CMTM6 status predicts survival in head and neck squamous cell carcinoma and correlates with PD-L1 expression

3.1 Patient and tumor characteristics

The study cohort consisted of 129 patients, all diagnosed with primary HNSCC, with surgical resection specimens available in 98% prior to adjuvant therapy. The clinicopathological characteristics are presented in Table 1. Most tumors were localized in the oral cavity and oropharynx (n = 42 and n = 46, respectively). Sixteen patients had hypopharyngeal cancer, 22 had laryngeal cancer, and regional lymph node metastases were observed in 3 cases. Approximately half of the patients were smokers (53.1%, ≥ 10 py), and one-third did not have critical alcohol consumption (33.1%). One third of the tumors (32.7%) were p16+ owing to previous HPV infection. In 11 p16+ cases, molecular HPV testing could not be performed because of poor DNA quality. These cases were classified as “HPV-like-p16+” because they were localized in the oropharynx (tonsils) and had any patient history of noxae. One case was finally classified as p16+/HPVneg. Tumors were obtained from all stages (T1–T4), with metastases in 7% of the cases. For three patients with available tumor material from primary and meta/synchronous metastatic/recurrent disease, the staining results from primary carcinomas were included for further analyses.

Table 1 Clinico-pathological characteristics of the HNSCC cohort3.2 Intra- and inter-individual heterogeneity of the TME

The TME was first examined using H&E-stained whole-mount sections to estimate the number and distribution of tumor-infiltrating leukocytes (TILs) (Figs. 1A, B).The TIL score at the invasion front was significantly higher than that at the tumor center (p < 0.0001; Fig. 1E, left). Subtype analysis of HPVpos and HPVneg cases revealed higher infiltration levels in the former, especially at the invasion front and tumor center (Fig. 1E, middle and right). Specific immunological subpopulations (T cells, macrophages, and natural killer [NK] cells) were then quantified (Figs. 1, 2, and Supplementary Fig. 2). The number of CD8+ cytotoxic T cells, as well as CD68+ and CD163+ macrophages, differed significantly between the invasive front and the tumor center (Fig. 1F, left; Figs. 2E, F, left). The number of CD8+ T cells was higher in HPV-associated cases than that in their HPV-unrelated counterparts (p < 0.01 (invasion front); p < 0.05 (tumor center); Fig. 1F, middle and right). In contrast, the number of macrophages (CD68+ and CD163+) was similar between the two subtypes (Figs. 2E, F, middle, and right). The number of CD56+ natural killer (NK) cells was low and showed a similar pattern in each compartment (Supplementary Figs. 2A, B, E, and G).

Fig. 1figure 1

Assessment of leukocyte infiltration score by morphology and CD8+ T cells by immunohistochemistry for the invasive front and the tumor center. Representative slides showing different infiltration patterns in individual tumors with absent (A) leukocyte infiltration (= Score 0; HE, 100x), moderate (B) leukocyte infiltration (= Score 3; HE, 100×), minimal (C) CD8+ cells (= Score 1, 200x) and high (D) CD8+ cells (= Score 4, 200x). Asterix = tumor center, green line = invasive front. E TILs at the invasive front and the tumor center (left, n = 63 invasive front; n = 66 center; ****p < 0.0001, Unpaired t-test (two-tailed)), as well as between HPVneg and HPVpos HNSCCs (n = 88 HPVneg; n = 36 HPVpos, ns–not significant). F CD8+ cells at the invasive front (left; *p < 0.05, one-way ANOVA (Tukey's multiple comparisons test)) and the tumor center (right; *p < 0.05, one-way ANOVA (Tukey's multiple comparisons test)), as well as between HPVneg and HPVpos HNSCCs (n = 88 HPVneg; n = 36 HPVpos, ns–not significant) (G, H) TILS and CD8.+ cells in different anatomical sites. n = 35 oral cavity; n = 36 oropharynx; n = 13 hypopharynx; n = 16 larynx; *p < 0.05, one-way ANOVA (Tukey's multiple comparisons test))

Fig. 2figure 2

Assessment of CD68+ and CD163+ cells by immunohistochemistry for the invasive front and the tumor center. AD Representative slides showing macrophage infiltration patterns (CD68+, CD163+) in individual tumors with minimal (A) CD68+ TAMs (= Score 1; 100x), high (B) CD68+ TAMs (= Score 4; 100x), minimal (C) CD163+ TAMs (= Score 1, 200x) and high (D) CD163+ TAMs (= Score 4, 200×). Asterix = tumor center, green line = invasive front. E CD68+ macrophages at the invasive front and the tumor center (left, n = 66 invasive front; n = 68 center; ****p < 0.0001, U-test (two-tailed)), as well as between HPVneg and HPVpos HNSCCs (n = 52 HPVneg; n = 16 HPVpos, ns–not significant). (F) CD163+ macrophages at the invasive front and the tumor center (left, n = 63 invasive front; n = 66 center; **p < 0.01, U-test (two-tailed)), as well as between HPVneg and HPVpos HNSCCs (n = 84 HPVneg; n = 31 HPVpos, ns–not significant). (G) CD68+ and (H) CD163+ macrophages at the invasive front (left; *p < 0.05, Kruskal Wallis test (Dunn’s multiple comparisons test)) and the tumor center (right; *p < 0.05, one-way ANOVA (Tukey's multiple comparisons test)) in different anatomical sites. CD68: n = 16 oral cavity; n = 27 oropharynx; n = 8 hypopharynx; n = 14 larynx; CD163: n = 34 oral cavity; n = 41 oropharynx; n = 14 hypopharynx; n = 20 larynx

Hypopharyngeal tumors had fewer TILs (HE, CD8+), both at the invasion front and tumor center (p < 0.05 vs. oropharynx; Figs. 1G, H). Conversely, oropharyngeal carcinomas showed the highest levels of TILs and CD8+ cells at the invasion front. The oral and laryngeal tumors had comparable TIL levels (Figs. 1G, H). The number of CD68+ and CD163+ macrophages differed significantly between the different anatomical sites. Again, hypopharyngeal cancers had the lowest number of infiltrating CD68+ and CD163+ macrophages in each compartment (CD68: p < 0.05 vs. oral cavity and CD163: p < 0.05 vs. larynx; Figs. 2G, H).

3.3 Impact of HPV status, anatomical site, TILs, and CD8+ T cells on overall survival

Morphological examination of HPVpos and HPVneg cancers revealed characteristic patterns in both subtypes with respect to differentiation, p53 staining, and Ki-67 proliferation index (Figs. 3A–H). We then focused on the prognostic impact of the TME on the overall survival (OS) of patients at the initial diagnosis. This analysis revealed a significant survival benefit for patients with HPV-associated HNSCC compared with their HPV-unrelated counterparts (p < 0.01; Fig. 3i; Table 2). In the HPV-related cohort, tobacco smoking (n = 8) and alcohol consumption (n = 4) did not adversely affect patient survival (p = 0.3). In this subgroup, 89% of the patients received adjuvant therapy (i.e., cisplatin, cetuximab, and/or radiotherapy). Four patients underwent surgery without the administration of adjuvant therapy. In this small group, two patients were still alive, and two died. The latter presented with a worse overall health status at diagnosis (ECOG 2 vs. ECOG 0) and advanced-stage disease. In the HPV-unrelated group, the choice of treatment had little effect on OS. When analyzing the impact of the anatomical site, oral cancer was found to have the best OS, while hypopharyngeal cancer had the worst outcome (Fig. 3J). Similarly, TILs and CD8+ T cells were positively associated with OS, independently of the treatment (Fig. 4). For TILs, the difference was statistically significant (invasion front). For CD8+ T cells, the survival benefit was independent of the tumor location according to the univariate analysis of Kaplan–Meier curves (invasion front: p < 0.001; center: p < 0.01, log-rank) and multivariate Cox regression (Table 2).

Fig. 3figure 3

Morphology, immunophenotyping and prognostic impact of HPV-negative and HPV-positive cancer. A moderate differentiated squamous cell carcinoma of the larynx (HE staining), p16INK4A-negative (B), p53-mutant (C) with moderate proliferation index (D). E Poorly differentiated squamous cell carcinoma of the right tonsil, F p16INK4A-positive with diffuse nuclear and cytoplasmic staining, G p53 wildtype and H higher proliferation index (200 × magnification). I Kaplan–Meier curves show a significant longer OS for HPV+ cases, n = 129; ** p < 0.01 Log-rank analysis. J Kaplan Meier survival curve of HNSCC patients depending on anatomical location revealed best OS for tumors located in the oral cavity; n = 126; Log-rank analysis, n.s.–not significant

Table 2 Uni- and Multivariate Cox regression analysis of clinic-pathological variablesFig. 4figure 4

Impact of TILs and CD8+ T cells on overall survival of HNSCC patients. A Prognostic relevance of TILs depending on the location, i.e. tumor invasion front or tumor center. Categorization: 0–1; > 1; n = 121 and n = 124, respectively; ** p < 0.01, * p < 0.05, Log-rank analysis. B Prognostic relevance of CD8+ cytotoxic T cells depending on the location, i.e. tumor invasion front or tumor center. Categorization: 0–1; > 1; n = 124 and n = 126, respectively; ** p < 0.01; ***p < 0.001, Log-rank analysis

3.4 PD-L1 and CMTM6 status and its prognostic impact

Although PD-L1 is an approved biomarker, we tested whether CMTM6 might have a comparable prognostic value (Figs. 5, 6). First, a consensus cutoff was established to determine positivity. All samples were categorized as CMTM6 high vs. low, with CPS ≥ 10 defined as high, according to the results of the receiver operating characteristic (ROC) curve analysis (area under the curve (AUC): 0.72; specificity > 65%). Using this cutoff, 67.2% of all samples were classified as CMTM6high, which is slightly below the PD-L1 positivity in this cohort (CPS ≥ 1: 74.4%, Table 1). However, the median CPS of CMTM6 was higher than the median CPS of PD-L1 (25 vs. 5). Representative images of individual HNSCC cases are shown in Figs. 5C-F.

Fig. 5figure 5

Combined Positive Score, Tumor Proportion Score and Immune Cell Score for PD-L1 and CMTM6. A Overall PD-L1 score (CPS, TPS, ICS), and CPS in HPVneg and HPVpos as well as in different anatomical sites. n = 125; *p < 0.05; ****p < 0.0001 Kruskal Wallis test (Dunn’s multiple comparisons test), ns–not significant). B Overall CMTM6 score (CPS, TPS, ICS), and CPS in HPVneg and HPVpos as well as in different anatomical sites. n = 119; ****p < 0.0001 Kruskal Wallis test (Dunn’s multiple comparisons test), ns–not significant). C Squamous cell cancer specimen with negativity for PD-L1 on the tumor cells showing one positive tumor infiltrating lymphocyte (= PD-L1 negative), D membranous positivity for PD-L1 on the tumor cells and abundant positive tumor infiltrating lymphocytes (= PD-L1 positive), E negativity for CMTM6 on the tumor cells and few positive tumor infiltrating lymphocytes (= CMTM6 low), F membranous positivity for CMTM6 on the tumor cells and abundant positive tumor infiltrating lymphocytes (= CMTM6 high; all 200x). For PD-L1, CPS ≥ 1 was set positive, for CMTM6, CPS ≥ 5 was set positive. CPS, Combined Positive Score; TPS, Tumor Proportion Score; ICS, Immune Cell Score

Fig. 6figure 6

Impact of PD-L1 and CMTM6 on overall survival of HNSCC patients. A, B Prognostic relevance of PD-L1 and CMTM6 within HPVneg and HPVpos cases. CPS: ratio of all marker positive TC, lymphocytes, macrophages to TC in the corresponding area multiplied by 100 (therefore containing no unit). n = 128; ** p < 0.01; ***p < 0.001 Log-rank analysis. C Prognostic relevance of PD-L1 and CMTM6 within HPVneg and HPVpos cases depending on the CPS. n = 128; * p < 0.05 Log-rank analysis. D Prognostic relevance of PD-L1 CPS according to HPV status. n = 128; *** p < 0.001 Log-rank analysis. EJ Immunohistochemistry. Paired Examples of cases being PD-L1-positive (E)/ CMTM6 high (F), PD-L1-negative (G, here with positive stained TILs to show reactivity)/CMTM6 high (H) and PD-L1- negative (I)/CMTM6 low (J). There were no tumors with high PD-L1 but low CMTM6-status

A closer look at PD-L1 and its spatial distribution revealed a strong correlation between CPS and TPS (Spearman’s r = 0.779, p < 0.001), CPS and ICS (Spearman’s r = 0.851, p < 0.001), and between ICS and TPS (Spearman’s r = 0.616, p < 0.001). For CMTM6, there was a strong correlation between CPS and TPS (Spearman’s r = 0.730, p < 0.0001). While CPS correlated weakly with ICS (Spearman’s r = 0.242, p < 0.01), there was no correlation between ICS and TPS (Spearman’s r = -0.041).

A weak correlation was observed between the CPS of both markers (Spearman’s r = 0.217; Supplementary Fig. 3A). Most samples with a CMTM6 CPS ≥ 10 were PD-L1 positive. For both markers, the CPS did not differ significantly between HPV-associated and HPV-unrelated HNSCCs (Figs. 5A, B, middle panel). Among the HPV-associated specimens, high-risk HPV subtypes (Table 1) were evenly distributed between the CMTM6low/high and PD-L1pos/neg groups. Neither UICC stage, smoking or alcohol consumption affected CMTM6 or PD-L1 status.

Likewise, anatomical location had little impact on PD-L1 or CMTM6 positivity (Figs. 5A, B, right). An exception was observed for PD-L1 CPS, which was generally lower in hypopharyngeal cancers than that in other sites (Fig. 5A, right).

To validate the immunohistochemical observations at the molecular level, PD-L1 and CMTM6 mRNA expression levels were detected in 29 tumors (Fig. 7). In this subgroup, PD-L1 and CMTM6 mRNA expression levels showed high similarity (Fig. 7A) and a weak correlation (Pearson’s r 0.22; p 0.12, Fig. 7B). For PD-L1, mRNA levels correlated with protein status in 22 of 29 samples analyzed. For CMTM6, mRNA levels correlated with protein status in 21 of 26 samples analyzed (Fig. 7C). Using TCGA data from from 612 primary HNSCC samples, the correlation between CMTM6 and PD-L1 mRNA expression was positive (Spearman's rho = 0.09845; p = 0.02141; supplementary Fig. 4A). Using either the median or interquartile range of expression, none of the markers had a significant impact on survival in the 250 analyzed samples (supplementary Fig. 4B, C), which were, however, not stratified accodirng to p16 status.

Fig. 7figure 7

mRNA expression levels of PD-L1 and CMTM6. A Analysis was done on 29 tumor samples, from which mRNA was available. B Pearson correlation for PD-L1 and CMTM6 mRNA expression levels showed a weak correlation. C Heatmap depicting correlation of mRNA levels with protein status in 21 of 26 samples analyzed

In our cohort, patients with PD-L1 CPS ≥ 1 and the subgroup with CMTM6 CPS ≥ 10 had a significantly longer OS (p < 0.0001 and p < 0.001, respectively; log-rank test, Figs. 6A, B, E, F). The Cox regression analysis revealed sex, grade, TNM, ECOG, HPV status, and CD8+ T cells as prognostic predictors. Multivariable Cox regression analysis also indicated a significantly better OS of the CMTM6 high versus CMTM6 low groups (p = 0.032; Table 2). Besides, tumor grade, p16 status, CD8+ T cells, and PD-L1, were statistically significant parameters for OS (Table 2). Notably, the significance of CMTM6 was comparable to that of PD-L1, confirming its prognostic relevance. The combination of both markers showed prolonged survival in the subgroup with PD-L1 expression ≥ 1 and synchronous CMTM6 expression ≥ 10 (p < 0.0001; log-rank test, Figs. 6C, E, F). In contrast, the CMTM6high/PD-L1neg and CMTM6low/PD-L1neg subgroups showed a shorter OS (median OS, 40 months; Figs. 6C, G–J).

The best outcome was seen for patients with PD-L1pos/HPVpos cancers with a 100% survival rate in a follow-up duration of > 80 months (Fig. 6D). Contrastingly, patients with PD-L1neg/HPVneg HNSCC had the worst outcomes (median OS, 39 months). A comparable outcome was observed with CMTM6 (Supplementary Fig. 3B). Notably, HPVpos patients without CMTM6 expression showed poor outcomes. Noteworthy, however, is the high variation between individual patient samples, identified by nested analysis for CMTM6, PD-L1 CPS, and CD8 positivity in HPVneg and HPVpos HNSCCs (supplementary Fig. 3C).

Additional prognostic markers were examined to obtain a more holistic view of the relationship between CMTM6 and PD-L1 expression. The number of central TILs showed a moderate correlation with the CPS of CMTM6 (Spearman’s r = 0.34, p < 0.0001). For PD-L1, a significant correlation was found between CPS and TILs at the invasive front (Spearman’s r = 0.266, p < 0.01). Both parameters correlated with the CD8+ T cell score (Spearman’s r = 0.22–0.34). There were no significant differences in PD-L1 or CMTM6 expression between HPVpos and HPVneg cases, although the latter tended to be more PD-L1 positive (p = 0.07; Fisher’s exact test).

Notably, patients with CMTM6high tumors and a CD8+ T-cell score > 1 showed the best OS (Supplementary Fig. 5A). The same positive prognostic effect was observed when CD8+ T-cells and PD-L1 CPS were included in the analysis (Supplementary Fig. 5B). Hence, CMTM6, in addition to PD-L1, is indicative of an inflammatory tumor microenvironment and predicts better patient outcomes.

3.5 Significance of TAMs, NK cells, and DKK1 expression

Neither the number of CD163+ and CD68+ macrophages nor the number of CD56+ NK cells had a significant influence on OS (Fig. 8, Supplementary Fig. 2). There was a highly significant correlation between CD68+ and CD163+ TAMs (Spearman’s r > 0.80 for both sites, p < 0.0001), with analogous staining patterns in the examined samples. Twelve patients with a CD68/CD163 ratio > 1 in the tumor center showed significantly longer OS than those without dominant M1-polarized macrophages (Fig. 8C). In the CD56-stained samples, 10% and 5% had a score ≥ 3 in each compartment. In the DKK1-stained subgroup, our scoring method showed that 71% of samples with DKK1-positive tumor cells had no clinicopathological relevance (Supplementary Figs. 2F, H). DKK1-positive carcinomas showed significantly more DKK1-positive cells in the surrounding stroma than DKK1-negative carcinomas (p = 0.0002; Fisher’s exact test).

Fig. 8figure 8

Impact of macrophages on overall survival of HNSCC patients. Prognostic relevance of A CD68+ and B CD163+ TAMs depending on the location, i.e. tumor invasion front or tumor center. Categorization: 0–1; > 1; n = 100; ** p < 0.01 Log-rank analysis. Categorization: 0–1; > 1; CD68: n = 68; CD163: n = 112; ** p < 0.01, Log-rank analysis. C The M1:M2 macrophage ratio was calculated by dividing the score of CD68+ M1 macrophages by the score of CD163+ M2 macrophages. Categorization: 0–1 n = 67; > 1; n = 12; * p < 0.05, Log-rank analysis

留言 (0)

沒有登入
gif