Mesothelin cohort demographics and expression distribution in primary tumors, metastatic site, tumor side, and CMS subtypes

The study population comprised 7446 patients representing the low and high MSLN expressing quartiles of the 14,892 total CRC patient samples in the Caris Life Sciences database, of which 6847 patients were assigned to the “MSS cohort” according to MSI status. Both the entire patient cohort and MSS cohort were dichotomized as MSLN Low or MSLN High. Shown in Table 1A, the MSLN low group included 3723 patients with 54.9% being male and 45.1% female. The MSLN high group included 3723 patients, with 52.9% being male and 47.1% female. The median age was 63 years and 62 years respectively. In Table 1B, the MSLN low cohort included 3377 patients with a median age of 62 years, with 56.5% being male and 43.5% female. The MSLN high group was comprised of 3470 patients, 53.7% of whom were male and 46.3% female. The median age was 62 years for both MSLN expression levels in the MSS cohort.

Table 1 Demographic patient data by mesothelin (MSLN) expression level in the entire cohort of patients (A) and MSS cohort of patients (B). Patients were categorized as either MSLN low or MSLN high by RNA TPM quartiles.

MSLN expression patterns were compared across primary and metastatic sites, metastatic location, CRC side of origin, and CMS subtypes (Fig. 1). Figure 1A, B shows MSLN expression in the primary tumor versus metastatic sites in all patients. Overall, tumor samples from metastatic sites expressed MSLN at significantly higher levels compared to the primary tumor site (q ≤ 0.01) (Fig. 1A). Samples from metastases to the skin (39.4 transcripts per million (TPM)), connective/soft tissue (11.9 TPM), and the peritoneum/retroperitoneum (11.7 TPM) exhibit the highest MSLN expression levels amongst metastatic sites (Fig. 1B). Shown in Fig. 1C, MSLN expression was highest in right-sided CRC tumors in both the entire cohort (6.1 TPM) and MSS cohorts (6.2 TPM) (q ≤ 0.01). Additionally, MSLN expression was high in transverse tumor locations in both patient cohorts. Left-sided tumors exhibited the lowest median MSLN expression levels at 4.8 TPM. Similar expression patterns between the entire and MSS cohorts were observed when comparing CMS subtypes (Fig. 1D). Both cohorts exhibited the highest MSLN expression in CMS1 (8.3 and 10.4 TPM, respectively, q ≤ 0.001) and CMS4 (10.1 and 10.0 TPM respectively, q ≤ 0.001). Moreover, we analyzed MSLN IHC protein expression in CRC normal paired/matched and CRC tumor tissues. Our data suggests a higher proportion of MSLN positivity in paired/matched tumor tissue (Supplementary Fig 1B, D, F) compared to normal tissue (Supplementary Fig 1A, C, E).

Fig. 1: Mesothelin expression in CRC is enhanced in metastatic tissue, right-sided CRC, and CMS1 and CMS4 molecular subtypes.

Median MSLN TPM expression was analyzed in primary and metastatic CRC (A), various sites of CRC metastasis (B), primary tumor sidedness across patient cohorts (C), and CMS subtypes across patient cohorts (D). Analysis was completed via WTS and subpopulations were tested for statistical significance using Chi-square and Mann–Whitney U tests, with p values adjusted for multiple comparisons. *q < 0.05; **q < 0.01; ***q < 0.001, ****q < 0.0001.

Mesothelin expression is associated with increased mutations in several cancer-associated genes

Numerous oncogenic driver mutation rates and their relation to MSLN gene expression were explored. Within the entire and MSS patient cohorts, comparisons were made between MSLN high-expressing tissues and MSLN low-expressing tissues (Fig. 2). High and low expression quartiles were determined as described in the methods above. The entire cohort of patients (Fig. 2A) displayed significant differences for numerous gene mutations between the MSLN high and low expression cohorts. APC (p = 1.7e-20), BCL9 (p = 0.0006), CREBBP (p = 8.5e−5), FLCN (p = 1.39e−7), NSD2 (p = 0.002), and EP300 (p = 1.42e−6) all had significantly higher mutation rates in MSLN low expressing tumor samples compared to MSLN high expressing tumor samples (Fig. 2A). Conversely, KRAS (p = 2.67e−90), FBXW7 (p = 1.29e−14), BRAF (p = 9.32e−8), GNAS (p = 2.54e−29), and SMAD2 (p = 3.93e−5) mutated tumors were significantly more common in MSLN high expressing tumor samples versus MSLN low expressing tumor samples (Fig. 2A). These mutation patterns were quite similar in the MSS cohort. APC (p = 8.35e−26), TSC2 (p = 0.001), FLCN (p = 3.77e−7), and MAP2K4 (p = 0.002) were significantly more likely to be mutated in MSLN low tumor samples whereas KRAS (p = 2.58e−89), FBXW7 (p = 8.38e−16), RNF43 (p = 4.06e−10), GNAS (p = 4.36e−29), SMAD2 (p = 4.7e−6), BMPR1A (p = 1e−4), and STK11 (p = 8.61e−6) were significantly more likely to be mutated in MSLN high tumor samples (Fig. 2B). Of note, APC, KRAS, and TP53, regardless of patient cohort, exhibited high mutation rates of at least 30% in each MSLN expression quartile. Full mutation data are presented in Supplementary Table 1.

Fig. 2: Mesothelin expression yields high rates of oncogenic driver mutations.

Mutation analysis was performed in CRC samples expressing low and high MSLN across the entire cohort (A) and MSS cohort (B). Mutation frequencies were calculated via NGS and statistical significance was determined using Chi-square and Mann–Whitney U tests, where *p < 0.05 is considered significant.

Mesothelin high tumors exhibit greater expression of PD-L1 and higher T-cell inflamed score, immune cell infiltration, and expression of immunosuppressive genes

Figure 3 illustrates the prevalence of immune markers across MSLN high and low tissue expression split into the two cohorts as mentioned previously. Across the entire cohort (Fig. 3A), MSLN low expression had a significant association with higher tumor mutation burden (TMB) and DNA mismatch repair (dMMR)/microsatellite instability-high (MSI-H) positivity in comparison to MSLN high expression. However, MSLN high expression exhibited an association with high PD-L1 expression via IHC (IHC-PD-L1) compared to MSLN low expression. In the MSS cohort, MSLN high tumor samples yielded significantly higher IHC-PD-L1 positivity compared to MSLN low expression (Fig. 3B). Figure 3C, D shows similar results regarding T-cell inflammation quantification and IFN-γ scores. Across both cohorts, it was found that MSLN high expression significantly correlated with both markers of T-cell inflammation and low IFN-γ scores.

Fig. 3: Microsatellite stable tumors with high MSLN expression yield high PD-L1 staining and T-cell inflammation.

Immune landscape characterization was quantified using tumor mutation burden (TMB), dMMR/MSI-high status, and PD-L1 percent positivity (A) T-cell inflamed score (B) and IFN-y score (C). Analyses and statistical comparisons were performed in MSLN low and high CRC, with adjustments made for multiple comparisons. *q < 0.05; **q < 0.01; ***q < 0.001.

Specific immune cell infiltration into the TME (Fig. 4A) and immune gene expression (Fig. 4B) were examined within each cohort in addition to the immune marker results presented in Fig. 3. Several immune cell types including B cells, M1 and M2 macrophages, neutrophils, natural killer (NK) cells, and regulatory T-cells (Tregs) were more prevalent in MSLN high expressing tumors across both the entire cohort and MSS cohort. Specifically, M1 and M2 macrophages were most significant within these groups. Macrophage infiltration-associated cytokine and growth factor expression was also increased in MSLN high CRC across both patient cohorts (Supp. Fig. 2A, B). Only dendritic cells (DC) were more prevalent in MSLN low tumors. Full immune cell fraction data values are presented in supplementary table 2. Immune marker gene expression levels, shown in Fig. 4B, were all higher in MSLN high tumors across both cohorts except for IL12A. Particularly, HAVCR2 (TIM-3), CD80, CD86, and IL1B expression were of the highest magnitude change between MSLN high and low-expressing tumors. In the entire cohort of patients, comparing MSLN high and MSLN low groups, HAVCR2 (15.33 TPM vs. 8.39 TPM, q ≤ 0.05), CD80 (4.14 TPM vs. 2.73 TPM, q ≤ 0.05), CD86 (7.07 TPM vs. 4.12 TPM, q ≤ 0.05), and IL1B (9.00 TPM vs. 8.76 TPM, q ≤ 0.05) were all significantly higher in the MSLN high group. A relationship in HAVCR2 expression was seen in the MSS cohort between MSLN high and low tissue (15.00 TPM vs 8.03 TPM, q ≤ 0.05). In the MSS cohort, MSLN high tissue again yielded increased gene expression in HAVCR2 (15.00 TPM vs. 8.03 TPM, q ≤ 0.05), CD80 (4.08 TPM vs. 2.66 TPM, q ≤ 0.05), CD86 (8.76 TPM vs. 3.97 TPM, q ≤ 0.05), and IL1B (8.76 TPM vs. 6.71 TPM, q ≤ 0.05).

Fig. 4: Significant immune cell fractions in MSLN high CRC are found in B cells macrophages, neutrophils, NK cells, and T-regs regardless of patient cohort.

RNA-seq using Quantiseq and subsequent deconvolution to estimate immune cell fraction (A) further characterized the tumor immune microenvironment. Immune-related gene expression (B) was analyzed in MSLN low and high CRC in the entire and MSS cohorts via WTS, where Chi-square and Mann–Whitney U tests were implemented for statistical analyses, with p values adjusted for multiple comparisons. *q < 0.05; **q < 0.01; ***q < 0.001.

Gene set enrichment analysis of MSLN high tumors

The association of immune cell recruitment, tumor microenvironment, and corresponding gene sets related to immune response can support the possibility of a gene target being a marker for tumor susceptibility to antigen-specific immunotherapy. Complementing the immune marker and microenvironment analyses, a gene set enrichment analysis (GSEA) was performed to evaluate the differences in MSLN-associated gene expression between patients with MSLN low and MSLN high tumors within each cohort. For all gene sets shown (Fig. 5A, B), a positive normalized enrichment score (NES) was observed, indicating higher gene enrichment for patients with MSLN high tumors. Each cohort of patients exhibited significantly high NES for gene sets related to immune response related to MSLN high expression, shown using red bars. Specifically, TNFα signaling via NFKβ (NES = 1.33, false discovery rate (FDR) = 0.09), IL2 STAT5 signaling (NES = 1.32, FDR = 0.08), IFN-γ response (NES = 1.33, FDR = 0.09), and IL6 JAK STAT3 signaling (NES = 1.29, FDR = 0.12) were three such pathway enrichments related to immune response that were significantly enriched in patients with MSLN high tumors in contract with those who had MSLN low tumors. In the MSS cohort, it must first be noted that one of the most highly enriched pathways was that of the inflammatory response (NES = 1.44, FDR = 0.01). Additionally, immune-related pathways including IFN-γ Response (NES = 1.44, FDR = 0.02), IL2 STAT 5 Signaling (NES = 1.37, FDR = 0.06), IFNα Response (NES = 1.37, FDR = 0.07), IL6 JAK STAT3 Signaling (NES = 1.36, FDR = 0.07), and TNFα Signaling via NFKβ (NES = 1.39, FDR = 0.09) were all highly significant in the MSS cohort. These pathways were more significant based on FDR in the MSS cohort than in the entire cohort, possibly suggesting a higher role of immune modulation in patients with MSS tumors.

Fig. 5: Microsatellite stable CRC with high MSLN exhibits significant gene enrichment for inflammation and immune response.

Gene set enrichment analysis (GSEA) was performed in MSLN low and high CRC entire cohort (A) and MSS cohort (B). Positive values for normalized enrichment score (NES) indicate enhanced pathway signaling for the indicated gene set. Where FDR ≤ 0.25 is considered significant between MSS and the entire cohort. Significantly different gene sets are indicated in red.

Patient outcomes in association with MSLN expression

Survival outcomes relative to MSLN expression levels were analyzed using Caris CODEai™. Patient outcomes in each cohort of patients (MSS cohort and entire cohort) have been compared across various treatment modalities based on chemotherapy (5-fluorouracil (FU) based) or immune checkpoint inhibitory (ICI) therapy (nivolumab (nivo), ipilimumab (ipi), and pembrolizumab (pembro). Both the entire cohort (Fig. 6A, Supp. Fig. 3A) and the MSS cohort of patients (Fig. 6B, Supp. Fig. 3B) displayed similar patterns in that MSLN high expression had an associated lower survival than patients whose tumors were MSLN low in the 5-FU based chemotherapy-treated group. In the entire cohort, nivolumab (18.1 months vs 11.4 months, HR = 1.433, 95% CI 0.854–2.403, p = 0.172), ipilimumab (Inf vs 21.1 months, HR = 2.596, 95% CI 0.798 – 8.444, p = 0.1), and the combination Ipi-Nivo (18.1 months vs 11.4 months, HR = 1.447, 95% CI 0.863–2.426, p = 0.16) were the exceptions in that patients with MSLN high tumors trended towards a longer median survival time compared to those with MSLN low tumors (Fig. 6A, Supp. Fig. 3A). In the MSS cohort, patients with MSLN high tumors treated with immunotherapy exhibited trends toward better survival. Patients with tumors exhibiting MSLN high expression survived a median of 13.0 months vs 4.6 months in those with MSLN low tumors when treated with nivolumab (HR = 1.731, 95% CI 0.957–3.13, p = 0.068, Supp Fig. 3B). Patients treated with pembrolizumab whose tumors exhibited MSLN high expression had an improved survival trend at a median of 10.6 months while patients with MSLN low tumors survived a median of 8.0 months (HR = 1.044, 95% CI 0.651–1.675, p = 0.86, Fig. 6B). The Ipi-Nivo combination therapy also resulted in an improved median survival trend for patients with MSLN high tumors of 13.0 months versus those with MSLN low tumors who had a median survival time of 4.6 months (HR = 1.762, 95% CI 0.975–3.186, p = 0.059, Supp. Fig. 3B). Finally, ICI treatment in MSLN high tumor patients yielded a median survival time of 12.2 months, while MSLN low tumor patients survived a median of 8.0 months (HR = 1.221, 95% CI 0.855–1.743, p = 0.272, Fig. 6B).

Fig. 6: Patient survival outcomes improve for MSLN high MSS CRC when treated with pembrolizumab and ICI.

Patient insurance claims data provided to Caris Life Sciences were used to generate Kaplan–Meier curves via the CODEai™ data portal. Curves depict survival from the time of tissue collection or first treatment to last contact for the entire cohort (A) and MSS cohort (B) of CRC patients with MSLN low (blue/cohort 1) compared to MSLN high (red/cohort 2) across several therapeutic interventions.

Survival outcomes in patients treated with immune checkpoint inhibitors, analyzed based on MSLN expression quartiles

Survival outcomes were again analyzed via Caris CODEai™ using the highest and lowest 25% MSLN expression quartiles to further evaluate the correlation of MSLN expression with treatment efficacy. Each cohort, the entire cohort (Fig. 7A, Supp. Fig. 4A) and the MSS cohort (Fig. 7B, Supp. Fig. 4B), were categorized by treatment intervention, all being immunotherapeutic in nature: nivolumab, pembrolizumab, Ipi-Nivo combination, and all ICI combined. For the entire cohort (Fig. 7A, Supp. Fig. 4A), patients with tumors in the lowest 25% MSLN expression (blue) had a trend toward longer median survival time than patients with tumors exhibiting the highest 25% MSLN expression (red) in pembrolizumab (20.6 months vs 18.8 months; HR 0.94, 95% CI 0.574–1.55, p = 0.816), nivolumab (21.1 months vs 15.6 months; HR 1.18, 95% CI 0.528–2.495, p = 0.73), Ipi-Nivo combination (21.1 months vs 15.6 months; HR 1.172, 95% CI 0.54–2.547), and the combination of each ICI (20.6 months vs 18.4 months; HR 1.08, 95% CI 0.724–1.625, p = 0.694) treatment groups. In the MSS patient cohort (Fig. 7B, Supp. Fig. 4B), the opposite effect of ICI appeared to be present. Across pembrolizumab, nivolumab, Ipi-Nivo, and all ICI, the patients with the highest 25% MSLN expression in their tumors (red) survived longer than the patients with tumors exhibiting the lowest 25% MSLN expression (blue). Median survival times for pembrolizumab were 8.0 months and 11.1 months for the bottom and top 25% MSLN expression groupings, respectively (HR = 1.164, 95% CI = 0.629–2.151, p = 0.629). In the nivolumab-treated group, median survival times were 3.9 months and 12.2 months for the bottom and top 25% MSLN expression cohorts, respectively (HR = 4.87, 95% CI 1.583–14.983, p = 0.003). The combined Ipi-Nivo treatment yielded survival times of 3.9 months and 12.2 months for the bottom 25% of the MSLN-expressing cohort and the top 25% of the MSLN-expressing cohort, respectively (HR = 5.128, 95% CI 0.67–15.747, p = 0.002). In the combined ICI curve, patients with tumors exhibiting the lowest 25% MSLN expression survived a median of 7.5 months. Those whose tumors had the highest 25% MSLN expression survived a median of 12.7 months (HR = 1.423, 95% CI 0.871–2.326, p = 0.157). Both the cohorts with high MSLN expression who were treated with nivolumab had significantly longer survival times at p = 0.003.

Fig. 7: Patient survival outcomes improve in MSLN high tumors with MSS status across all immunotherapeutic treatment strategies.

Patient insurance claims data provided to Caris Life Sciences were used to generate Kaplan–Meier curves via the CODEai™ data portal. Curves depict survival from the time of tissue collection or first treatment to last contact for the entire cohort (A) and MSS cohort (B) of CRC patients by the lowest 25% MSLN expression quartile (blue/cohort 1) and highest 25% MSLN expression quartile (red/cohort 2) across several therapeutic interventions.