Role of Elevated PHIP Copy Number as a Prognostic and Progression Marker for Cutaneous Melanoma

This article is featured in Highlights of This Issue, p. 4057

Melanoma represents the fifth most common malignancy in the United States (1). As the clinical behavior of melanoma is unpredictable, the development of prognostic markers has represented an important area of research. Histologic prognostic factors are the mainstay of prognostic assessment for primary cutaneous melanoma (2). Although many factors may play a role in refining melanoma prognosis, tumor thickness and ulceration are the only factors currently incorporated within the melanoma staging classification for primary melanoma (3).

Molecular factors represent the next frontier in prognostic factor research. The development of molecular prognostic factors for melanoma has been hampered largely by the lack of sufficient validation of promising markers in independent tissue sets. In addition, the role of a putative marker in melanoma progression has rarely been explored in the context of the same patient. Finally, given the identification of dominant mutational drivers of melanoma, information regarding the role of a given marker in the context of the known mutational landscape of melanoma would be important (4). As a result, no molecular markers are routinely utilized in the prognostic assessment of patients with melanoma, and none is incorporated into its staging classification.

Our previous studies identified both a functional and a biomarker role for pleckstrin homology domain-interacting protein (PHIP) in melanoma progression, and as a molecular prognostic factor for melanoma. PHIP was identified as the top gene overexpressed in melanoma metastases when compared with primary tumors using cDNA microarray analysis (5). Subsequent studies showed an independent prognostic role for PHIP in primary melanoma (6, 7). Functional studies demonstrated that increasing PHIP expression enhanced the distant metastatic potential of melanoma (6, 7). Here, we aimed to confirm the prognostic role of PHIP in primary melanoma, to identify the molecular subtypes of melanoma in which PHIP is enriched, and to determine whether PHIP is selected for during the progression from primary to metastatic melanoma.

We developed three nonoverlapping patient cohorts to assess PHIP's biomarker role: (i) cohort No. 1—unstained primary melanoma tissue sections obtained from 204 stage I/II patients in the German Cancer Consortium (see Table 1, for characteristics); (ii) cohort No. 2—tissue microarrays (TMA) from a previously described cohort of 130 patients with stage III melanoma in the MD Anderson Cancer Center (8); and (iii) cohort No. 3—37 matched primary and metastatic melanomas obtained from the same patient (N = 15) who developed both regional lymph node and distant metastasis, identified from the CPMC Center for Melanoma Research and Treatment archives (see Supplementary Table S1, for characteristics). In cohort No. 3, 21 samples were from seven patients with triple-matched primary melanoma, regional lymph node and distant metastasis, and 16 samples were from eight patients where the primary melanoma was matched either with a regional lymph node or distant metastasis. The mean follow-up of cohort No. 1 was 4.58 years. Information regarding mitotic rate was not available in cohort No. 1. These analyses were complemented by analysis of PHIP expression levels obtained from The Cancer Genome Atlas (TCGA; ref. 4), available from 66 primary tumors and 264 metastases. The investigations described here were conducted in accordance with recognized ethical guidelines (i.e., Declaration of Helsinki), were approved by the Sutter Health Institutional Review Board, and informed consent was obtained from each subject.

FISH for PHIP copy number was performed as previously described (6, 7) using bacterial artificial chromosome (BAC) clones RP11-767O1 and CTD-2297E14 to detect the PHIP locus and clones RP11-26M18 and RP11-136K2 to detect 6q11.1 and 6p11.1, respectively (interpreted as chromosome 6 centromere). FISH for PTEN copy number was performed using BAC clones RP11-79A15 and RP11-813O3. For a chromosome 10-specific probe, BAC clone RP11-96F8 was used to detect 10p11.1 (all BAC clones were obtained from the Children's Hospital Oakland Research Institute). The quality and mapping of all probes were verified by hybridization to normal metaphase spreads in combination with a commercially available centromeric probe for chromosomes 6 and 10 (Thermo Fisher Scientific) before tissue analysis. Z-stacked images were acquired using a Zeiss Axio Image Z2 microscope controlled by AxioVision software (Zeiss). At least 30 nuclei from each case were evaluated, and the signals were interpreted according to guidelines described previously (9). Signals from BAC clones detecting 6q11.1, 6p11.1, and 10p11.1 were interpreted as the centromeric signal for chromosome 6 and 10, respectively.

The impact of elevated PHIP copy number on melanoma progression was analyzed in several ways. In cohort No. 1, consistent with our previous analysis (6), elevated PHIP copy number was measured by recording the percentage of cells harboring three or more copies of the PHIP locus. The cut-point utilized to define high PHIP copy number was that which maximized the average of sensitivity and specificity for predicting DSS (determined to be greater than or equal to 19%). The same cut-point was uniformly applied to analyses of DMFS and ulceration status. Subsequently, a previously identified cut-point for PHIP copy number (greater than or equal to 16%; ref. 6) was applied to the analysis of DMFS, DSS, and ulceration in this cohort. The association between elevated PHIP copy number and ulceration status was determined using logistic regression analysis. The association between elevated PHIP copy number and survival outcomes (i.e., DMFS and DSS) was determined using Kaplan–Meier analysis, and univariate and multivariate Cox regression analyses. In cohort No. 2, the enrichment of PHIP in various molecular subtypes of melanoma was assessed using the T test or the ANOVA. In cohort No. 3, for the analysis of seven triple-matched tissue specimens, the significance of monotonically increasing PHIP copy number in melanoma progression was assessed using the Friedman two-way analysis of variance by ranks test. The significance of increasing PHIP copy number between primary and metastatic melanoma in the eight double-matched specimens was assessed using the binomial sign test. As a control, identical analyses were performed for PTEN copy number in the same 15-patient matched cohort. Except for the directional analyses just described, all P values reported are two-sided.

In this study, we aimed to perform a comprehensive analysis of the role of PHIP levels in distinct, sequential stages of melanoma progression, and to confirm its role as a prognostic marker for primary melanoma. Initially, we aimed to replicate the prognostic role of elevated PHIP copy number (assessed as percentage of cells harboring three or more copies of PHIP) in primary melanoma in cohort No. 1 (Fig 1A, B). We assessed the potential association between PHIP copy number and survival using univariate Cox regression. Increasing PHIP FISH scores (as a continuous variable, using the entire scale) were significantly predictive of distant metastasis-free survival (DMFS, P < 0.03) and disease-specific survival (DSS, P = 0.025). By Kaplan–Meier analysis, high PHIP scores (assessed as a dichotomous variable) were also significantly predictive of DMFS (P = 0.0017, Fig. 2A) and DSS (P = 0.0002, Fig. 2B). In addition, we assessed the association between PHIP FISH scores and both tumor thickness and ulceration status (6). Increasing PHIP copy number (using the entire scale) was associated separately with increased tumor thickness (P = 0.0001) and with a higher prevalence of ulceration (P < 0.0001) by univariate logistic regression analysis. High PHIP FISH scores (assessed as a dichotomous variable) were significantly associated with increased prevalence of ulceration (P < 0.00005).

We then assessed the independent impact of elevated PHIP copy number on survival associated with melanoma using multivariate stepwise Cox regression analysis with backward elimination. We included in the model the following covariates included in the legacy prognostic analyses performed by the American Joint Committee on Cancer (AJCC) melanoma committee (10): tumor thickness (assessed by T category), ulceration status, Clark level, patient age, patient gender, and tumor site. In the analysis of DMFS, elevated PHIP copy number (as a dichotomous variable) emerged as significantly predictive of survival (P = 0.025), along with tumor thickness and gender (Table 2). In the analysis of DSS, PHIP FISH scores were also significantly predictive of survival (P = 0.01), along with tumor thickness (Table 2). Thus, elevated PHIP copy number was independently predictive of both DMFS and DSS, replicating our original analysis (6).

In addition, given that the cut-point identified in this analysis (19%) was very close to that identified in our original analysis (16%; ref. 6), we examined whether use of the previously-identified cut-point would be predictive of outcome in cohort No. 1. High PHIP scores using the original cut-point were also significantly predictive of DMFS (P = 0.03), DSS (P = 0.0069), and ulceration status (P < 0.00005) by univariate analysis.

Next, we assessed the role of PHIP in melanoma nodal metastases in cohort No. 2. We initially assessed the potential prognostic role of elevated PHIP copy number in this cohort. There was no significant association between PHIP copy number (assessed as a continuous variable) in metastatic tumors and survival in stage III melanoma. The dataset from this cohort contained information regarding three important molecular markers in melanoma: BRAF and NRAS mutation status, and PTEN expression level (8). Our analysis identified a significant association between elevated PHIP copy number and NRAS mutation status in melanoma. The mean percentage of melanoma cells harboring ≥3 copies of PHIP was 45.7% in NRAS-mutant melanomas, compared with 19.3% in BRAF-mutant melanomas, and 15.4% in BRAF/NRAS-wild-type melanomas (P < 0.00005, ANOVA; Fig. 3A). Separately, our analysis indicated that PHIP copy number was significantly higher in melanomas with intact PTEN protein expression versus those with PTEN loss. Specifically, the mean percentage of melanoma cells harboring ≥3 copies of PHIP increased from 13.5% in melanomas with PTEN loss to 22.6% in PTEN-expressing melanomas (P = 0.004, T test; Fig. 3B). In addition, in BRAF-mutant melanomas, there was a trend toward increased PHIP copy number in PTEN-expressing melanomas versus those with PTEN loss (P = 0.086, T test; Fig. 3C). Finally, cases with a mean PHIP copy number of ≥3 were present in each molecular subtype of melanoma investigated, including in wild-type BRAF/wild-type NRAS/PTEN-expressing melanomas. Taken together, these results indicate that PHIP levels are enriched in NRAS-mutant and PTEN-expressing melanoma, and that PHIP copy-number elevations are observed in both BRAF-mutant and BRAF wild-type/NRAS wild-type/PTEN-expressing melanomas.

Subsequently, we assessed the role of PHIP as a progression marker in melanoma. We performed FISH analysis of PHIP copy number on a matched cohort of 15 patients from whom the primary tumor and at least one metastatic tumor were available (cohort No. 3; Supplementary Table S1). Among the seven triple-matched patients (in whom the primary tumor, a lymph node metastasis, and a distant metastasis were available for analysis), PHIP copy number increased monotonically in the transition from primary to lymph node to distant metastasis in four patients (Fig. 4). In two patients, the lymph node metastasis had a lower copy number than the primary tumor, but the distant metastasis had a higher copy number than both the primary tumor and the lymph node metastasis. In one patient, the lymph node metastasis had a slightly higher copy number than the distant metastasis, but both the lymph node and distant metastasis had higher copy numbers than the primary tumor. There was a significant increase in PHIP copy number in the transition from primary tumor to lymph node to distant metastasis (P = 0.008, Friedman test). The mean percentage of cells with elevated PHIP copy number increased monotonically from 13.5% in the primary tumor to 21.6% in the lymph node metastases, and to 55.5% in the distant metastases. In the matched-pair analysis, for each of the eight cases, PHIP copy number was higher in the metastatic lesion when compared with the corresponding primary tumor (P = 0.0039, binomial sign test; Fig. 4). To determine whether this systematic change was unique to PHIP, we performed an identical analysis of PTEN copy number on the same matched-specimen cohort. There was no significant change in PTEN copy number in the transition from primary tumors to either lymph node or distant metastases or both (Supplementary Fig. S1). By comparison, the mean percentage of cells with elevated PTEN copy number was 9.1% in the primary tumor, 12.7% in the lymph node metastases, and 11.8% in the distant metastases (P = 0.19, Friedman test applied to the same seven matched triplets).

Finally, we examined several of these associations by analyzing PHIP transcript levels in the TCGA cutaneous melanoma cohort (4). Increasing PHIP expression (assessed as a continuous variable) in the primary tumor was associated with significantly reduced relapse-free (RFS, P = 0.02) and overall survival (OS, P = 0.018) by univariate Cox regression analysis. By multivariate analysis, PHIP expression levels remained significantly predictive of RFS (P < 0.05) and OS (P = 0.015), even with the inclusion of AJCC stage. Similar to the FISH analysis, there was no significant association between increasing PHIP expression levels in metastatic tumors and survival. In addition, mean PHIP expression levels were significantly (4.3-fold) higher in NRAS-mutant melanomas when compared with NRAS-wild-type cases (P = 0.02, Mann–Whitney test; Supplementary Fig. S2A). Finally, mean PHIP transcript levels were significantly (3.7-fold) higher in melanoma metastases when compared with primary tumors (P < 0.00005; Supplementary Fig. S2B).

We have performed a rigorous analysis to elucidate the role of PHIP copy-number elevation in sequential stages of melanoma progression, and as a prognostic marker for primary melanoma. We analyzed the role of PHIP copy number in three independent patient cohorts from different centers. In cohort No. 1, we show that elevated PHIP copy number is an independent prognostic marker for primary cutaneous melanoma in a cohort of patients derived from a distinct population in Germany, and confirm a significant association with ulceration status, an important prognostic factor incorporated into the AJCC melanoma staging system. In cohort No. 2, we identify distinct molecular subtypes of melanoma in which PHIP is enriched. In cohort No. 3, we determine that the degree of PHIP copy-number elevation is uniformly increased in the progression from primary to metastatic melanoma. Several of the key findings from these three cohorts were replicated using TCGA cohort analysis.

The independent prognostic role of PHIP levels in primary cutaneous melanoma has now been shown using FISH analysis in two distinct cohorts, replicated at the protein level using immunohistochemical analysis (7), and replicated at the RNA level using TCGA cohort analysis. In addition, our prior studies demonstrated a significant correlation between PHIP copy number and expression in primary melanoma (6). There are few (if any) molecular markers whose prognostic impact has been demonstrated both in distinct cohorts of patients with melanoma and using several distinct platforms. These studies advance PHIP as an important molecular marker of melanoma outcome, although additional studies in larger cohorts will be required in order for it be routinely used as a prognostic factor.

Beyond its independent prognostic role, our studies of cohort No. 1 confirmed a powerful association between PHIP copy number and ulceration. Although the prognostic significance of ulceration is well appreciated, the biologic basis for ulceration development in melanoma has been poorly understood. Our previous studies showed that PHIP mediates some of its effects on melanoma metastasis by virtue of activation of the IGF1R–PI3K pathway. PHIP is a potent activator of AKT (7), which drives both glucose metabolism and angiogenesis (11, 12). Our studies in melanoma cell lines showed that PHIP gene silencing results in decreased glycolysis (including reduced LDH expression) in vitro and decreased angiogenesis in vivo (6). These results are consistent with a model in which PHIP activation, in part through elevated copy number, promotes glycolysis and angiogenesis, resulting in increased risk of development of ulceration and distant metastasis. Intriguingly, this model suggests that a common signaling pathway may underlie the development of and explain the biological basis of ulceration, a key prognostic marker in stage I to III melanoma, and LDH levels, a prognostic marker for stage IV disease.

In addition, our analyses of cohort No. 2 identified the molecular subtypes of melanoma in which PHIP copy-number elevations are present, and in which they are specifically enriched. Elevated PHIP copy number was present in each molecular subtype of melanoma examined, including melanomas with BRAF mutation, NRAS mutation, as well as in melanomas without any alterations in these markers. Moreover, we found that PHIP copy number was significantly higher in melanomas with NRAS mutation or with intact PTEN expression. The observation that PHIP participates in the IGF1R–PI3K signal transduction pathway in which PTEN also operates provides a rational basis for these results, suggesting that either molecular alteration (PHIP copy-number gain or PTEN loss) may be sufficient to activate this pathway in melanoma. Given the importance of PTEN loss specifically in BRAF-mutant melanomas (4), our analysis also identified a trend toward higher PHIP copy number in PTEN-expressing BRAF-mutant tumors, introducing PHIP as a novel potential driver of this molecular subtype of melanoma. An intriguing and unexpected finding was the enrichment of PHIP copy number in NRAS-mutant melanoma. Given that NRAS mutations are thought to activate both the MAPK and PI3K pathways, our results suggest the intriguing possibility of novel biochemical functions of PHIP beyond its participation in the IGF1R–PI3K axis. However, our results are consistent with the observation that PTEN loss is essentially mutually exclusive with NRAS mutation. In addition, they suggest PHIP targeting as a potential strategy for the treatment of NRAS-mutant melanoma, which has represented an important therapeutic challenge given the modest activity of MEK inhibitors (13). Recent studies describing the development of a small molecular inhibitor of PHIP (14) demonstrate its druggability, thereby highlighting its role as a potential therapeutic target.

Our studies of cohort No. 3 analyzed the role of PHIP in the progression cascade of melanoma by comparing copy-number changes in matched primary versus metastatic tumors from the same patient. Remarkably, the matched-specimen analysis revealed elevated PHIP copy number in at least one metastatic lesion in every case analyzed when compared with its corresponding primary tumor. In many instances, the copy number in the metastasis was extremely high, with three or more copies of PHIP observed in greater than 50% of cells in 40% of the metastases, including 56% of distant metastases examined. In addition, in the majority of cases where we had access to the primary tumor, and both lymph node and distant metastases, there was a monotonic increase in PHIP copy-number elevation in the transition from primary melanoma to lymph node metastasis, and then to distant metastasis. These results strongly suggest that elevated PHIP copy number is selected for in melanoma progression, supporting our original cDNA microarray analysis in which PHIP was identified, and which was further supported by TCGA cohort analysis, showing significantly increased PHIP expression in metastatic versus (unmatched) primary melanomas. We are unaware of any similar studies analyzing the role of a given molecular marker in successive stages of melanoma progression using tissues from the same patient. Similarly, a recent analysis of colon cancer assessed the somatic variants in hypermutable DNA regions in a cohort of 19 patients with primary colon cancer and matched lymph node and distant metastasis, enabling a conclusion that two-thirds of distant metastases have a profile distinct from that of the lymph node metastasis (15). Our analysis underscores the potency of these matched-tissue comparisons to determine the role of a given gene in cancer progression within individual patients.

In conclusion, our studies to date have shown that increased PHIP levels are predictive of melanoma distant metastasis, promote distant metastasis in mouse models of melanoma, and are enriched in melanoma metastases. The presence of elevated PHIP copy number in distant melanoma metastases as shown in this study indicates that this molecular target is present and enriched in the setting of advanced melanoma. Taken together with functional studies demonstrating an important role for PHIP in promoting melanoma progression, these studies describe an unprecedented role for PHIP in melanoma progression, suggesting it as a promising target for therapy.

J.E. Gershenwald is a consultant/advisory board member for Merck, Syndax, and Castle Biosciences. M.A. Davies reports receiving commercial research grants Bristol-Myers Squibb; Roche/Genentech; GSK; Sanofi-Aventis; Oncothyreon; AstraZeneca; and Merck, and is a consultant/advisory board member for Bristol-Myers Squibb; Novartis; GSK; Nanostring; Roche/Genentech; Syndax; and Vaccinex. M. Kashani-Sabet reports receiving commercial research grants from Merck, speakers bureau honoraria from Cepheid, and holds ownership interest (including patents) in Melanoma Diagnostics. No potential conflicts of interest were disclosed by the other authors.

Conception and design: V. Bezrookove, J.R. Miller III, D. Schadendorf, M. Kashani-Sabet

Development of methodology: V. Bezrookove, M. Nosrati, J.R. Miller III

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): V. Bezrookove, M. Nosrati, D. De Semir, A.A. Dar, E. Vosoughi, E. Vaquero, A. Sucker, A.J. Lazar, J.E. Gershenwald, M.A. Davies, D. Schadendorf

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. Nosrati, J.R. Miller III, D. De Semir, A.A. Dar, E. Vosoughi, J.E. Gershenwald, M. Kashani-Sabet

Writing, review, and/or revision of the manuscript: V. Bezrookove, M. Nosrati, J.R. Miller III, D. De Semir, A.A. Dar, E. Vosoughi, A.J. Lazar, J.E. Gershenwald, M.A. Davies, D. Schadendorf, M. Kashani-Sabet

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M. Nosrati, E. Vaquero, A. Sucker

Study supervision: M. Nosrati, M. Kashani-Sabet

This work was supported by a grant from the NIH (R01CA175768) to M. Kashani-Sabet. The collection of the tissue set at MD Anderson was supported by the following: The Melanoma Specialized Program of Research Excellence of the University of Texas MD Anderson Cancer Center (P50 CA093459); and philanthropic contributions from the Dr. Miriam and Sheldon G. Adelson Medical Research Foundation, the AIM at Melanoma Foundation, and the MD Anderson Melanoma Moon Shot Program.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.