Melanoma-specific Expression of the Tumor Suppressor Proteins p16 and PTEN is a Favorable Prognostic Factor in Established Melanoma Brain Metastases

Abstract

Objective:

PTEN and p16 frequently undergo (epi)genetic aberrations in melanoma resulting in decreased, or absent, protein levels. We investigated the prognostic significance of these tumor suppressor genes in melanoma brain metastases (MBMs).

Methods:

Immunohistochemical analysis was performed on archived tissue sections from craniotomies. Expression of PTEN and p16 was semiquantitatively scored (0–3 scale) in melanoma cells, glia, TILs, and endothelial cells of tumor-associated vessels and was compared among the different brain tumor cell compartments. Overall survival (OS) analysis was performed according to PTEN and p16 expression in melanoma cells.

Results:

58 patients (median age 56, 37 males) underwent craniotomy for MBMs before February 2014. The OS of patients with decreased, or absent, expression (0, 1+) of PTEN and p16 in melanoma cells was significantly shorter compared to that of patients with high (2+, 3+) expression (median OS 2.40 vs. 10.75 months and 4.1 vs. 8.1 months, respectively; Gehan-Breslow-Wilcoxon test p=0.026 and p=0.037, respectively). PTEN and p16 protein expression was significantly lower in TILs compared to melanoma cells (Mann-Whitney test p=0.023 and p<0.0001, respectively).

Conclusions:

Low/absent protein expression of PTEN/p16 is an adverse prognostic factor in MBMs. Surprisingly, expression of both PTEN and p16 proteins was significantly lower in TILs compared to melanoma cells. Proliferating (p16 absent/low) TILs within the brain with or without an active PI3K-Akt pathway (PTEN absent/low) may represent a favorable host response in MBMs. Thus, treatment of patients with MBMs with CDK4/6 or PI3K inhibitors may result in an unfavorable, bystander off-target effect on host immune response.

INTRODUCTION

A better understanding of the biology of melanoma brain metastases (MBMs) has recently revised the concept that MBMs are a uniformly devastating disease. While there are several clinical and radiographic prognostic factors for patients with established MBMs, only a handful of tumor tissue-based factors assess the overall prognosis and response to systemic treatments following craniotomy. In line with this, we have previously shown that patients who underwent craniotomy for MBMs have the most prolonged survival if their brain tumor specimen shows a high density of tumor-infiltrating lymphocytes (TILs) and low intratumoral hemorrhage.[1] More recently, we showed that patients with a high density of TILs in craniotomy specimens from MBMs trend to have the most significant clinical benefit to treatment with immune checkpoint inhibitors following craniotomy.[2] In this report, we explored the prognostic significance of PTEN and p16 protein expression in craniotomy specimens from patients with MBMs for the following reasons: First, both PTEN and p16 are crucial tumor suppressor genes frequently dysregulated in melanoma via genetic (mutations or gene deletions) or epigenetic (hypermethylation) mechanisms;[3, 4] Second, their expression has been previously associated with TIL density in melanoma;[5, 6] Third, the corresponding pathway can be pharmacologically targeted; Fourth, loss of PTEN protein expression has been associated with earlier development of MBMs.[7]

METHODS

Under institutionally review board-approved protocols, patients who had undergone craniotomy for MBMs at the University of Pittsburgh Medical Center (UPMC) or the University of North Carolina at Chapel Hill (UNC-CH) were considered for this study. Information regarding the demographics, melanoma subtype, overall survival from craniotomy to last follow-up, status at last follow-up (dead or alive), local and systemic treatments prior and following craniotomy were available.[1, 2] Information regarding BRAFV600 status, tumor volume of the craniotomy specimen, and degree of peritumor edema based on the brain MRI that was performed immediately prior to craniotomy (absent, present-low, present-high) from the recently published UNC-CH craniotomy cohort were also available, as we have recently published.[2] We performed single-color immunohistochemistry (IHC) in 5μm-thick sections obtained from formalin-fixed, paraffin-embedded tumor tissues. We carried out tissue IHC for p16 (E6H4, 705–4713; Roche, Tucson, AZ) and PTEN (6H2.1, 04–035 Millipore, Billerica, MA) in the Leica Bond-III fully automated staining system (Leica Microsystems, Inc., Leica Biosystem, IL).[8] Slides were dewaxed in Bond Dewax solution (AR9222, Leica Biosystem, IL) and hydrated in Bond Wash solution (AR9590). Epitope retrieval of p16 protein was performed for 20 min for p16 at 100ºC in Bond Epitope Retrieval Solution 1, pH 6.0 (AR9961), and in Solution 2, pH 9.0 (AR9640) for antigen retrieval for PTEN (30 min). After pretreatment, slides were incubated with primary antibodies against p16 (ready-to-use, 30 min) and PTEN (1:3,000, 4 hours). Single-color detection and hematoxylin counterstain were performed using the Bond Polymer Refine Red Detection kit (DS9390, Leica Biosystem, IL). Both positive and negative controls (no primary antibody) were included for each stain. A neuropathologist (DGT) scored expression of PTEN and p16 protein in melanoma cells and other components of interest (e.g., adjacent glia, TILs, and endothelial cells of tumor-associated blood vessels) using the 0–3 scale where 0 = no expression, 1 = mild (1–25%) expression, 2 = moderate (26–74%) expression, and 3 = extensive (75% or greater) expression. If a particular cell type or component was not present in the tissue, a score of “not applicable” was given. PTEN and p16 protein expression were evaluated across different cells or cellular compartments using the Wilcoxon matched-pairs signed-rank test (paired analysis) and Mann-Whitney test (unpaired analysis). We investigated the correlation between protein expression of various proteins by melanoma cells and previously published data on TIL density [1, 2] using the Kendall rank correlation statistic. Protein expression for PTEN and p16 in melanoma cells were dichotomized by IHC as high expression (2+, 3+) or low/absent expression (0, 1+). We then performed overall survival (OS) univariate analysis to assess the prognostic significance of PTEN and p16 proteins in MBMs using the Kaplan-Meier method. Given the significantly higher number of early events (deaths) in patients with MBMs, we used the Gehan-Breslow-Wilcoxon test and performed statistical analysis using Prism 8 (GraphPad Software, version 8.3.1, San Diego CA). Cox regression for survival analysis was performed using important variables from univariate analysis.

RESULTS

The combined UPMC/UNC-CH cohort includes 58 patients (median age 56, range 20–82; 37 males; 44 cutaneous, 12 unknown primary, one mucosal, one unknown histology) who underwent craniotomy between August 1995 and February 2014. Of note, 17 patients underwent craniotomy after December 2010. Information about the tumor volume at craniotomy and degree of peritumoral edema was available for 28 patients. Of the 28 patients who underwent BRAFV600E mutation testing using sequencing or an immunohistochemical assay, ten had a BRAFV600E mutation. Eighteen patients did not receive any systemic therapy before craniotomy, and four had received radiation therapy to the brain. The most frequent form of systemic treatment before craniotomy was chemotherapy (N=19), followed by adjuvant high-dose interferon therapy (N=15). Only three patients had received targeted therapy before craniotomy. Following a craniotomy, 33 patients did not receive any systemic therapy. Immunotherapies (ipilimumab and high dose bolus IL-2) were administered to only five patients, whereas targeted therapy was administered to only four patients following craniotomy. At a median follow-up of 5.5 months (range 0.1–80 months), only six patients were alive.

Expression of both PTEN and p16 proteins in melanoma cells was absent or low in more than half of brain tumor specimens analyzed (28/52 37/56, respectively). Figure 1, panels, a and b, shows representative images of tissue sections corresponding to MBMs with absent PTEN and p16 expression. Neither expression of PTEN nor that of p16 protein correlate with the degree of peritumoral edema seen on brain MRI immediately prior to craniotomy. There were not significant differences in the tumor of craniotomy specimens according to expression of PTEN and p16 in melanoma cells (low/absent vs. high). However, expression of PTEN was significantly higher in melanoma cells compared to that in TILs (Mann-Whitney test p=0.023), but significantly lower compared to endothelial cells (p=0.0001, Figure 1c). Expression of PTEN in melanoma cells did not correlate with TIL density. Patients with absent or low (0, 1+) expression of PTEN in melanoma cells had shorter OS compared to patients with high expression of PTEN (2+, 3+; median OS was 2.40 vs. 10.75 months; Gehan-Breslow-Wilcoxon test p=0.026, Figure 1e). Expression of p16 was significantly higher in melanoma cells compared to that in TILs (p<0.0001), glia (p=0.0075), and endothelial cells (p=0.0036, Figure 1d). Expression of p16 protein in melanoma cells moderately but significantly correlated with TIL density (Kendal tau 0.26, 2-sided p=0.027). Patients with absent or low expression of p16 protein in melanoma cells had a significantly shorter OS compared to patients with high expression of p16 (4.1 vs. 8.1 months, p=0.037, Figure 1f). Patients with absent or low expression of both PTEN and p16 proteins in melanoma cells had a significantly shorter OS compared to patients with high expression of both PTEN and p16 in melanoma cells (1.9 vs. 12.8 months, p=0.032, Figure 1g). However, a Cox regression model that included PTEN/p16 protein expression (both low/absent vs. other) by melanoma cells and adjusted for other covariates (age, and sex) did not show any significant impact in OS.

Figure 1.

Expression of PTEN (upper panel) and p16 proteins (lower panel) in melanoma brain metastases. (a), (b). Digital images (20 X magnification) corresponding to representative tissue sections obtained from craniotomy specimens that were immunohistochemically stained with an antibody against PTEN (a) and p16 [(b); bond polymer refine red detection, red). Counterstain is with hematoxylin. Examples of melanoma cells that do not stain for PTEN or p16 (thick black arrows). Please note expression of PTEN and p16 in reactive glia (purple thin arrow) and vascular endothelial cells (blue thick arrow). (c), (d) Boxplots showing semiquantitative expression of PTEN (c) and p16 (d) in different cell compartments (melanoma cells, glia, endothelial, TILs). P values correspond to unpaired analysis using the Mann–Whitney test between melanoma cells and other brain compartments. Numbers indicate number of observations. (e), (f), (g) Overall survival analysis (Kaplan–Meier method) of patients who underwent craniotomy according to PTEN (e), p16 (f) status (high 2+, 3+; low/absent 0, 1+) in melanoma cells. OS, overall survival; TIL, tumor-infiltrating lymphocyte clusters; P value <0.001***, 0.001–0.01**, and 0.01–0.05*.

DISCUSSION

Although patients with MBMs have the worst prognosis among all sites of metastasis, as reflected in the most recent revision of the AJCC staging system, they exhibit considerable differences in overall prognosis. There are no other tumor tissue-based prognostic factors for patients with established brain metastases, beyond host immune response and intratumoral hemorrhage. For example, although hotspot mutations in BRAFV600 and NRASQ61 can predict local and distant brain failure, they are not associated with OS.[9]

Our craniotomy cohort is unique given that patients were largely treated before the era of immune checkpoint inhibitors and targeted therapies; therefore, the impact of post-craniotomy systemic treatments on OS was minimal. Also, only 4/58 patients had received radiation therapy to the brain, a potential confounding factor that may affect histopathology of melanoma brain tumors. In this cohort, we showed that low/absent expression of p16 and PTEN proteins in melanoma cells, alone or combined, constitute an adverse prognostic factor in patients who underwent craniotomy for MBMs. Similar to a recent report, we confirmed that low or absent expression of PTEN was not associated with the degree of host immune response.[10] Given that astrocyte-conditioned medium may stimulate PI3K-Akt signaling pathway activation in melanoma cells and increase invasiveness in vitro,[11] we postulate that low/absent expression of PTEN in MBMs is associated with worse OS by means of more aggressive melanoma biology and not by immune evasion. In contrast, p16 expression in melanoma cells is more directly associated with host immune response, which we have previously shown to be an adverse prognostic factor in MBMs.

In addition to analyzing the expression of PTEN and p16 proteins in melanoma cells from brain tumor specimens, we also assessed the expression of these two proteins in TILs. PTEN signaling regulates various immune cell subsets, including both T-cell activation and T regulatory function.[12] Regarding p16, its expression in T cells is associated with a non-proliferative state, anergy, senescence, and immunosuppression.[13] Although we did not perform additional immunostains to specify which tumor-infiltrating immune cell subsets express PTEN or p16, we were surprised to see that PTEN and p16 protein expression in TILs was significantly lower compared to that of the same-patient’s melanoma cells (Wilcoxon matched-pairs signed-rank test, p=0.0019 and p<0.0001, respectively), suggesting that TILs are proliferating and have a more active PI3K-Akt-PTEN pathway as opposed to melanoma cells themselves. The low/absent expression of PTEN and p16 proteins in TILs may have significant treatment implications for patients with MBMs since the pathways activated within melanoma cells —and are actually associated with adverse prognosis— maybe even more active in TILs. In that case, active proliferating TILs within the brain with or without an active PI3K-Akt-PTEN pathway could be a favorable host immune response and an unfavorable, bystander off-target effect if pharmacologically targeted. Our data may in part explain the lack of antitumor response seen in patients with metastatic brain tumors treated with abemacyclib, a brain-penetrant CDK4/6 inhibitor.[14] Similar studies with buparlisib, a brain-penetrant pan-class I PI3K inhibitor, are underway (clinicaltrials.gov NCT02452294).

In summary, p16 and PTEN protein expression in metastatic melanoma cells in the brain are promising prognostic factors in patients who undergo craniotomy for MBMs in univariate analysis. We propose that semiquantitative assessment of TILs and immunostains for PTEN and p16 would be useful tumor tissue-based biomarkers to incorporate into standard neuropathology synoptic assessments that are currently limited to diagnostic confirmation of MBMs. Analysis of tumor specimens from prospective clinical trials in patients with active MBMs may provide insights as to whether PTEN and p16 expression in melanoma cells is also a predictor of response to systemic therapies. However, the expression of none of these proteins in melanoma cells was associated with any radiographic features, such as peritumor edema and size of craniotomy tumors. Therefore, these potentially interesting tumor tissue biomarkers cannot be proposed as surrogates to commonly used radiographic fiindings in patients with MBMs.

Abbreviations:

MBMs

melanoma brain metastases

TILs

tumor-infiltrating lymphocytes

IHC

immunohistochemistry

OS

overall survival