As in most solid tumors, EOC cells downregulate MHC I expression as an immune evasion mechanism. Indeed, expression of MHC I genes is impaired in up to 60% of ovarian tumors ^[1][2][3][2,28,29]. Several EOC subtypes including serous, clear cell, endometrioid and mucinous, are immunogenic tumors capable of recruiting T cells into the tumor microenvironment (TME), resulting in positive prognoses ^[4][5][6][7][30,31,32,33]. Indeed, the presence of T cells specific to neoantigens expressed by EOC cells is strongly associated with increased survival ^[8][9][34,35] and the mechanisms related to immune cell infiltration are dependent on the antigen processing and presentation machinery (APM) components and MHC-I and -II status ^[10][11][36,37]. Nevertheless, the heterogeneity of HLA I allotype expression in a healthy cell is ultimately lost as tumors evolve to express fewer allotypes or completely lose HLA I expression ^[1][6][10][2,32,36].

The mechanisms by which MHC I expression is suppressed during tumor development has a major impact on the response to cancer immunotherapy ^[12][38]. Cancer cells lose or downregulate MHC I molecules because of the loss or decreased transcription of MHC I related genes or defects in APM components ^[13][39]. These defects can be classified as either “hard” or “soft”, depending on whether they are irreversible or reversible, respectively, by gene regulators or cytokines ^[13][39]. In healthy cells and cancer cells with soft defects, APM and MHC I genes can be induced by the IFN regulatory factor 1 (IRF-1), NF-κB, and the NOD-like receptor family caspase recruitment domain-containing 5 (NLRC5) in response to stimulatory cytokines such as TNF-α and IFN-γ ^[14][15][16][40,41,42].

EOC immunogenicity has been measured with humoral and cellular antitumor immune response markers detectable in peripheral blood, tumor sites, and ascites derived from EOC patients ^[17][43]. Goodell and colleagues were able to detect p53 antibodies in serum from 104 EOC patients, whose levels were positively correlated with overall survival ^[17][43]. Importantly, the presence of neoantigen-reactive T cells in patients with EOC can improve survival ^[8][9][34,35]. Brown et al. analyzed TCGA RNA-seq data from six EOC tumor sites in 515 patients, and identified mutational epitopes presented by the autologous HLA-A alleles that predicted tumor immunogenicity. These mutational epitopes triggered higher CTL content in the tumor niche and were associated with increased patient survival. However, tumors devoid of CTL infiltration lacked these mutational epitope signatures ^[8][34]. Wick and colleagues analyzed T cell reactivity towards 79 Tumor Associated Antigens (TAAs) originating from non-synonymous mutations identified by whole exome sequencing of autologous tumors, using T cells from the tumors of three EOC patients. A robust and specific CD8+ T-cell response to the mutated hydroxysteroid dehydrogenase-like protein 1 (HSDL1)^L25V was detected in one patient at different levels over the course of disease recurrence, highlighting the evolving expression of neoantigens and the limit of naturally occurring antitumoral immunity recognition over EOC progression ^[9][35].

In fact, ovarian tumors generally possess intermediate or low mutational burdens as a consequence of a very low incidence of naturally processed and presented neoantigens that could generate a significant antitumoral response ^[18][44]. Nonetheless, TAA presentation is the pivotal factor enabling CTL-tumor cell recognition and killing ^[19][45]. The following section will elucidate the current understanding of the expression of HLA class I allotypes and their intricate association with tumor burden and survival outcomes.

Downregulation of classic MHC I is a prevalent immune evasion mechanism used by tumor cells to escape antitumor T-cell-mediated immune responses ^[20][46]. Under physiological conditions, classic HLA class I molecules are expressed by virtually all cell types, allowing for NK or T cell recognition to achieve immunosurveillance. A tissue microarray of 339 EOC samples stained for MHC I and β2M established a positive correlation between HLA I expression and increased patient survival independent of age, stage, level of cytoreduction, and exposure to chemotherapy ^[21][47]. Although specific allotypes, such as the HLA-A*02 subtype, correlate with poor prognosis in advanced-stage serous EOC ^[22][8], the HLA-B allotype is a positive predictor of the immune response to cancer testis’ TAAs ^[23][48]. While MHC I gene expression in EOC cells can be downregulated as a consequence of somatic mutations, these mutations are not common in EOC. Shukla et al. analyzed 7930 samples across 20 different tumor types and found that ovarian carcinoma, glioblastoma, and breast cancer largely lacked somatic mutations in HLA genes, being present in only 0–0.6% of the tumor samples ^[24][49].

Despite the lack of HLA mutations, differences related to total HLA I and II expression in ovarian tumors occur. Using RNA-seq, immunohistochemistry, and flow cytometry analysis on 27 EOC samples, Schuster et al. revealed that most ovarian tumors display strong HLA I expression, and to some extent HLA II expression. However, only the EpCAM+ population was considered in the cancer cell subset which may not necessarily represent most EOC cells ^[25][50], and the degree of immune infiltration of the EOC samples was not included in the analysis, potentially resulting in an overestimated HLA expression in highly infiltrated tumors. In a more recent study, tissue sections from 30 untreated high-grade serous ovarian cancers (HGSC) were analyzed for MHC I staining and showed sub-clonal loss in 7/30 (23%), including areas of retained MHC class I expression immediately juxtaposed with areas of negative staining ^[26][51]. Neither of these studies classified the overall diversity of HLA class I allotypes being retained in the tumor tissue, which may turn out to be a notable weakness, as non-classic HLA class I expression may negatively impact patient survival as described in the following section.

As cancer cells downregulate classic HLA I molecules to avoid CTL recognition, they can also circumvent detection and elimination by NK cells through alternative means. Non-classic HLA I molecules are less polymorphic and display distinct expression patterns in developing and adult tissues, exerting functions in both the innate and adaptive immune systems ^[27][28][52,53]. In many malignancies, non-classic HLA I allotypes are aberrantly expressed, perhaps as a consequence of the proximity of genes such as HLA-E, -F, and -G to the class I region on chromosome 6 ^[29][54]. Indeed, aberrant expression of non-classical HLA I molecules in tumors contributes to inhibition of NK cells, rendering tumor cells resistant to NK cell-mediated lysis ^[30][55]. The following sections summarize key studies underlining the potential effects of non-classic HLA molecules in EOC.

HLA-E is expressed in most healthy human tissues, including placenta, but with a weak expression pattern on the cell surface ^[31][56]. HLA-E/peptide complexes are recognized by the CD94 receptor in conjunction with the inhibitory NKG2A or the stimulatory NKG2C molecule, expressed on the majority of NK cells and some activated CTLs ^[32][33][34][16,57,58]. In several malignancies, HLA-E can compensate for the loss of classic HLA I expression. Indeed, HLA-E expression is upregulated concurrently with the downregulation of classic HLA I allotypes and the presence of free β2M in the cytoplasm of tumor cells ^[30][55]. Tumor cells possessing an imbalance in heavy chain and β2M expression also possess this unique inverse expression pattern. HLA-E/β2M complexes are weaker compared to classic HLA I complexes, and when the latter are absent the HLA-E complexes become prevalent ^[30][55].

Nonetheless, in other cases, HLA-E overexpression can unbalance pre-established antitumoral immunity. In a study including 150 cervical and 270 EOC samples, HLA-E was found to be expressed at higher levels than healthy tissue and positively associated with expression of APM components, classical HLA I molecules, and CTLs in 80% of the samples ^[35][59]. In situ analysis revealed that HLA-E interacts with the inhibitory CD94/NKG2A receptor predominantly expressed on intraepithelial CTLs. Notably, the favorable prognostic effect of infiltrating CTLs in EOC was neutralized by high expression of HLA-E on the surface of the cancer cells, suggesting that HLA-E impedes antitumoral CTL activity in the TME ^[35][59].

Interestingly, a recent multivariate analysis from the phase III AGO-OVAR-12 trial involving 103 HGSC patients suggested that the HLA-E/CD94-NKG2A/2C axis is a potential target to improve antitumoral activity, particularly in the group of patients with homologous recombination deficiency (HRD). Similar to the overexpression of HLA-E in unstable microsatellite tumors in colorectal cancer ^[36][60], HLA-E was preferentially overexpressed in HRD HGSC, although the germline or somatic BRCA mutation status was not explored in a study ^[37][61]. HGSC patients with a high fraction of intratumoral CD3+ T lymphocytes had longer progression-free survival (PFS) as well as high HLA-E expression on tumor cells, along with an HRD profile which showed improved overall survival ^[37][61]. Moreover, the authors found that HLA-E-overexpressing tumors were highly enriched in Tregs (FOXP3+, ICOS+) and IgG, and, similar to findings by Gooden and colleagues ^[35][59], the survival benefit driven by T cell infiltration was lost with high HLA-E expression. This study provides insights into the impact of HRD lesions enhancing genomic instability which also influence tumor immunogenicity, tumor immune infiltration, and, potentially, HLA expression. Although, it is still not clear if genomic instability can directly affect HLA-E expression as a consequence of DNA damage ^[38][62], or if these findings are a consequence of type II IFN (IFN-γ) production in the TME, as HLA-E can be induced by IFN-γ ^[39][63].

Genetic variations in HLA-E alleles can influence their role in tumor immunosurveillance. Only two alleles (HLA-E*0101 and HLA-E*0103) have been reported, and they potentially accomplish different functions ^[40][64]. Zheng et al. studied 85 primary serous EOC tumors compared to 100 healthy tissues and found a high frequency of HLA-E*0103 expression at the transcriptional and protein levels in serous EOC. This allele improved the transfer of the HLA-E molecule to the cell surface, rendering the HLA-E/peptides complex more stable and increasing its capability to inhibit NK cell cytolysis ^[40][64]. However, it is still unclear how the HLA-E*0103 allele is preferentially expressed in EOC tumor cells and how both alleles are affected and regulated during tumor development.

HLA-F is the smallest of the HLA I molecules and is expressed in the skin, the developing fetal liver, to a lesser extent in the placenta and extra-placental tissues, and also in monocytes and lymphocytes such as NK, T, and B cells ^[41][65]. HLA-F is expressed as an empty heterodimer devoid of peptide in the cytosol, and acts as a ligand for several intracellular proteins such as TAP and calreticulin ^[42][66], and immune specialized receptors including immunoglobulin (Ig)-like transcript 2 (ILT2), ILT4 ^[43][67], KIR three Ig domains and long cytoplasmic tail 2 (KIR3DL2), KIR two Ig domains and short cytoplasmic tail 4 (KIR2DS4) ^[44][68], and KIR three Ig domains and short cytoplasmic tail 1 (KIR3DS1) ^[45][69]. Interestingly, HLA-F can also participate as a chaperone in the cytosol to stabilize classic HLA I open conformers (without peptide) on activated monocytes and lymphocytes, cooperating in the exogenous cross-presentation pathway independent of the TAP and tapasin proteins ^[46][70]. Collectively, the evidence suggests that HLA-F is an immune regulatory molecule that acts as a stabilizer; however, its binding partners are unknown.

HLA-F mRNA is overexpressed in glioblastoma compared to healthy tissue ^[47][71]; however, to date, there is no known link between HLA-F and EOC. Nevertheless, Fang et al. recently found that the long noncoding RNA HLA-F-AS1 is overexpressed in EOC cells and attenuates EOC development in vivo and in vitro by targeting the miR-21-3p/PEG3 axis ^[48][72]. Since HLA-F plays several immune regulatory roles, more studies are needed to better understand its potential role in EOC tumorigenicity.

Similar to HLA-E, HLA-G is expressed in extra-embryonic tissues during gestation, especially in the placental trophoblasts where it participates in the establishment of an immunotolerant state during pregnancy ^[31][49][56,73]. HLA-G can also be found in the cornea, nail matrix, pancreas, erythroid and endothelial precursors, and stem cells ^[50][51][52][74,75,76]. In the thymus, HLA-G appears to participate in the development of the T cell repertoire, potentially explaining T cell tolerance to HLA-G ^[53][77]. In healthy tissues, HLA-G plays a protective immunosuppressive role, whereas, under neoplastic conditions, HLA-G allows tumor progression by being overexpressed in cancer cells ^[54][78]. This allotype can undergo alternative splicing of its primary transcript to produce seven HLA-G protein isoforms (HLA-G1 to -G7), three of which can become soluble proteins (HLA-G5 to -G7) ^[55][79]. HLA-G possesses a heterogeneous and focal expression pattern, and can be expressed at the cell surface, secreted, or associated with tumor-derived exosomes. In all these forms, HLA-G exerts immune-modulatory functions by binding to CD8, LILRB1 (Leukocyte Immunoglobulin-Like Receptor B1, expressed by monocytes, DCs, B cells, and NK cells), LILRB2 (expressed only by monocytes), and KIR2DL4 (expressed by placental NK cells) ^[55][79]. HLA-G acts as a tolerogenic molecule by inhibiting the immune cell functions of APCs, NK cells, and CD4+ and CD8+ T cells by directly binding with their inhibitory receptors or indirectly through trogocytosis, leading to T cell anergy and rendering T lymphocytes more regulatory and immunosuppressive ^[22][8]. sHLA-G also induces apoptosis in NK cells and antigen specific CD8+ T lymphocytes ^[56][80].

In view of the numerous immune-regulatory functions of HLA-G, it is not surprising to find that its expression is associated with a worse clinical outcome in patients with solid tumors, including mesothelioma and breast carcinoma ^[57][58][81,82], but not in hematological malignancies ^[54][78]. Interestingly, in vitro studies have shown that just 10% of HLA-G+ tumor cells is sufficient to protect the rest of the tumor cells from elimination by CTLs ^[59][83], highlighting the strong regulatory capabilities of HLA-G in tumor promotion. When Lin et al. transfected the EOC cell lines HO-8910 and OVCAR-3 with the HLA-G gene, the cells acquired higher invasion potentials compared to parental cells. Moreover, when introduced into Balb/c nu/nu mice, EOC cells overexpressing HLA-G developed widespread metastasis, conferring poor survival ^[60][84]. HLA-G is also involved in tuning the immune response, as when HLA-G+ cells were cultured with PBMCs, the immune response was more accentuated towards a Th-2 cytokine profile ^[61][85]. This is the opposite of the actions of sHLA-G, which favors an anti-inflammatory environment induced by the release of IL-10 ^[62][86].

HLA-E and HLA-G allotypes can be co-expressed in EOC tissues, with a semi-quantitative analysis of 62 EOC revealing high HLA-E expression associated with the serous subtype and advanced stages ^[63][87]. In another study, Andersson et al. analyzed non-classic HLA I expression in primary tumors from 72 patients with advanced-stage serous EOC and in metastatic cells derived from ascites from eight patients ^[22][8]. The site-specific downregulation of classical MHC I allotypes alongside the focal cell expression of HLA-G and HLA-E correlated with poor survival and worse prognosis in patients harboring the HLA-A*02 subtype, but not with different HLA genotypes. Interestingly, metastatic lesions had a higher expression of HLA-G compared to primary tumors, which was inversely correlated with the frequency of TILs and increased immunosuppression ^[22][8]. Furthermore, sHLA-G may be a potential marker of malignant ascites in EOC ^[64][88] which could be used to assess the progression and recurrence of the disease ^[65][89]. HLA-G may also regulate vascular remodeling in tumors, pointing towards the strong capability of this molecule to influence the EOC TME ^[66][90].

In contrast, Rutten et al. showed that HLA-G expression was correlated with longer PFS and overall survival and an improved response to chemotherapy in 169 HGSC patients ^[67][91]. Importantly, serum sHLA-G levels did not correlate with protein or gene expression levels in the tumors or survival ^[67][91]. Discrepancies regarding the pro-tumoral or anti-tumoral effects of HLA-G remain controversial and yet to be clarified. According to the available studies, HLA-G is frequently expressed in high-grade ovarian tumors, especially at advanced stages, and in rare cases in low-grade tumors ^[68][92]. Contradictory findings can be due to differences in staining techniques, gene expression vs. protein expression, scoring scale, or the definition of positive expression in each study. In addition, the role of HLA-G could be a consequence of the heterogeneity, unique to each EOC, being influenced by the tumor mutational burden (TMB) and the immune composition, in some scenarios, compensating for a lack of classic HLA I expression, and in other scenarios being aberrantly co-expressed with HLA-E. Moreover, since there are several spliced forms of HLA-G, post-translational regulation within the TME may play a dominant role in protein expression which could ultimately change the impact of HLA-G expression (membrane-bound or soluble) and function in the tumor niche.

NLRC5 (also known as CITA) is a critical regulator of MHC I genes, as well as some related genes involved in MHC I-dependent APM via the formation of CITA enhanceosomes ^{[15][69][70][71]}[41,93,94,95]. NLRC5 induces the expression of both classical and non-classical class I molecules, but also the main components of the APM pathway like β2M, immunoproteasome components (PSMB9, i.e., LMP2), and TAP1 ^[15][41]. The expression of MHC I and APM components strongly correlates with NLRC5 gene expression in multiple cancers such as lung, melanoma, thyroid, breast, prostate, uterine, and EOC. Defects in NLRC5 expression found in human tumors include genetic modifications such as copy number loss, somatic mutations, and promoter methylation, which strong downregulate MHC I expression ^[72][73][96,97]. An analysis of the NLRC5 gene in multiple cancer types revealed that EOC patients (n = 489) displayed the highest frequency of copy number loss at 72.2%. This loss was associated with the reduced expression of NLRC5 and MHC I and related genes, including HLA-A, HLA-B, HLA-C, B2M, LMP2, and LMP7 ^[72][96]. Recently, itwe is found that overall reduced NLRC5 expression in OC tumors correlates with poor survival, NLRC5 gene expression being mainly found in the immune compartment of infiltrated OC tumors and correlated with antitumoral and IFN gene signatures (Rodriguez, 2023 doi: 10.3389/fimmu.2023.1295208). NLRC5-induced overexpression in a murine model of OC was capable of rescuing MHC I expression, shaping the tumor immune composition, and increasing the production and recognition of TAAs by circulating T cells (Rodriguez, 2023 doi: 10.3389/fimmu.2023.1295208).

In summary, classic HLA I expression is associated with better survival for EOC patients while non-classic HLA I is more pronounced in aggressive and more advanced EOCs. Many unknown aspects related to non-classic HLA I, such as ligands, polymorphisms, and post-translational regulation, still need to be explored.