Differential distribution of immune checkpoints across molecular subtypes of colorectal cancer

ABSTRACT

The recent introduction of immune checkpoint inhibitor therapy has significantly improved outcomes for patients with colorectal cancer (CRC). The most pronounced clinical benefits were observed in patients with immunogenic microsatellite instable (MSI)/deficient MMR (dMMR) tumors. However, emerging evidence indicates that a subset of patients with microsatellite stable tumors may also respond to therapy. Finding predictive markers to identify these patients is critical. In this study, we analyzed the immunohistochemical expression of immune checkpoints CTLA-4, PD-1, and PD-L1 using multispectral imaging in 151 CRC patients with defined molecular characteristics. Consistent with prior reports, MSI tumors had higher levels of all immune checkpoints analyzed than microsatellite stable tumors. Notably, distinct patterns of immune checkpoint expression were associated with KRAS and BRAF mutation status. KRAS-mutated tumors showed lower, and BRAF-mutated tumors higher, expression of immune checkpoints compared to wild-type/wild-type tumors. The strongest association with KRAS and BRAF mutations was observed for PD-L1 expression. The relationship between PD-L1 and KRAS/BRAF-mutational status was validated in a second cohort of 527 CRC patients, finding the association for PD-L1 expression in both stroma and in tumor cells. Furthermore, the role of BRAF mutation on immunity in CRC was found to be partly independent of MSI status. The strongest prognostic role was found for PD-L1 in stroma, underscoring the clinical significance of this marker. In conclusion, our findings suggest that KRAS and BRAF mutations, alongside MSI, may serve as valuable biomarkers for identifying CRC patient subgroups likely to benefit from immune checkpoint blockade in CRC.

Introduction

Colorectal cancer (CRC) is one of the most common cancers worldwide and is a leading cause of cancer-related deaths.¹ CRC is a heterogeneous disease with both genetic and environmental components. The carcinogenic pathways of CRC include the chromosomal instability pathway (approximately 85%) and microsatellite instability (MSI) pathway (10–15%). These pathways are partly driven by activating mutations in tumor oncogenes such as KRAS (30–40%) and BRAF (5–15%). BRAF mutation has also been linked to the MSI pathway in CRC.²

More recent attempts to classify molecular subtypes of CRC are based on analyses of the transcriptome and describe the different Consensus molecular subtypes -– CMS1 (hypermutated, MSI, BRAF mutated, CIMP high, immune activation), CMS2 (epithelial, marked WNT and MYC signaling activation), CMS3 (epithelial and evident metabolic dysregulation, KRAS mutation), and CMS4 (prominent TGFβ activation, stromal invasion and angiogenesis).³ The CMS1/MSI subtype is of particular interest because it has been shown to be highly immunogenic and linked to a better patient survival.³ An MSI/deficient MMR (dMMR) status defines hypermutated tumors with alterations in the DNA mismatch repair system, which can be identified using PCR-based or immunohistochemistry-based techniques, respectively. MSI/dMMR status is an approved marker for the response to immune checkpoint inhibitor therapy in colorectal cancer.^4,5

The immune response is a strong positive prognostic factor in CRC and is suggested to be even stronger than tumor stage and tumor MSI status.^6,7 These findings suggest that there are important immune cell infiltrated subgroups of CRC, besides MSI, with a potential response to immunotherapy. MSS/pMMR CRCs, the majority of CRCs, have been suggested to display an immunosuppressive tumor microenvironment with increased neutrophils and M2 macrophages, along with exhausted lymphocytes.⁸ However, several studies have reported subsets of MSS CRCs with an active anti-tumor immune response, including but not exclusive to hypermutated POLE-mutated CRCs.^9–12 In addition, the expression of immune checkpoints such as PD-1/PDCD1 and PD-L1/CD274 are also found in MSS tumors.^13,14

PD-L1 may be expressed by tumor cells, as well as stromal cells, mainly tumor-infiltrating macrophages.^14,15 The interaction of PD-1 on T cells with the ligand PD-L1 negatively regulates T cell proliferation and activity.¹⁶ The blocking of the PD-1/PD-L1 pathway using immune checkpoint inhibitors can therefore reactivate cytotoxic T cells to combat the tumor. CTLA-4 is another immune checkpoint that regulates T cell activity and can be targeted for immunotherapy.¹⁷ The use of monoclonal antibodies directed against PD-1/PD-L1 and CTLA-4 has thus far only been approved by the FDA for metastatic MSI/dMMR CRCs. Given recent promising findings for neoadjuvant immunotherapy in dMMR CRC,¹⁸ identifying reliable predictive markers beyond MSI status to better identify patients with a potential for response to immunotherapy could yield major clinical benefits.

In a recent study, we evaluated the tumor infiltration of cytotoxic T cells, T helper 1 cells, T regulatory cells, B cells and macrophages in a cohort of patients with CRC, demonstrating an opposing role of KRAS and BRAF mutations on immune cell infiltration.¹⁹ While BRAF mutated tumors were found to be more infiltrated by cytotoxic T cells and T helper 1 cells, compared to BRAF wild-type tumors, the opposite relationship was found for KRAS mutated tumors. We concluded that a combined evaluation of MSI status, BRAF mutation, KRAS mutation and immune infiltration (cytotoxic T cells) could potentially improve the prediction of response to immunotherapy in CRC. In this follow-up study, we analyzed the expression of the immune checkpoints CTLA-4, PD-1, and PD-L1 in the same CRC patients, further evaluating the potential of KRAS and BRAF mutational status as predictive markers for decisions on immunotherapy.

Materials and methods

CRC patient cohorts

This study included stage I-IV CRC patients diagnosed in Umeå between the years 2010–2014, recruited within the longitudinal Uppsala-Umeå Comprehensive Cancer Consortium (U-CAN).²⁰ U-CAN collects samples of tumor tissue, blood, and feces, as well as radiological and clinical data from patients. This study included 257 patients with CRC originally selected based on the availability of fecal samples for the study of fecal microbial markers in relation to CRC characteristics.²¹ The cohort was extensively characterized according to molecular subtypes (BRAF^V600E mutation, KRAS mutations in codon 12–13, and PCR-based MSI status), clinicopathological variables, and survival, as previously described.¹⁹

Retrospectively collected formalin-fixed paraffin-embedded (FFPE), resected tumor specimens were used to build a tumor tissue microarray (TMA) with the TMA GRAND Master instrument (3DHISTEC, Budapest, Hungary) and a 1 mm punch needle. The TMA was constructed from 151 patients with available FFPE tumor tissue at that time and included two cores from the tumor front and one core from the tumor center from each patient. Immune cell infiltration (CD8, FoxP3, T-bet, CD68 and CD20) was previously assessed on this TMA using multispectral imaging with the VECTRA system.¹⁹

A set of CRC patients from the population-based Northern Sweden Health and Disease Study (NSHDS) was used for validation.²² This study included 527 stage I-IV CRC patients diagnosed between the years 2000–2016 who had not received neoadjuvant radiotherapy, and for which retrospectively collected FFPE resected tumor specimens were available for immunohistochemical evaluation of PD-L1 expression, as well as tumor molecular data.

The study was conducted in accordance with the Declaration of Helsinki, and all procedures were approved by the Ethical Committee of Umeå University/Regional Ethical Review Board, including the procedure through which patients provided written informed consent (dnr 2014/321-31 and dnr 2016/219-31).

Multiplex immunohistochemistry and multispectral imaging for in situ analysis of immune checkpoints using tissue microarrays

Analysis of immune checkpoints was performed by multispectral quantitative automated pathology imaging using the VECTRA 3® system (Akoya Biosciences, Marlborough, Massachusetts, USA). TMAs were sectioned into 4 µm sections and sequentially stained for different epitopes (CTLA-4, PD-1, FoxP3 and PD-L1). The details of the antibodies, Opal dyes, concentrations, conditions, and respective staining order can be found in Supplementary Table S1. Slides were mounted using Prolong Diamond Antifade Mountant (ThermoFischer, Waltham, MA, USA). Staining was optimized according to the Opal Multiplex IHC Assay Development guide and included monoplex stainings for comparison and inspection of potential cross-talk and interference, in alignment with recent guidelines.²³ The slides were imaged using the VECTRA 3 Quantitative Pathology Imaging System (Akoya Biosciences) as previously described.¹⁹ Spectral unmixing was performed using a spectral library and autofluorescense control with the inForm® software (Akoya Biosciences). Image analysis was blinded to patient information and supervised by two experienced pathologists (AL and RP). The inForm® software was used to train machine learning algorithms to segment tissues and cells, as well as to phenotype cells positive for different immune markers. Tissues were segmented into tumor epithelial area, stromal area, tumor debris, and no tissue. Cells were phenotyped based on the expression of one or more immune markers, identifying cells positive for CTLA-4, CTLA-4/FoxP3, PD-L1, or PD-1, and tumor cells positive for cytokeratin (CK)/PD-L1. The number of positive cells per mm² within each tissue area was recorded for each TMA core. TMA cores for which a large portion or the entire core was lost, as well as cores with poor staining quality, lack of tumour epithelial or stromal area, or heavy necrosis, were excluded. The final dataset included data from the tumour front of 142 patients (of which 22 had data from only one core), and tumour center of 145 patients. Patients were further classified into groups based on the high or low numbers of cells positive for the indicated markers, using the median number of positive cells among all patients as the cutoff, excluding those who had received neoadjuvant radiotherapy.

Immunohistochemical evaluation of PD-L1 on whole tumor tissue sections

PD-L1 was stained on 4 µm whole-slide tumor-tissue sections from patients of the NSHDS cohort on the Ventana Benchmark Ultra (Ventana Medical Systems, Tucson, AZ), with CC2 pre-treatment and the OptiView DAB detection kit (Ventana Medical Systems) for visualization. The antibody anti-PDL1 clone SP142 was used at a dilution of 1:50 (Abcam, Cambridge, UK). The slides were counterstained with hematoxylin. PD-L1 expression was evaluated in the tumor front by a pathologist blinded to patient characteristics (BG). Partial membrane staining was required to be considered as a positive cell.²⁴ The numbers of PD-L1 positive cells in the stroma and PD-L1 positive tumor cells were counted in five x200 high-power fields, resulting in a total area of 4.75 mm². A representative immunohistochemical staining can be found in Supplementary Figure S1. Twenty patients were excluded due to poor tissue/staining quality, resulting in a total of 507 patients with immunohistochemical evaluation of PD-L1.

Statistics

Statistical analyses were conducted using IBM SPSS Statistics version 29. Correlations between continuous variables were analyzed using the Spearman´s rank correlation test. The Mann-Whitney test or Kruskal-Wallis test was used to compare continuous variables between two or more categories, respectively. Cancer-specific survival was estimated for surgically resected CRC using Kaplan-Meier survival curves and differences in 5 year cancer-specific survival (to fulfill the assumption of proportional hazards) between groups were assessed using the log-rank test. Cancer-specific mortality was defined as death due to cancer in patients with disseminated or recurrent disease. Patients were followed up from the time of surgery until the time of death or end of follow up (October 2021 for U-CAN, May 2022 for NSHDS). Univariable and multivariable survival analyses were conducted for patients with stage I-III CRC disease from the NSHDS cohort using Cox proportional hazards models. This was done to exclude the heterogeneously treated group of stage IV patients with a known poor prognosis. A P value < 0.05 was considered statistically significant.

Results

Distribution of immune checkpoints according to intratumoural location

The expression of immune checkpoints (CTLA-4, PD-1, and PD-L1) was evaluated on TMAs from a cohort of CRC patients using multiplex immunohistochemistry and multispectral imaging (Figure 1). Data were collected from both the stromal area and the tumor epithelial area and presented as the number of positive cells per mm² tissue area. Analyses included tumour front of 142 patients and tumour center of 145 patients, and the distribution of immune checkpoints was found to be comparable between the two tumour areas (Table 1). CTLA-4 positive cells were mainly found in the stromal area and rarely observed in the tumour epithelial area. Regulatory T cells have been reported to constitutively express CTLA-4,²⁵ and the main fraction of CTLA-4 positive cells was also positive for FoxP3 and defined as regulatory T cells (CTLA-4/FoxP3). PD-L1 positive cells were mainly found in the stromal area, often in close proximity to the tumour epithelium (Figure 1A,B). Some tumour cells also expressed PD-L1 (CK/PD-L1) (Figure 1A,B). PD-1 positive cells were found mainly in the stromal area but were occasionally present in the tumour epithelial area. For further statistical analyses, we focused on CTLA-4 and PD-L1 positive cells in the stromal area, and CK/PD-L1 positive cells in the tumour epithelial area, while PD-1 positive cells were analyzed in both the stromal and tumor epithelial areas.

Figure 1.

Multispectral imaging of immune checkpoints in the tumor front and center of CRC. Shown are multiplex stainings displaying CTLA-4 (red), PD-L1 (yellow), PD-1 (cyan), FoxP3 (orange), and cytokeratin (magenta), in tumor tissue microarrays (1 mm) taken in (a) the tumor front or in (b) the tumor center of CRC tumors at x20 magnification. Left panel displays composite images after applying the spectral library. Right panel displays tissue segmentation maps with tumor area (magenta), stromal area (blue), tumor debris (yellow), and no tissue (brown), and cell phenotype maps displaying CTLA-4 (red), CTLA-4/FoxP3 (orange), PD-L1 (yellow), PD-1 (cyan) and CK/PD-L1 (green).

Table 1.

Distribution of immune checkpoints in the tumor front and center in CRC.

Shown are mean numbers of cells/mm² stromal area or tumor epithelial area positive for the indicated marker in the tumor front and center.

We assessed the correlations between immune checkpoints in the tumor front, as well as between the immune checkpoints and markers for specific immune cells previously analyzed in this cohort: cytotoxic T cells (CD8), T helper 1 cells (T-bet), regulatory T cells (FoxP3), B cells (CD20) and macrophages (CD68).¹⁹ Positive correlations were observed between most of the immune markers (Supplementary Table S2), including a strong positive correlation between PD-L1 in the stromal area and CK/PD-L1 in the tumour epithelial area (p < 0.001). The strongest correlation was, not surprisingly, found between CTLA-4/FoxP3 positive cells and FoxP3 positive cells identifying the same cell type but on separate multispectral analysis panels run on subsequent TMA sections (r_s = 0.791, p < 0.001). Data on FoxP3 in relation to molecular subtypes and survival have been previously published in this cohort.¹⁹ Therefore CTLA-4 single-positive cells, but not CTLA-4/FoxP3 positive cells, were further studied here.

Immune checkpoints in relation to clinicopathological characteristics

The strongest associations between immune checkpoints and clinicopathological traits were found in the tumor front (Table 2). PD-L1 and PD-1 expression in the stromal area was positively associated with higher age, right-sided tumor location, and lower tumor stage. CTLA-4 positive cells in the stromal area and CK/PD-L1 positive cells in the tumor epithelial area were also positively associated with right-sided tumor location (Table 2). Pre-operative radiotherapy was administered to 59.2% of rectal cancer patients and was linked to lower tumor expression of immune checkpoints (Table 2). For further analyses, patients with pre-operatively irradiated rectal cancer were excluded, as previously described.¹⁹ The associations between immune checkpoints in the tumor center and clinicopathological characteristics are shown in Supplementary Table S3.

Table 2.

Associations between immune checkpoints in the tumor front and clinico-pathological characteristics of CRC patients.

Shown are mean numbers of cells/mm² stromal area or tumor epithelial area positive for the indicated marker in the tumor front. *Indicates significant P values (P < 0.05) according to Mann-Whitney U tests or Kruskal-Wallis H tests.

Immunohistochemical evaluation of PD-L1 in the tumor front of whole tumor tissue sections from patients in the NSHDS cohort revealed a similar distribution of PD-L1 positive cells in relation to clinicopathological characteristics (Supplementary Table S4).

Immune checkpoints in relation to tumor molecular characteristics

The associations between immune checkpoints and tumor molecular traits in patients from the U-CAN cohort were next studied (Table 3 and Supplementary Table S5). PD-L1 expression was lower in the stromal area of the tumor front in KRAS mutated tumors than in wild-type tumors, whereas the opposite was found in BRAF-mutated tumors, showing higher expression of PD-L1 compared to wild-type tumors (Table 3). A similar pattern was observed also for CTLA-4 expression (Table 3). No associations were found between molecular traits and PD-1 (Table 3 and Supplementary Table S5). CK/PD-L1 positive cells in the tumor epithelial area followed the same pattern and were significantly increased in BRAF mutated tumors when analyzed in the tumor center (Supplementary Table S5). Furthermore, CTLA-4 and PD-L1 in the stromal area, along with CK/PD-L1 in the tumor epithelial area were positively associated with MSI tumors (Table 3 and Supplementary Table S5). The highest expression of PD-L1 and CTLA-4 was observed in patients with MSI tumors, regardless of BRAF mutational status, whereas for CK/PD-L1, BRAF-mutated MSI tumors showed higher expression than all other combinations (Table 3 and Supplementary Table S5).

Table 3.

Associations between immune checkpoints in the tumor front and molecular characteristics of CRC tumors.

For multiplex IHC analyses on patients from the U-CAN cohort, data is presented as mean numbers of cells/mm² stromal area or tumor epithelial area positive for the indicated marker in the tumor front. For conventional IHC analyses on patients from the NSHDS cohort, data is presented as mean numbers of positive cells in stroma or positive tumor cells in the tumor frount. Pre-operatively irradiated rectal cancers were excluded. *Indicates significant P values (P < 0.05) according to Mann- Whitney U tests or Kruskal-Wallis H tests. Abbreviation: MSI, microsatellite instability.

The opposing associations between KRAS and BRAF mutations and PD-L1 were validated by analyses in the NSHDS cohort of 527 patients (Table 3). Lower numbers of PD-L1 positive cells in stroma and PD-L1 positive tumor cells, were found in KRAS-mutated versus KRAS wild-type tumors, whereas higher numbers of PD-L1 positive cells in stroma and PD-L1 positive tumor cells, were found in BRAF-mutated versus BRAF wild-type tumors, as well as in MSI versus MSS tumors (Table 3). Furthermore, BRAF mutation was associated with higher numbers of PD-L1 positive cells, compared to BRAF wild-type, in patients with MSI tumors and patients with MSS tumors (Table 3).

Immune checkpoints and survival

We assessed immune checkpoint expression in the tumor front in relation to cancer-specific survival of the CRC patients in the U-CAN study cohort (Figure 2). Patients with high numbers of PD-L1 positive cells in the stromal area had better survival rates, compared to those with low numbers (Figure 2B). PD-1 positive cells in the tumor epithelial area were also linked to a better survival (Figure 2E). For the remaining immune markers in the stromal and tumor epithelial areas, no significant associations with survival were observed (Figure 2). The prognostic role of PD-L1 was further assessed in the NSHDS validation cohort, which revealed a supporting significant increase in survival in patients with a higher expression of PD-L1 in the stroma (Figure 2F). The prognostic role of PD-L1 was observed in patients with wild-type/wild-type, as well as KRAS and BRAF mutated CRC (Figure 3). The independent prognostic role of PD-L1 was evaluated in patients with stage I-III CRC from the NSHDS cohort using Cox regression analyses. The prognostic role of PD-L1 in the stroma was significant in a multivariable model including sex, age, tumor localization, tumor stage, BRAF mutation, KRAS mutation and MSI status (Supplementary Table S6). The prognostic role of PD-L1 positive tumor cells was less clear (Figure 2D,G, and Supplementary Table S6).

Figure 2.

The prognostic importance of immune checkpoints in the tumor front of patients with CRC. Shown are Kaplan-Meier plots of cancer-specific survival in non-irradiated stage I-IV CRC patients in (A-E) the U-CAN study cohort (n = 94), and (F-G) the NHSDS validation cohort (n = 507). Survival was assessed in U-CAN patients with low or high numbers of cells in the stromal area positive for (A) CTLA-4, (B) PD-L1, or (C) PD-1, or low or high numbers of cells in the tumor epithelial area positive for (D) CK/PD-L1, or (E) PD-1. Survival was assessed in NSHDS patients as negative or positive for PD-L1 in (F) cells in stroma, or (G) tumor cells. Log-rank tests were used to calculate differences in 5-year cancer-specific survival between groups. p < 0.05 was considered statistically significant.

Figure 3.

The prognostic importance of PD-L1 positive cells in the stromal area of the tumor front in patients with molecular subtypes of CRC. Shown are Kaplan-Meier plots of cancer-specific survival in non-irradiated stage I-V CRC patients in relation to numbers of PD-L1 positive cells in the stromal area of the tumor front in wild-type/wild-type, KRAS mutant or BRAF mutant tumors, as indicated, in (a) the U-CAN study cohort and in (b) the NSHDS validation cohort. Log-rank tests were used to calculate differences in 5-year cancer-specific survival between groups. p < 0.05 was considered statistically significant.

Discussion

Recent advancements in cancer treatment using immune checkpoint inhibitors have had an undeniable impact. Although applications for CRC have been limited, the most notable results have been observed in patients with immunogenic MSI/dMMR tumors.^4,5 However, not all patients with MSI/dMMR tumors respond to treatment and emerging evidence suggests that a subset of patients with MSS tumors may respond to immune checkpoint inhibitors.²⁶ There is an urgent need for predictive markers to accurately define these patients. In a recent study, we demonstrated the opposing role of KRAS and BRAF mutation on immune cell infiltration in CRC, suggesting possible implications for immunotherapy.¹⁹ In this follow-up study, we further investigated the expression of the immune checkpoints CTLA-4, PD-1 and PD-L1 in the same CRC cohort, finding a similar opposing effect by KRAS and BRAF mutation on immune checkpoint positive cells, strengthening our previous conclusion of a possible predictive power of these molecular traits.

Immune checkpoints were primarily expressed in right-sided colon cancers, where highly immunogenic MSI tumors mainly reside.²⁷ A pre-dominant right-sided localization has previously been reported for PD-1/PD-L1.¹³ Immune checkpoints were further linked to tumor stage, with the weakest infiltration of immune checkpoint positive cells found in stage IV CRCs. This finding suggests a possible value of neo-adjuvant immune checkpoint inhibitor therapy in patients with stage I-III CRC. In fact, neo-adjuvant immune checkpoint inhibitor therapy has shown a pathological response in early-stage dMMR (20/20) and pMMR (4/15) CRC patients.¹⁸ Neo-adjuvant immune therapy has been the focus of several ongoing clinical trials, and has shown promising potential.²⁸

In this study, the expression of immune checkpoints was found to be dependent on mutations in KRAS and BRAF, with the most conclusive results found for PD-L1 expression, in both stroma and in tumor cells. In line with our previous study including cytotoxic T cells and Th1 cells, the expression of PD-L1 was found decreased in KRAS mutated CRCs, while it was increased in BRAF mutated CRCs, compared to wild-type/wild-type tumors. The opposing role of KRAS and BRAF mutations on PD-L1 expression was validated in a second larger CRC cohort using whole tumor tissue sections, where PD-L1 positive cells were manually counted, reinforcing our method and our conclusions. Furthermore, CTLA-4, PD-L1, and CK/PD-L1 were found to be more prevalent in MSI CRCs, as previously reported, with a suggested importance in the maintenance of immune balance.²⁹ Indeed, we found a strong positive correlation between most immune markers (both activating and inhibiting), which supports their parallel enrichment. The parallel presence of activating and inhibiting mechanisms is also demonstrated by the fact that PD-L1 positive macrophages were shown to mainly display M1 polarization and enhanced colocalisation with tumor cells.¹⁵ Stratifying immune checkpoint infiltration in MSI tumors by BRAF mutation status further suggested possible independent roles of MSI status and BRAF mutation, which will require further investigation. We investigated both the stromal and tumor epithelial areas and observed a generally higher density of immune checkpoints in the stromal area, a finding consistent with previous studies using similar approaches, which reported a higher overall number of tumor-infiltrating immune cells in the stromal area compared to the tumor epithelium.^19,30 The findings in the stroma and tumor epithelium were largely consistent, however with stronger associations in stroma. In the larger NSHDS cohort, associations between PD-L1 and KRAS and BRAF mutation were observed for PD-L1 expression in both stroma and in tumor cells.

Reduced immune cell infiltration and suppression of KRAS mutated CRC tumors have been previously recognized.^31–35 Preclinical models have exhibited a poor response to immunotherapy in patients with KRAS mutated CRC.³⁴ Accordingly, a decreased benefit of pembrolizumab (PD-1 inhibitor) was found in patients with KRAS-mutated metastatic MSI/dMMR CRC.³⁶ An association between BRAF mutation and increased immunity has also been found in several studies, including increased expression of PD-1, PD-L1 and CTLA-4.^31,37–40 Studies by Bolzacchini et al. and Barras et al. further suggested that the immunity evoked by BRAF mutated tumors was partly independent of tumor MSI/MMR status.^37,38 An improved response to pembrolizumab in BRAF mutated treatment resistant MSS CRCs has been reported.⁴¹ Additionally, first-line chemotherapy and nivolumab (PD-1 inhibitor) treatment for MSS CRCs in the METIMMOX trial, showed response in a small subset of patients, with a complete response observed in BRAF-mutated CRCs (n = 3) devoid of POLE/POLD mutations.⁴² Further studies are needed to establish the therapeutic importance of KRAS and BRAF mutations in immunotherapy for both MSI and MSS CRCs.

A positive prognostic role was found for the higher expression of PD-L1 in the stromal area and PD-1 within the tumor epithelium. The prognostic role of PD-L1 was further supported by analyses in the validation cohort. Furthermore, the prognostic importance of PD-L1 remained in patients with KRAS or BRAF mutated CRCs, underscoring the clinical significance of this marker in molecular subtypes of CRC. The prognostic role of PD-L1 in tumor cells was less clear in this study. The prognostic role of PD-L1 and PD-1 in CRC has been vigorously studied, however conclusions have been hampered by the different protocols used for staining and evaluation, and the different compartments analyzed. A recent meta-analysis including prognostic studies with immunohistochemical evaluation of PD-1 and PD-L1 in CRC, concluded that the predominant share of studies reported a beneficial effect on prognosis by PD-1 and PD-L1 expression in stroma, while the prognostic role of PD-L1 in tumor cells remains controversial.⁴³

PD-L1 expression in tumor cells has been proposed as a marker for predicting the response to immunotherapy by PD-1/PD-L1 checkpoint inhibitors, and this approach is used across various cancers with different scoring techniques.⁴⁴ For example, PD-L1 positive tumor cells are used as a predictive marker for pembrolizumab in non-small cell lung cancer.⁴⁵ In contrast, the expression of PD-L1 on immune cells predicts the response to atezolizumab (PD-L1 inhibitor) in triple-negative breast cancer.⁴⁶ Pembrolizumab is the standard first-line treatment for patients with metastatic MSI/dMMR CRCs, but recent findings suggest that the immunogenicity of CRC may extend beyond tumor hypermutability (MSI and POLE mutations).^12,47–49 Consensus has yet to be reached in terms of the potential of PD-L1 as a predictive biomarker for immune checkpoint inhibitor therapy in CRC, and a standardized scoring system is still lacking.^44,50 Nevertheless, a subset of patients with metastatic MSS tumors (12.5%) was found to have increased infiltration of cytotoxic T cells and engagement with tumor cells, and all these patients had PD-L1 positive tumors.⁵¹ Numerous clinical trials are currently underway to assess the efficacy of immune checkpoint inhibitors and combination therapies in both MSI and MSS CRCs, in early-stage operable cases and non-operable metastatic cases.^52,53

The strengths of this study include the simultaneous and precise evaluation of different immune checkpoints in distinct compartments of CRC specimens according to molecular subtypes. However, we cannot fully discriminate between different types of immune cells expressing PD-1 and CTLA-4. Although immune checkpoint regulation has gained most attention in T cells, recent findings suggest regulatory roles of immune checkpoints also in NK cells and B cells.^54,55 Furthermore, even though multiplex IHC has emerged as a very potent histopathological tool, there are some pitfalls and challenges to address before it can be translated into clinical practice.⁵⁶

This study was limited to patients who had undergone primary surgery. Further weaknesses included a limited patient sample size in the study cohort and the use of TMAs, which may not represent the whole tumor. The infiltration of immune checkpoints was though found similar in the tumor front and center, validating the use of TMAs for evaluation. In this study, the main findings were validated by conventional IHC in a larger patient cohort by evaluations on whole tumor tissue sections using a manual scoring system. Another limitation was the restricted evaluation of BRAFV600E and KRAS mutations (codon 12, 13). Further studies including additional BRAF/KRAS and NRAS mutations are needed. The patients in this study were diagnosed before the initiation of immunotherapy; therefore, future studies should consider these findings in the context of immunotherapy. Additionally, the importance of the immune contexture in the metastatic setting of CRC is gaining increased attention,^57,58 highlighting the need for further studies of the distribution of immune checkpoints across different tumor compartments within the metastatic lesions.

In conclusion, our findings suggest differences in the immunogenicity of KRAS- and BRAF-mutated CRC tumors which may inform clinical decision-making for immunotherapy. The potential predictive value of KRAS and BRAF mutations, along with tumor MSI status, could be advantageous, given that these molecular markers are already used in the clinical management of CRC patients.

Funding Statement

This study was supported by the Swedish Cancer Society, the Swedish Research Council, Lion´s Cancer Research Foundation, the Cancer Research Foundation in Northern Sweden, Cutting edge grant from the County Council of Västerbotten, the Sjöberg Foundation, and a regional agreement between Umeå University and Region Västerbotten (ALF).

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

Data are available upon meeting relevant legal and ethical requirements. Further information can be obtained from the corresponding author (RP).

Supplementary Information

Supplemental data for this article can be accessed online at https://doi.org/10.1080/2162402X.2025.2546406