Tetraspanin8 expression predicts an increased metastatic risk and is associated with cancer-related death in human cutaneous melanoma

We previously showed that Tspan8 expression is restricted to invasive melanoma cell lines [13]. However, when Tspan8 expression was studied in the sixty-three cell lines from the Cancer Cell Line Encyclopedia (CCLE), very low levels of TSPAN8 mRNA and protein were observed (sup. Figure 1a-c). Heterogeneous but easily detectable Tspan8 expression levels were observed in vitro only in short-term cultures derived from metastatic pleural effusions from patients developing a primary invasive melanoma [14] (unpublished results; sup. Figure 1d). We can surmise that classic establishment of cell lines, that favors cell proliferative capacities rather than invasiveness, leads to a low Tspan8 expression, which is not representative of Tspan8 expression observed in situ in primary melanoma samples [13]. To overcome this in vitro problem, we performed our studies in patient samples.

We first analyzed TSPAN8 mRNA expression in the data set from The Cancer Genome Atlas (TCGA; https://​cancergenome.​nih.​gov/​) and observed a heterogeneous expression. To decipher the inner structure of the TSPAN8 expression distribution across the cohort, we applied Gaussian finite mixture models (mclust R package) and highlighted 2 groups of tumors: 1/3 of samples showed absent/low (in black) TSPAN8 mRNA expression and 2/3 of samples displayed a higher one (in red) (Fig. 1a). These two groups have been used for further analyses.

No correlation was found with CL, neither on the totality of the 414 samples nor on the 100 primary samples only (Fig. 1b). However, BI analysis surprisingly revealed that melanoma samples displaying a high TSPAN8 expression were more frequently thin melanomas (Fig. 1c, left panel). This highly significant effect was completely lost when only primary melanomas were analyzed (Fig. 1c, right panel), that can be explain by the fact that, in the TCGA cohort, mRNA expression data from primary melanomas are available only for thick melanomas (sup. Figure 2a), probably due to the small amount of tumor material for thin melanomas. Nevertheless, in this cohort, all thin melanomas gave rise to metastases, which could link TSPAN8 mRNA expression to metastatic properties.

Accordingly, high TSPAN8 expression was observed in 44.4% of regional cutaneous tissues, 41.9% of regional lymph nodes and 42.1% of distant metastases, versus 26% for primary tumors (Fig. 1d). Overall, we observed that only 26% of primary melanomas displayed high TSPAN8 expression, whereas local or distant melanoma dissemination was significantly associated with a more frequent TSPAN8 expression (Fig. 1e), consistently with the correlation of Tspan8 expression with invasiveness previously demonstrated in vitro [3, 4, 6, 13].

To evaluate the impact of Tspan8 expression at the protein level, we performed Tspan8 immunohistochemical staining analysis in a cohort of 100 primary melanoma samples: 73% of primary melanomas were positive for Tspan8 protein expression (sup. Figure 2b), without any significant differences depending on patient age or gender (sup. Figure 2c,d). We analyzed Tspan8 immunostaining in details, and sections were scored positive if any reactive areas were seen in melanoma lesions with growing intensity from 0 (no staining) to 4 (strong staining) (Fig. 2a, four first panel). Tspan8 staining is often clonal with some positive clones neighboring some negative areas (Fig. 2a, third panel), but could also be detected in the invasive front of the tumor (Fig. 2a, fifth panel) and, even if the majority of melanoma cells harbored a cytoplasmic staining, some rare cells in some samples presented a stronger staining at the membrane (Fig. 2a, sixth panel).

In accordance with the TCGA cohort, no correlation with CL (sup. Figure 2e) or BI emerged (sup. Figure 2f,g). However, Tspan8 was more frequently expressed in samples from patients who died during the 5-year period after the melanoma diagnosis than in samples from patients who remained alive, and especially for thin melanomas, inferior to 1 mm (Fig. 2b). Tspan8 protein seemed more frequently expressed in the thinnest melanomas from patients dead during the 5-year period after diagnosis (92.3% of samples). Moreover, the thinnest primary melanomas seemed to express the highest levels of Tspan8 protein (immunointensity score 4 exclusively in < 1 mm samples), especially in patients’ dead 5-years after diagnosis (60% of immunointensity scores 3–4; Fig. 2c).

Besides, only 65% of primary tumors that did not give metastases expressed Tspan8 whereas in primary tumors that disseminated, 84.6% expressed Tspan8 (Fig. 2d), demonstrating that Tspan8 expression in primary tumors was significantly correlated with the propensity to metastasize. Interestingly, in thin melanomas (less than 1 mm) especially, 11 among 12 samples (91.7%) from patients who died during the 5 years after the melanoma diagnosis gave rise to metastasis and 10 of them (90.9%) expressed Tspan8 protein, strengthening the correlation between Tspan8 protein expression and metastatic dissemination in thin melanomas. These results are particularly interesting since it is known that patients bearing thinner melanomas with no deep cutaneous invasion have generally a better survival rate [15] but that some thin melanomas (< 0.75 – 1 mm) could acquire some aggressive properties and lead to patient death. Indeed, metastases in sentinel lymph node were detected in more than 5% of patients who had a sentinel lymph node biopsy after the detection of a thin melanoma (< 1 mm) [16]. Our data showed that Tspan8 immunostaining could thus be a biomarker to assess the individual risk of metastasis, especially for thin melanomas, and potentially linked with patient poor survival since Tspan8 expression seemed also correlated with the vital status of the patients. The proportion of Tspan8-expressing primary melanomas increased from 66.1% for alive patients to 82.9% for dead patients, 5 years after the diagnosis (Fig. 2e), and the 5-years overall survival curve (Fig. 2f) showed that the presence of Tspan8 protein in primary melanomas tended to represent an increase risk of death.

In melanoma cell lines in vitro, we were not able to highlight a possible correlation between TSPAN8 mRNA expression and the presence of a BRAF or a NRAS mutation (sup. Fig. 3a), that are the most frequent genetic alterations detected in melanoma [17]. In the TCGA cohort as well, no statistical difference could be observed, neither for BRAF nor NRAS mutation (sup. Fig. 3b,c), but in our cohort of primary melanomas sample, a systematic screening of BRAF^V600E mutation by immunohistochemistry (sup. Fig. 3d) revealed a significant correlation between Tspan8 protein expression and the presence of the BRAF^V6OOE mutation (Fig. 2g). These data emphasize the functional interaction between Tspan8 and BRAF that we previously showed in vitro, suggesting that Tspan8 is a downstream effector of the RAF/MEK/ERK signaling pathway [3].