The PD-1/PD-L1 interaction is crucial for the physiological development of immunological tolerance, which involves monitoring immune system over-reactions that could result in tissue destruction and/or the development of autoimmune diseases, potentially resulting in serious consequences 
. PD-L1 expression can therefore be either constitutive or inducible 
. In the first scenario, PD-L1 is expressed at a certain level on resting lymphocytes, antigen-presenting cells (APCs), as well as in corneal, syncytiotrophoblastic, and Langerhans cells where it contributes to promoting an intratissue balance that orchestrates and regulates inflammatory responses 
. As a result, PD-L1 is able to confer a “privileged immune” status on some tissues, including the placenta, testis, and anterior chamber of the eye, allowing for the tolerization of external antigens without eliciting an inflammatory immune response 
. On the other hand, haematopoietic, endothelial, and epithelial cells are stimulated to produce PD-L1 as a suppression signal during an infectious or inflammatory phase 
. Thus, PD-L1 plays a role in the processes of central and peripheral tolerance, which are the negative selection of self-reactive lymphocytes in primary (central tolerance) and secondary (peripheral tolerance) lymphatic organs, respectively. It also plays a role in the so-called “immune exhaustion” process, which is the progressive decline of effector T-cell function after persistent antigen presentation, a physiological mechanism that prevents tissue destruction in chronic infections, and finally, it is also used by cancer cells to suppress the immune system’s activity in the regulation of the antitumour immune response 
. Due to the latter feature, some cancer types use PD-L1 expression as a strategy to elude immune system cells, which may be related to a worse prognosis. The expression of PD-L1 on tumour cells assessed by immunohistochemistry (IHC) was initially identified as a biomarker for predicting response to treatment with anti-PD-1/anti-PDL1 therapies, and this topic has been extensively studied on different tumour types with conflicting results. Malignant melanoma represents one of the most “immunogenic” neoplasms in existence and this has allowed the testing, development, and application of immunotherapeutic drugs specifically in metastatic melanoma patients.
2. History of PD-L1 in Malignant Melanoma
One of the earliest papers in literature concerning the immunohistochemical expression of PD-L1 was by Yang W. et al. 
, in which the authors evaluated the expression of this protein in nine primary and five metastatic uveal melanomas (UM). In addition to the use of cell lines, flow cytometry, and reverse-transcription polymerase chain reaction (RT-PCR), the authors used the MIH1 e-Bioscience clone of the monoclonal anti PD-L1 antibody. In all the cases they analysed, no PD-L1 expression was detected in situ, but together with the data from the cell lines and RT-PCR, the authors hypothesised that T-cells, by processing IFN-γ at the level of UM liver metastases, promoted the initiation of PD-L1 expression, which in turn reduced T-cell proliferation by hindering the production of interleukin-2 (IL-2). Some time later, Krönig H. et al. 
conducted a study on the expression of PD-1 and its interaction with PD-L1 using 100 peripheral blood samples from stage I-IV MM patients and simultaneously performed immunohistochemical reactions on 37 primary/metastatic melanoma samples, testing Melan-A, PD-L1 and PD-1. All IHC investigations for PD-L1 were performed with polyclonal rabbit antibody ProSci, Poway, CA, USA) and expression was quantified at 400× magnification. In the case of primary melanomas, PD-L1 was analysed at the edges of the invasive front where the signal appeared most intense. Combining data from the peripheral blood mononuclear cells of HLA-A2+ patients and immunohistochemical results, the authors found that, compared to the entire CD8+ T-cell population, PD-1 expression from A2/Melan-A + CD8+ T cells was over-represented in stages III and IV, but although Melan-A + PD1 + T cells were elevated, this did not impact OS, while a positive correlation between PD-1 expression on MM cells and longer OS was described. The data reported by these authors were contrasted with a study by Hino et al. 
in which it was reported that a high PD-L1 expression was an independent prognostic factor for a worse prognosis, although the clone used in the study by Hino was different (clone 27A2; MBL; Nagoya, Japan). On the other hand, the study by Gadiot et al. 
also seemed to go in the same direction as Krönig.
In 2015, an interesting paper by Berghoff A.S. et al. 
addressed the issue of PD-1 and PD-L1 immunoexpression in a sample of MM metastatic to the brain. Among other markers analysed (such as CD3+, CD8+, and CD45 RO), the authors reported PD-L1 expression in 22 samples of the 43 analysed and in 9/22 cases, PD-L1 was observed in >5% of the neoplastic cells. Furthermore, PD-L1 expression was associated with a higher density of PD-1, CD3, and FoxP3 TILs infiltration and these data were the basis for suggesting an implementation of therapeutic regimes in patients with brain metastases from MM.
A paper by Madore J. et al. 
reported an interesting analysis of 139 samples from 58 MM patients (43 primary melanomas and 96 metastatic melanomas) and, when possible, studied individual patient samples longitudinally at various stages of disease, i.e., primary melanoma (PM), loco-regional metastases (LR), and distant metastases (DM). All analysed samples demonstrated a significant heterogeneity for PD-L1 expression both intra- and interpatient, using the cutoff >/=1% for PD-L1 positivity/negativity. Interestingly, when comparing longitudinal samples from each individual patient, there was no significant intrapatient concordance between PD-L1 status in the primary lesion and loco-regional malignancy, nor between a primary melanoma and distant metastases. Finally, there was no concordance of PD-L1 status in LR and DM. This paper clearly demonstrated that it was preferable, if possible, (1) to determine PD-L1 on excisional biopsies instead of incisional/punch biopsies; (2) to use the patient’s most recent metastatic specimen as PM was not a reliable source of predictive PD-L1 immunoexpression status in metastases; (3) to keep in mind the possibility of patients with atypical/outlier PD-L1 expression as they are candidates for a further molecular analysis to elucidate any mechanisms involved. In 2014, Massi D. et al. 
reported on 81 consecutive cases of MM whose PD-L1 expression was studied using the antibody ab58810 polyclonal (Abcam, Cambridge, United Kingdom) and 5H1 monoclonal. The concordance data were very high between the two antibodies used and, in total, 40.3% of the metastatic melanoma samples were positive for PD-L1, compared to 14% of the primary melanomas. Using cell lines and pRT-PCR, it was shown that PD-L1 expression was an independent negative prognostic factor.
The paper by Kakavand H. et al. 
was very interesting, in which the authors studied 93 tumours from 40 patients treated with a BRAF inhibitor (BRAFi, n
= 28) or a combination of BRAF and MEKi (Combi, n
= 12) whose samples were excised before treatment, early during treatment, and at progression. In addition to IHC staining for CD4, CD8, CD-68, LAG-3, and PD-1, staining for PD-L1 was performed with Merck’s monoclonal antibody 22C3, at a 1:1000 dilution. The authors noted that the lesions of patients who were positive for PD-L1 at baseline showed a significant decrease at progression, whereas the opposite was true (lesions negative for PD-L1 at baseline showed a significant increase at progression), regardless of treatment with BRAFi or Combi. Overall, PD-L1 expression was highly correlated with the presence of TILs. In another paper by Kakavand H. et al. 
, the authors tried to understand whether PD-L1 expression in MM cells present in the sentinel lymph node of positive patients could have any impact on their management and potential use of PD-1/PD-L1 inhibitors in the adjuvant setting. The metastasis-positive sentinel lymph nodes of 60 treatment-naive patients were analysed and, in addition to CD3, CD4, CD8, FoxP3, and PD-1, Merck’s monoclonal antibody 22C3, at a 1:1000 dilution, was used. Although the tumour expression of PD-L1 was present in 26/60 cases and did not correlate with outcomes, the authors showed that there was a positive correlation between recurrence-free/overall survival and the number of CD3+, CD4+, and CD8+, along with a negative correlation with the number of PD1+ T-lymphocytes in peritumoural distribution. It was very interesting to note that a certain microenvironment in the LN could predict the patients’ outcome.
Sunshine J.C. et al. 
analysed very clearly the diagnostic performance of five different antibodies used in studies conducted on 34 MM samples from as many patients and broken down as 7 primary melanomas, 1 loco-regional localisation, and 27 metastatic melanomas. They used the clones 5H1, SP142, 28-8 (Dako PharmDx, Santa Clara, CA, United States), 22C3, and SP263, and the results showed a strong correlation between the staining patterns and the use of the different clones. When differences occurred, the spatial heterogeneity of the melanoma tissue section was responsible for the discrepancies rather than the variations in the PD-L1 antibody staining features. A strong correlation between PD-L1 intensity/H-scores and the proportion of PD-L1(+) cells was also found, and their findings further contradicted the idea that chromogenic PD-L1 IHC tests should include an intensity/H-score.
Kaunitz J.G. et al. 
conducted an important study on the expression of immunohistochemistry for PD-L1 on 200 formalin-fixed paraffin-embedded (FFPE) specimens from patients with acral (n
= 16), mucosal (n
= 36), uveal (n
= 103), and chronic sun-damaged (CSD) (n
= 45) melanomas (24 lentigo maligna, 13 “mixed” desmoplastic, and 8 “pure” desmoplastic melanomas) in order to understand whether there was a differential expression of PD-L1 and CD8+ according to different MM histotypes.
The extent of the presence of CD8+ TILs was classified as mild, moderate, or severe, and the association between PD-L1 expression and location was examined, and in 31% of acral melanomas, 44% of mucosal melanomas, 10% of uveal melanomas, and 62% of CSD melanomas, PD-L1 expression was found. The proportion of PD-L1(+) lesions was lower in uveal and greater in CSD melanomas, although PD-L1 expression in the acral and mucosal subtypes was comparable to a precedent paper by the same authors. All subtypes of PD-L1 expression showed a moderate-to-severe CD8+ TIL grade correlation, supporting an adaptive mechanism of expression. The pure desmoplastic subtype of CSD melanomas, which expressed PD-L1 at lower levels than other subtypes, was associated with distinct tumour microenvironments. PD-L1 expression was not correlated with the existence of lymphoid aggregates; however, it was significantly higher in PD-L1(+) cases with spindle-cell shape than in patients with a nested phenotype in those cases.
In this field, numerous attempts have been made to develop standardised immunostaining protocols in order to succeed as much as possible in reducing the differences between the different assays 
In a paper by Ren M et al. 
, the analysis of 78 primary acral melanoma samples made it possible to study the expression characteristics of PD-L1 by correlating them with clinicopathological and survival parameters. The authors demonstrated that the expression of PD-L1 occurred at the tumour–stroma interface in tumour cells and TILs, which was consistent with the main pattern of TIL distribution. A high PD-L1 expression in cancer cells was also linked to the presence of peritumoral TILs. Furthermore, there was a strong correlation between the expression of PD-L1 in TILs. However, there was a lack of association among clinicopathological features and either PD-L1 expression in cancer cells or that in TILs. Cases with PD-L1-positive TILs showed a significantly worse survival in a univariate analysis than those with PD-L1-negative TILs. While PD-L1 expression in cancer cells was not substantially connected with survival in a univariate analysis or a multivariate analysis (p
= 0.354), it was an independent predictor for poor prognosis in a multivariate analysis for TILs.