Correlation of MET and PD-L1 Expression in Malignant Melanoma

The proto-oncogene MET, the hepatocyte growth factor (HGF) receptor, is a transmembrane receptor tyrosine kinase (RTK) with a prominent role in tumor metastasis and resistance to anti-cancer therapies. Melanoma demonstrates relatively frequent MET aberrations, including MET gene amplification. Concurrently, programmed death-ligand 1 (PD-L1), with its ability to evade anti-tumor immune responses, has emerged as a prominent therapeutic target in melanoma and other malignancies and its expression is used as a predictive biomarker of response to immunotherapy. We performed immunohistochemistry analysis of MET and PD-L1 in 18 human melanoma cell lines derived from both primary and metastatic lesions, and in a human melanoma tissue microarray containing one hundreds melanocytic lesions, including primary cutaneous melanomas, primary mucosal melanomas, metastatic melanomas and benign melanocytic nevi as controls. After color deconvolution, each core was segmented to isolate staining and calculate the percentage of positive cells. Overall, MET expression was higher in tumors with increased PD-L1 expression. Moreover, a robust correlation between MET and PD-L1 expression was found in samples from metastatic melanoma and not in primary cutaneous or mucosal melanoma. These data suggest that relative expression levels of these proteins in combination is a marker of advanced disease and testing for expression of these markers should be considered in patients with melanoma.


Background
Melanoma is the leading cause of death in patients with cutaneous malignancies and its incidence has been rapidly rising over the past 30 years. According to the American Cancer Society, 91,270 new cases of melanoma and 9,320 deaths related to melanoma were estimated in 2018 [1]; moreover, the incidence of melanoma is estimated to increase by 7.7% in 2019. While the ve-year survival rate is up to 98% if melanoma is diagnosed at an early stage, this rate drops to about 20% if the disease is diagnosed at a late stage and has metastasized to distant organs. While a better understanding of the molecular basis of this cancer and its microenvironment has resulted in relatively successful novel therapeutic options, resistance ensues and there is still need for a better way to select for patients who will more likely bene t from these novel therapies.
The proto-oncogene MET, the hepatocyte growth factor (HGF) receptor, is a transmembrane receptor tyrosine kinase (RTK) with a prominent role in tumor metastasis and resistance to anti-cancer therapies [2], and dysregulation of the HGF/MET signaling pathway has been demonstrated in a wide range of malignancies, including malignant melanoma [3,4]. A recent large whole genome sequencing (WGS) analysis of melanomas has demonstrated relatively frequent MET aberrations, including MET gene ampli cation, single nucleotide variations/deletions, and structural variants [5].
The pleiotropic effect of the HGF/MET signaling pathway include also a role in modulation of the immune response, including involvement in dendritic cell function [6] and neutrophilic antitumoral response [7]. This function has been postulated to be involved in the potential acquisition of resistance to immunotherapy treatments [6].
Programmed death-ligand 1(PD-L1) is a transmembrane protein encoded by the CD274 gene, located on chromosome 9, and is expressed by antigen presenting cells (APCs) and tumor cells [8]. PD-L1, with its ability to evade the anti-tumor immune response, has emerged as a prominent therapeutic target in melanoma and other malignancies and its expression is used as a predictive biomarker of response to immunotherapy [9]. PD-L1 expression can be induced either by cytokines (INF-gamma) or by activation of an oncogene. For example mutations in receptor tyrosine kinase pathways, such as epidermal growth factor receptor (EGFR), have been shown to induce PD-L1 expression in lung tumors [10] and overexpression of PD-L1 on cancer cells can block anti-tumor immunity, resulting in immune escape [11], which can be overcome by PD-1/PD-L1 inhibition, restoring tumor-speci c T-cell immunity.
MET expression has been shown to promote upregulation of PD-L1 in renal cancer cells [12] and expression of PD-L1 and PD-L2 are upregulated in MET-ampli ed gastric and lung tumor cells [13].
However, the relationship between MET and PD-L1 expression in human malignant melanoma is not well characterized.
In the present study, we surveyed by immunohistochemistry MET and PD-L1 expression in melanoma cell lines and in a human tissue microarray of cutaneous melanomas, mucosal melanoma, and metastatic melanomas, with the goal of analyzing the expression pattern and to explore a possible correlation between expression of these two proteins involved in tumor progression and immune evasion.  Table 1.

Methods
To prepare cell pellet blocks, cells were grown in T-75 asks to near con uence then washed 3 times with PBS and xed with 10% neutral-buffered formalin (NBF) in asks overnight. After xation, cells were gently scraped and transferred to a conical tube and centrifuged to ~300 x g for 5 min at room temperature (RT). Formalin supernatant was then removed and pellets were washed with PBS. Finally, cell pellets were re-suspended in 80% ethanol and para n embedded in cell blocks. A tissue microarray (TMA), with 1.5 cm cores for each cell line, was then derived from the cell blocks. Two TMA slides were stained for MET and two were stained for PD-L1. The average percentage of positively stained cells and FastRed mean intensity were calculated.

Tissue Microarray and Patient Characteristics
TMA slides with 100 cores were purchased from Biomax (Cat. # ME1004g; US Biomax, Inc., Derwood, MD). Available clinicopathological characteristics are summarized in Supplemental Table 2. The mean age of the patients from this TMA was 50 years (range 0.5 -84 years). The study included 58% (58) males and 42% (42) females. Staging (TNM and clinical staging) was only provided for 48 patients which include 42 cases of primary cutaneous melanoma and 6 cases of mucosal melanoma. Overall, the TMA included 42 cases with cutaneous malignant melanoma, 20 mucosal melanomas (including malignant melanoma from vulva, rectum, stomach and esophagus), 21 cases were obtained from metastatic sites including lymph nodes and 17 cases were benign melanocytic nevi.
Two TMA slides were stained for MET and two were stained for PD-L1. The average percentage of positively stained cells and FastRed mean intensity were calculated.

Immunohistochemistry (IHC)
Unstained TMA sections (4 µm) were de-para nized and rehydrated using standard methods. For antigen retrieval, slides were incubated in 6.0 pH buffer (Reveal Decloaking reagent, Biocare Medical, Concord, CA) in a steamer for 30 min at 95-98°C, followed by a 20 min cooldown period. Subsequent steps were automated using an immunohistochemical staining platform (Intellipath, Biocare). Endogenous peroxidase activity was quenched by slide immersion in 3% hydrogen peroxide solution (Peroxidazed, Biocare) for 10 min followed by TBST rinse. A serum-free blocking solution (Background Sniper, Biocare Medical, Concord, CA) was placed on sections for 10 min. Blocking solution was removed and slides were incubated in primary antibody diluted in 10% blocking solution/90% TBST for 60 minutes at room temperature. Rabbit monoclonal anti-MET (clone D1C2 XP(R)(Cell Signaling, Denver, MA;1:50), followed with a TBST rinse and detection with Novocastra Novolink Polymer Kit (Leica Microsystems Inc., Buffalo Grove, IL) using the manufacturer's speci cations. Slides then proceeded with TBST rinse and detection with diaminobenzidine (DAB) (Covance, Dedham, MA). Slides were incubated for 5 min followed by TBS rinse then counterstained with CAT Hematoxylin (Biocare, Concord, CA) for 5 min.
Finally, slides were dehydrated and coverslipped.
Rabbit monoclonal anti-PD-L1 (clone 28-8, Cell Marque, Rocklin, CA, 1:200) was followed by a TBST rinse and biotinylated anti-rabbit (Vector Labs, Burlingame, CA,1:200) was applied for 30 minutes. The slides were again rinsed with TBST and 4+ Streptavidin-Alkaline Phosphatase label (4+SA-AP) (Biocare Medical, Concord, CA, RTU) was applied for 30 minutes at room temperature. Slides proceeded to a TBST rinse and detection with WARP Red Chromagen (Biocare Medical, Concord, CA) according to manufacturers' speci cations. Following detection, slides were rinsed well in running tap water and counterstained for 1 minute in CAT Hematoxylin (1:2) (Biocare Medical, Concord, CA). Tap water rinse and 2 minutes in PureView PH Blue (Cancer Diagnostics, Durham, NC), 5 minute tap water rinse and then slides were air dried. When completely dry, slides were dipped in xylene and coverslipped with Permount mounting medium (Fisher, Fair Lawn, NJ).
Image analysis TMA slides were scanned using an Aperio scanner, with a 40X objective. The high resolution images were analyzed using QuPath [14] version 0.1.2. The work ow consisted of (i) color deconvolution, (ii) identifying the TMA cores, (iii) segmenting the tissue region in each core, (iv) isolating nuclear and cytoplasmic regions of interest, (v) estimating the abundance of the deconvolved red component in each cell and nally (vi) calculating the percentage of positive cells in each core. 3-color deconvolution was performed using vectors calibrated visually on the image data, following the procedure outlined in the software documentation. All images were deconvolved using the same stain vectors. Steps (iii) to (v) were performed using the default algorithms in QuPath. For step (vi), each cell was considered positive if the red component intensity was above a threshold, calculated independently for each TMA slide as the average between the 5th and 95th percentile of red intensities across the slide. All algorithms and parameters for the analysis in QuPath were recorded in a script for repeatability (see Supplementary Material).

Statistical analysis
Statistical analysis was primarily descriptive. Correlation between percentage of MET-and PD-L1-positive cells was calculated using Pearson's correlation coe cient (r) and its 95 % con dence interval calculated by the 2.5 and 97.5 percentiles of 500 bootstrap resamplings.

Expression of MET and PD-L1 in melanoma cell lines
We analyzed the expression of MET and PD-L1 in eighteen human melanoma cell lines in a tissue microarray derived from seven primary and eleven metastatic lesions. Immunohistochemical staining for MET and PD-L1 revealed a wide range of expressions in these cell lines. Expression of PD-L1 was higher in metastatic melanoma cell lines compared to primary melanoma cell lines ( Figure 1A). Median (interquartile range) PD-L1 was 32 (29, 50) for metastatic and 24 (19,27) for primary cell lines (Wilcoxon rank sum test p=0.02), and the highest value was in a metastatic cell line. Expression of MET was more similar, with a median of 24 for both primary and metastatic, although the interquartile range was slightly higher for metastatic (17, 41) than primary (11,29) (Wilcoxon rank sum test p=0.44), and the highest value of MET was in a metastatic cell line.
MET and PD-L1 values were correlated among primary melanoma cell lines, with Pearson's r = 0.73, with a notable wide con dence interval. For metastatic cell lines, the Pearson's correlation score was higher, with a r = 0.89 ( Figure 1B) Expression of MET and PD-L1 in a human melanoma tissue microarray One-hundred melanocytic lesions were evaluated for expression of MET and PD-L1 in a human TMA with benign nevi (17 patients), cutaneous melanomas (42 patients), mucosal melanomas (20 patients), and metastatic melanomas (21 patients). The characteristics of these patients are summarized in Supplemental Table 2. Brie y, the median age was 26 for nevi , 52 years for cutaneous melanoma, 57 years for mucosal melanoma and 51 years for metastatic melanoma (range 0. 5  We then calculated the correlation between MET and PDL1 expression in each category of melanocytic lesions. As shown in Figure 3, there was modest correlation for benign nevi and cutaneous primary melanoma, with Pearson's correlation coe cients of 0.46 and 0.49, respectively, and no correlation in mucosal melanoma (r = -0.02). In contrast, MET and PD-L1 expressions were highly correlated in metastatic melanoma (r = 0.74).

Discussion
Malignant melanoma remains a major cause of death among patients with cutaneous malignancies, despite the introductions of novel therapeutic approaches such as immunotherapy. The PD-1/PD-L1 axis [15] have emerged as a major immune checkpoint target explored successfully for immunotherapy. Resistance to drugs targeting these proteins eventually emerges [16] through several mechanisms, some of which have been characterized. For example, activation of canonic oncogenic signaling pathways, such as those driven by receptor tyrosine kinases, are well known. Thus, a viable strategy to overcome resistance is to combine immunotherapy with conventional targeted therapies, such as inhibitors of receptor tyrosine kinases. Assessment of expression of these targets in tumor tissue sections is an important strategy to improve patient selection and increase e cacy of potential drug combinations.
Several lines of evidence point out a prominent role of the MET/HGF axis in tumor progression and resistance to therapy of several malignancies, including malignant melanoma. For example, in a case of acral melanoma with KIT mutation, targeting MET with a selective inhibitor successfully overcame resistance to KIT inhibition, as con rmed also in cell line studies [17]. Another study has established that MAPK pathway inhibition following BRAF inhibitor treatment induced rapid increases in MET and GAB1 expression [18] and MET ampli cation was also observed to co-exist with BRAF hotspot mutations [5]. Moreover, MET appears to have a role in the regulation of immunity, as demonstrated by the key role of its ligand, HGF, in the regulation of autoimmunity and in ammation [19]. More recently a subpopulation of CD8 positive cytotoxic T-cell has been found to express MET, further linking its pathway to a role in tumor immunity [20]. Moreover, HGF has been linked to increased expression of PD-L1 in dendritic cells and CTLA-4 in T-cells, with a role in the induction of immune tolerance [6,21].
In the era of personalized medicine, the prediction of patient's drug response has become an important prerequisite for administration of targeted therapies [22] [23]. Previous reports demonstrated PD-L1 expression to be associated with better prognosis in patients with lung, colorectal, breast, and malignant melanoma cancers [24][25][26][27][28][29]. Our study demonstrates that both MET and PD-L1 expressions are increased in metastatic melanocytic lesions when compared with primary melanoma and benign melanocytic nevi in patients' tissue samples, suggesting that in advanced lesion both proteins are potential biomarkers and viable targets for selective combined inhibition.
Currently, no de nitive cut-off for PD-L1 positivity is universally recognized, in part because of variability between different antibody clones and staining platform used. In our current study we used clone 28.8 (from Cell Marque). This antibody was developed in conjunction with the rst immune checkpoint inhibitor evaluation in patients with cancer, namely nivolumab, a human monoclonal antibody targeting PD-1 binding to PD-L1 and PD-L2. Measurable positive effects on overall survival were observed in patients with melanoma, NSCLC, renal cell carcinoma and ovarian cancer. This is a "complementary" diagnostic, so it has not been included as a mandatory test in their regulatory labeling, and many studies have validated its use in PD-L1 immunohistochemistry assays.
In our study we took advantage of software analysis to avoid a qualitative or semiquantitative assessment of percentage of positive cells. High resolution images were used in the assessment with a recently described open source software (QuPath) for digital pathology and image analysis.
The methodology used in our assessment, which includes color deconvolution of high resolution images, allows for a sensitive assessment of each individual staining and is able to detect low levels of signal that may be otherwise below the limit of detection by the human eye in a qualitative or semiquantitative assessment. Hence, he use of deconvolution allows for a more sensitive and quantitative assessment of the immunohistochemistry staining, in conjunction with the use of digital imaging analysis.
Our results in the established cell lines show a good correlation between MET and PD-L1 expression, meaning that degree of MET and PD-L1 expressions follow a similar trend, in both primary and metastatic cell lines. In contrast, our data on tumor biopsies from patients show a relatively modest correlation between these two proteins in primary lesions, but a much higher correlation in metastatic disease. These nding further highlights discrepancies we frequently see experimentally between cell lines and patient's samples, stressing the importance to always correlate in vitro ndings with actual clinical specimens. However, one limitation of our study is the lack of outcomes and follow-up clinical information which was not available for the current TMA.
Authors's contributions A.G., K.Y.S , S.D., T.P. and R.S. contributed to the design and implementation of the research, and to the analysis of the results. A.G. and K.Y.S. wrote the main manuscript. K.Y.S., S.D. carried out most of the experiments. All authors read and approved the nal manuscript.  MET and PD-L1 expression levels in benign nevi, cutaneous melanoma, metastatic melanoma and mucosal melanoma.