The tumour immune landscape and its implications in cutaneous melanoma

The field of tumour immunology has rapidly advanced in the last decade, leading to the advent of effective immunotherapies for patients with advanced cancers. This highlights the critical role of the immune system in determining tumour development and outcome. The tumour immune microenvironment (TIME) is highly heterogeneous, and the interactions between tumours and the immune system are vastly complex. Studying immune cell function in the TIME will provide an improved understanding of the mechanisms underpinning these interactions. This review examines the role of immune cell populations in the TIME based on their phenotype, function and localisation, as well as contextualising their position in the dynamic relationship between tumours and the immune system. We discuss the function of immune cell populations, examine their impact on patient outcome and highlight gaps in current understanding of their roles in the TIME, both in cancers in general and specifically in melanoma. Studying the TIME by evaluating both pro‐tumour and anti‐tumour effects may elucidate the conditions which lead to tumour growth and metastasis or immune‐mediated tumour regression. Moreover, an in‐depth understanding of these conditions could contribute to improved prognostication, more effective use of current immunotherapies and guide the development of novel treatment strategies and therapies.


| INTRODUC TI ON
In recent years, it has become generally accepted that the immune system influences the development, progression and outcome of many cancers (Fridman et al., 2012;Koebel et al., 2007). Cells of both the innate and adaptive immune systems have been associated with tumours and found to localise intratumourally, at the tumour margins, and in the peritumoural stroma (Fridman et al., 2012;Galon et al., 2006). The immune system can perform anti-tumour activity, leading to eradication of cancer cells, and likely destroys some tumours before they are ever clinically apparent. The importance of this continuous immune surveillance becomes evident in patients who suffer from the absence or dysregulation of these immune mechanisms (Dunn et al., 2002;Grivennikov et al., 2010). As part of adaptive immunity, T cells and B cells can recognise antigens expressed by tumours, coordinate their specific targeting and cause tumour cell death. Innate immune cells work in tandem with the adaptive immune system and use a number of mechanisms to kill tumour cells, including phagocytosis, antibody-dependent cell-mediated cytotoxicity (ADCC) and the release of cytokines including IFN-γ. Cells involved in the anti-tumour response generally display an inflammatory phenotype (Thorsson et al., 2018).
However, the immune system can also act to promote the growth and development of the tumour. This can be through the immunosuppressive activity of many tumour-associated immune cells, promotion of angiogenesis and cell proliferation, and through promotion of an anti-inflammatory response.
An immune cell's functional response to a tumour can be modulated by the tumour itself. The tumour microenvironment consists of not only the tumour cells, but also all of the tissue surrounding the tumour, including fibroblasts and immune cells associated with the tumour. Several factors, including the cytokine milieu and genetic context, modify the phenotype and function of cells within the tumour microenvironment so that it becomes more conducive to tumour growth. This extends to the immune cells, which may begin as anti-tumour but take on a more pro-tumour function as the tumour develops. When considering the combined effect of all of the above functions, it is evident that the interaction between cancer and the immune system is dynamic and complex (Fridman et al., 2012).
When assessing the role of immune cells in cancer, it is important to consider the phenotype, function and localisation of different types of immune cells. All of these factors contribute to shaping the host immune response and can ultimately impact cancer-specific survival in many common malignancies, such as cutaneous melanoma. Although the mechanisms remain incompletely understood, immune responses to melanoma can result in rare cases of spontaneous regression (Botella-Estrada & Kutzner, 2015;Bulkley Gregory et al., 1975) and can be modified by systemic immunotherapies to improve outcomes in patients with advanced disease (Luke et al., 2017). A deeper understanding of immune cells and their interactions in the tumour microenvironment is required to expand the clinical benefits of therapies that modulate the immune system to a greater proportion of patients with cancer.

| Melanoma and the immune infiltrate
Melanoma has the highest prevalence of somatic mutations of any major cancer type (Alexandrov et al., 2013;Chalmers et al., 2017), with a mutational signature closely associated with UV radiation (Hayward et al., 2017). It has been implied that this high mutational burden gives rise to the generation of a high number of neoantigens, thus allowing for the generation of immune responses specific to melanoma. Melanoma is considered to be highly immunogenic (Schumacher & Schreiber, 2015), with the immune system playing a prominent role in its development and progression. Some of this immunogenicity has been attributed to recognition of antigens expressed by melanocytes (Faure et al., 2018).
One of the first clues that melanoma outcome is directly affected by the activity of the immune system was the observation of spontaneous melanoma regression (Bulkley Gregory et al., 1975). Recent work has identified T-cell (CD4 + ) and dendritic cell subsets (CD123 + pDCs) as diffusely infiltrating spontaneously regressing melanomas (Botella-Estrada & Kutzner, 2015). Over the years, the immune system has been implicated in melanoma progression and outcome, as pathological assessment frequently identifies tumour-infiltrating lymphocytes (TILs) in primary and metastatic melanomas.
In recent years, the use of TILs as a prognostic marker in melanoma has become the subject of increasing interest. TILs in melanoma are usually graded by quantifying their density within the tumour, as well as observing their pattern of distribution. Multiple studies have found positive correlations between TIL grade and clinicopathological characteristics, including Clark level, SLN status, Breslow thickness, mitotic rate, ulceration and satellitosis. A high TIL grade has also been correlated with a lower risk of melanoma recurrence and improved patient survival (Azimi et al., 2012;Erdag et al., 2012;Taylor et al., 2007;Weiss et al., 2016). However, differences between the methodology of these studies make their findings difficult to compare. Most methods will incorporate a measure of TIL density and distribution (Wilmott et al., 2012); however, some quantify TIL density alone. Further complicating this is the fact that TIL distribution cannot always be stratified into definite categories, and evidence suggests that the predictive capability of TILs is higher in thick melanomas than in thin melanomas (Azimi et al., 2012).
Other studies find less clear links between TILs and survival. One of these studies in nodal metastatic melanoma found that the absence of TILs correlated with poorer distant metastasis-free survival and melanoma-specific survival, but not with disease-free survival (Murali et al., 2011). Another study in primary melanoma found that in a cohort of 476 patients, 6 patients with TILs all survived beyond 5 years, but the absence of TILs was not significantly linked to outcome (Barnhill Raymond et al., 1998). In primary melanomas, the presence of TILs has even been associated with an increased risk of nodal metastasis (Ma et al., 2012). All of these studies reported presence or absence of TILs alone, rather than describing TIL density or distribution.
An intense TIL infiltrate in primary and metastatic melanoma is a positive prognostic indicator, yet a high degree of TIL infiltration does not always guarantee a good clinical outcome. This may be due to the high heterogeneity of the immune infiltrate in cancer and the broad repertoire of pro-tumour and anti-tumour immune functions occurring within the tumour microenvironment. Examining the composition of TILs may therefore provide a deeper insight into how different types of immune responses relate to clinical outcomes in melanoma ( Figure 1) (Weiss et al., 2016). Most work in this area has focused on metastatic melanoma, and it is only recently that studies have begun to investigate the immune composition of primary melanoma in detail.

| T cells T cells comprise the majority of immune cells infiltrating tumours
and have been characterised in a wide variety of cancers (Fridman F I G U R E 1 Immune infiltrates in primary cutaneous melanoma. This invasive melanoma demonstrates brisk tumour-infiltrating lymphocytes. The diversity of immune cells associated with the same tumour can be appreciated using multiplex immunohistochemistry and image analysis. Although tumour-infiltrating CD8 + T cells have been shown to be positively correlated with improved clinical outcomes In melanoma, the importance of the relative composition of all immune cells and their functional phenotypes is yet to be elucidated  Grivennikov et al., 2010). T-cell infiltration into primary tumours has been correlated with higher survival rates in melanoma and in carcinomas of the lung, breast (Her2-positive and triple-negative subtypes), liver, colorectum, ovary, endometrium and bladder (Azimi et al., 2012;Flecken et al., 2014;Galon et al., 2006;de Jong et al., 2009;Liakou et al., 2007;Pagès et al., 2005;Reynders & Ruysscher, 2016;Stanton & Disis, 2016;Zhang et al., 2003). Tumourinfiltrating T cells are comprised of diverse subsets (Chevrier et al., 2017), some of which have opposing effects on the immune response and cancer survival (summarised in Table 1). The presence of CD8 + T cells in the tumour microenvironment has been correlated with improved survival in most of the above cancers, suggesting that they play a protective role against cancer (Erdag et al., 2012;Flecken et al., 2014;Ling et al., 2014). Since CD8 + T-cell effector function is dependent on cell-cell contact, the activity of CD8 + T cells is impacted by their location in relation to tumour cells. Multiple studies have found the highest populations of CD8 + T cells can be found at the margins of solid tumours, but they also comprise a significant proportion of tumour-infiltrating lymphocytes (TILs) Tumeh et al., 2014). Current evidence suggests that the presence of tumour-infiltrating CD8 + T cells is associated with improved survival, which corresponds with better access to the tumour (Galon et al., 2006;Ling et al., 2014). Yet, studies assessing the effector function of tumour-infiltrating CD8 + T cells in hepatocellular carcinoma have found that they elicit a weaker response than non-infiltrating CD8 + T cells (Flecken et al., 2014). This observation suggests that the localisation and density of CD8 + T cells can be related to their activation and effector function, and may vary between cancer types.

| CD8 + T cells
In addition to the circulating effector CD8 + T-cell population, tissue-resident CD8 + T cells are also highly significant due to their potential to provide long-lasting anti-tumour immunity  and their high expression of immune checkpoint receptors such as PD-1 and CTLA-4, which allows them to be reactivated using immunotherapy (Edwards et al., 2019). These tumour-resident T cells are usually characterised by the expression of CD69, which is believed to be an activation marker; and CD103, an integrin which binds to E-cadherin expressed by epithelial and other cells (Cepek et al., 1994;Woon et al., 2016). CD103 + tumour-resident memory CD8 + T cells have been shown to be positively correlated with survival in a wide range of solid tumours of different types and stages (Djenidi et al., 2015;Ling et al., 2007;Webb et al., 2014;Workel et al., 2016). These cells have also been associated with increased TILs and are thought to play an important role as early immune effectors in the tumour microenvironment.

CD8 + T cells in melanoma
CD8 + T cells have frequently been found to be one of the best indicators of survival in melanoma (Erdag et al., 2012;Kakavand et al., 2015;Kaufman et al., 2014). Furthermore, CD8 + T cells in stage III melanoma display high CD69 expression, indicating that they are highly activated (Jacquelot et al., 2016). Recent work has focused on the role of tumour-resident CD69 + CD103 + CD8 + T cells in melanoma (Edwards et al., 2018). In mice, CD103 + CD8 + tumour-resident T cells were found to remain resident and kill melanocytes following immune challenge with melanoma, confirming the anti-tumour role of these tumour-resident T cells (Malik et al., 2017) and more recent studies suggest that CD69 + CD103 + tumour-resident CD8 + T cells could have an important role in maintaining immune equilibrium with melanoma to prevent tumour escape . In humans, tumour-resident CD103 + CD8 + T cells have been associated with melanoma-specific survival and improved responses to anti-PD-1 therapy. Notably, one study has found that CD103 + CD8 + T cells have a better prognostic value than CD8 + T cells alone (Edwards et al., 2018). Another study identifies an effector memory CD8 + T-cell population, phenotypically characterised as EOMES + CD69 + CD45RO + , as comprising a significantly higher proportion of CD8 + T cells in patients who respond to combined anti-CTLA-4 and anti-PD-1 immunotherapy, but not in patients responding to anti-PD-1 therapy (Gide et al., 2019).
Evidently, tumour-resident CD8 + T cells are a vital component of the immune response to melanoma, regardless even of the treatment context.

| Helper CD4 + T cells
Helper CD4 + T cells are also activated following antigen presentation, but in contrast to CD8 + T cells, they are restricted to recognising antigens presented by class II MHC molecules present on professional antigen-presenting cells such as DCs, monocytes and B cells. They are known to proliferate upon recognition of tumourassociated antigens and have been located in the immune infiltrate of melanoma, clear cell renal cell carcinoma and other tumours (Veatch et al., 2018;Zhang et al., 2015). Multiple functionally distinct subtypes of helper CD4 + T cells (Th) exist, including Th1, Th2 and Th17. Due to a lack of surface immune biomarkers distinguishing helper T cell subtypes in tissues, these cells are usually characterised by the transcription factors T-bet, GATA-3 or RORγt for Th1, Th2 and Th17, respectively, or by the cytokines they produce. Th2 can be particularly difficult to detect in cancer tissues as its transcription factor, GATA-3, is expressed by multiple tumour types (Miettinen et al., 2014 (Galon et al., 2006) and the expression of Th1 genes (Tosolini et al., 2011) to be associated with a better prognosis, while boosted Th1 function has been associated with reduced tumour burden in leukaemic cutaneous T-cell lymphoma (Guenova et al., 2013). Th2 cells, which can not only regulate Th1 responses but also activate regulatory macrophages and B cells, are more involved in extracellular immunity. Usually, either a Th1 or Th2 immune response will predominate, although plasticity between these two subtypes has been observed. Studies that identify Th1 and Th2 by either their cytokine profile or expression of transcription factors have generally found that the Th2 response is more dominant in cancers, including melanoma and colorectal cancer (Halim et al., 2017). One possible explanation for the dominance of Th2 responses is that tumours and tumour-associated fibroblasts release cytokines which condition APCs to promote Th2 cells (Bellone et al., 1999;Tatsumi et al., 2002). Media from pancreatic cancer cell lines has been shown to induce a Th2 cytokine pattern in PBMCs in vivo, and condition DCs to activate Th2 polarised CD4 + T cells (Bellone et al., 1999;De Monte et al., 2011). Further studies in pancreatic cancer found that a high Th2/Th1 ratio was associated with reduced survival (De Monte et al., 2011). Th2 identified in the blood of colorectal cancer and melanoma patients were found to produce the highest levels of the immunosuppressive cytokine IL-10 and the Th1-suppressive cytokine IL-4 of the Th subsets studied (Halim et al., 2017).
Th17 cells release the cytokine IL-17 and are considered to be pro-inflammatory. Studies in both mice and in human ovarian cancer found that Th17 cells were likely to be recruiting CD8 + T cells and NK cells to the tumour site. Furthermore, the presence of Th17 cytokines was found to correlate with both patient survival and the presence of Th1 cytokines, and multiple cancers were found to show similar Th17 cytokine profiles (Kryczek et al., 2009;Martin-Orozco et al., 2009).

Helper CD4 + T cells in melanoma
Th1 has generally been considered the main anti-tumour CD4 + T cell subset, and this has to some extent been observed in melanoma.
Studies of Th1 and Th2 cytokine gene expression have identified high Th1 cytokine numbers in regressing primary melanomas, although Th2 cytokines were also present in these tumours implying that polarisation was not absolute (Lowes et al., 1997). Furthermore, higher expression of Th2 than Th1 cytokine genes was identified in melanoma lymph node metastases, indicating that there may be some association between melanoma progression and Th subtype (Grotz et al., 2015). Studies of melanoma in mice have found that Th17 cells actually had increased anti-tumour activity, mediated by the production of IFN-γ (Muranski et al., 2008). Characterising CD4 + T cells by CD4 positivity alone has been less demonstrative, with a study in mice finding CD4 + T cells to drive tumour regression (Bellavance et al., 2011), whereas studies in human stage III melanoma have correlated CD4 + T cells with a poorer outcome (Jacquelot et al., 2016). This lack of clarity could be due to the fact that CD4 is expressed by all helper T cell subsets and by Tregs, thus grouping together all of these cell populations in findings reported by these studies. Considering their role in the suppression of CD8 + T cell activity, it F I G U R E 2 Overview of known interactions between immune cells in the context of the tumour microenvironment. Cells marked with * indicate populations which have been characterised in mice, but not in humans would be expected that Tregs negatively impact patient survival.

| Regulatory T cells
Indeed, regulatory T cells have been found to reduce patient survival in multiple different cancer types including ovarian, oesophageal, and hepatocellular carcinomas (Curiel et al., 2004;Fu et al., 2007;Kono et al., 2006). In hepatocellular carcinoma, Tregs have also been demonstrated to reduce tumour-infiltrating CD8 + T cell proliferation, activation and function (Flecken et al., 2014;Fu et al., 2007).
Yet, in colorectal cancer Treg numbers at the tumour margins and centre were found to positively correlate with patient survival, more strongly than even CD8 + T cells (Haydn et al., 2014). Importantly, colorectal cancer develops in the context of the intestinal microbiome, and it has been proposed that the regulatory T cells detected within colorectal cancer are actually functioning to generate immune tolerance to the microbiome. In this way, they are acting to reduce inflammation and tissue damage related to the microbiome, rather than suppressing tumour-reactive CD8 + T cells per se.
However, as well as being used to identify Tregs, the transcription factor FoxP3 can also be expressed by activated helper CD4 + T cells . As such, this result could indicate CD4 + T cell activation in response to the tumour rather than a Treg-mediated response to the microbiome. Similar results were observed in patients with gastric cancer, with a high numbers of FoxP3 + Tregs at the tumour margins correlating with increased survival . However, it is unclear whether this was associated with inflammation caused by Helicobacter pylori infection, or a Tregmediated tolerogenic response to the gut microbiota. Due to their ability to act in both a cell/cell-dependent manner and through the release of immunosuppressive cytokines, Treg localisation in relation to tumours is not as strongly predictive of patient outcome as it is for CD8 + T cells. Nevertheless, FoxP3 is upregulated in Tregs infiltrating hepatocellular carcinoma and suggests that the tumour microenvironment does impact regulatory T-cell characterization and function, and thus, Treg localisation should not be disregarded (Fu et al., 2007). Helper CD4 + T cells are highly heterogenous in their function, yet the impact of the Th subtypes on cancer development is not yet well understood.

Regulatory T cells in melanoma
Studies of the role of Tregs in melanoma are conflicting. FoxP3 + Tregs have been associated with poorer prognosis and a reduced number of CD8 + and CD4 + T cells, as expected considering their role in immune suppression (Jacquelot et al., 2016;Ma et al., 2012).
Conversely, high Treg populations have also been associated with a high number of CD8 + T cells, and yet other studies find that Tregs do not significantly impact survival (Spranger et al., 2013;Weiss et al., 2016). This disparity could in part be due to the impact of skin microbiota on inflammation in cancer, which has been implicated in the development of tumours in other barrier tissues. While this role for the skin microbiota remains to be seen, Langerhans cells-a skin-resident dendritic cell subset-are known to induce the proliferation of highly activated Tregs residing in human skin to promote tolerance to skin microbiota (Seneschal et al., 2012). Furthermore, a population of resident memory Tregs was recently identified as having a significantly different TCR repertoire to circulating Tregs, further supporting the possibility that Tregs in melanoma could be suppressing inflammation related to an immune reaction to the skin microbiota (Sanchez Rodriguez et al., 2014). If this were the case, it could explain why the exact role of Tregs in melanoma has not yet been elucidated.

| B cells
Although B cells are only thought to comprise a small percentage of the immune infiltrate in cancer, they have a broad range of functions in the tumour microenvironment and various impacts on tumour outcome (Chevrier et al., 2017). Foremost of these is their ability to produce antibodies reactive to tumour antigens. Even though B cells can produce tumour-reactive antibodies from distant sites in the body, B cells producing anti-tumour antibodies have long been known to localise within the tumour microenvironment (Punt et al., 1994). Studies in breast cancer have found that B cells infiltrating tumours often cluster in tertiary lymphoid structures (TLSs).
TLSs form close to tumours, often at the margin or within the tumour stroma. Their structure and composition can resemble that of secondary lymphoid organs (Kroeger et al., 2016), generally comprising of a cluster of B cells with features of a germinal centre surrounded by CD4 + and CD8 + T cells, interspersed with CD21 + follicular dendritic cells and located in the vicinity of high endothelial venules (HEVs). TLSs are associated with favourable patient outcomes in the majority of cancers in which they are found, including ovarian cancer (Kroeger et al., 2016), colorectal cancer (Posch et al., 2018) and pancreatic cancer (Hiraoka et al., 2015). While the structure and composition of TLSs varies between different cancers, this is more dependent on the tissue in which the TLS is located rather than whether the tumour is a primary or a metastatic tumour (Sautès- Fridman et al., 2019). B cells in TLSs have been found to be clonal, and to proliferate in response to the tumour (Cipponi et al., 2012).
Both of these findings suggest that B cells represent part of an ongoing immune response to tumour antigens; however, the mechanisms underpinning this have not yet been adequately investigated.
Furthermore, since the B-cell response from TLSs resembles a memory-driven B-cell response originating from the secondary lymphoid organs, it is possible that memory B cells drive antibody-mediated responses to tumours. However, the role of B cell memory in cancer has not yet been well investigated (Coronella et al., 2002;Nzula et al., 2003). These studies also identified B cells in close proximity to T cells within tumours. B cells are known to present tumour antigens to CD4 + T cells, generating tumour-specific helper CD4 + T cells and further expanding the anti-tumour immune response (Lapointe et al., 2003). Through the release of antibodies, B cells also contribute to the cytotoxic response to tumours. Antibody opsonisation of tumour cells can recruit macrophages, NK cells and neutrophils to directly kill tumour cells in ADCC. Macrophages binding to antibodies opsonising tumour antigens are also able to phagocytose tumour cells, although this function is affected by the size of the tumour (Overdijk et al., 2015). All of the above functions indicate an association between B cell infiltration and better tumour control, a phenomenon which has been observed in medullary breast cancer (Nzula et al., 2003).
However, B cells have also been found to promote tumour progression. B cell functions that are normally part of a robust immune response can actually be harmful in the context of the tumour microenvironment. For example, lymphotoxin produced by B cells is known to have tumour-promoting effects (Ammirante et al., 2010), and studies in mice have suggested that IgM opsonisation of tumour cells blocks the access of CD8 + T cells to the tumour, limiting T cell-mediated cytotoxicity against tumours (Manson, 1991).
However, many pro-tumour B cell functions have been attributed particularly to B cells with either a resting or a regulatory phenotype.
In mice, resting B cells have been found to inhibit CD8 + T cell effector function promoted by dendritic cell (DC) vaccination against tumour antigens, although this was reversible upon Treg depletion and B-cell binding to CD40L (Watt et al., 2007). Regulatory B cells (Bregs) are characterised as CD19 + CD69 high CD25 high in mice. Bregs use a range of mechanisms to suppress anti-tumour immunity and promote metastasis. Specifically, IL-10 and TGF-β suppress the activity of T cells and NK cells, and TGF-β promotes the conversion of resting helper CD4 + T cells to CD25 + FoxP3 + regulatory T cells (Olkhanud et al., 2011). Furthermore, CD154 blockade was found to limit Breg proliferation and IL-10 and TGF-β production, suggesting a two-way positive feedback-type interaction between Bregs and tumour cells in which the promotion of Breg activity by the tumour augments Breg contribution to tumour progression (Shao et al., 2014). In human, hepatocellular carcinoma Bregs generally localise to the tumour margin, where they are able to interact with carcinoma cells directly through the interaction of CD40 and CD154 (Shao et al., 2014). However, the study of Bregs in humans is fairly limited due to a lack of distinguishing markers. Indeed, many studies of B cells in humans limit their B cell characterisation to the pan-B cell marker CD20, which identifies all B cells before differentiation.
While some studies include CD138 to identify plasma cells, markers of B cell memory, residency and regulatory function are generally not included in studies of B cells in the tumour microenvironment.
This likely contributes to the difficulties in clearly ascribing a prognostic role to B cells in the tumour microenvironment.

| B cells in melanoma
B cell lineage cells in melanoma have not been well studied, but in line with their various functions in other cancers they can have differing impacts on patient outcomes. In metastatic melanoma, immunohistochemical staining demonstrated that a high density of intratumoural CD20 + B cells and CD138 + plasma cells correlated with improved survival in metastatic melanoma (Erdag et al., 2012).
B cells can also form TLSs at the tumour margins of melanoma, as observed in other solid tumours (Cipponi et al., 2012;Garg et al., 2017).
B cells in primary melanoma TLSs have been observed co-localizing with CD25 + T cells, although the significance of this is not clear as the phenotype of these T cells was not further clarified (Ladányi et al., 2011). Plasma cells are not often present in melanoma, particularly in primary melanoma, but are correlated with poor survival and clinicopathological markers indicating poor prognosis (Bosisio et al., 2016). This could be associated with the production of IgA by plasma cells in the context of melanoma, which have been identified in two separate studies (Bosisio et al., 2016;Cipponi et al., 2012).
It has been proposed that IgA limits the immune response to melanoma by physically blocking immune cell and IgG antibody access to the tumour; however, this has not yet been proven. IgA-producing plasma cells were identified in the TLSs of metastatic melanoma, but not in primary melanoma-instead, they were identified in the melanoma-draining lymph nodes. Isotype switching to IgA is caused by TGF-β, which is produced in the melanoma microenvironment, potentiating a scenario in which IgA-producing plasma cells in the context of primary melanoma could migrate to draining lymph nodes (Bosisio et al., 2016). From this, it appears that B cells have conflicting roles in the immune response to melanoma; however, this is likely due to their wide range of potential functions and the limited examination of different B-cell phenotypes. Due to the lack of distinguishing B-cell biomarkers in humans, some studies have instead opted to investigate the B-cell cytokine profile. One such study which did so in metastatic melanoma implicated the release of IGF-1 by B cells as a mechanism of resistance to BRAF and MEK inhibitors (Somasundaram et al., 2017). Further studies using similar techniques could eventually lead to an improved stratification of B-cell function in regards to their phenotype and function in melanoma.

| Natural killer cells
Although NK cells only make up a small proportion of the immune infiltrates in tumours, they have been found to positively correlate with patient survival in a number of cancers including gastric cancer, colorectal cancer and lung cancer (Al-Shibli et al., 2009;Chevrier et al., 2017;Coca et al., 1997;Ishigami et al., 2000;Villegas et al., 2002). These studies generally characterised NK cells solely by the presence of CD57 expression. However, CD57 expression is only gained by NK cells as they mature, so these studies would not have identified immature NK cells. Immature NK cells are characterised as the phenotype CD56 bright , while mature NK cells are CD56 dim CD57 + and also gain the expression of killer cell immunoglobulin-like receptors (KIRs). As NK cells mature, they become less responsive to cytokines and chemokines, lose expression of the activation markers NKp46 and NKp30, and also become less proliferative (Björkström et al., 2010). There is some evidence to suggest that tumours promote recruitment of an immature NK cell phenotype, as CD56 bright CD16 -NK cells were highly enriched in lung cancer tissue (Carrega et al., 2008).
Activating receptors expressed by NK cells include NKG2A, DNAX Accessory Molecule-1 (DNAM-1) and natural cytotoxicity receptors (NCRs). NK cells are able to secrete cytotoxic granules onto the surface of cells, including tumours, expressing ligands recognised by these activating receptors (Lakshmikanth et al., 2009). Humans possess a diverse range of KIRs, each of which is able to bind to a different HLA-I allele to inhibit NK cell-mediated killing of host cells (Uhrberg et al., 1997). Tumours can downregulate class I MHC to escape from CD8 + T cells, but in doing so tumours become vulnerable to NK cell cytolysis (Seo et al., 2017;Turcotte et al., 2013).
Rather than localising to the tumour margin or intratumourally, NK cells are generally located in the tumour stroma (Al-Shibli et al., 2009;Ishigami et al., 2000). The exact role of NK cells in the tumour environment is unclear. NK cells associated with lung cancer were found to have reduced cytolytic function, and together with the fact that NK cells do not localise near tumour cells, it could be argued that they are unlikely to directly kill tumour cells. This is further supported by the fact that NK cells in lung cancer were found to contribute to tumour rejection by secreting the anti-tumour cytokines IFN-γ and TNF-α (Carrega et al., 2008).  (Carrega et al., 2008), this population of cells secreted little TNF-α or IFN-γ yet they had a higher cytotoxic activity. They also expressed the activation markers CD16 and CD69, and were more responsive to cytokines produced by melanoma including CXCL8, IL-6 and CCL2 (Ali et al., 2014). Furthermore, NK cells lose the activation markers PD-L1 and CD137L as a higher number of lymph nodes are invaded, and that the expression of NKG2D is associated with a better prognosis in patients with BRAF-mutant melanoma (Jacquelot et al., 2016). CD16 and granzyme B expression by NK cells have also been correlated with an improved response to anti-PD-1 therapy (Lee et al., 2019).

| Dendritic cells
The professional antigen-presenting capabilities of dendritic cells allow them to initiate T cell responses against cancer. As such, while they may not directly act upon tumour cells, they do have an important role to play in initiating and modulating T cell responses to tumours.
One of the major dendritic cell characteristics affecting immune control mechanisms in cancer is maturity. Immature dendritic cells, broadly characterised as CD1a + , are well suited to phagocytosing antigens but are poor stimulators of T cell activity. Mature dendritic cells, characterised by expression of markers including CD80, CD83 and CD86, are potent activators of T cells, and are associated with a stronger T-cell response. Studies in breast, colon (Gulubova et al., 2012) and lung (Baleeiro et al., 2008) cancers have found that immature dendritic cells are more likely to infiltrate the tumour, while mature dendritic cells remain confined to the tumour margins and peritumoural areas. Furthermore, mature dendritic cells at the tumour margin often form part of TLSs, from which they are able to mount an ongoing immune response to tumours (Bell et al., 1999;Sautès-Fridman et al., 2019). CD21 + follicular dendritic cells have been located in these structures along with B cells, as have CD4 + and CD8 + T cells (Coronella et al., 2002;Nzula et al., 2003). In consideration of the previously discussed finding that T cells at the tumour margins had higher activity than tumour-infiltrating T cells, it is possible that mature DCs at the tumour margin contribute to this Cross-presentation, the presentation of exogenous antigens by class I MHC by DCs, is the only way in which CD8 + T cells can become activated. As outlined above, since CD8 + T cells are considered to be the main anti-tumour effector immune cells the process of cross-presentation is critical to the anti-tumour immune response.
In contrast, current evidence suggests that pDCs limit the anti-tumour immune response and therefore are likely to play a more regulatory role. In breast cancer and ovarian cancer, a BDCA2 + CD123 + subset of pDCs were found to infiltrate the tumour better than cDCs. Furthermore, the tumour microenvironment was found to cause IFN-α production by pDCs, indirectly promoting Treg expan-

| Dendritic cells in melanoma
The impact of dendritic cells on melanoma outcome is dependent on the subtype of dendritic cell associated with the tumour. cDCs are thought to promote an anti-tumour immune response in primary melanoma, as DC-LAMP + DCs and CD11c + DCs have separately been found to be elevated in melanoma patients without nodal metastasis (Ma et al., 2012). As observed in other cancers, DC-LAMP expression in melanoma was mostly present in the peritumour (Jensen Trine et al., 2011). Of the two cDC subtypes, cDC1 activity has been most closely associated with melanoma outcome. Recent work by Nsengimana et al. found that downregulation of genes in the β-catenin pathway is associated with a poor outcome in human primary melanomas (Nsengimana et al., 2018). This mechanism is known to impair cDC1 trafficking in mice, and thus could be working in the same way to limit cDC1 trafficking in humans. Further supporting this is the finding that CCR7 + cDC1s are the main activators of CD8 + T cell immunity against melanoma in mice (Roberts et al., 2016). Combined, these studies suggest a crucial role for cDC1s in activating anti-tumour immunity against human primary melanomas. The tumour microenvironment has also been found to impact dendritic cell function in melanoma, in which CD123 + CD86 + pDCs exist in a positive feedback loop with Tregs and the release of IL-10 by melanoma cells inhibits the action of cDCs (Ma et al., 2012).
CD123 + pDCs have also been associated with a poor outcome in primary melanoma (Jensen Trine et al., 2011).

| Macrophages
The role of macrophages in the control of tumours is highly contentious and reports linking their numbers with patient survival are conflicting. Studies in some cancers find that macrophages promote survival while in other cancers, macrophages have been found to negatively impact survival and yet in some others report that macrophages have no impact on tumour progression (Al-Shibli et al., 2009;Forssell et al., 2007;Kataki et al., 2002;Tsutsui et al., 2005;Wang et al., 2011). This would suggest that macrophages have different roles in different patients with different cancer types, although opposing roles have also been described for macrophages even within the same type of cancer (Lissbrant et al., 2000;Shimura et al., 2000). This could be because many of the studies correlating macrophages to survival use a very limited definition of macrophages, often only characterising them based on the expression of CD68, whereas it is now known that there is a spectrum of macrophage phenotypes (Awad et al., 2018). Aside from tissue-resident macrophages, most tumour-associated macrophages (TAMs) are derived from CD14 + monocytes. As they enter peripheral tissues monocytes differentiate into macrophages, and it is from here that their function and phenotype become highly diversified. Broadly, studies in mice have characterized macrophages as either M1 proinflammatory macrophages, or M2 anti-inflammatory macrophages (also known as alternatively activated macrophages (AAM)), with research generally finding that M1 activity is anti-tumour while M2 activity promotes tumour growth (Fridman et al., 2012;Thorsson et al., 2018). Macrophage differentiation into either M1 or M2 is thought to be highly dependent on the environment that they enter.  (Awad et al., 2018;Chevrier et al., 2017;Elliott et al., 2007;Mantovani et al., 2004;Stromnes et al., 2014).
TAMs are generally influenced by the tumour microenvironment to become polarised towards a pro-tumour function. These pro-tumour macrophages are characterized by the release of anti-inflammatory cytokines IL-4, IL-10, IL-13 and TGF-β, and promote angiogenesis (Fridman et al., 2012;Varney et al., 2005). Tregs also release immunosuppressive cytokines IL-4, IL-10 and IL-13 to promote pro-tumour macrophage function, causing macrophages to upregulate the markers CD206, CD163, CCL18 and reduce the expression of HLA-DR and CD86 and increase their phagocytic capacity (Tiemessen et al., 2007).
Anti-tumour (or pro-inflammatory) macrophages can also exist in the context of the tumour microenvironment, and are characterised by the secretion of pro-inflammatory cytokines IL-1, IL-6 and TNF-α, as well as reactive oxygen and reactive nitrate species (Fridman et al., 2012). Another potential function of anti-tumour macrophages contributing to the anti-tumour response is their antigen-presenting capacity. A study in mice identified a CD169 + CD11c + macrophage population performing cross-presentation to activate CD8 + T cells, thus promoting anti-tumour immunity (Asano et al., 2011). While this study suggested that these macrophages were confined to the lymph nodes, studies in humans have identified a CD169 + CD86 + HLA-DR + macrophage population infiltrating hepatocellular carcinoma and stimulating CD8 + T cell effector function. Even though these macrophages acted in an anti-tumour capacity, described as being indicative of an M1 phenotype, they were also concurrently modulated by the immune microenvironment to produce TGF-β. TGF-β produced by these macrophages was found to act in an autocrine manner, downregulating CD169 so that fewer CD169 + macrophages were detected within the tumour (Zhang et al., 2016). This demonstrates the limitations of the M1/M2 model in tumours, in that TAMs cannot always be neatly classified as M1 or M2.
The impact of macrophage localization in relation to tumours is unclear. Studies in ovarian cancer did not find an association between localisation and survival (Zhang et al., 2014), yet CD68 + macrophages located in the tumour infiltrate or margins were correlated with improved survival in gastric cancer and colon cancer, respectively (Forssell et al., 2007;Wang et al., 2011). This is explained in part by the different methods used to characterise macrophages in these studies: the ovarian cancer study used HLA-DR and CD163 to classify M1 and M2, respectively, whereas the gastric and colon cancer studies defined all macrophages as CD68 + .

| Macrophages in melanoma
Studies of macrophage function in melanoma have mostly been confined to mice. Tumour progression has been associated with a transition from an M1 to an M2-polarised macrophage population in multiple mouse models. TAMs associated with melanomainducing tumour-initiating cells exhibited a gene expression profile consistent with that of M2 macrophages, and supported their growth through the secretion of TGF-β (Tham et al., 2014).
Another macrophage effector function associated with tumour development is the expression of adrenomedullin, which promotes angiogenesis and melanoma growth. Adrenomedullin was found to act in an autocrine manner to promote M2 macrophage polarisation, and indeed, the M2 marker CD206 was upregulated in later-stage tumours (Chen et al., 2011). Studies in human tissue have found that patients who respond to BRAF inhibitors have higher numbers of macrophages in their melanoma pretreatment and that macrophage numbers inversely correlate with TIL grade in primary melanoma ( Wang, Xiao et al., 2015). Co-culturing of melanoma and macrophage cell lines found that treatment with BRAF inhibitors actually caused macrophage growth through activation of the MAPK pathway. These TAMs contributed to tumour growth through the secretion of VEGF, a growth factor which promotes angiogenesis. Furthermore, studies of tissue collected from patients treated with BRAF inhibitors found that a high number of CD163 + Ki-67 + macrophages before treatment correlated with early relapse and a shorter progression-free survival (Wang, Xiao et al., 2015). The multitude of functions in which macrophages have been found to drive melanoma progression warrants careful investigation of macrophage cell numbers and functions, especially in the context of treatment.

| Neutrophils
While neutrophils perform a multitude of functions that could have an anti-tumour effect, their impact on survival is currently unclear.
CD66b + neutrophils have been associated with poorer patient survival in clear cell renal cell carcinoma and hepatocellular carcinoma (Jensen et al., 2009;Li et al., 2011). Yet, characterisation of neutrophils in early stage lung cancer found that neutrophils were CD54 high CD62 low , indicating high activation, and promoted T-cell proliferation. They were also highly responsive to cytokines and promoted IFN-γ production from both CD8 + and CD4 + T cells.
Consequently, this contributed to a more Th1 polarised helper CD4 + T-cell response, suggesting that neutrophil activity could limit tumour growth (Eruslanov et al., 2014). While neutrophils in early tumours may exhibit an anti-tumour function, tumour-associated neutrophils (TANs) are influenced by the tumour microenvironment.
Similar to macrophages, studies in mice have identified two polarised forms of neutrophils designated as N1, which are anti-tumour, and N2 which are pro-tumour. In mouse models of mesothelioma and lung cancer, the gene expression profile of neutrophils was found to shift from an N1 profile to an N2 profile as the tumour progressed.
N2 neutrophils in more progressed tumours were found to have reduced cytotoxicity, and removal of N2 caused a reduction in tumour growth (Mishalian et al., 2013). Due to the short lifespan of neutrophils, it was suggested that early tumours in mice recruit N1 TANs and more progressed tumours recruit N2 TANs. However, other studies have suggested that cytokines in the tumour microenvironment increase neutrophil lifespan, so that TANs in early tumours could possibly convert from an N1 phenotype to an N2 phenotype over the course of tumour growth (Eruslanov et al., 2014).
Studies of TAN functionality in mice remain divided. There is evidence to suggest that that neutrophils play a major role in antibody-mediated killing of tumour cells in both B16 melanoma and BT474-M1 breast cancer models, although it is unclear whether this activity is phagocytic or cytotoxic. Furthermore, the extent to which these findings are applicable to human treatment settings is uncertain (Albanesi et al., 2013).

| Neutrophils in melanoma
Like in other cancers, the role of neutrophils in melanoma is poorly understood. Briefly, the presence of CD66b + neutrophils in human primary melanoma has been correlated with poor prognosis and activation of the anti-apoptotic protein STAT3 by melanoma cells (Jensen Trine et al., 2011). Studies in mice have yielded further contradictory findings: depleting TANs in a melanoma model was found to slow tumour growth (Nabizadeh et al., 2016), yet findings that neutrophils play a major role in antibody-mediated killing of tumour cells could potentially be applicable in an immunotherapy treatment setting (Albanesi et al., 2013). Clearly, more work is needed to accurately determine the roles of neutrophils in the immune response to tumours, including in the context of mAb immunotherapy (Ferrucci et al., 2015).

| Myeloid-derived suppressor cells
Myeloid-derived suppressor cells (MDSCs) are immature myeloid cells which, while absent in healthy individuals, can proliferate under conditions of chronic inflammation such as that of the tumour microenvironment (Kumar et al., 2016;Tcyganov et al., 2018;Umansky et al., 2014). They are induced through ongoing weak stimulation of the immune system, which is currently though to prevent immature myeloid cells from fully maturing and leads them to acquire a variety of immunosuppressive functions (Condamine et al., 2015). are known to be able to differentiate into macrophages (Tcyganov et al., 2018). As such, many of the immune functions attributed to myeloid cells polarised towards an immunosuppressive function may actually be caused by MDSCs. Gr-1 + CD11b + MDSCs infiltrating tumours have been observed in mice (Blattner et al., 2018), and MDSC depletion has been associated with improved survival outcomes in multiple mouse studies Sinha et al., 2007;Vincent et al., 2010). pancreatic, gastric and oesophageal cancers (Gabitass et al., 2011) although the phenotypic definition of MDSCs is not uniform between these studies. Suppression of T cell function specifically is the most frequent immunosuppressive activity associated with MDSCs. This is carried out through the release of NO, ROS and arginase-1 and expression of membrane inhibitory ligands including PD-L1, galectin-9 and ADAM17 (Chen et al., 2017;Condamine et al., 2015).
MDSCs can also promote angiogenesis (Gabrilovich, 2017) and release immunosuppressive cytokines, including IL-6, IL-10 and TGF-ß, to suppress a wide range of immune cells. This has most consistently been found to inhibit DC function and proliferation, as well as cause TAM suppression (Chen et al., 2017;Gabrilovich, 2017;Sinha et al., 2007). MDSCs can also promote the development of other immunosuppressive cell phenotypes. TGF-ß released by MDSCs has been associated with FoxP3 upregulation in T cells and subsequent differentiation of Tregs (Chen et al., 2017). Recent studies further suggest that MDSCs play a vital role in the development of premetastatic niches Gabrilovich, 2017), which are accumulations of cells conducive to tumour growth that can develop at a distant metastatic site before tumour cells even enter the bloodstream.

| Myeloid-derived suppressor cells in melanoma
Investigation of MDSCs in melanoma has mostly been performed within the context of immunotherapy. These have identified that MDSCs in melanoma are generally associated with a poorer response to immunotherapy (Umansky et al., 2014). Mouse models of melanoma have highlighted a CCR5 + MDSC population as being particularly immunosuppressive due to its increased PD-L1 expression, production of ARG-1, NOS, RO and ability to traffic to the tumour site through binding to CCR5 ligands expressed by melanoma (Blattner et al., 2018;Richmond et al., 2009;Weber et al., 2018).
Depletion of MDSCs in mice has also been associated with improved responses to immunotherapy, and conversely, immunotherapy has been implicated in the diminishment of MDSC generation Weber et al., 2018).
In humans, higher circulating MDSCs and MDSC-associated miR-NAs prior to immunotherapy or targeted therapy has been associated with a lack of response, as well as with more advanced disease (Huber et al., 2018;Meyer et al., 2014).

| Summary of the anti-tumour response
Although tumours originate from host cells, they express antigens that can be recognised by the adaptive immune system. These include antigens derived from mutated self-proteins, overexpressed proteins and tissue differentiation antigens (Vigneron et al., 2013).
Metastatic tumours will also carry tissue-specific antigens from the tissue they originated from, which can also be recognised by the adaptive immune system at the metastatic tumour site.
The main effectors of the immune anti-tumour response are CD8 + T cells, which recognise these antigens upon presentation by DCs migrating to the lymph nodes. These DCs also activate helper CD4 + T cells, and both CD8 + and CD4 + T cells migrate to the site of the tumour. While dendritic cells are responsible for initiating immune responses against tumours, macrophages and B cells with antigen-presenting capabilities have also been located in the vicinity of tumours (Asano et al., 2011;Zhang et al., 2016). CD8 + T cells are able to directly lyse tumour cells, while antibodies released by B cells may be significant in phagocytosis and antibody-dependent cell cytotoxicity performed by anti-tumour neutrophil and macrophage populations (Awad et al., 2018;Fridman et al., 2012;Thorsson et al., 2018).

| Tumour immune evasion and immunosuppression
Since recognition of tumour cells by CD8 + T cells is dependent on tumour expression of class I MHC, tumours can evade cytotoxic activity from these cells by downregulating class I MHC. Doing so leaves tumours vulnerable to recognition and killing by tumour-associated NK cells. Indeed, while multiple studies in melanoma have found that high expression of class I MHC is associated with improved patient survival in primary melanoma (van Houdt et al., 2008) and response to anti-PD-1 therapy in metastatic melanoma, it has also been found that patients were able to respond to anti-PD-1 therapy despite having low class I MHC expression if they also had high intratumoural levels of NK cells (Lee et al., 2019). Due to the interplay of these factors, the impact of class I MHC downregulation on patient survival should be examined within the immune context in which it occurs (Kakavand et al., 2017).
The tumour microenvironment also functions to suppress the anti-tumour immune response. Tumour cells themselves release the immunosuppressive cytokines TGF-β, IL-10 and VEGF, which suppress a number of inflammatory cells (Awad et al., 2018;Okita et al., 2014). Furthermore, the tumour microenvironment promotes the development of Tregs and Bregs, which specifically act to suppress adaptive immunity as well as releasing more immunosuppressive cytokines, such as IL-10 (Fu et al., 2007;Shao et al., 2014). Tumours also promote the polarisation of macrophages, neutrophils and Th towards more pro-tumour cell phenotypes (Awad et al., 2018). In addition, Gr + CD11b + myeloid-derived suppressor cells are also elevated in tumours, and promote angiogenesis (Yang et al., 2004). These interactions are summarised in Figure 3.
Immune cells responding to tumours can also become exhausted, as observed in a decreasing CD8 + T cell response throughout later stages of hepatocellular carcinoma (Flecken et al., 2014). This is in part driven by metabolism, as tumour cells rapidly take up glucose.
Depleting glucose from the tumour microenvironment in this way prevents uptake of glucose by immune cells, thus reducing their metabolism. Exhausted T cells upregulate PD-1, which causes reduced T-cell effector functions and proliferation upon binding to its ligand, PD-L1 (Parry et al., 2005). PD-L1 is upregulated by tumours, primarily at the tumour margins where CD8 + T cells are present (Tumeh et al., 2014). Blockade of PD-1 on T cells using mAb has been found to re-stimulate T cells to kill tumour cells (Tumeh et al., 2014). Immune checkpoint inhibitors, which block T-cell inhibitory molecules such as PD-1 and CTLA-4, have been shown to improve disease-specific survival in cutaneous melanoma and other cancer types.

| Determinants of the immune landscape of melanoma
As evaluated above, the types of immune cells which make up the TIME are highly diverse with wide-ranging influences on tumour progression. This diversity is observed not only between different types of cancers, but can also be seen between cancers of the same type. In the past few years, researchers have begun to investigate the determining factors which shape the TIME, allowing for further classification of cancers into subtypes based on their genetic and immunological context. Particularly, investigation of the impact of this genetic contexture on the TIME has been facilitated through the study of specific genes and gene signalling pathways; and the profiling of tumour gene expression signatures.
In some cases, it is possible to associate specific oncogenic mutations or changes in the expression of a single gene with immune cell activity in the TME. Mutations in BRAF have been well studied, with the V600E and V600K mutations being the most frequent and well known in melanoma. Recent study has found that V600E-mutant and V600K-mutant melanomas have distinct molecular gene expression F I G U R E 3 Overview of known mechanisms of immune evasion or suppression in melanoma. Cells marked with * indicate populations that have been characterised in mice, but not in humans profiles, and that V600K-mutant melanomas respond better to anti-PD-1 therapy , suggesting that there may be differences in the immune microenvironment of these two tumour subtypes. While these mutations alone are not the only drivers of patient response to immunotherapy, they suggest that even a single amino acid mutation in one gene can cause differences in the tumour immune contexture. Another well-studied example of this in melanoma is the impact of aberrant β-catenin expression on the TIME. In humans, overexpression of and gain-of-function mutations in β-catenin negatively correlate with CD8 + T cell infiltration into melanomas, and with reduced CD8 + T-cell cytotoxicity (Nsengimana et al., 2018;Spranger et al., 2015). Further studies in mice revealed that this reduced cytotoxicity was associated with a loss of CD103 + DC (equivalent to cDC1 in humans) recruitment into melanomas.
This in turn was linked to the downregulation of cytokines involved in DC migration. Moreover, mice with β-catenin + melanomas exhibited a limited response to immune checkpoint inhibitors (Spranger et al., 2015). were also observed (Nsengimana et al., 2018;Spranger et al., 2015).
CD103 + DC depletion in these β-catenin-high tumours was facilitated by downregulation of CCL4, mediated by ATF3 upregulation.
More broadly, many oncogenes regulate signalling pathways that control immune function. For instance, increased cytolytic activity in the TME across multiple cancer types has been correlated with mutations in multiple oncogenes (Rooney et al., 2015). Upregulation of the oncogene MYC, located in the β-catenin signalling pathway, has effects spanning multiple immune regulatory pathways. MYC is able to suppress pro-inflammatory signalling pathways, promote immune checkpoint receptor expression and alter cytokine expression to reduce the infiltration of B cells, T cells and NK cells into tumours (Wellenstein & de Visser, 2018). NF-kB expression is regulated by p53 and PTEN, such that loss of either of these genes can stimulate the NF-kB pathway and the subsequent pro-inflammatory response (Wellenstein & de Visser, 2018). However, PTEN loss concurrent with BRAF mutation in melanoma is also known to induce P13K activation, resulting in T cell exclusion and a poorer response to anti-PD-1 immunotherapy (Peng et al., 2016), highlighting that these mutations can have wide-ranging impacts on the TIME.
Many of the gene signalling pathways specific to melanoma described above are also involved in the developmental and differentiation programs of melanocytes. Since oncogenic mutations in these pathways have measurable effects on the tumour immune infiltrate, melanocyte differentiation status in melanomas can also impact the TIME. Indeed, this has been investigated through the application of different gene expression signatures to melanocyte differentiation.  (Lee et al., 2020). More specifically, the oestrogen receptor GPER promotes melanocyte differentiation, and the long-term protective effects of oestrogen against melanoma development have been attributed to GPER in mice. The latter study further identified that loss of GPER expression in melanoma caused a measurable loss of tumour-infiltrating lymphocyte and macrophage populations, as well as a worse response to anti-PD-1 immunotherapy (Natale et al., 2018). In this way, melanocyte differentiation status is closely tied to the overall genetic contexture of melanoma and as such can be associated with the resulting TIME.
Oncogenic mutations and the aberrant expression of oncogenes have been linked to a higher tumour mutational burden (TMB) (Wellenstein & de Visser, 2018), which in turn often leads to the generation of neoantigens. Recognition of tumour antigen-including neoantigen-by adaptive immune cells is crucial for immune-mediated tumour clearance, and the subsequent recruitment of these cells can shape the TME. A high neoantigen load has been associated with many factors of the anti-tumour immune response, including higher cytolytic activity (Rooney et al., 2015), a high infiltration of leucocytes, CD8 + T cells, CD4 + T cells, M1 macrophages and a lower Treg infiltration (Thorsson et al., 2018). Melanoma has one of the highest TMBs and neoantigen loads of any cancer type, which has been considered to be causative of its high responsiveness to immunotherapy (Schumacher & Schreiber, 2015;Wellenstein & de Visser, 2018). Interestingly, BRAFV600K mutant melanomas typically occur in the head and neck region in older patients and are associated with a higher degree of solar elastosis and higher TMB than BRAFV600E melanomas (Menzies et al., 2012). It is therefore not surprising that BRAFV600K mutant melanoma patients have better responses to immunotherapy than BRAFV600E mutant melanoma patients (Pires da Silva et al., 2019). However, it is important to note that many patients can respond well to immunotherapy without having a high TMB, and a high neoantigen load can, in some cases, negatively correlate with progression-free survival (Thorsson et al., 2018). Consequently, while the TMB is a good indicator of the extent of oncogenic mutation, it does not always directly correspond to the immune response to the tumour.
Genetic determinants of the tumour immune contexture have generally been studied by investigating a single oncogene or genetic pathway at a time. Furthermore, these studies usually take place in mouse models and only investigate specific immune cells, making them difficult to translate directly to the complex environment of human tumours (Wellenstein & de Visser, 2018). With recent advances in transcriptomics, the associations between gene expression and TIME can now be studied more in-depth. These studies will generally subtype cancers based on their expression of immune genes.
In this way, associations can be made not just between a particular gene or pathway and a single immune cell, but rather between multiple signalling pathways and an overall immune contexture. One such study in primary melanoma created six ʽconsensus immune clusters', which were defined by their expression of genes associated with different immune cells and immune signalling pathways. Of these six clusters, a distinctive high β-catenin/low immune signature was observed in the cluster with the worst melanoma-specific survival.
Methylation, mutation and copy number variation of genes in the β-catenin signalling pathway were studied across all six clusters allowing for the generation of a genetic score that distinguished the clusters better than expression data alone (Nsengimana et al., 2018).
A more in-depth study spanning 33 cancer types including cutaneous melanoma also generated six distinct tumour subtypes categorised by immune cell signatures. These were defined as wound healing, IFN-γ dominant, inflammatory, lymphocyte depleted, immunologically quiet and TGF-β dominant, respectively. Mutations in 33 oncogenic driver genes were found to specifically associate with different immune subtypes, while 34 genes were associated with changes in leukocyte fraction within the tumour. This study also found that the impact of an increased neoantigen load was dependent on the immune subtype of the tumour. A high neoantigen load in wound healing and IFN-γ-dominant tumours was associated with improved progression-free survival, while it was correlated with worse survival in inflammatory, lymphocyte depleted and immunologically quiet tumours (Thorsson et al., 2018). Such complexities are best revealed by these comprehensive transcriptomic studies, which will continue to provide insights into in the relationship between the TIME and the tumour genetic contexture.

| CON CLUS IONS
In this review article, we find that the overall impact of the immune response on cancer progression is highly dependent not only on the presence of particular immune cells, but also the context in which they function. Different immune cells within the tumour microenvironment can influence each other's activity, such that immune cells can contribute to the pro-tumour or anti-tumour immune response without directly interacting with the tumour. Furthermore, tumours also influence the phenotype of surrounding immune cells, facilitating the transformation of the immune microenvironment such that it supports tumour growth. Cell localisation in the tumour microenvironment is an important consideration for cell function, as cells that are expected to have anti-tumour functions are, upon observation, not able to carry out these functions if they are not located in the correct position to do so. In melanoma specifically, the best characterised immune cell response is that of CD8 + T cells, due to their unique ability to target and kill melanoma cells. Yet, since the CD8 + T cell response is initiated and regulated by many tumour-associated immune cells, it is important to elucidate the mechanisms by which all immune cells in the melanoma immune microenvironment contribute to melanoma development and outcome. Doing so could potentially inform prognostic and treatment decisions at all stages of melanoma, as well as provide directions for research into maximising the effectiveness of immune therapies.