Comprehensive Approach to Genomic and Immune Profiling: Insights of a Real-World Experience in Gynecological Tumors

Gynecological cancer accounts for an elevated incidence worldwide requiring responsiveness regarding its care. The comprehensive genomic approach agrees with the classification of certain tumor types. We evaluated 49 patients with gynecological tumors undergoing high-throughput sequencing to explore whether identifying alterations in cancer-associated genes could characterize concrete histological subtypes. We performed immune examination and analyzed subsequent clinical impact. We found 220 genomic aberrations mostly distributed as single nucleotide variants (SNV, 77%). Only 3% were classified as variants of strong clinical significance in BRCA1 and BRCA2 of ovarian high-grade serous (HGSC) and uterine endometrioid carcinoma. TP53 and BRCA1 occurred in 72% and 28% of HGSC. Cervical squamous cell carcinoma was entirely HPV-associated and mutations occurred in PIK3CA (60%), as well as in uterine serous carcinoma (80%). Alterations were seen in PTEN (71%) and PIK3CA (60%) of uterine endometrioid carcinoma. Elevated programmed death-ligand 1 (PD-L1) was associated with high TILs. Either PD-L1 augmented in deficient mis-matched repair (MMR) proteins or POLE mutated cases when compared to a proficient MMR state. An 18% received genotype-guided therapy and a 4% immunotherapy. The description of tumor subtypes is plausible through high-throughput sequencing by recognizing clinically relevant alterations. Additional concomitant assessment of immune biomarkers identifies candidates for immunotherapy.


Introduction
Among gynecological cancer, three types account for the vast majority of new cases that are included in the top ten most common female cancers worldwide in 2020: cervix uteri (6.5%), corpus uteri (4.5%) and ovary (3.4%) [1]. Recent data from Spain estimates a 12% rise in new cases in gynecological cancer, thus indicating an elevated incidence and related mortality requiring our attention for an appropriate management and care [2].
Since the implementation of high-throughput sequencing in the clinical setting, several cancer types have improved predictive therapeutic interventions, hence resulting in better prognosis. In some malignancies, such as in lung or in colorectal cancer [3,4], the clinical effects of multiple biomarkers analysis in a single assay have led to the development of numerous targeted treatments. However, in gynecological cancer, the role of next-generation sequencing has brought understanding on its molecular pathogenesis but is still limited

FFPE Tissue Collection
We compiled data from 49 patients with gynecological tumors (including ovary, uterus, cervix and vagina) who were asked for high-throughput sequencing testing between 2017 and 2019 during a routine molecular diagnostic practice at the Fundación Jiménez Díaz University Hospital. We retrieved archived formalin-fixed, paraffin-embedded (FFPE) tissue material from the Biobank Fundación Jiménez Díaz (PT20/00141) that belongs to the Spanish National Biobank network. Each donor gave written consent. All investigations followed standard operating procedures with the approval of the Fundación Jiménez Díaz University Hospital Ethics and Scientific Committee (PIC209-21) and were performed in accordance to the principles outlined in the Declaration of Helsinki.

DNA and RNA Isolation
FFPE tissues were sectioned 3-µm thick for hematoxylin and eosin staining (Dako coverstainer, Agilent, Santa Clara, CA, USA) to ensure appropriate tumor-cell content. We used consecutive sections to extract both genomic DNA and RNA with Recoverall total nucleic acid isolation kit according to the manufacturer's instructions (ThermoFisher Scientific, Waltham, MA, USA). We then quantified the purified DNA and RNA with respectively Qubit dsDNA BR assay kit (ThermoFisher Scientific, Waltham, MA, USA) and RNA Screentape Analysis kit (Agilent, Santa Clara, CA, USA).

High-Throughput Sequencing
The multi-biomarker Oncomine Comprehensive Assay (OCA) v3 (ThermoFisher Scientific, Waltham, MA, USA) was used to prepare DNA and RNA libraries in agreement with the manufacturer's protocol. The panel enabled the detection of relevant single nucleotide variants, copy number variations, gene fusions, and INDELs from 161 cancer-associated genes (Table S1). Briefly, DNA libraries generation used the Ion Ampliseq library kit 2.0 (ThermoFisher Scientific, Waltham, MA, USA) by incorporating barcode, adapters and RNA libraries employed the Ion Ampliseq RNA library kit (ThermoFisher Scientific, Waltham, MA, USA). Both pooled libraries templates were prepared on the Ion Ampliseq library Ion Chef system with the Ion 540 Chef kit (ThermoFisher Scientific, Waltham, MA, USA). Subsequently, multiplexed libraries templates were sequenced using Ion 540 chips on the Ion Torrent S5 XL platform (ThermoFisher Scientific, Waltham, MA, USA) as indicated by the manufacturer's protocol. Data analysis for variant calling and annotation were performed with the Ion Reporter software v.5.10. Mapped reads used the hg19 reference genome, a minimum sequencing depth of 500× was considered appropriate and variants with an allelic frequency (AF) less than 5% were filtered and excluded before review, except for certain variants where a 1% AF was contemplated. We considered intergenic fusions with a depth superior than 40× and intragenic fusions with 100×, whereas the limit of copy number variation was set up to more than five copies. Variant categorization was assessed as previously described [23,24]. Briefly, Tier IA and IB categories corresponded to variants with strong clinical significance, Tier IIC and IID to variants with potential clinical significance, and Tier III to variants with unknown clinical significance.

Immunohistochemistry
Serial FFPE tissue sections of 3 µm thick were employed for immunohistochemical studies in order to determine the protein expression of ERBB2, MLH1, MSH2, MSH6, PD-L1, PMS2, and p53 as depicted in Table S2. Recommendations of the international TILs working group were followed to evaluate TILs [27,28]. Slides were incubated 1 h at 56 • C and deparaffinized. High pH EnVision FLEX Target Retrieval solution was used for antigen retrieval in the PT-link instrument (Dako, Glostrup, Denmark) for 20 min at 95 • C. Staining with primary antibodies was automatically achieved on the Autostainer Link 48 with the EnVision FLEX System-HRP kit (Dako, Glostrup, Denmark) after endogenous peroxidase activity blocking with the EnVision FLEX peroxidase-blocking reagent for 10 min. Preparations were finally incubated 10 min on chromogen 3,3-diaminobenzidine, automatically counterstained and mounted in the Coverstainer platform (Dako, Glostrup, Denmark). Two different pathologists independently assessed the presence or absence of tumor cell immunoreactivity and positive controls. Evaluation of PD-L1 was done by calculation of the combined positive score (CPS = number of PD-L1 staining cells (tumor cells, lymphocytes, macrophages)/total number of viable tumor cells × 100), as reported earlier [29]. Assessment of p53 was scored by 0-normal or wild-type, 1-abnormal overexpression, 2-abnormal cytoplasmatic expression, and 3-complete absence, as previously described [30]. ERBB2 was graded as 0-no staining, 1+ incomplete membrane staining in <10%, 2+ strong complete membrane staining in ≤30% or incomplete in ≥10%, and 3+ strong staining in >30% of tumor cells [31].

ERBB2 Fluorescent In Situ (FISH) Hybridization
We used additional sequential FFPE tissue sections of 3-µm thick to identify ERBB2 amplification on tumors with a 2+ immunohistochemical score. It also served as the orthogonal approach that confirmed ERBB2 amplifications resulting from the high-throughput sequencing panel. We performed FISH according to the manufacturer's instructions (Pathvision, Abbott molecular, Des Plaines, IL, USA), as previously described [32]. We achieved the analysis on a fluorescent microscope DM5500B (Leica Microsystems, Wetzlar, Garmany) using Cytovision software (Leica Microsystems, Wetzlar, Garmany). Assessment was done by counting red signals by nucleus and green signals by nucleus respectively corresponding to ERBB2 gene and chromosome 17 centromere (CEP17). According to the established criteria, we considered amplification of ERBB2 if the ratio of ERBB2/ CEP17 ≥ 2 or ERBB2/ CEP17 < 2 and the mean signals of ERBB2 by nucleus ≥6 [31].

Statistical Analysis
Statistical analysis used SPSS version 21.0 software for Windows (IBM, New York, NY, USA) and GraphPad Prism version 5.0 software (GraphPad Software, Inc., La Jolla, CA, USA). Descriptive data were expressed as median and interquartile range. Associations were made by using Spearman's rho rank correlation coefficient and comparisons between multiple groups used Kruskal-Wallis tests with Bonferroni correction of post hoc Mann-Whitney U tests. p < 0.05 was considered significant.

Clinicopathological Characteristics of Cohort
From the 49 female patients included in the study, we identified almost half of them (n = 22) with ovarian cancer, n = 15 with uterine cancer, n = 11 with cervical cancer, and n = 1 with vaginal cancer. Only 16% of the patients exhibited metastatic disease at the time of diagnosis and 26% had family history of malignancy. The major histological subtype representation corresponded to high-grade serous ovarian carcinoma (37%), followed by endometrial endometrioid carcinoma (12%) and cervical squamous cell carcinoma (12%). Thirty-five patients (71%) showed advanced-stage disease. Taxane plus platinum treatment was the chemotherapy mostly administered in the cohort (59%) whereas adjuvant radiotherapy was given in 34% of the patients. About 79% underwent surgery debulking and experienced disease recurrence. Additional clinical data displaying radiological response to therapy by tumor site are depicted in Table 1.

Detection of Genomic Alterations across the Cohort
Evaluation of the 49 patients permitted the recognition of 220 variants distributed as single nucleotide variants (SNV, 77%), copy number variation (CNV, 13%), insertiondeletion (INDEL) mutations (7%), gene fusions (2%), and multiple nucleotide variants (MNV, 1%). No variants were detected in 12% of the patients. SNV exhibited a median coverage of 1995× (1333-1999) and a variant allele frequency of 34% , whereas INDEL mutations and MNV were covered 1973× (1356-1992). CNV alterations showed a median copy number of 7 (6-11) and gene fusions presented a median count of 131 (75-185) reads. Further sequencing raw data information of detected variants can be accessed in Table S3. Only 3% of detected variants were classified as variants of strong clinical significance (Tier I), 47% were variants of potential clinical significance (Tier II), and 50% were variants of uncertain significance (Tier III). Tier IA variants were found in BRCA1 and BRCA2 genes from three ovarian high grade serous carcinomas and one uterine endometrioid adenocarcinoma, respectively ( Figure 1A). Tier I and tier II alteration types were more represented in the TP53, PIK3CA, ERBB2, CCNE1, PTEN, and STK11 genes ( Figure 1B). The number of genomic alterations ranged from 0 to 15 in the whole cohort. Uterine serous and endometrioid carcinomas presented an elevated number of genomic alterations, 7 (4-12) and 6.5 (5-9), respectively, in comparison with ovarian HGSC, 3.5 (3)(4)(5), p = 0.01 and with cervical HPV-associated carcinoma, 1.5 (0-5), p = 0.01 ( Figure 1C).  We also compared the frequencies of single somatic mutations-affected cases in the studied cohort against those cases listed in the TCGA database for each tumor site. The TP53 gene was more frequently altered in the ovary (64% against 91% in TCGA), whereas PIK3CA (47% against 51%), PTEN (33% against 67%), and TP53 (27% besides 40%) genes presented more altered variants in the uterus. Cervix exhibited the PIK3CA gene as the more altered gene, (33% of the cases against 29% of the TCGA) (Table S4).

Characterization of Histological Subtypes
We conducted a brief literature review in order to define the molecular biomarkers characterizing precise histological subtypes included in the cohort. Each of them was assigned to a particular signaling pathway. We assessed those cancer-associated genes included in our high-throughput sequencing panel. We were unable to identify mutations in EMSY, BRIP, BARD1, and FOXM1 CNV in high-grade ovarian carcinoma as well as PTEN loss in clear cell carcinoma, LRPB1 loss and SOX17 amplification in uterine serous carcinoma (      We then analyzed the mutational profiles and protein expression of the histological subtypes. We found that alterations in the TP53 and BRCA1 genes occurred in 72% and 28% of the HGSC cases, respectively. Other top altered genes in HGSC included CCNE1 (17%), FANCD2 (17%), POLE (17%), and PMS2 (17%). Only one case showed a PIK3CA CNV (5%) (Figure 2A). All HGSC cases exhibited abnormal P53 expression except in one case ( Figure 3) and were proficient MMR or MSS. Although ovarian clear cell carcinoma was represented in a minor proportion, PIK3CA and ARID1A genes were found mutated in two of the three studied cases. Only case 403, an ovarian clear cell carcinoma, presented dMMR with a concrete pattern of expression loss in MSH6 (Figure 4). Regarding uterine serous carcinoma, the proportion of cases showing altered genes was PIK3CA (80%), TP53 (40%), ERBB2 (40%), FBXW7 (40%) and with less representation PPP2R1A (20%) and FGFR2 (20%). From the two cases showing SNV in ERBB2, one revealed concomitant ERBB2 amplification (20%). Furthermore, as illustrated in Table S5 and Figure S1, additional histological subtypes exhibited ERBB2 amplification such as ovarian HGSC and cervical adenocarcinoma. Even though, we observed dissimilarities with the orthogonal approach in cases 287 and 393, probably explained by the limit of detection applied in the assessment of the copy number variation from the high-throughput panel. In fact, we just found a 10% of the cohort presenting ERBB2 SNV and 6% CNV, whereas conventional testing identified up to 8% of ERBB2 amplification. Concerning endometrial endometrioid carcinoma, cases presenting genes with alterations in PTEN (71%), PIK3CA (60%), and NOTCH1 (43%) were the most observed. About 28% of affected cases presented alterations in CTNNB1, KRAS, POLE, SETD2, PIK3R1, FANCI, and ARID1A genes ( Figure 2B). On the other hand, squamous cell carcinoma cases of the cervix were entirely HPV-associated and mutations occurred in the PIK3CA (60%) gene, whereas mucinous carcinomas were HPV-independent and mostly exhibited alterations in the STK11 gene ( Figure 2C and Figure S2).

Immune Characterization of Gynecological Tumors and Detected Altered Genes
Immune microenvironment along the diverse histological subtypes was studied by PD-L1 and TILs presence. We then correlated them with the genotypes obtained with the high-throughput panel. We assessed the number of genomic alterations for each patient with either PD-L1 or TILs scores in order to confirm whether those tumors with an elevated quantity of alterations may present a high score of immune markers. We did not find any association with the PD-L1 score (rho = 0.07, p = 0.68), nor with the TILs value (rho = 0.13, p = 0.41) ( Figure 5A,B). We also evaluated the PD-L1 and TILs in each histological subtypes so we could seek any evidence of an immune phenotype in a particular histology. Ovarian HGSC, uterine endometrioid carcinoma, and cervical HPV-associated cases exhibited the more elevated PD-L1 and TILs scores ( Figure 5C-F). An association was found between PD-L1 score and TILs in the entire cohort (rho = 0.48, p < 0.01) as depicted in Figure 5G. We finally compared the immune markers of those samples presenting a mismatch repair proficiency with those showing a deficiency state and an alteration in the POLE gene. We found elevated PD-L1 scores for dMMR, altered POLE, and the sum of both in comparison to the pMMR cases ( Figure 5H).  261  294  11  189  436  393  100  317  99  370  349  333  159  14  277  332  144  178  353  403  394  77   TP53  1   TBL1XR1::PIK3CA   1   Histology   Patient  375  103  65  272  163  287  97  201  325  176  384  245  437  388  237  1   Histology   Patient  146  444  137  121  36  276  246  93  37  213  390    The altered POLE cases presented a median TILs score value of 90% (38-94) and the dMMR cases, 75% . Both kind of cases expressed an elevated score (p < 0.01 and p < 0.01) in comparison to the pMMR state group which exhibited a 10% (1-60) score (Table 3).

Clinical Impact
We finally studied the utility of clinically relevant alterations and administered targeted therapies to assess the clinical impact of the routine high-throughput sequencing application in gynecological cancer patients (Table 4).

Immune Characterization of Gynecological Tumors and Detected Altered Genes
Immune microenvironment along the diverse histological subtypes was studied by PD-L1 and TILs presence. We then correlated them with the genotypes obtained with the high-throughput panel. We assessed the number of genomic alterations for each patient with either PD-L1 or TILs scores in order to confirm whether those tumors with an elevated quantity of alterations may present a high score of immune markers. We did not find any association with the PD-L1 score (rho = 0.07, p = 0.68), nor with the TILs value (rho = 0.13, p = 0.41) (Figure 5A,B). We also evaluated the PD-L1 and TILs in each histological subtypes so we could seek any evidence of an immune phenotype in a particular histology. Ovarian HGSC, uterine endometrioid carcinoma, and cervical HPV-associated cases exhibited the more elevated PD-L1 and TILs scores ( Figure 5C-F). An association was found between PD-L1 score and TILs in the entire cohort (rho = 0.48, p < 0.01) as depicted in Figure 5G. We finally compared the immune markers of those samples presenting a mismatch repair proficiency with those showing a deficiency state and an alteration in the POLE gene. We found elevated PD-L1 scores for dMMR, altered POLE, and the sum of both in comparison to the pMMR cases ( Figure 5H). About 18% of the cases received a targeted therapy based on PARP inhibitors (PARPi). Out of the 8% of cases presenting a Tier IA alteration in the BRCA1 and BRCA2 genes, only 4% were administered Olaparib treatment resulting in a partial radiological response. Another 10% received PARPi including Olaparib or Niraparib yet only a 6% presented a Tier III alteration in genes of the HR pathway. About 12% of the cases had the PD-L1 score evaluated at the time of diagnosis whereas the rest of the cases were calculated prospectively. From them, only a 4% received pembrolizumab treatment with a successive partial radiological response.
Clinical benefit was considered for those patients achieving complete or partial response whereas non-clinical benefit was reflected as progressive disease. From the five patients presenting POLE mutations, three radiological partial responses were observed whereas in two of them it remained lost to follow-up.

Total Cost and Cost Difference versus NGS
We performed an evaluation of the conventional testing strategy versus the highthroughput panel used for those more clinically relevant biomarkers in high-grade serous ovarian carcinoma. The cost per sample, the testing strategy, total cost, and cost difference are depicted in Table S6. The application of the profiling protocol including the comprehensive NGS panel and the identification of immune biomarkers (PD-L1 and TILs) resulted more profitable than the use of conventional or single gene testing.

Discussion
In this study, we showed how comprehensive high-throughput sequencing currently established in the clinical care practice enabled the characterization of mutational profiles from histological subtypes of gynecological cancer through the identification of precise alterations in cancer-associated genes. Furthermore, the use of such a comprehensive panel was able to recognize clinically relevant genomic alterations leading to genotype-guided therapy.
In our study cohort, we described histological subtypes with a major representation at the molecular level. Alterations in the TP53 and BRCA1 genes occurred in 72% and 28% of the HGSC cases. In fact, it is reported that about 20-30% of HGSC presented a BRCA mutation [33]. Standard guidelines identified germline mutations in BRCA1 and BRCA2 genes in 15% of women affected with ovarian cancer, whereas somatic mutations in an additional 7% [34]. Instead, a much-elevated proportion is described for TP53 mutations, ranging from 94% to 97% as a probably observed frequency in previous studies [35]. Our rate in TP53 mutation (72%) was similar to the observed abnormal protein expression of p53 (78%). We detected 17% of discordant cases with a wild-type genotype exhibiting an aberrant protein expression of p53 that may be explained by the sensitivity of the panel, a mutant allelic frequency less than 5% was excluded for variant review. Considering either TP53 mutation or p53 abnormal expression analysis, 89% of the HGSC cases were found to be altered, thus complying with the elevated rate widely described in the literature. Another top changed gene in this particular histological subtype contemplated amplifications in the CCNE1 gene, more observed in cases developing resistance to platinum-based chemotherapy with a frequency of 28% [36]. CCNE1 amplification in HGSOC is mutually exclusive of homologous recombination pathway mutations, including BRCA alterations, resulting in an ineffective PARP inhibition [37]. Our comprehensive analysis detected less copy number variations, only in 17% of the cases. As well, diverse alterations were encountered in genes of the HRR pathway, which is in line with other studies that described alterations in RAD51C, PALB2, or genes of the Fanconi pathway [38]. In fact, ovarian HGSC and endometrial serous carcinomas share similar molecular alterations [39], nearly each case of endometrial serous carcinoma harbors an alteration in the TP53 gene or aberrant expression of p53 [40]. However, in our cohort, only 40% of endometrial serous carcinoma exhibited a somatic alteration and 60% an aberrant protein expression. In contrast, an 80% of the cases showed an alteration in the PIK3CA gene which is associated with metastasis and is reported in a much lower frequency, between 30% and 60% [41]. Another biomarker characterizing the endometrial serous carcinoma considered amplifications in the ERBB2 gene in an approximately 25-40% rate and we found between 20 and 40% to be amplified or mutated [42]. Indeed, endometrial serous carcinoma has been found as the histology showing the highest frequency of ERBB2 amplification, up to 38%, followed by other subtypes likewise clear cell or endometrioid carcinoma as pointed out by a study evaluating 2042 endometrial carcinomas [43]. These kinds of cases were found to be high-grade and associated with TP53 alterations or CNV of MDM2 which is concordant with our two cases that expressed ERBB2 amplification concomitantly to TP53 or MDM2. In fact, current guidelines recommend ERBB2 testing in advance stage or recurrent serous endometrial cancer [44]. Despite of the low prevalence of ERBB2 alterations in gynecological tumors, the good responses observed to these therapies warrant the need to determine this genomic aberration. In addition, a few representations of altered genes FBXW7, PPP2R1A, and FGFR2 were encountered in the serous histological subtype. Recent studies demonstrate the relevance of PPP2R1A somatic alterations that may contribute to poor prognosis in patients with advance stage endometrial cancer independent of the histological subtype [45]. On the other hand, the endometrial endometrioid carcinoma was mainly characterized by somatic mutations in the PTEN gene (71%) which is the most common genomic aberration in this concrete subtype, occurring between 63% and 82% of endometrioid affected cases [14,46]. The PIK3CA gene was situated as the second more altered gene (60%), which is in accordance with described molecular genetics data in endometrial carcinomas [47]. An unexpected finding concerned the SNV of the NOTCH1 gene occurring in 48% of the endometrial endometrioid cases. Precisely, each detected variant for this gene was categorized as Tier III or variant of uncertain significance. NOTCH1 variants were called in a retrospective study conducted in 299 gynecological cancers identifying somatic mutations in the clear cell carcinoma subtype [48]. Other altered genes labelling the endometrioid cancer were CTNNB1, KRAS, POLE, and ARID1A genes, each of them accepted as genetic biomarker of the endometrioid tumors [49]. We found two cases of low-grade endometrioid carcinoma with altered CTNNB1 that is thought to show prognostic significance with worsen recurrence-free survival in early-stage low-grade endometrial carcinoma. It has recently been reported that this specific subtype with altered CTNNB1 could be incorporated in the future as the fifth histomolecular entity [39]. Finally, cervical HPV-associated cases, particularly those exhibiting a squamous cell carcinoma subtype were more altered for the PIK3CA gene (60%). Indeed, the PIK3CA gene is described as the top altered gene in cervical carcinoma along with the MTOR, KMT2D and FAT1 genes [50]. Overall, such a molecular characterization seems to define the distinct studied gynecological cancers.
Although we conducted the study in a modest cohort of 49 patients, we compared only frequencies of single somatic mutations affected cases with those listed in the TCGA database. Only PIK3CA from uterus and cervix were reasonably close whereas TP53 and PTEN were somehow distant from the TCGA frequencies.
Despite of obtaining a 2% Tier I variants in the BRCA1 and BRCA2 genes, as little as 4% of the patients received a PARPi treatment, however another 10% received PARPi exhibiting Tier III alteration in genes of the HRR pathway that summed up to a 14% of the cases. Such a scarce number of Tier I alterations found in our study raise some uncertainty about the utility of a comprehensive panel in the clinical practice of gynecological cancer. In fact, some studies have reported that 15% of patients undergoing comprehensive genotype profiling received genotype-guided therapies in solid tumors [51]. Alterations in the POLE gene are associated with hypermutated tumors and improved prognosis [52,53]. Most of the POLE mutated patients were classified as Tier III except one case interpreted as Tier II. A study analyzing 453 advanced tumors with aberrant POLE mutations also described an elevated number of variants of uncertain significance, about 69% of the studied cohort presented POLE variants of uncertain significance [54]. Despite of missing medical records regarding the radiological response it remains essential to consider a comprehensive sequencing panel including the analysis of the POLE gene. Considering the endometrial endometrioid case was at an initial stage no recurrence was seen, which is a good prognostic marker. On the other hand, the PIK3CA mutation in SSC, apart from its utility in molecularly defining the precise tumor type, could also be used as a molecular target as has been seen in another solid tumor such as in breast cancer and its associated guided-therapy alpelisib plus fulvestrant [55].
While immune checkpoint inhibitors (ICIs) have been successfully incorporated into the main therapeutic guidelines as the first line of treatment in some solid tumors such as in lung carcinoma and melanoma [56], they are not currently considered front-line in any gynecologic carcinoma types.
One of the main difficulties faced by ICIs is that there is not an established gold standard biomarker to select the most responsive patients to this kind of treatments [57]. In our study, we have analyzed two of the most used biomarkers to choose patients who are best candidates for ICIs [58], the percentage of TILs and the protein expression of PD-L1.
In endometrial cancer, we have observed an association between these two biomarkers, with higher levels, simultaneously, in the carcinomas considered to be of greater "immunogenicity", in the dMMR-hypermutated and in the POLE-ultramutated. This has already been reported in diverse studies [58][59][60] and suggest that TILs, PD-L1 [61], and tumor mutational burden (TMB) [62] could be excellent alternatives as biomarkers, especially in the POLE-ultramutated subtype, since determining the POLE mutational status may require more complex and expensive molecular techniques. In contrast, in dMMR carcinomas, a subtype which already has an FDA-approved ICIs in advanced dMMR/MSI-H carcinomas [57], the mismatch repair immunohistochemistry is an acceptable assay to select these carcinomas [56]. However, further investigation would be required to elucidate whether the response rates are higher in dMMR/MSI-H tumors with higher levels of PD-L1 and/or TILs.
In endometrial cancer, certain genetic signatures are investigated as alternative biomarkers for ICIs, such as ARID1A alterations, associated with higher TILs, supporting the view that deficient ARID1A might be a potential predictor factor for ICIs efficacy [59]. On the contrary, in non-inmunogenic endometrial cancer, recent studies have associated alterations such as CTNNB1 or PIK3CA genomic aberrations and CMYC amplification, with a low neoantigen load, that could lead to predict poor effects of immunotherapy response. So far, they could be used as potential candidates for ICIs. In our study, those endometrial cancers with an ARID1A mutation showed elevated levels of TILs and PD-L1 scores, and in contrast, cases with CTNNB1 mutated showed low levels of these two biomarkers, which would support this hypothesis.
On the other hand, we have also seen high levels of these two biomarkers in cervical squamous carcinoma, also reflected in the literature [63]. At present, among gynecological cancer, only in advanced cervical carcinomas there is an FDA-approved ICI in which PD-L1 immunostaining is enlisted for checkpoint inhibitor access based on PD-L1 CPS status ≥1 as stated in the phase 2 Keynote 158 [64] and phase 3 Keynote 826 trials [65].
Among the ovarian carcinomas from our study, we observed the highest levels of PD-L1 and TILs in HGSC which is consistent with data reported in other studies [66][67][68]. The treatment of this subtype is based on PARPi, and it has been suggested that these therapies induce the formation of neoantigens generating an immune response. This would support this subtype as a good candidate to receive ICI in combination with PARPi and in consequence determining TILs and PD-L1 would be good candidates for patient eligibility [68]. Although in lower percentages than in endometrial cancer, in ovarian cancer there are also cases exhibiting a POLE mutation and dMMR status [66,68], thereby the presence of these alterations could determine a niche to offer ICIs in ovarian cancer.
Although all these contributions need further research in gynecological tumors, it is unreasonable to think of a scenario contemplating an analysis of several ICIs biomarkers performed simultaneously instead of one of them at once. That would be to consider the analysis at the time of diagnosis of TILs and PD-L1 scores as well as the MMR, POLE, and ARID1A mutational status aimed to determine precisely which patients could be more appropriate to respond to these treatments. In our study, we demonstrate the feasibility to perform this kind of study in the clinical practice. In fact, there are many clinical trials underway in gynecological cancers and the approval of checkpoint inhibitors in combination with other treatments should be forthcoming [69]. The selection of patients more sensitive to immunotherapy remains vital and for this reason the study of immune markers concomitant to molecular biomarkers urge to be studied in patients affected by gynecological cancer.
Several limitations may be recognized in this study. The compilation of data in 49 patients probably explained the limited description of additional tumor types in gynecological cancer. The illustration of further subtype such as ovarian low-grade serous carcinoma or uterine carcinosarcoma, among others, is lacking from the performed analysis. In fact, the major number of cases appeared in HGSC as it is the most common subtype of the five principal ovarian subtypes [70]. Although the application of the comprehensive high-throughput panel permitted a confident molecular characterization of certain histology, some biomarkers are not present in the used panel such as PTEN CNV, TMB or large deletions. Commercially available tests for molecular profiling and widely spread through diagnostic laboratories include this kind of alterations that are lacking from our study [71]. As well, the limited sample size of uterus and cervix cases could probably have explained the dissimilarity found between the TCGA frequencies and those from our analysis regarding the TP53 and PTEN genes. An additional limitation of our study concerned the heterogeneity of the selected cohort. Although we were able to characterize diverse histological subtypes by the analysis of both molecular and immune biomarkers, we could not point out whether genomic alterations originated as a result of chemotherapy treatment or disease progression.

Conclusions
Lastly, the new version of the 2020 WHO classification in female genital tumors has given a special value in key molecular events as well as integrated both morphological and molecular features to attain a more refined classification of gynecological tumors [72]. In this sense, we may conclude that a molecular characterization in the clinical setting would be able to define, at least, the distinct studied gynecological tumor types by using comprehensive high-throughput sequencing. Moreover, it also permitted the identification of relevant genomic alterations related to a specific genotype-guided therapy. Finally, a routine concomitant analysis of immune biomarkers may provide a better selection of candidates for immunotherapy.