of CD137-expressing T cells in the presence (or not) of the target antigen MART-1_ELA (Supplementary Fig. S4B). Additionally, we confirmed the competence of T cells to react in the presence of the 12 HLA-A2⁺ PDOs (four MSI and eight MSS), when loaded with different concentrations of the Mart-1_ELA antigen, the MHC affinity–enhanced version of Mart-1_25–35, which is specifically recognized by the transgenic TCR. T-cell activation was evaluated by measuring surface expression of CD137 and release of IFNγ (Supplementary Figs. S5 and S6, respectively).

To distinguish between the MSI and MSS tumors’ intrinsic capabilities of being recognized by the immune system, we selected organoids with comparable HLA-A2 expression levels and engineered them to present the same target antigen (MART-1_ELA; Fig. 1B). Engineering of PDOs was obtained via peptide loading with a concentration of 0.02 µmol/L of MART-1_ELA antigen, thus preventing saturation across all samples. This decision was supported by the findings presented in Supplementary Fig. S5 and S6, where saturation was reached with 0.1 µmol/L of MART-1_ELA in several samples (MSS_16, MSS_17, MSS_21, MSS_22, MSS_24, and MSS_25 in Supplementary Fig. S5; MSI_03 and MSS_22 in Supplementary Fig. S6). Notably, when T cells were placed in the presence of MSI tumors, they secreted similar amounts of IFNγ. However, when the same T cells were exposed to MSS tumors, there was a consistent 60% to 85% reduction in IFNγ secretion (Fig. 1C). Similar secretion patterns were observed for other effector molecules, such as tumor necrosis factor α (TNFα; Fig. 1D; Supplementary Fig. S7A), granzyme-A (Fig. 1E; Supplementary Fig. S7B), and granzyme-B (Fig. 1F; Supplementary Fig. S7C). T-cell reactivity assessed by CD137 expression led to analogous results. In the presence of MSI tumors, TCR-transduced T cells showed comparable levels of CD137 expression. However, in the majority of the MSS PDOs, exposure to T cells led to a significantly diminished reactivity level (Fig. 1G and H). These observations held true also when the wild-type, non–MHC affinity–enhanced, variant MART-1_EEA was used (Supplementary Fig. S8A and S8B).

Next, a 24-hour killing coculture assay (22, 23) was used to assess how effectively the different PDOs were eliminated by T cells. Both Far Red signaling and flow cytometry–based quantifications were used to assess the amount of surviving tumor cells. As shown in Fig. 1I and J, MSI PDOs showed a notable reduction of the count of live cells (∼50%) compared with controls, whereas MSS PDOs remained largely unaffected, with no killing detected (Fig. 1I and J).

We reasoned that these results pointed to the existence of mechanisms which extend beyond mere mutational (neo)antigenic burdens, limiting the response of MSS colorectal cancer to ICB.

To explore the immunomodulatory milieu during antigen recognition, we profiled the supernatants from MSI and MSS tumor organoids, each loaded with the MART-1_ELA peptide and cocultured with autologous T cells, using a 96-plex immuno-oncology panel (Fig. 2A; Supplementary Fig. S9; 24). Comparison of MSI versus MSS cocultures revealed multiple statistically significant differences in protein abundance. In particular, hallmark effector cytokines of T-cell activation, including IFNγ and TNFα, were markedly elevated in the presence of MSI organoids, and a distinct array of chemokines and inflammatory mediators involved in immune cell recruitment and intercellular communication were differentially expressed—findings that agree with our earlier functional assays. Strikingly, further inspection of the significant hits revealed that several mediators were not solely T cell–derived but were also secreted by the tumor organoids themselves (Fig. 2A).

This dual origin suggests the existence of tumor-intrinsic circuits that may contribute to the immune modulatory phenotype we observe. We therefore hypothesized that tumor-secreted factors play a direct role in shaping the secretome and may constitute an alternative mechanism of immunoresistance in our system.

To investigate whether factors secreted by colorectal cancer cells (the secretome) played a role in restricting their recognition by T cells, we utilized conditioned media (CM) from MSI and MSS PDOs, collected 24 hours after IFNγ stimulation (Fig. 2B). We set up a T-cell recognition assay which involved measuring intracellular IFNγ expression levels upon coculture with MSI tumor cells in the presence of PDO-derived CM. MSI tumor cell recognition was not affected by CM derived from MSI tumor organoids. Strikingly, when MSS-derived CM was used, T-cell recognition of MSI tumor cells was markedly impaired (>50% reduction) in all cases (Fig. 2C and D). The killing efficacy of T cells was similarly impaired in the same experimental conditions (Fig. 2E and F).

To further delve into the functional mechanisms constraining T-cell functionality, we evaluated the impact of PDO-derived CM on T cells by examining three key parameters: viability, proliferation, and target recognition capability. T-cell viability was determined using an ATP release assay. Following 24 hours of culture in the presence of different CM, no significant alterations in T cells’ viability were observed (Supplementary Fig. S10A). Next, T-cell proliferation, both unstimulated and stimulated by the MART-1ELA antigen, was measured by flow cytometry. The presence of CM from MSS tumor organoids did not significantly affect T-cell proliferation (Supplementary Fig. S10B). Finally, to assess T-cell target recognition, we utilized professional antigen-presenting cells (APC) loaded with the MART-1ELA peptide. T cells were cocultured for 5 hours with the APCs in the presence of different CM, and intracellular IFNγ expression was measured. Consistently with previous data, no significant differences were observed across the different CM used (Supplementary Fig. S10C).

These results show that the CM derived from MSI and MSS cocultures had no detectable impact on T-cell viability, proliferation, or activation, indicating the absence of a direct suppressive effect on T-cell functionality. Notably, professional APCs loaded with the same MART-1_ELA peptide were efficiently recognized by T cells even in the presence of tumor-derived CM. These findings suggest that the impaired recognition observed in tumor organoid–T cell cocultures is not due to a general inhibition of T-cell responsiveness but rather reflects a mechanism that specifically interferes with T cell–tumor cell interactions.

Given that the observed impairment was specific to tumor–T cell interactions, we next sought to functionally assess the quality of this interaction under different conditions. Specifically, we focused on two parameters: the degree of T-cell engagement and the strength of the resulting interactions, reflecting the stability and avidity of T cell–tumor contacts. These metrics allowed us to quantitatively evaluate how CM influences the formation and robustness of the tumor–T cell binding.

To quantify T-cell engagement with tumor cells, we performed high-content microscopy on cocultures of labeled T cells and tumor organoids. The proportion of T cells in direct contact with tumor cells was calculated and used as a measure of engagement efficiency across conditions (Fig. 3A). First, to assess whether tumor-intrinsic properties influence the ability of T cells to establish contact, independently of antigen presentation, we used this assay to compare T-cell engagement across different MSI and MSS tumor organoids, all presenting the MART-1_ELA antigen (Fig. 3B). Next, to isolate the impact of MSS tumor-derived secretome on the physical interaction between T cells and tumor cells, we focused on two different MSI tumors singularly and evaluated engagement in the presence of different CM (Fig. 3C). These experiments revealed that T-cell engagement varies significantly between MSI and MSS tumor organoids, despite presentation of the same cognate antigen. For MSI tumors, we observe that approximately 60% of T cells established contact with tumor cells, whereas with MSS tumors engagement dropped to as low as 20% (Fig. 3B). A similar reduction in engagement was seen when MSI tumor organoids were placed with T cells in the presence of MSS-derived CM, resulting in a decrease of approximately 50% in the proportion of engaged T cells (Fig. 3C).

To complement our engagement analysis, we next assessed the strength of T cell–tumor interactions using a single-cell force-based assay (25). This approach revealed marked differences in the stability of T cell–tumor contacts between MSI and MSS tumors, mirroring what we assessed with the engagement data (Fig. 3D). Notably, when the binding strength to MSI tumors was measured in the presence of MSS tumor-derived CM, we noticed a significant reduction (Supplementary Fig. S10D).

These findings suggest that MSS tumor-derived soluble mediators can modulate the efficiency of T cell–tumor interactions, impairing the quality of the immune synapse.

To understand the mechanisms behind our observation, we first analyzed the surface expression of four key molecules critical to our system. We exposed four different MSI tumor organoids to CM collected from both MSI and MSS tumor organoids. Following this exposure, we assessed the surface expression levels of MHC-I, HLA-A2, MHC-II, and PD-L1. No significant differences were observed in the expression of these molecules across the different CM treatments (Supplementary Fig. S10E).

We next sought an unbiased view of how tumor-derived secretomes remodel the tumor-cell surface. To this end, we performed mass spectrometry (MS)–based membrane proteome on MSI tumor organoids that had been preactivated with IFNγ and then cultured for 24 hours in the presence of CM from either MSI or MSS tumors. Unsupervised clustering of the resulting membrane–protein intensities revealed a striking separation between the two treatment groups (Fig. 3E), demonstrating that tumor-secreted factors profoundly reprogram surface-protein distribution (all detected proteins and statistically significant differences are reported in Supplementary Tables S1–S6).

Pathway analysis of differentially abundant proteins revealed that exposure to MSS CM leads to substantial molecular reprogramming (Fig. 3F; Supplementary Fig. S11A). Affected pathways were primarily associated with immune regulation, metabolic activity, and posttranslational modifications, suggesting that the MSS-associated secretome exerts a broad modulatory effect on mediators of immune cell function and intercellular communication. Among the altered pathways, we specifically investigated those that could reflect tumor-intrinsic mechanisms actively contributing to immune evasion within a short timeframe (5–24-hour-assay). Given the evolutionary conservation and the well-established role of glycosylation in shaping immune escape, remodeling the tumor microenvironment, and modulating T cell–mediated recognition and immunotherapy response (26–31), we focused on this process. We therefore analyzed the subset of genes that were both significantly altered and annotated within the enriched pathways, with particular attention to those functionally linked to glycosylation. This analysis is summarized in the alluvial plot (Fig. 3G), which illustrates the connection between enriched pathways and glycosylation-related genes, their respective gene functions, and broader functional macrocategories. Strikingly, a substantial proportion of the altered genes exhibited direct or indirect associations with glycosylation, underscoring the prominent role of this modification in mediating the proteomic effects of MSS-derived secreted factors. The intersection between significantly regulated membrane proteins and glycosylation-related pathways across individual PDOs is shown in Supplementary Fig. S11B.

To functionally validate the impact of glycosylation on immune recognition, we repeated the MSI T-cell recognition assay in the presence of CM from both MSI and MSS sources (Fig. 2B–D). In this setting, we cotreated organoids with the O-glycosylation inhibitor benzyl-α-GalNAc and the glycosyltransferase inhibitor lithocholic-O-aspartate (Supplementary Fig. S12). Remarkably, inhibition of glycan elongation fully restored T-cell recognition in all MSI samples, reaching levels comparable with untreated controls (Supplementary Fig. S12). These results demonstrate that tumor-derived soluble factors can induce glycosylation-dependent changes that directly interfere with T cell–tumor interface, potentially promoting immune escape.

To further determine whether glycosylation directly modulates tumor cell susceptibility to T-cell recognition, we performed additional coculture experiments in the absence of CM using antigen-loaded PDOs. Consistent with our previous observations, MSS organoids induced significantly lower T-cell activation compared with MSI organoids. Importantly, pharmacologic inhibition of glycosylation had minimal impact on MSI samples but restored T-cell activation in MSS organoids to levels comparable with MSI controls (Fig. 3H), confirming a direct role of glycosylation in limiting T-cell responsiveness in MSS tumors.

Cumulatively, these findings indicate that MSS colorectal cancers are capable of actively disrupting T-cell recognition and activation, revealing a novel tumor cell–intrinsic mechanism of immune escape. Our data identify the tumor secretome as a key regulator of immune recognition dynamics, specifically impairing T-cell engagement and the formation of a stable tumor–immune synapse. Surface proteomic profiling further demonstrated that tumor-derived factors extensively remodel the tumor membrane landscape, with glycosylation emerging as a central, functionally relevant modification. Together, these results highlight a previously underappreciated pathway of immune evasion mediated by secretome-driven modulation of the tumor-cell surface.

Discussion

Patients diagnosed with MSI colorectal cancer often benefit from therapies based on ICB, whereas those with MSS tumors generally do not (5, 10, 11, 13). The prevalent view is that MSI tumors are recognized by the immune system owing to their elevated mutational/neoantigen repertoire. However, emerging evidence suggests that factors beyond neoantigen count play a pivotal role in modulating the therapeutic response. In the current study, we aimed to dissect the biological mechanisms responsible for the different immune response in MSS compared with MSI tumors.

To this end, we first developed an experimental platform wherein both MSS and MSI tumors presented the same antigen (MART-1_ELA) at an identical concentration via the same presenting molecule (HLA-A2) to a homogeneous T-cell population engineered to express 1D3-TCR. This design ensures that the main variables in the tumor–host interaction are the intrinsic features of the cancer cells. MSS tumors consistently displayed a reduced recognition by the immune system. Impaired immune activation was accompanied by a common biological profile as defined by distinct effector molecules and activation markers such as IFNγ, TNFα, granzyme-A, granzyme-B, and CD137. Functional assessment of T-cell cytotoxic capacity in the presence of MSS and MSI cells revealed that although MSI tumors were efficiently killed by the engineered T cells, MSS tumors remained unaffected.

The experimental platform we implemented in this study can be broadly used for understanding how distinct cancer subtypes modulate T-cell reactivity. Moreover, these results show that MSS tumors are intrinsically recalcitrant to T cell–mediated recognition and killing. How tumor-intrinsic factors modulate tumor–host interaction, immune recognition, and the ensuing killing have remained largely unrecognized. Our findings have a strong clinical relevance: even if the TMB of MSS tumors was successfully augmented by experimental strategies, for example, with mutagenic/alkylating agents, the native resistance mechanisms of MSS tumors could still hamper an effective immune-mediated control, as shown in our current work. The same would apply if a strongly immunogenic neoantigen is successfully introduced in MSS colorectal cancer. In this regard, genetic or pharmacologic approaches designed to disable the MMR pathway rather than simply increasing the TMB could be advantageous, as we and others have shown (16–18, 32–34).

After establishing that tumor-intrinsic factors are instrumental for resistance and response to immune surveillance, we set out to characterize the biological basis for this phenotype.

We focused on the tumor secretome as a potential contributor to the observed differences. Proteomic profiling revealed distinct differences in the soluble factors released by MSI and MSS tumors during antigen-specific T-cell engagement. Notably, MSS tumors shaped a secretory milieu that seemed less supportive of immune activation, characterized by reduced lymphocyte recruitment and weakened proapoptotic signaling, suggesting a shift toward an immunosuppressive environment. To functionally test whether tumor-derived soluble factors directly affect T-cell activity, we performed secretome functional experiments. The results were striking: CM derived from MSS tumor organoids significantly reduced T-cell recognition of MSI tumor cells. Furthermore, exposure of engineered T cells to MSS CM was sufficient to impair their cytotoxic ability against MSI targets. These findings support the notion that MSS tumors engage tumor-intrinsic secretory programs capable of interfering with effective immune recognition.

Importantly, the effects mediated by MSS CM were not due to a generalized impairment of T-cell function. T cells retained their viability, proliferative capacity, and remained fully responsive when stimulated by professional APCs, even in the presence of tumor-derived soluble factors. This indicates that the observed phenotype is not the result of broad immunosuppression but rather reflects a mechanism that specifically interferes with the physical interaction between T cells and tumor cells.

By analyzing the dynamics of T cell–tumor engagement, we observed a reduced frequency and strength of T-cell contact in the presence of MSS tumors or MSS-derived CM, suggesting that the impaired recognition involves active modulation of the tumor–immune interface.

To further confirm our findings, we used an unbiased surface proteomic approach, which uncovered a widespread remodeling of the tumor membrane in response to MSS-derived factors. Among the pathways altered by MSS CM, we primarily observed changes in metabolism, immune-related molecules, antigen presentation, and cell–cell communication. Glycosylation, although not explicitly highlighted as a top enriched pathway in our analysis, emerged from the profile of proteins involved as a likely modification affecting key molecules at the immune synapse. Our data, together with previous reports emphasizing the importance of metabolic rewiring, especially glycosylation, in immune escape, support the hypothesis that this posttranslational modification reshapes the tumor-cell surface (30, 31, 35, 36). Indeed, glycosylation is an evolutionarily conserved mechanism through which tumor cells remodel their surface to evade immune detection, with aberrant glycan patterns shown to alter receptor–ligand interactions, modulate antigen presentation, and disrupt immune synapse formation across multiple cancer types (26–31). Our findings build on this established framework and uncover a previously unrecognized layer of complexity: in MSS colorectal cancer, tumor-secreted factors are sufficient to induce glycosylation-dependent surface remodeling on neighboring tumor cells, extending immune evasion beyond cell-autonomous mechanisms.

Given the known role of glycosylation in immune modulation, we tested its functional relevance in our system. Strikingly, pharmacologic inhibition of glycosylation fully rescued T-cell recognition of MSI tumors treated with MSS CM, restoring recognition to control levels. Critically, glycosylation blockade also significantly enhanced T-cell activation against native MSS PDOs, restoring it to levels comparable with those observed with MSI targets, implicating glycosylation-dependent immune evasion as an intrinsic feature of MSS tumor cells themselves. These results suggest that tumor-secreted factors can modulate the tumor-cell surface in trans through glycan remodeling, impairing the formation of a productive immune synapse and ultimately weakening immune recognition. Importantly, this effect operates primarily at the tumor-cell surface rather than through direct modulation of T-cell function. Siglec receptors, lectin-type receptors known to dampen immune responses upon recognition of sialylated glycans (29), were virtually absent on TCR-transduced T cells and were not induced by glycosylation inhibitor exposure. Consistently, pharmacologic glycosylation blockade at the concentrations used in coculture assays left T-cell viability, proliferation, and activation capacity fully intact. The glycosylation-dependent immune evasion we describe is therefore a tumor-intrinsic phenomenon, acting at the interface between cancer cells and T cells. Although pharmacologic inhibition proved effective, we also explored enzymatic deglycosylation using PNGase F as a complementary approach. Although PNGase F efficiently removed glycans from denatured protein extracts, it failed to act on intact viable PDOs. This is consistent with the known biochemical requirements of glycan-remodeling enzymes, which depend on protein denaturation to access glycosylation sites, rendering them inherently incompatible with live-cell systems in which proteins retain their native conformation. This underscores the value of pharmacologic inhibitors as the most suitable strategy to interrogate glycosylation-dependent mechanisms in intact live model systems.

The identity of the upstream secreted mediator(s) responsible for triggering glycan remodeling remains an open and important question. Targeted secretome profiling of MSS PDOs using a 96-plex immuno-oncology panel did not converge on a single dominant factor, pointing instead toward a combinatorial or noncanonical signal, potentially involving extracellular vesicles, metabolites, or lipid mediators, that falls outside the scope of antibody-based panels. In line with this, analysis of publicly available colorectal cancer RNA sequencing datasets revealed no coordinated transcriptional upregulation of glycosylation pathways in MSS versus MSI tumors, suggesting that the observed surface remodeling reflects posttranscriptional or posttranslational regulation rather than a hardwired transcriptional program. Dissecting these upstream mechanisms will be a key priority for future work.

It is therefore conceivable that, beyond lacking the high neoantigen load characteristic of MMRd tumors, MSS tumors may actively shape an immunosuppressive secretome that impairs immune recognition. Our findings highlight the tumor secretome as a critical modulator of immune responsiveness and suggest that targeting secreted factors from MSS tumors may provide new opportunities to overcome resistance in MSS cancers. In addition to informing new therapeutic strategies, this work provides a rationale for integrating secretome profiling into clinical decision-making. By identifying secretome-driven mechanisms that influence immune recognition, we shed light on new strategies for patient stratification, enabling the design of combination treatments tailored to the tumor's immunomodulatory profile. Such an approach could help extend the benefits of immunotherapy to patients with MSS tumors, who currently show limited responses to ICIs.

Importantly, although our study implicates glycosylation-dependent changes as key contributors to immune evasion, these mechanisms are unlikely to be exclusive. Other intrinsic pathways may similarly mediate resistance to T-cell pressure in MSS tumors, despite the presence of targetable antigens. This highlights a central insight of our work: MSS tumors possess an active, biologically encoded resistance to immune attack, beyond antigenicity alone. Crucially, our experimental strategy provides a framework to uncover such intrinsic immune escape mechanisms, offering a broadly applicable approach to dissecting tumor immunoresistance across cancer types and not limited to the MSS/MSI colorectal cancer context.

In conclusion, our data provide new transforming knowledge on how tumor cells interact and modulate the immune system of the host. Combinatorial treatments, designed to elevate neoantigen loads while at the same time neutralizing tumors’ intrinsic resistance, should be explored to foster immune surveillance of colorectal cancer tumors which are currently not eligible for immunotherapeutic treatment.

Methods

Supplementary Material

Data Availability

All data generated and/or analyzed during this study are included in this published article. Additional information can be requested to the corresponding authors. The proteomics data have been deposited in the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD066343.

Authors’ Disclosures

G. Mauri reports personal fees from Amgen outside the submitted work. S. Siena reports other support from Agenus, AstraZeneca, Bayer, Bristol Myers Squibb, Checkmab, Daiichi Sankyo, GSK, Guardant Health, INGM, MSD, Merck, Novartis, and T-One Therapeutics outside the submitted work. G. Germano reports grants, personal fees, nonfinancial support, and other support from NeoPhore during the conduct of the study, as well as a patent for PCT/GB2017/051062 licensed to NeoPhore. A. Bardelli reports grants, nonfinancial support, and other support from NeoPhore during the conduct of the study, as well as other support from Kither Biotech outside the submitted work. No disclosures were reported by the other authors.

Authors’ Contributions

C.M. Cattaneo: Conceptualization, data curation, formal analysis, supervision, visualization, methodology, writing–original draft, writing–review and editing. S. Scardellato: Data curation, formal analysis, visualization, methodology, writing–original draft, writing–review and editing. G. Mauri: Resources, writing–review and editing. V. Matafora: Formal analysis, methodology. V.K. Ojala: Formal analysis, methodology. C. Cannariato: Formal analysis, methodology. R. Chilà: Methodology. Z.I. Magnani: Methodology. W. Scheper: Resources, writing–review and editing. L. Lazzari: Resources, writing–review and editing. E.E. Voest: Resources, writing–review and editing. M.C. Bonini: Resources, writing–review and editing. A. Bachi: Resources. S. Siena: Resources, writing–review and editing. S. Marsoni: Resources. G. Germano: Supervision, writing–review and editing. A. Bardelli: Conceptualization, resources, supervision, funding acquisition, writing–review and editing.