ABSTRACT
Objective
To validate c‐MET protein expression in colorectal cancer (CRC) and assess whether joint MET and hepatocyte growth factor (HGF) transcript expression provides descriptive information on tumor microenvironment features beyond MET alone.
Methods
c‐MET protein expression was evaluated by immunohistochemistry in 51 paired CRC and adjacent tissues. The Cancer Genome Atlas (TCGA) and TNMplot datasets assessed MET mRNA expression and clinicopathological associations; Kaplan–Meier Plotter evaluated unadjusted survival; and TIMER (Tumor Immune Estimation Resource) examined purity‐adjusted immune correlations. In TCGA colon adenocarcinoma (COAD) and rectal adenocarcinoma (READ) cohorts, samples were divided using cohort‐specific median MET and HGF expression. ESTIMATE (Estimation of STromal and Immune cells in MAlignant Tumor tissues using Expression data) and CIBERSORT (Cell‐type Identification by Estimating Relative Subsets of RNA Transcripts) compared inferred tumor purity, immune/stromal scores, and immune‐cell composition. These computational estimates were not experimentally validated in this study.
Results
MET mRNA was elevated in CRC datasets, and c‐MET protein H‐scores were higher in CRC than in paired adjacent tissues. MET expression was not associated with the tested clinicopathological variables. High MET expression was associated with shorter overall and relapse‐free survival in unadjusted analyzes. In both cohorts, the MET‐high/HGF‐low group generally showed higher estimated tumor purity and lower immune and stromal scores, whereas the MET‐low/HGF‐high group showed the converse pattern. Individual immune‐cell subset differences were cohort dependent.
Conclusion
MET mRNA and c‐MET protein are elevated in CRC. Joint MET/HGF expression may complement MET‐only analysis by retaining discordant patterns associated with inferred tumor‐cell and immune/stromal composition. These retrospective, algorithm‐derived findings do not establish independent prognostic value, HGF/c‐MET pathway activation or causality, or a basis for clinical patient stratification.
Keywords: c‐MET, colorectal cancer, hepatocyte growth factor, immunohistochemistry, MET, tumor immune microenvironment
1. Introduction
Colorectal cancer (CRC) is one of the most frequently diagnosed malignancies worldwide and remains a major cause of cancer‐related death [1, 2]. Despite advances in screening, surgery, and systemic treatment, recurrence and survival vary substantially among patients, including those with similar anatomical stage [3, 4]. Conventional TNM staging therefore does not fully capture the molecular and microenvironmental heterogeneity of CRC, and additional biomarkers are needed to improve risk stratification and guide future individualized studies.
The MET proto‐oncogene encodes c‐MET, a receptor tyrosine kinase whose ligand is hepatocyte growth factor (HGF). HGF binding induces receptor dimerization and tyrosine phosphorylation, leading to activation of RAS/MAPK, PI3K/AKT, STAT3, and other signaling networks involved in proliferation, survival, migration, epithelial–mesenchymal transition, and therapeutic resistance [5, 6, 7]. Importantly, MET overexpression, amplification, abnormal splicing, and receptor cross‐talk can also enhance signaling independently of ligand abundance. Thus, MET/HGF transcript levels cannot be interpreted as direct measurements of pathway activation.
Previous studies have reported increased c‐MET protein expression and adverse prognostic associations in CRC, although the findings are not fully consistent across cohorts, antibodies, scoring systems, and disease stages [8, 9, 10]. The tumor immune microenvironment comprises immune cells, stromal cells, vascular components, and extracellular matrix and has a central role in tumor progression and treatment response [11]. Wan et al. recently linked MET expression to an acidic microenvironment, metastasis, and multiple immune‐cell populations in CRC [12]. Consequently, elevated MET expression or single‐gene immune correlations alone cannot be considered novel. Whether joint MET/HGF expression patterns provide additional information on tumor purity and immune/stromal composition remains an exploratory question.
Accordingly, the present study had two objectives: first, to validate c‐MET protein expression in 51 paired CRC and adjacent tissues; and second, to evaluate four joint MET/HGF transcript‐expression groups separately in TCGA colon adenocarcinoma (COAD) and rectal adenocarcinoma (READ) cohorts using ESTIMATE and CIBERSORT. The study was designed to identify associations that may inform future mechanistic work, not to establish independent prognostic value or a causal immune‐regulatory mechanism.
The four‐group classification examines the two‐dimensional MET/HGF transcript‐expression space rather than treating MET expression as a single continuum. Its potential incremental value is the explicit comparison of discordant MET‐high/HGF‐low and MET‐low/HGF‐high states, in addition to the double‐high and double‐low states. This may help distinguish tumor‐cell‐dominant from non‐tumor‐rich bulk‐transcriptome profiles, but it does not identify the cellular source of either transcript, quantify c‐MET phosphorylation, or define a biologically or clinically actionable subtype.
2. Materials and Methods
2.1. Clinical Specimens
Colorectal cancer specimens and paired adjacent morphologically normal tissues were obtained from 51 patients who underwent surgical resection at The First Hospital of Tsinghua University between January 2019 and December 2022. All cases were histopathologically confirmed primary colorectal carcinomas. Specimens were fixed in 4% neutral buffered formaldehyde and paraffin embedded. Histologic classification followed the fifth edition of the World Health Organization classification of tumors of the digestive system, and clinical staging followed the eighth edition of the American Joint Committee on Cancer TNM system. None of the patients received preoperative radiotherapy, chemotherapy, immunotherapy, or other antitumor treatment.
2.2. Reagents
A rabbit polyclonal anti‐c‐MET antibody (catalog no. 18321; 100 μg) was obtained from Shanghai Ruibosi Biotechnology Co. Ltd. Sodium citrate antigen‐retrieval solution (ZLI‐9069), phosphate‐buffered saline, an enzyme‐labeled goat anti‐mouse/rabbit secondary antibody (PV‐6000D), and DAB chromogen were obtained from Beijing Zhongshan Jinqiao Biotechnology Co. Ltd.
2.3. Immunohistochemistry and H‐Score Assessment
Paraffin‐embedded tissue sections (4 μm) were baked at 60°C for 30 min, deparaffinized, rehydrated, and subjected to heat‐induced antigen retrieval in citrate buffer (pH 6.0) for 20 min. Endogenous peroxidase activity was blocked with 3% hydrogen peroxide for 10 min, followed by serum blocking for 30 min.
Sections were incubated with the c‐MET primary antibody at 4°C overnight and with the secondary antibody at room temperature for 1 h. Staining was visualized with DAB, and nuclei were counterstained with hematoxylin. c‐MET staining was assessed in the cell membrane and/or cytoplasm.
The H‐score was calculated as the sum of the percentage of cells at each staining intensity multiplied by the corresponding intensity score (0, 1, 2, or 3), yielding a total score from 0 to 300. An H‐score ≥ 100 was used to define positive c‐MET protein expression for the descriptive paired positivity analysis. This pragmatic dichotomization was not intended as a clinically validated diagnostic, prognostic, or treatment‐selection cutoff; continuous H‐scores were retained for the paired Wilcoxon analysis. Paired tumor and adjacent tissues were evaluated using the same criteria.
2.4. TCGA Analysis
MET mRNA expression profiles and corresponding clinical data for TCGA‐COAD and TCGA‐READ were obtained through the UCSC Xena platform [13]. MET mRNA expression was compared between tumor and adjacent morphologically normal tissues. Within each cohort, patients were divided into high‐ and low‐expression groups at the cohort‐specific median tumor MET mRNA level. The median was selected as a reproducible, data‐driven split for exploratory comparison and to avoid an externally unsupported clinical cutoff; it should not be interpreted as a therapeutic threshold. Clinicopathological variables were compared between the groups.
2.5. TNMplot Analysis
MET mRNA expression in CRC and normal tissues was compared using microarray and RNA‐sequencing data available through TNMplot [14].
2.6. TIMER and Joint MET/HGF Expression Analyzes
TIMER was used to evaluate purity‐adjusted Spearman correlations between MET expression and inferred immune‐cell infiltration separately in COAD and READ [15]. The analyzed immune‐cell populations included B cells, CD8+ T cells, CD4+ T cells, macrophages, neutrophils, and dendritic cells.
Within each cohort, samples were classified as MET‐high/HGF‐high, MET‐high/HGF‐low, MET‐low/HGF‐high, or MET‐low/HGF‐low according to the cohort‐specific median tumor expression values of MET and HGF. This median‐based partition was selected to make the two transcript dimensions explicit and to provide sufficiently sized exploratory groups within each cohort; it was not selected as a biological activation threshold or clinical decision cutoff. ESTIMATE was used to calculate immune, stromal, and total ESTIMATE scores and inferred tumor purity [16]. CIBERSORT was used to estimate the relative fractions of 22 immune‐cell populations [17]. These joint transcript‐expression groups were treated as exploratory strata and not as direct measures of HGF/c‐MET pathway activation.
2.7. Kaplan–Meier Plotter Analysis
Kaplan–Meier Plotter for colon cancer was used to evaluate the associations between MET expression and overall survival (OS) and relapse‐free survival (RFS) [18]. Because the online analysis did not provide complete adjustment for individual clinical covariates, these results were interpreted as unadjusted prognostic associations.
2.8. Statistical Analysis
Paired c‐MET positivity rates in tumor and adjacent tissues were compared using McNemar's test, and paired H‐scores were compared using the Wilcoxon signed‐rank test. Independent continuous variables were analyzed using Student's t‐test or the Mann–Whitney U test, as appropriate; categorical variables were analyzed using the chi‐square test or Fisher's exact test. Four‐group comparisons were performed using the Kruskal–Wallis test. Correlations were assessed using Spearman's method, and survival curves were compared using the log‐rank test. Effect magnitudes were reported as available from each analysis: proportions and the absolute percentage‐point difference for the paired immunohistochemical positivity comparison, medians and interquartile ranges for H‐scores, Spearman correlation coefficients for TIMER analyzes, and hazard ratios with 95% confidence intervals for Kaplan–Meier Plotter analyzes. All tests were two sided, and p < 0.05 was considered statistically significant. Analyzes were conducted in R version 4.2.0. No multivariable Cox model was performed; therefore, this study did not assess whether MET was an independent prognostic factor.
3. Results
3.1. MET mRNA Expression in CRC Tissues From Public Databases
TCGA analysis showed that MET mRNA expression was higher in COAD tumors (n = 478) than in adjacent tissues (n = 93) and higher in READ tumors (n = 173) than in adjacent tissues (n = 19) (both p < 0.001; Figures 1 and 2). TNMplot microarray data likewise showed higher MET expression in CRC tissues (n = 1450) than in normal tissues (n = 377) (p < 0.001; Figure 3). Consistent results were obtained from TNMplot RNA‐sequencing data, in which CRC tissues (n = 469) had higher MET expression than normal tissues (n = 315) (p < 0.001; Figure 4).
FIGURE 1.
MET mRNA expression in TCGA colon adenocarcinoma (COAD) tumors and adjacent morphologically normal tissues. Each point represents one sample; boxplots summarize the distribution of the displayed MET expression values (central line, median; box, interquartile range; whiskers, spread as displayed). Tumor and normal groups are indicated in red and blue, respectively. The p‐value shown in the panel is for the between‐group comparison.
FIGURE 2.
MET mRNA expression in TCGA rectal adenocarcinoma (READ) tumors and adjacent morphologically normal tissues. Each point represents one sample; boxplots summarize the distribution of the displayed MET expression values (central line, median; box, interquartile range; whiskers, spread as displayed). Tumor and normal groups are indicated in red and blue, respectively. The p‐value shown in the panel is for the between‐group comparison.
FIGURE 3.
MET mRNA expression in colorectal cancer (CRC) and normal tissues in the TNMplot microarray dataset. Points represent samples, and the overlaid boxplots summarize the distribution of displayed expression values. The normal and tumor groups are shown in green and red, respectively; the displayed p‐value is for the between‐group comparison.
FIGURE 4.
MET mRNA expression in colorectal cancer (CRC) and normal tissues in the TNMplot RNA‐sequencing dataset. Points represent samples, and the overlaid boxplots summarize the distribution of displayed expression values. The normal and tumor groups are shown in green and red, respectively; the displayed p‐value is for the between‐group comparison.
The direction of the differential MET mRNA expression was therefore consistent across databases and analytical platforms.
3.2. Immunohistochemical Validation of c‐MET Protein Expression
c‐MET protein was predominantly localized to the tumor‐cell membrane and cytoplasm (Figure 5). Among the 51 paired specimens, c‐MET positivity (H‐score ≥ 100) was observed in 82.4% (42/51) of tumor tissues and 17.6% (9/51) of adjacent tissues (McNemar's test, p < 0.001), corresponding to an absolute between‐tissue percentage‐point difference of 64.8%. The median tumor H‐score was 180 (interquartile range, 120–240), compared with 40 (20–60) in adjacent tissues (Wilcoxon signed‐rank test, p < 0.001); the difference between these reported medians was 140 H‐score units. These paired descriptive results should not be interpreted as estimates of prognostic effect or as a validation of a clinically actionable H‐score threshold.
FIGURE 5.
Immunohistochemical detection of c‐MET protein in paired colorectal cancer and adjacent non‐neoplastic tissues. (A) Low‐power overview (4×) showing adjacent mucosa (upper region) and tumor (lower region); boxes identify the areas enlarged in panels (B and C). (B) Tumor glands with strong membranous and cytoplasmic c‐MET staining (20×). (C) Adjacent non‐neoplastic mucosa with weak staining (20×). Brown chromogen indicates c‐MET immunoreactivity, and blue hematoxylin counterstains nuclei. Images were generated from institutional specimens in this study.
These findings confirm increased c‐MET protein expression in the institutional CRC specimens.
3.3. Association Between MET mRNA Expression and Clinicopathological Characteristics
In TCGA‐COAD and TCGA‐READ, median‐based high and low MET expression groups did not differ significantly in age, sex, history of colorectal polyps, history of previous malignancy, T/N/M stage, overall clinical stage, preoperative carcinoembryonic antigen level, or maximum tumor diameter (all p > 0.05; Tables 1 and 2). Thus, MET expression did not show a stable association with the conventional clinicopathological variables examined in these datasets.
TABLE 1.
Association between MET mRNA expression and clinicopathological characteristics in TCGA‐COAD.
| Group | High MET expression | Low MET expression | Statistical method | p |
|---|---|---|---|---|
| 235 | 236 | |||
| n (%) | ||||
| Age | Mann–Whitney U test | 0.210 | ||
| Median (interquartile range) | 68 (57.25, 77) | 69 (60, 78) | ||
| Unknown | 1 (0.43) | 1 (0.42) | ||
| Sex | χ 2 test | 0.247 | ||
| Female | 105 (44.87) | 118 (50.21) | ||
| Male | 129 (55.13) | 117 (49.79) | ||
| Unknown | 1 (0.43) | 1 (0.42) | ||
| History of colon polyps | χ 2 test | 0.989 | ||
| None | 131 (64.53) | 127 (64.47) | ||
| Yes | 72 (35.47) | 70 (35.53) | ||
| Unknown | 32 (13.62) | 39 (16.53) | ||
| History of malignant tumors | χ 2 test | 0.598 | ||
| None | 205 (87.61) | 202 (85.96) | ||
| Yes | 29 (12.39) | 33 (14.04) | ||
| Unknown | 1 (0.43) | 1 (0.42) | ||
| T stage | χ 2 test | 0.319 | ||
| T1 | 6 (2.56) | 5 (2.14) | ||
| T2 | 47 (20.09) | 33 (14.1) | ||
| T3 | 151 (64.53) | 168 (71.79) | ||
| T4 | 30 (12.82) | 28 (11.97) | ||
| Uncertain | 1 (0.43) | 2 (0.85) | ||
| N stage | χ 2 test | 0.684 | ||
| N0 | 134 (57.26) | 143 (60.85) | ||
| N1 | 57 (24.36) | 50 (21.28) | ||
| N2 | 43 (18.38) | 42 (17.87) | ||
| Uncertain | 1 (0.43) | 1 (0.42) | ||
| M stage | χ 2 test | 0.893 | ||
| M0 | 171 (83.82) | 172 (84.31) | ||
| M1 | 33 (16.18) | 32 (15.69) | ||
| Uncertain | 31 (13.19) | 32 (13.56) | ||
| Clinical stage | χ 2 test | 0.147 | ||
| I | 46 (20.35) | 32 (13.79) | ||
| II | 80 (35.4) | 103 (44.4) | ||
| III | 67 (29.65) | 65 (28.02) | ||
| IV | 33 (14.6) | 32 (13.79) | ||
| Uncertain | 9 (3.83) | 4 (1.69) | ||
| Preoperative or pre‐treatment CEA level | Mann–Whitney U test | 0.535 | ||
| Median (interquartile range) | 3 (1.89, 6.82) | 3.2 (1.51, 7.55) | ||
| Unknown | 79 (33.62) | 95 (40.25) | ||
| Histologic category | Fisher's exact test | 0.608 | ||
| Adenoma and adenocarcinoma | 200 (85.11) | 199 (84.32) | ||
| Complex epithelial neoplasms | 0 (0) | 2 (0.85) | ||
| Cystic, mucinous, and serous tumors | 35 (14.89) | 34 (14.41) | ||
| Epithelial tumors | 0 (0) | 1 (0.42) | ||
| Maximum diameter | Mann–Whitney U test | 0.378 | ||
| Median (Interquartile range) | 1.1 (0.9, 1.5) | 1 (0.9, 1.5) | ||
| Unknown | 34 (14.47) | 89 (37.71) | ||
TABLE 2.
Association between MET mRNA expression and clinicopathological characteristics in TCGA‐READ.
| Group | High MET expression | Low MET expression | Statistical method | p |
|---|---|---|---|---|
| 83 | 84 | |||
| n (%) | ||||
| Age | t‐test | 0.899 | ||
| Mean (standard deviation) | 64.68 (10.23) | 64.45 (13) | ||
| Unknown | 1 (1.2) | 0 (0) | ||
| Sex | χ 2 test | 0.767 | ||
| Female | 38 (46.34) | 37 (44.05) | ||
| Male | 44 (53.66) | 47 (55.95) | ||
| Unknown | 1 (1.2) | 0 (0) | ||
| History of colon polyps | χ 2 test | 0.285 | ||
| None | 59 (81.94) | 56 (74.67) | ||
| Yes | 13 (18.06) | 19 (25.33) | ||
| Unknown | 11 (13.25) | 9 (10.71) | ||
| History of malignant tumors | χ 2 test | 0.749 | ||
| None | 74 (90.24) | 77 (91.67) | ||
| Yes | 8 (9.76) | 7 (8.33) | ||
| Unknown | 1 (1.2) | 0 (0) | ||
| T stage | Fisher's exact test | 0.857 | ||
| T1 | 4 (4.94) | 5 (6.02) | ||
| T2 | 12 (14.81) | 16 (19.28) | ||
| T3 | 58 (71.6) | 55 (66.27) | ||
| T4 | 7 (8.64) | 7 (8.43) | ||
| Uncertain | 2 (2.41) | 1 (1.19) | ||
| N stage | χ 2 test | 0.635 | ||
| N0 | 41 (51.9) | 43 (51.81) | ||
| N1 | 24 (30.38) | 21 (25.3) | ||
| N2 | 14 (17.72) | 19 (22.89) | ||
| Uncertain | 4 (4.82) | 1 (1.19) | ||
| M stage | χ 2 test | 0.354 | ||
| M0 | 58 (81.69) | 68 (87.18) | ||
| M1 | 13 (18.31) | 10 (12.82) | ||
| Uncertain | 12 (14.46) | 6 (7.14) | ||
| Clinical stage | χ 2 test | 0.712 | ||
| I | 12 (15.58) | 18 (22.78) | ||
| II | 27 (35.06) | 24 (30.38) | ||
| III | 26 (33.77) | 25 (31.65) | ||
| IV | 12 (15.58) | 12 (15.19) | ||
| Uncertain | 6 (7.23) | 5 (5.95) | ||
| Preoperative or pre‐treatment CEA level | Wilcoxon test | 0.502 | ||
| Median (interquartile range) | 4.21 (1.7, 12.49) | 2.99 (1.41, 10.6) | ||
| Unknown | 32 (38.55) | 23 (27.38) | ||
| Histologic category | χ 2 test | 0.062 | ||
| Adenoma and adenocarcinoma | 79 (95.18) | 73 (86.9) | ||
| Cystic, mucinous, and serous tumors | 4 (4.82) | 11 (13.1) | ||
| Maximum diameter | Wilcoxon test | 0.863 | ||
| Median (interquartile range) | 1 (0.9, 1.4) | 1 (0.8, 1.5) | ||
| Unknown | 20 (24.1) | 13 (15.48) | ||
3.4. Association Between MET Expression and Survival Outcomes
Kaplan–Meier Plotter analysis showed that the MET‐high group (n = 754) had a median OS of 32.05 months, compared with 47 months in the MET‐low group (n = 307) (log‐rank p = 0.0033; HR = 1.46, 95% CI, 1.13–1.88; Figure 6). The MET‐high group (n = 840) also had a shorter median RFS than the MET‐low group (n = 496) (34 vs. 119 months; log‐rank p = 0.029; HR = 1.29, 95% CI, 1.02–1.61; Figure 7).
FIGURE 6.
Kaplan–Meier analysis of overall survival according to MET expression in colon cancer. Black and pink curves indicate the MET‐low and MET‐high groups, respectively; tick marks indicate censored observations, and the numbers at risk are shown below the plot. The hazard ratio (HR), 95% confidence interval (CI), and log‐rank p‐value are displayed in the panel. This is an unadjusted online analysis and does not establish independent prognostic value.
FIGURE 7.
Kaplan–Meier analysis of relapse‐free survival according to MET expression in colon cancer. Black and pink curves indicate the MET‐low and MET‐high groups, respectively; tick marks indicate censored observations, and the numbers at risk are shown below the plot. The hazard ratio (HR), 95% confidence interval (CI), and log‐rank p‐value are displayed in the panel. This is an unadjusted online analysis and does not establish independent prognostic value.
These online, unadjusted analyzes support an association between higher MET expression and poorer survival outcomes but do not establish MET as an independent prognostic factor.
3.5. Correlation Between MET Expression and Immune‐Cell Infiltration
Purity‐adjusted TIMER analysis showed weak‐to‐moderate correlations between MET mRNA expression and inferred infiltration of B cells, CD8+ T cells, CD4+ T cells, macrophages, neutrophils, and dendritic cells in COAD (all p < 0.05; Figure 8); the largest coefficient was observed for CD4+ T cells (r = 0.259, p < 0.001). In READ, MET mRNA expression correlated with tumor purity and inferred infiltration of B cells, CD8+ T cells, macrophages, and neutrophils (all p < 0.05; Figure 9); the largest coefficient was observed for CD8+ T cells (r = 0.326, p < 0.001). The effect sizes were therefore modest and quantify association, not immune‐cell abundance measured experimentally.
FIGURE 8.
Purity‐adjusted TIMER correlations between MET mRNA expression and inferred tumor purity or immune‐cell infiltration in TCGA‐COAD. Panels show tumor purity, B cells, CD8+ T cells, CD4+ T cells, macrophages, neutrophils, and dendritic cells. Each dot represents a tumor sample; the blue line and gray band show the fitted trend and its confidence band as displayed by TIMER. Red annotations report the partial correlation coefficient and p‐value. Infiltration values are computational estimates and not experimental cell counts.
FIGURE 9.
Purity‐adjusted TIMER correlations between MET mRNA expression and inferred tumor purity or immune‐cell infiltration in TCGA‐READ. Panels show tumor purity, B cells, CD8+ T cells, CD4+ T cells, macrophages, neutrophils, and dendritic cells. Each dot represents a tumor sample; the blue line and gray band show the fitted trend and its confidence band as displayed by TIMER. Red annotations report the partial correlation coefficient and p‐value. Infiltration values are computational estimates and not experimental cell counts.
These algorithm‐derived correlations do not demonstrate that MET directly promotes or suppresses infiltration of any specific immune‐cell population.
3.6. Joint MET/HGF Expression and Tumor Microenvironment Features
After stratification by cohort‐specific median MET and HGF expression, inferred tumor purity and ESTIMATE‐derived scores differed among the four groups in COAD (Figure 10). The MET‐high/HGF‐low group generally showed the highest inferred tumor purity and the lowest immune, stromal, and total ESTIMATE scores, whereas the MET‐low/HGF‐high group showed lower tumor purity and higher immune and stromal scores. The double‐high and double‐low groups were generally intermediate. A similar overall ordering was observed in READ (Figure 11). These are distributional comparisons of retrospective bulk‐transcriptome estimates; the boxplots display the observed effect magnitude and between‐sample variability, whereas the omnibus p‐values do not establish a mechanistic or clinical effect.
FIGURE 10.
ESTIMATE‐derived tumor microenvironment features across joint MET/HGF transcript‐expression groups in TCGA‐COAD. MET and HGF were dichotomized at cohort‐specific medians; figure labels using “cMET” denote the MET transcript group, not c‐MET protein. (A) Sample‐level ESTIMATE, immune, stromal, and inferred tumor‐purity components. (B) Inferred tumor purity by group. (C) ESTIMATE, immune, and stromal scores by group. In boxplots, center lines indicate medians, boxes indicate interquartile ranges, whiskers show the displayed spread, and dots indicate individual samples/outliers as plotted. Asterisks indicate nominal between‐group p‐values as displayed (*p < 0.05; **p < 0.01; ***p < 0.001). All measures are computational estimates from bulk transcriptomes.
FIGURE 11.
ESTIMATE‐derived tumor microenvironment features across joint MET/HGF transcript‐expression groups in TCGA‐READ. MET and HGF were dichotomized at cohort‐specific medians; figure labels using “cMET” denote the MET transcript group, not c‐MET protein. (A) Sample‐level ESTIMATE, immune, stromal, and inferred tumor‐purity components. (B) Inferred tumor purity by group. (C) ESTIMATE, immune, and stromal scores by group. In boxplots, center lines indicate medians, boxes indicate interquartile ranges, whiskers show the displayed spread, and dots indicate individual samples/outliers as plotted. Asterisks indicate nominal between‐group p‐values as displayed (*p < 0.05; **p < 0.01; ***p < 0.001). All measures are computational estimates from bulk transcriptomes.
These results suggest that joint MET/HGF transcript‐expression groups may reflect differences in tumor‐cell content and non‐tumor microenvironmental components. They do not support interpreting the double‐high group as uniformly immune depleted or as direct evidence of ligand‐dependent c‐MET activation.
CIBERSORT analysis identified differences in selected dendritic‐cell, macrophage, natural‐killer‐cell, CD8+ T‐cell, and regulatory T‐cell fractions among the four COAD groups (Figure 12). Fewer immune‐cell subsets differed significantly in READ (Figure 13).
FIGURE 12.
CIBERSORT‐inferred fractions of 22 immune‐cell populations across joint MET/HGF transcript‐expression groups in TCGA‐COAD. MET and HGF were dichotomized at cohort‐specific medians; figure labels using “cMET” denote the MET transcript group, not c‐MET protein. Each dot represents a sample, and boxplots summarize the distribution of the inferred relative fraction for each cell type (center line, median; box, interquartile range; whiskers, displayed spread). Asterisks indicate nominal between‐group p‐values as displayed (*p < 0.05; **p < 0.01; ***p < 0.001). Estimated fractions should not be interpreted as experimentally measured immune‐cell abundance.
FIGURE 13.
CIBERSORT‐inferred fractions of 22 immune‐cell populations across joint MET/HGF transcript‐expression groups in TCGA‐READ. MET and HGF were dichotomized at cohort‐specific medians; figure labels using “cMET” denote the MET transcript group, not c‐MET protein. Each dot represents a sample, and boxplots summarize the distribution of the inferred relative fraction for each cell type (center line, median; box, interquartile range; whiskers, displayed spread). Asterisks indicate nominal between‐group p‐values as displayed (*p < 0.05; **p < 0.01; ***p < 0.001). Estimated fractions should not be interpreted as experimentally measured immune‐cell abundance.
The identity and direction of the differing immune‐cell subsets were not fully concordant between COAD and READ. Accordingly, the four expression groups could not be assigned a reproducible, universal continuum from immune ‘cold’ to immune ‘hot’ tumors.
Joint MET/HGF expression is therefore better interpreted as an exploratory framework for describing microenvironmental heterogeneity than as a validated immune subtype classification.
4. Discussion
By integrating paired‐tissue immunohistochemistry with several public datasets, this study confirmed increased MET mRNA and c‐MET protein expression in CRC and identified an unadjusted association between higher MET expression and shorter OS and RFS. Because these findings broadly agree with previous work [8, 9, 10, 19], the principal incremental contribution is not the repeated demonstration of MET overexpression or single‐gene immune correlations, but the exploratory assessment of whether joint MET/HGF expression further characterizes tumor purity and immune/stromal composition in COAD and READ.
The reported frequency and prognostic relevance of c‐MET protein expression in CRC vary with antibody clone, membrane‐versus‐cytoplasmic scoring, H‐score threshold, disease stage, and treatment context. Several studies have associated increased c‐MET expression with advanced disease or poorer survival [9, 20, 21, 22, 23], whereas others have not found stable relationships with recurrence or survival [7, 19]. Our TCGA analysis likewise found no significant association between MET expression and TNM stage. These discrepancies support the need for standardized assays, prespecified thresholds, and independent validation rather than universal conclusions from a single cohort.
The Kaplan–Meier Plotter results in the present study were univariable and did not comprehensively adjust for age, sex, stage, molecular subtype, or treatment. The original description of c‐MET as an independent prognostic biomarker was therefore not statistically justified and has been removed. A definitive assessment would require individual‐level survival data, univariable screening followed by a multivariable Cox model incorporating MET expression and clinically relevant covariates, and reporting of adjusted hazard ratios, 95% confidence intervals, p‐values, and the proportional‐hazards assessment.
Wan et al. reported MET overexpression, prognostic associations, and relationships with CD8+ T cells, dendritic cells, macrophages, regulatory T cells, and tumor mutational burden in CRC and also examined acid‐induced epithelial–mesenchymal transition and metastasis experimentally [12]. We therefore do not claim that MET–immune correlations are novel. The difference in the present study is limited to four‐group ligand–receptor transcript stratification, separate analyzes of COAD and READ, and joint evaluation of tumor purity, ESTIMATE scores, and 22 CIBERSORT‐derived immune‐cell fractions. This remains exploratory evidence and does not replace functional experiments.
Conceptually, the four groups preserve information lost by MET‐only dichotomization. In particular, the contrast between MET‐high/HGF‐low and MET‐low/HGF‐high tumors distinguish a bulk profile enriched for estimated tumor‐cell content from one enriched for estimated immune/stromal content in both cohorts. The double‐high and double‐low states did not form a reproducible biological continuum, and the classifier is not proposed as a pathway‐activation score. Thus, its present contribution is hypothesis generation concerning the relationship between ligand–receptor transcript context and tissue composition, rather than an incremental diagnostic, prognostic, or predictive biomarker claim.
The joint‐expression results also require a more cautious biological interpretation than that presented in the original manuscript. In both cohorts, the MET‐high/HGF‐low group was closer to a tumor‐cell‐dominant state with lower immune and stromal scores, whereas the MET‐low/HGF‐high group had more inferred non‐tumor components; the double‐high group was not uniformly the least infiltrated. Because HGF can be produced by stromal cells [5, 6, 7, 24], high HGF transcript abundance may partly reflect stromal content rather than tumor‐cell c‐MET activation. The current data therefore support an association between joint expression and tissue composition, not a causal claim that ligand‐dependent c‐MET signaling creates an immune‐cold microenvironment.
Clinically, joint MET/HGF stratification may help generate hypotheses for future biomarker‐stratified studies of MET‐directed and immune‐based therapies [24, 25, 26, 27, 28]. Recent CRC reviews also emphasize the interconnectedness of signaling networks and the tumor microenvironment, but neither validates the present MET/HGF classifier [29, 30]. The present findings are insufficient for patient selection, prediction of response to immune‐checkpoint inhibition, or treatment decisions. Any future clinical application requires prespecified thresholds and prospective, adequately powered, multicenter validation in independent cohorts with relevant molecular, treatment, and outcome data.
This study has several limitations. First, the immunohistochemical cohort was small (51 paired cases), single center, and used only to validate differential c‐MET protein expression; it therefore has limited statistical power and precision for subgroup, clinicopathological, or outcome analyzes, and its findings may not generalize to other patient populations, laboratories, antibodies, or treatment settings. Complete survival validation was not performed in this cohort. Second, the online survival analysis was unadjusted, and no multivariable Cox model was available. Third, ESTIMATE and CIBERSORT provide computational estimates based on bulk transcriptomes, and the median‐based expression cutoffs may be cohort dependent. Fourth, MET/HGF mRNA levels do not substitute for cell‐specific HGF, total c‐MET, phosphorylated c‐MET, or direct pathway‐activity measurements. Finally, the study lacked gene perturbation, co‐culture, spatial, or in vivo experiments and therefore cannot establish a causal mechanism of immune‐microenvironment remodeling.
Future work should validate the joint‐expression patterns in independent, prospectively assembled, multicenter clinical cohorts with prespecified expression thresholds and multivariable outcome models. To define cellular context, paired tumor sections should be assessed by multiplex immunohistochemistry or immunofluorescence for HGF, total c‐MET, phosphorylated c‐MET, epithelial markers, and immune‐cell markers, with flow cytometry used where fresh tissue is available. Single‐cell and spatial transcriptomic analyzes would help identify HGF‐producing stromal populations and c‐MET‐expressing tumor or immune cells. To test causality, MET knockout/knockdown or pharmacologic inhibition and HGF overexpression, depletion, neutralization, or fibroblast–tumor co‐culture experiments should be coupled to assays of c‐MET phosphorylation, downstream signaling, cytokine production, and immune‐cell migration or cytotoxicity; relevant findings should then be assessed in organoid or in vivo models. Such studies will be required before MET/HGF‐based microenvironmental stratification can be considered for translational application. Figure 14 summarizes the observed associations and the working hypothesis while explicitly distinguishing them from unvalidated causal links.
FIGURE 14.
Working model derived from the observed associations. The solid upper boxes summarize the cohort‐level patterns observed after exploratory joint MET/HGF transcript stratification: MET‐high/HGF‐low was associated with higher inferred tumor purity and lower immune/stromal scores; MET‐low/HGF‐high showed the opposite pattern; double‐high and double‐low profiles were generally intermediate. Arrows converge on a candidate interpretation concerning tumor‐cell dominance and microenvironmental composition. The lower box emphasizes that pathway activation, biological mechanisms, causal inference, and clinical application require independent prospective validation and functional experiments. The figure presents a hypothesis‐generating summary rather than a confirmed molecular pathway.
5. Conclusion
MET mRNA and c‐MET protein were elevated in CRC, and higher MET expression was associated with poorer OS and RFS in unadjusted analyzes. Joint MET/HGF expression groups were associated with inferred tumor purity and immune/stromal composition, with MET‐high/HGF‐low tumors showing a more tumor‐cell‐dominant profile and MET‐low/HGF‐high tumors showing greater non‐tumor content. These findings are exploratory and require independent prospective multicenter cohort validation, multivariable survival analysis, protein‐level and spatial assessment, and functional perturbation experiments before any biomarker or patient‐stratification application can be considered.
Author Contributions
Shuai Luo: writing – original draft, funding acquisition, methodology, software, project administration, resources, conceptualization, investigation, writing – review and editing, formal analysis, validation, data curation, visualization. Yanli Cheng: supervision, resources, project administration, funding acquisition. Xiangli Li: validation. Hongwei Guo: supervision. Yiyan Long: software. Zepeng Xin: software.
Funding
This work was supported by Beijing Key Laboratory of Helicobacter pylori Infection and Upper Gastrointestinal Diseases, BZ0371.
Ethics Statement
The use of human tissue samples was approved by the Ethics Committee of The First Hospital of Tsinghua University (Approval No. 2024‐021‐02), and the requirement for informed consent was waived.
Conflicts of Interest
The authors declare no conflicts of interest.
Acknowledgements
This research was supported by the Open Project Program of the Beijing Key Laboratory of Helicobacter pylori Infection and Upper Gastrointestinal Diseases (No. BZ0371).
Data Availability Statement
The datasets analyzed in this study are publicly available from TCGA, TNMplot, and TIMER. Clinical sample and IHC data are available from the corresponding author upon reasonable request due to ethical restrictions.
References
- 1. Bray F., Laversanne M., Sung H., et al., “Global Cancer Statistics 2022: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries,” CA: A Cancer Journal for Clinicians 74, no. 3 (2024): 229–263, 10.3322/caac.21834. [DOI] [PubMed] [Google Scholar]
- 2. Han B., Zheng R., Zeng H., et al., “Cancer Incidence and Mortality in China, 2022,” Journal of the National Cancer Center 4, no. 1 (2024): 47–53, 10.1016/j.jncc.2024.01.006. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Ebner D. W., Fendrick A. M., Kisiel J. B., et al., “Evaluating Benefit‐To‐Burden Ratios of Established and Emerging Colorectal Cancer Screening Strategies,” Journal of the National Cancer Institute 117, no. 12 (2025): 2653–2660, 10.1093/jnci/djaf209. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Shalata W., Gluzman A., Man S., et al., “Colorectal Cancer in Elderly Patients: Insights Into Presentations, Prognosis, and Patient Outcomes,” Medicina (Kaunas, Lithuania) 60, no. 12 (2024): 1951, 10.3390/medicina60121951. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Comoglio P. M., Trusolino L., and Boccaccio C., “Known and Novel Roles of the MET Oncogene in Cancer: A Coherent Approach to Targeted Therapy,” Nature Reviews. Cancer 18, no. 6 (2018): 341–358, 10.1038/s41568-018-0002-y. [DOI] [PubMed] [Google Scholar]
- 6. Organ S. L. and Tsao M. S., “An Overview of the c‐MET Signaling Pathway,” Therapeutic Advances in Medical Oncology 3, no. 1 Suppl (2011): S7–S19, 10.1177/1758834011422556. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Yao H. P., Tong X. M., Hudson R., and Wang M. H., “MET and RON Receptor Tyrosine Kinases in Colorectal Adenocarcinoma: Molecular Features as Drug Targets and Antibody‐Drug Conjugates for Therapy,” Journal of Experimental & Clinical Cancer Research 39 (2020): 198, 10.1186/s13046-020-01711-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Parizadeh S. M., Jafarzadeh‐Esfehani R., Fazilat‐Panah D., et al., “The Potential Therapeutic and Prognostic Impacts of the c‐MET/HGF Signaling Pathway in Colorectal Cancer,” IUBMB Life 71, no. 7 (2019): 802–811, 10.1002/iub.2063. [DOI] [PubMed] [Google Scholar]
- 9. Lee S. J., Lee J., Park S. H., et al., “C‐MET Overexpression in Colorectal Cancer: A Poor Prognostic Factor for Survival,” Clinical Colorectal Cancer 17, no. 3 (2018): 165–169, 10.1016/j.clcc.2018.02.013. [DOI] [PubMed] [Google Scholar]
- 10. Gao W., Bing X., Li M., et al., “Study of the Critical Role of c‐Met and Its Inhibitor SU11274 in Colorectal Carcinoma,” Medical Oncology 30, no. 2 (2013): 546, 10.1007/s12032-013-0546-3. [DOI] [PubMed] [Google Scholar]
- 11. Binnewies M., Roberts E. W., Kersten K., et al., “Understanding the Tumor Immune Microenvironment for Effective Therapy,” Nature Medicine 24, no. 5 (2018): 541–550, 10.1038/s41591-018-0014-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Wan X., Li Z., and Zhou L., “C‐MET Is an Important Marker for Acid‐Driven Metastasis and Anti‐Immune in Colorectal Cancer,” International Immunopharmacology 142, no. Pt A (2024): 113153, 10.1016/j.intimp.2024.113153. [DOI] [PubMed] [Google Scholar]
- 13. Cancer Genome Atlas Network , “Comprehensive Molecular Characterization of Human Colon and Rectal Cancer,” Nature 487, no. 7407 (2012): 330–337, 10.1038/nature11252. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Bartha Á. and Győrffy B., “TNMplot.com: A Web Tool for the Comparison of Gene Expression in Normal, Tumor and Metastatic Tissues,” International Journal of Molecular Sciences 22, no. 5 (2021): 2622, 10.3390/ijms22052622. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Li T., Fu J., Zeng Z., et al., “TIMER2.0 for Analysis of Tumor‐Infiltrating Immune Cells,” Nucleic Acids Research 48, no. W1 (2020): W509–W514, 10.1093/nar/gkaa407. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Yoshihara K., Shahmoradgoli M., Martínez E., et al., “Inferring Tumour Purity and Stromal and Immune Cell Admixture From Expression Data,” Nature Communications 4 (2013): 2612, 10.1038/ncomms3612. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Newman A. M., Liu C. L., Green M. R., et al., “Robust Enumeration of Cell Subsets From Tissue Expression Profiles,” Nature Methods 12, no. 5 (2015): 453–457, 10.1038/nmeth.3337. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Lánczky A. and Győrffy B., “Web‐Based Survival Analysis Tool Tailored for Medical Research (KMplot): Development and Implementation,” Journal of Medical Internet Research 23, no. 7 (2021): e27633, 10.2196/27633. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Armstrong G. R., Khot M. I., Tiernan J. P., et al., “The Utility of c‐Met as a Diagnostic Tissue Biomarker in Primary Colorectal Cancer,” International Journal of Experimental Pathology 102, no. 3 (2021): 172–178, 10.1111/iep.12395. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. de Oliveira A. T., Matos D., Logullo A. F., et al., “MET Is Highly Expressed in Advanced Stages of Colorectal Cancer and Indicates Worse Prognosis and Mortality,” Anticancer Research 29, no. 11 (2009): 4807–4811. [PubMed] [Google Scholar]
- 21. Resnick M. B., Routhier J., Konkin T., et al., “Epidermal Growth Factor Receptor, c‐MET, Beta‐Catenin, and p53 Expression as Prognostic Indicators in Stage II Colon Cancer,” Clinical Cancer Research 10, no. 9 (2004): 3069–3075. [DOI] [PubMed] [Google Scholar]
- 22. Gayyed M. F., Abd El‐Maqsoud N. M., El‐Hameed El‐Heeny A. A., et al., “C‐MET Expression in Colorectal Adenomas and Primary Carcinomas With Corresponding Metastases,” Journal of Gastrointestinal Oncology 6, no. 6 (2015): 618–627. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Ginty F., Adak S., Can A., et al., “The Relative Distribution of Membranous and Cytoplasmic MET Is a Prognostic Indicator in Stage I and II Colon Cancer,” Clinical Cancer Research 14, no. 12 (2008): 3814–3822, 10.1158/1078-0432.CCR-08-0180. [DOI] [PubMed] [Google Scholar]
- 24. Xia Y., Huang C., Zhong M., et al., “Targeting HGF/c‐MET Signaling to Regulate the Tumor Microenvironment: Implications for Counteracting Tumor Immune Evasion,” Cell Communication and Signaling: CCS 23, no. 1 (2025): 46, 10.1186/s12964-025-02033-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Sun Z. J., Wu Y., Hou W. H., et al., “A Novel Bispecific c‐MET/PD‐1 Antibody With Therapeutic Potential in Solid Cancer,” Oncotarget 8, no. 17 (2017): 29067–29079, 10.18632/oncotarget.16173. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Bardelli A., Corso S., Bertotti A., et al., “Amplification of the MET Receptor Drives Resistance to Anti‐EGFR Therapies in Colorectal Cancer,” Cancer Discovery 3, no. 6 (2013): 658–673, 10.1158/2159-8290.CD-12-0558. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Raghav K., Morris V., Tang C., et al., “MET Amplification in Metastatic Colorectal Cancer: An Acquired Response to EGFR Inhibition, Not a de Novo Phenomenon,” Oncotarget 7, no. 33 (2016): 54627–54631, 10.18632/oncotarget.10365. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Joosten S. P. J., Mizutani T., Spaargaren M., Clevers H., and Pals S. T., “MET Signaling Overcomes Epidermal Growth Factor Receptor Inhibition in Normal and Colorectal Cancer Stem Cells, Causing Drug Resistance,” Gastroenterology 157, no. 4 (2019): 1153–1155. [DOI] [PubMed] [Google Scholar]
- 29. Kumari S., Peela S., Bramhachari P. V., Mallikarjuna K., Nagaraju G. P., and Srilatha M., “Raf‐Kinase Inhibitor Protein (RKIP): A Therapeutic Target in Colon Cancer,” Biochimica Et Biophysica Acta. Reviews on Cancer 1880, no. 5 (2025): 189387, 10.1016/j.bbcan.2025.189387. [DOI] [PubMed] [Google Scholar]
- 30. Dariya B., Girish B. P., Merchant N., Srilatha M., and Nagaraju G. P., “Resveratrol: Biology, Metabolism, and Detrimental Role on the Tumor Microenvironment of Colorectal Cancer,” Nutrition Reviews 82, no. 10 (2024): 1420–1436, 10.1093/nutrit/nuad133. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Data Availability Statement
The datasets analyzed in this study are publicly available from TCGA, TNMplot, and TIMER. Clinical sample and IHC data are available from the corresponding author upon reasonable request due to ethical restrictions.