Background
Uveal melanoma (UM) originates from the uveal tract of the eye [
1], and differs from the cutaneous melanoma in risk factors, biological behavior, epidemiology, prognostic features, and molecular profiles. Indeed, UM has an extremely low mutational burden contrary to cutaneous melanoma [
2], which harbors a high rate of C > T transitions and double CC > TT mutations induced by UV exposure, and hot-spot mutations in
BRAF or
RAS or loss of function mutations in
NF1 [
3]. On the other hand, the vast majority (85–95%) of metastatic UM harbors monosomy 3,
GNA11 or
GNAQ mutations [
4]. Despite the improvement and effectiveness of local tumor control (80% at 5 years), the high tendency to metastasize has not changed [
5] and still up to 50% of the patients develop distant metastases within 10 years after diagnosis. The liver is the most common site (80–90%), followed by lung (29%) and bone (17%) [
6,
7]. Once metastases are present, the disease course is generally aggressive and the prognosis remains poor, with a median overall survival (OS) of 13.4 months [
8], and a 2-year OS rate of only 8% [
9].
Liver metastasis is a relevant aspect of clinical course, and liver failure is almost the exclusive cause of death even when other visceral sites are involved. The median survival of patients who develop UM liver metastasis ranges between 6 to 12 months, as compared to 19–28 months for patients who first metastasize in other sites [
6,
10,
11]. In the metastatic stage, UM systemic therapy largely derives from approaches effective against cutaneous melanoma. Additionally, a variety of local liver-directed treatment options have been investigated, including surgical resection, hepatic artery embolization, hepatic arterial chemotherapy infusion, and radiofrequency, but none of them has resulted in an improved survival in metastatic disease [
12]. Further, CTLA-4 (ipilimumab) and PD-1 (nivolumab, pembrolizumab) inhibitors as monotherapy in sequence or combined, and targeted therapies with anti-angiogenic and kinase inhibitors [
13] have been also tested, but with disappointing results or only marginal success to date [
14‐
16]. Recently, the adoptive transfer of in vitro expanded autologous tumor-infiltrating lymphocytes (TILs) has been reported to mediate objective tumor regression in some patients with metastatic UM, thus fostering further investigation on the role of immune cells in this challenging disease [
17].
Assays using a variety of molecular techniques have the ability to analyze the primary tumor to predict prognosis and the risk of metastasis [
18,
19]. Differently from what observed in other cancer types [
20], evaluation of the prognostic impact of immune system in primary UM has revealed that high densities of tumor-associated macrophages (TAMs) and TILs are associated with a poor prognosis, and a high risk of metastasis [
21‐
23]. However, few studies have focused on deciphering the immunomodulatory features of metastatic UM microenvironment [
24,
25], largely due to the difficulty in acquiring specimens that often derive from percutaneous biopsies. Moreover, to date there are no prognostic models for clinical use in newly diagnosed metastatic disease [
26,
27]. Circulating UM cells that enter the liver encounter a unique immune system, as liver also acts as an immune-modulating organ devoted to quickly defeat gastrointestinal-derived pathogens, and at the same time to maintain tolerance against harmless food antigens [
28]. Therefore, interaction between liver immune system and cancer cells provide a complex tumor microenvironment that could help UM cells to evade an antitumor immune response. Thus, efforts must be put in place to understand the delicate interplay that occurs between tumor and immune cells acting at the site of metastasis, to allow the identification of prognostic/predictive factors which could facilitate the tailored management of patients and improve survival outcomes.
We designed a study aimed at analysing the density and the spatial distribution of immune cell subpopulations in a cohort of patients with hepatic and extra-hepatic metastasis from UM. The final goal was to identify immune biomarkers able to capture the immune contexture of tumor microenvironment that could stratify patients with better prognosis, in order to guide patient care and to facilitate future rational trial design to target appropriate metastatic UM patient subgroups.
Discussion
Currently, the present study comprises one of the largest sample cohort in which the metastatic UM immune microenvironment has been quantitatively analyzed, and in which the frequency and composition of immune cell infiltrate is correlated with patient outcome. Moreover, also the cell topography and thereby the probability of cell-to-cell interactions has been investigated, with additional correlations to clinical and prognostic parameters.
The eye is considered an immune-privileged site, and the immunobiology of primary UM has been already object of extensive investigation [
21,
23,
35,
36]. However, only few studies have focused on the characterization of the immune infiltrate and microenvironment of metastatic UM [
24,
25,
37], mainly due to the limited availability of biological samples, and because the vast majority of patients are not qualified for surgical resection due to the number or distribution of lesions [
38]. This aspect is critical for single-marker IHC studies, which often lead to incomplete quantitative data due to the exhaustion of the FFPE blocks [
39]. Multispectral imaging performed on metastatic UM samples allowed us to objectively assess seven markers simultaneously, to precisely quantify the number of cells with a specific phenotype and to determine their cartographic coordinates on a single 4-μm FFPE tissue section, thus consuming very few amounts of the precious metastatic UM sample.
Our observations on immune cells infiltrating UM metastases are consistent with recently published studies [
24,
25,
37], but we have additionally found that the immune cell subsets composition differs according to patient response. Indeed, in UM metastases from PD patients we identified immune features suggestive of an impaired antitumor immune response, such as a relevant presence of pro-tumorigenic M2 macrophages and T
reg cells, a reduced intra-tumoral CTL density and a lower stromal CD8/CD4 ratio. These results are in line with the observation that
NRP1 gene, which is involved in the immune-modulation of T
reg cells and M2-polarized macrophages, is upregulated in metastatic UM patients with an OS less than 1 year [
40]. Moreover, the spatial context of immune cells has been shown to be critical for cancer development [
41], since effector cells require close contacts with target cells to exert their cytotoxic antitumor functions. Our observations that the percentage of melanoma cells close to T cells was higher in CD than PD patients and that the majority of CD8 + Granzyme B+ T cells could be detected within the intra-tumoral region, suggest a specific anti-tumor effector role of CTL that may perform important biological functions in metastatic UM.
The potential importance of this cell subset is also supported by the observation that the percentage of activated CTL acts as a prognostic indicator able to stratify metastatic UM patients with better OS, while the mere density of total CD8+ T lymphocytes did not associate with patient outcome. Of note, when discriminating between the intra-tumoral and peri-tumoral regions, only the percentage of CD8 + Granzyme B+ T lymphocytes within the tumor masses retained the prognostic value, supporting the importance of effector cell localization in metastatic UM. All together, these observations go beyond the bias of previous studies that focused only on the overall cell counts using single marker IHC [
24,
25], highlighting the clinical relevance and possible functional importance of T cell infiltration for metastatic UM control.
TLS may support the activation of CTL against tumor cells, as the presence of TLS in melanoma patients was associated with improved outcome [
42]. However, in tumors arising in immunologically privileged sites, such us the brain (glioblastoma) and the eye (uveal melanoma), TLS are usually infrequent [
43]. In agreement with this observation, we found TLS only in a small proportion of metastatic UM patients, and no prognostic value was observed.
The complexity of mechanisms orchestrating the immune response against metastatic UM is underlined by the observation that a delicate equilibrium exists in patients with better outcome between lymphoid and myeloid cell responses within the tumor region, but not in the peri-tumoral stroma. Our data are in line with what observed by Massi et al. in a cohort of 158 metastatic cutaneous melanoma patients treated with MAPK inhibitors [
44]. Paradoxically, high densities of TAMs and TILs in primary UM are associated with a poor prognosis [
21,
45], likely because of the immunoregulatory influence of the intraocular microenvironment and the macrophage-mediated regulation of angiogenesis and cancer cell migration, which could promote tumor growth. Thus, the association of low densities of both T lymphocytes and macrophages with a prolonged OS in UM metastases, might be reminiscent of the primary tumor microenvironment. These results suggest that the combined evaluation of the density and spatial distribution of CD3+ and CD68+ cells in metastatic UM patients can be used as a prognostic indicator in metastatic UM.
The Immunoscore has been reported to overcome the classical TNM system in predicting disease-free survival (DFS) and OS in colorectal cancer (CRC) [
46]. The definition of an Immunoscore in cutaneous melanoma is still challenging, even though it is currently under evaluation in lymph node metastases from stage III melanoma patients [
47] and in metastatic tissues from individuals treated with Ipilimumab (the MISIPI study) [
48]. In this scenario, we tried to transfer the Immunoscore concept to metastatic UM as a potential prognostic marker. Despite the limited dimension of our cohort, the Immunoscore stratified patients in three distinct groups with statistically significant differences in terms of OS. However, while a low Immunoscore associates with the shortest DFS and OS in CRC patients, were the metastatic UM patients with I0 and I4 to exhibit a significantly increased OS as compared with those having an intermediated Immunoscore. Reasons for this unexpected trend require further investigations in a larger cohort.
Patients with hepatic UM metastases usually experience a worse outcome as compared to individuals with extra-hepatic UM metastases only [
10], thus suggesting a role for the microenvironment in UM metastatic progression. In this study, we had the chance to collect not only hepatic but also extra-hepatic UM metastases, and this allowed a comparative analysis of the tumor immune infiltrate between different UM metastatic sites. The liver is considered an immuno-modulatory organ, whose microenvironment could promote UM metastatic growth by protecting melanoma cells from immune surveillance [
49]. Differently from what described by Qin et al. [
39], we observed differences in the tumor immune cell contexture between hepatic and extra-hepatic UM metastases, supporting the delicate balance between immune elements with anti- or pro-tumorigenic functions in liver UM metastases. Indeed, the high density of intra-tumoral CD8+ T lymphocytes in LM was counterbalanced by an analogous elevated density of CD163+ M2-polarized macrophages, and by a low percentage of CD8 + Granzyme B+ activated CTL. Moreover, although we detected a higher percentage of tumor cells in close proximity to T lymphocytes as compared to EM, more than 60% of CTL in LM were in contact with M2-polarized macrophages, thus suggesting a role of TAM in suppressing cytotoxic CD8+ T cell functions. Furthermore, we observed that a high percentage of UM cells in liver were adjacent to pro-tumorigenic CD163+ TAMs. In melanoma cells, the expression of particular molecules, such as colony-stimulator factor 1 (CSF1) or CD47, in response to T-cell derived cytokines represents a conserved and adaptive resistance mechanism involved in disease progression [
34]. Indeed, the interaction between CSF1 on melanoma cells and its CSF1 receptor on macrophages shapes the tumor myeloid cell compartment toward immunosuppression by inducing the differentiation and accumulation of M2 TAM. Besides, the binding of CD47 on cancer cells with the inhibitory receptor signal regulatory protein alpha (SIRPα) on TAM suppresses the ability of macrophages to detect and phagocytose tumor cells [
50]. Collectively, these features are highly suggestive of a key role of TAM to impair the antitumor immune response in UM liver metastases by inhibiting the activation of CTL that are recruited to the tumor site. These observations could explain the worse outcome of patients with UM-derived LM and could have implications in the treatment modalities of these patients, through the depletion or re-education of TAM [
51].
Finally, we also explored the immune populations indicative of a better response to immunotherapy. Despite the low response rate and the negligible impact on survival of checkpoint inhibitors in metastatic UM [
52], in our cohort the patients with a better outcome disclosed a lower percentage of T
reg lymphocytes and a lower T
reg/CTL ratio, thus indicating that the amount of CD4 + FoxP3+ T
reg cells in metastatic UM might be considered a predictive biomarker for the response to immunotherapy. In support of this hypothesis, the gene expression analysis performed by Qin et al. on pre-treatment samples from six immunotherapy-treated metastatic UM patients revealed an upregulation of genes encoding cytokines and molecules of the pro-inflammatory signal network regulated by IL-13, IL-4 and NF-κB in non-responding individuals, and the upregulation of IFN-γ-regulated genes (
SOCS1 and MHC) in responding patients [
39].
We are aware that this study has some limitations. First, although being one of the biggest collections examined to date, the cohort is still limited. This aspect, together with the most recent case history, could have influenced the longer median OS observed in our cohort, as compared to what found in literature [
8,
9]. Second, patients were differently treated after metastatic UM diagnosis due to the different available treatments during the accrual; therefore, a possible influence of treatments on survival could not be ruled out. However, the absence of standard effective therapies should exclude or limit the treatment effect. Third, the expression of immune checkpoint molecules was not assessed, mainly due to the very limited availability of material samples. In this regard, however, there is increasing evidence that UM metastases are characterized by reduced levels of PD-1+ lymphocytes and PD-L1-expressing cells as compared to cutaneous melanoma metastases [
37,
39,
53], and this can provide a potential explanation for the failure of immunotherapy in UM [
27,
52]. On the other hand, molecular profiling at single cell-resolution on a limited number of UM liver metastases, showed the expression of TIM-3, LAG-3, and to some extent, TIGIT receptors on TILs, thus suggesting that alternative immune checkpoints may play a role in T cell response inhibition [
36,
54].
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.