Prediction of long term survival outcomes for melanoma patients is not only critical clinically for patient management but also highly valuable for patient risk stratification in evaluation of new melanoma therapies in clinical trials. For patients with primary melanoma (PM), the most widely used current prognostic factors include Breslow tumor thickness, Clark level of invasion for thin melanomas, mitotic index, and ulceration status. For metastatic melanoma (MM) patients, there are no well established prognostic systems, although the extent of metastasis (local regional versus visceral) and serum lactate dehydrogenase (LDH) levels (Balch et al., 2009) were known to influence patients’ survival. In general, the prognostication achieved using these methods are imprecise, especially for patients with lymph node and distant metastasis.
Alpha 1 antichymotrypsin, a serine protease inhibitor (serpin), was found to be aberrantly expressed in tissues of MM during gene expression profiling experiments (Zhou et al., 2005), and in serum of xenografted mice carrying human MM cell line (Kondo and Ohsawa, 1982). However, its clinical significance in melanoma has not been previously investigated. A 68 kDa secreted protein that normally is involved in modulating activities of proteases such as chymase, ACT is unique among serpins harboring DNA binding activities (Baumann et al., 1991; Naidoo et al., 1995), although the exact molecular and cellular consequences of this property remain to be elucidated. The current high-throughput immunohistochemistry study was performed using clinically annotated tissue biopsies to evaluate the clinical correlation and prognostic significance of ACT protein in patients with melanoma.
In western blot analysis (Appendix S1), ACT antibody detected a specific protein band with the expected molecular weight of ACT (68 kDa) in a MM cell line, KZ-28, but not in cultured normal human epidermal melanocytes (Figure 1A). In immunohistochemistry analysis, ACT was found to be specifically expressed in the cytoplasm and the intercellular space immediately adjacent to the melanoma cells (Figure 1B–D), with little expression in the nucleus. To examine if the levels of ACT protein are correlated with clinical phenotypes of melanoma patients, we performed immunohistochemistry using tissue microarrays to quantify expression of ACT in clinically annotated tissue biopsies of benign nevi (BN) (N = 21), dysplastic nevi (DN) (N = 55), melanoma in situ (MIS) (N = 11), invasive PM (N = 165), and MM (N = 51). The demographic and histopathological parameters of these patients are summarized in Tables S1 and S2. In the semi-quantitative analysis of the expression intensities using a scoring system previously reported (Tang et al., 2006), the ACT protein expression was found to be absent or weak in BN, DN and MIS. In contrast, in samples of invasive PM a significant increase in ACT protein expression was observed (P = 0, X2 = 67.924, Chi square test). A further increase of ACT expression was observed in MM biopsies compared with PM (P < 0.043, X2 = 4.109, Chi square test) (Figure 2). These observations suggest that ACT protein may be involved in promoting the malignant properties of melanoma cells during these two critical steps of melanoma progression, namely, tissue invasion and metastasis, although the molecular mechanism involved remains to be investigated.
To test if there were correlations between ACT expression levels and the clinical pathological variables and patient survival for melanoma, separate analyses were performed for patients with primary and MM due to different survival determinants for these two patient populations. In patients with PM no correlation was found between ACT expression levels and all parameters evaluated, including tumor thickness or ulceration status (Table S1). Survival analyzes using Kaplan–Meier and multivariable Cox regression methods did not demonstrate significant prognostic value of ACT expression in patients with PM (P = 0.65, log rank test, Figure 3A and Table S3).
Among the 51 patients with MM (Table S2), 39 patients had Stage III melanoma, and 12 had Stage IV melanoma. For Stage III patients, the ACT expression dramatically affected the patient’s survival in Kaplan–Meier Analysis (Figure 3B, P = 0.049, log rank test). Specifically, the ACT-low Stage III patients survived 43.4 months compared with 18.5 months for the ACT-high Stage III patients (P = 0.048, Student t-test (Table S4)). To test if the prognostic value is independent of other clinical pathological factors including lymph node stages, multivariable Cox regression analysis was performed. It was demonstrated that ACT’s prognostic value (HR = 9.63, 95% confidence interval: 1.38–67.0, P = 0.022) was independent of other variables, such as the patients’ nodal stages (N1, N2 or N3), serum LDH levels or corresponding primary tumors’ ulceration status and Breslow thickness (Table S1). Similarly, for Stage IV patients, Kaplan–Meier analysis demonstrated a prognostic significance of ACT (P = 0.0137, log rank test), with the ACT-low group surviving 35.51 months, 3.86 times longer than the ACT high group (which survived 9.20 month, P = 0.039, Student t-test). However, the small sample size (N = 12) did not permit vigorous multivariable Cox regression analysis to evaluate if this was an independent prognostic factor.
The combination of gene expression profiling experiments and high throughput immunohistochemistry analysis has resulted in a large number of prognostic histological markers for PM patients (Reviewed in (Gould Rothberg et al., 2009)). However, these markers have not been demonstrated to influence survival in patients with MM. As most deaths resulting from melanoma are caused by metastasis, achieving accurate prognostication for patients with MM has important clinical applications, including risk stratifications in disease management and in selecting patients for clinical trials. To date, the most widely used risk stratification system is the AJCC staging, which takes into account several factors, such as extent of nodal involvement (N1, N2, and N3), extent of distant metastasis (M1a, M1b, and M1c), and the status of serum LDH levels (Balch et al., 2009). In addition, several reports suggested that high risk features in the MM patients’ corresponding PM samples (such as advanced Breslow tumor thickness and ulceration status) also influence survival of Stage III MM patients (Homsi et al., 2005). Furthermore, several investigational serum markers have been reported, including MIA (Bosserhoff, 2005), MMP1 (Nikkola et al., 2005), and S100B in the adjuvant setting of high-risk melanoma patients (Tarhini et al., 2009). The only histological marker of prognostic significance in MM previously reported is β1 integrin (Nikkola et al., 2004), which was shown to influence patients’ survival in a small cohort of patients that did not involve multivariable analysis. Thus, it is difficult to evaluate if β1 integrin is independent of other prognostic factors. Therefore, ACT reported here, to our knowledge, is the only independent histological prognostic marker for MM demonstrated in a multivariable analysis.
In summary, our results demonstrated that ACT is an independent histological prognostic marker for patients with MM, especially those with Stage III melanoma. Further investigations are warranted to fully evaluate clinical application potential of ACT and to understand its biological actions in melanoma cells.