Homologous Recombination Deficiency and Cyclin E1 Amplification Are Correlated with Immune Cell Infiltration and Survival in High-Grade Serous Ovarian Cancer

Simple Summary Ovarian cancer is the deadliest gynecological cancer in developed countries of which high-grade serous ovarian carcinoma (HGSOC) is the most common subtype. How the tumor’s genetic characteristics are associated with the tissue surrounding the tumor; the tumor microenvironment (TME), is incompletely understood. Our study assessed the TME and genetic profiles of HGSOC and their associations with survival. 347 patients with HGSOC were categorized in the following profiles: BRCA mutation (BRCAm) (30%), non-BRCA mutated homologous recombination deficiency(HRD) (19%), CCNE1-amplification (13%), non-BRCAmut HRD and CCNE1-amplification (double classifier) (20%), and no specific molecular profile (NSMP) (18%). BRCAm profile showed the best survival and CCNE1 and double classifier the worst. Higher immune cell densities showed a favorable survival, also within the molecular profiles. Furthermore, immune cell densities differed per molecular profile with BRCAm profile tumors showing the highest and CCNE1 lowest densities. Our study showed that HGSOC is not one group but is grouped by different molecular profiles and TME. Abstract Background: How molecular profiles are associated with tumor microenvironment (TME) in high-grade serous ovarian cancer (HGSOC) is incompletely understood. Therefore, we analyzed the TME and molecular profiles of HGSOC and assessed their associations with overall survival (OS). Methods: Patients with advanced-stage HGSOC treated in three Dutch hospitals between 2008–2015 were included. Patient data were collected from medical records. BRCA1/2 mutation, BRCA1 promotor methylation analyses, and copy number variations were used to define molecular profiles. Immune cells were assessed with immunohistochemical staining. Results: 348 patients were categorized as BRCA mutation (BRCAm) (BRCAm or promotor methylation) (30%), non-BRCA mutated HRD (19%), Cyclin E1 (CCNE1)-amplification (13%), non-BRCAmut HRD and CCNE1-amplification (double classifier) (20%), and no specific molecular profile (NSMP) (18%). BRCAm showed highest immune cell densities and CCNE1-amplification lowest. BRCAm showed the most favorable OS (52.5 months), compared to non-BRCAmut HRD (41.0 months), CCNE1-amplification (28.0 months), double classifier (27.8 months), and NSMP (35.4 months). Higher immune cell densities showed a favorable OS compared to lower, also within the profiles. CD8+, CD20+, and CD103+ cells remained associated with OS in multivariable analysis. Conclusions: Molecular profiles and TME are associated with OS. TME differs per profile, with higher immune cell densities showing a favorable OS, even within the profiles. HGSOC does not reflect one entity but comprises different entities based on molecular profiles and TME.


Introduction
Worldwide approximately 295,000 women are diagnosed with ovarian cancer annually, and 185,000 women die due to the disease, making ovarian cancer the most deadly gynecologic malignancy in developed countries [1]. High-grade serous ovarian carcinoma (HGSOC) is the most common subtype of epithelial ovarian cancer (EOC) (approximately 70% of all EOCs) and accounts for 70-80% of all ovarian cancer deaths [2]. Even though HGSOC initially shows good response rates to platinum-and taxane-based chemotherapy, disease recurrence is frequent and often chemotherapy-resistant [2].
Large-scale genomic and epigenomic studies revealed that HGSOC is characterized by extensive copy number variations (CNV), high genomic instability, and clonal diversity [3][4][5]. The genetic makeup of HGSOC is associated with both distinct clinical and biological characteristics.
Mutational and functional alterations in genes that are involved in homologous recombination repair (HRR) mechanisms are found in approximately 50% of HGSOC [2,3,6]. The majority of these homologous repair deficient (HRD) tumors exhibit BRCA1 and BRCA2gene deficiencies [3]. BRCA1 and BRCA2 mutations are associated with higher response rates to platinum-based chemotherapy and PARP inhibitors (PARPi), and with longer survival compared to their wild-type counterparts [7][8][9]. Platinum-based therapy and PARPi exploit HRD, platinum-based therapy by inducing double-strand breaks in DNA, and PARPi by impeding tumor DNA repair via synthetic lethality. It is hypothesized that HRD tumors exhibit a high mutational load resulting in higher levels of neo-antigens. This, in turn, increases tumor-cell recognition by T-cells, facilitating an effective lymphoid immune response and also resulting in favorable survival [9].
HGSOC can also exhibit a histological and prognostic phenotype similar to the phenotype seen in BRCA1/2 mutation carriers called "BRCA-ness" [10]. BRCA-ness refers to the phenotypic characteristics of tumors lacking BRCA1/2 germline mutations that exhibit defects in HRR mimicking BRCA loss, per instance due to alterations in RAD51 or epigenetic silencing of BRCA1 [3,10,11]. Another way through which tumors can resemble BRCA-mutated tumors, is by epigenetic inactivation of the BRCA gene without alterations to its DNA sequence. Normally, the regulatory region of the full active BRCA1 gene is de-methylated. Methylation of the BRCA1 promotor leads to the incapability of BRCA1 gene transcription and therefore inactivation of the gene [12]. Hypermethylation of the BRCA1 gene-promotor occurs in 10 to 20% of EOCs [13], and such patients show a superior survival compared to patients with an unmethylated BRCA1 gene-promotor [14]. Hypothetically, BRCA-methylated EOCs could be a new subset of cancers with impaired BRCA function [14]. However, the 2020 ESMO recommendation stated that not enough evidence is available to determine the clinical validity of BRCA1 promoter methylation yet [15].
Another important genetic subgroup of HGSOC is represented by focal gene amplification of Cyclin E1 (CCNE1), present in approximately 20% of HGSOC [3]. CCNE1 is involved in tumorigenesis via induction of chromosomal and genetic instability [16]. Remarkably, previous studies demonstrated that CCNE1 amplification and BRCA mutations are largely mutually exclusive [3,6]. HGSOC with CCNE1 amplification is correlated with poor survival and primary resistance to platinum-based chemotherapy [6,17,18].
How the genetic makeup of HGSOC influences immune cell infiltration is not completely understood. To establish new prognostic biomarkers for patient outcome, as well as identify potential future therapeutic targets for different subtypes of HGSOC, knowledge of the underlying influences forming these subtypes is a prerequisite. Therefore, our study aims to assess the association of HRD and CCNE1 amplification in HGSOC with infiltration of T-cells (CD8 and CD103), B-cells (CD20), and macrophages (CD68) that have previously been associated with survival in HGSOC [19][20][21][22][23]. In addition, we analyzed whether molecular profile and immune cell infiltration are independently correlated with survival.

Patient and Tumor Selection
Patients with HGSOC International Federation of Gynecology and Obstetrics (FIGO) stage IIb-IV, who were treated with primary cytoreductive surgery (PDS) and adjuvant chemotherapy in one of three Dutch tertiary referral hospitals (Netherlands Cancer Institute-Antoni van Leeuwenhoek Hospital (NKI-AVL), Maastricht University Medical Centre (MUMC) and Amsterdam University Medical Centre (AUMC)), between January 2008 and December 2015, were eligible for the present study. Furthermore, patients from NKI-AVL treated with neoadjuvant chemotherapy (NACT) followed by interval debulking surgery (IDS) and adjuvant chemotherapy, were also included. Patients were excluded in case no tumor tissue was available for immune cell analyses or molecular analyses.
Clinical data were extracted from the Netherlands Cancer Registry (NCR) and histopathological data from the Dutch Pathology Registry (PALGA). The NCR is a nationwide registry managed by the Netherlands Comprehensive Cancer Organization (IKNL) and covers all primary malignancies in the Netherlands since 1989. The following parameters were extracted from the NCR; performance status, germline BRCA status, treatment sequence (NACT, NACT-IDS, or PDS), surgery outcome (complete with no visible disease; optimal with ≤1 cm residue or sub-optimal with >1 cm residue), and data on progression. Progression of disease was defined in case of symptoms combined with increased serum CA-125 levels, radiological signs of progression, or histological or cytological confirmation of recurrent disease. Vital status and date of death were obtained by the NCR via linkage with the municipal population registration. Pathological data and tumor tissue blocks were obtained from the nationwide network PALGA, which registers all records of histopathology and cytopathology with full coverage since 1991 [24].

Tissue Samples and Tissue Microarrays (TMA)
Formalin-fixed, paraffin-embedded (FFPE) tissue blocks from all patients with primary HGSOC were obtained. Tissue blocks originated from samples retrieved during debulking surgery, which subsequently resulted in pretreated tumor blocks in case a patient received NACT. All cases underwent pathological re-review based on conventional morphological examination of sections stained with hematoxylin and eosin (H&E) by three dedicated pathologists (K.V.d.V., H.H., J.S.). Tumor grade was designated according to the binary grading classification (to exclude low-grade serous ovarian carcinoma) [25].
The paraffin tissue blocks were organized into TMAs. Representative areas of the center and peripheral invasive margin of the ovarian tumor were selected on whole-tissue FFPE H&E stained slides for immune and tumor cell scoring. In case ovarian tumor tissue was not available a representative tumor block from another location (peritoneum, omentum) was selected. In each tumor four cores were selected, optimally representing tumor and peripheral stroma containing immune cell infiltrate. TMAs with one mm-sized cores were constructed using a tissue microarrayer (Grand Master, Sysmex Europe GmbH, Norderstedt, Germany). To enable adhesion of the cores to the recipient paraffin block, the block was melted at 70 • C for nine minutes and cooled down overnight.

Molecular Analyses
DNA isolation from FFPE tissue blocks with a minimal tumor percentage of 20% was performed fully automated according to standard protocols using the Qiacube (Qiagen, Hilden, Germany). Ten serial sections of 10 µm thickness of each tumor were taken using a Finesse ME+ microtome (Thermo Fisher Scientific, Waltham, MA, USA) and deparaffinized using the ST5020 multi-stainer (Leica Microsystems, Wetzlar, Germany). DNA isolation was performed with AllPrep RNA/DNA FFPE Kit (Qiagen, Hilden, Germany). The manufacturer's instructions were followed for DNA isolation using the Qiacube (Qiagen, Hilden, Germany). Quantification of the concentration and purity of all DNA extracts was performed using a NanoDrop-8000 spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA). Measurement of double-stranded DNA (dsDNA) yield was performed with a Qubit dsDNA High Sensitivity Assay Kit (Thermo Fisher Scientific, Waltham, MA, USA).

Molecular Analyses
DNA isolation from FFPE tissue blocks with a minimal tumor percentage of 20% performed fully automated according to standard protocols using the Qiacube (Qia Hilden, Germany). Ten serial sections of 10 μm thickness of each tumor were taken u a Finesse ME+ microtome (Thermo Fisher Scientific, Waltham, MA, USA) and depa finized using the ST5020 multi-stainer (Leica Microsystems, Wetzlar, Germany). DNA lation was performed with AllPrep RNA/DNA FFPE Kit (Qiagen, Hilden, Germany). manufacturer's instructions were followed for DNA isolation using the Qiacube (Qia Hilden, Germany). Quantification of the concentration and purity of all DNA extracts performed using a NanoDrop-8000 spectrophotometer (Thermo Fisher Scientific, W tham, MA, USA). Measurement of double-stranded DNA (dsDNA) yield was perfor with a Qubit dsDNA High Sensitivity Assay Kit (Thermo Fisher Scientific, Waltham, USA).
Molecular profiles were determined in a stepwise manner. Germline BRCA muta information was obtained from the NCR database. In all patients without a kn germline BRCA mutation, tumor BRCA mutation was determined. In cases without BR mutation, BRCA1 promotor methylation status was determined. Patients without a BR mutation or BRCA1 promotor methylation were further analyzed with CNV sequen for non-BRCA mutation (non-BRCAmut) HRD profile and CCNE1 amplification base low-coverage whole-genome sequencing data. This stepwise manner was used as BR methylation and CCNE1 amplification are mutually exclusive with BRCA mutation as BRCA-ness refers to non-BRCA mutated tumors mimicking BRCA loss [3,10,13]. In remaining patients in which none of the aforementioned molecular profiles were fo were categorized as "no specific molecular profile" (NSMP).

Somatic BRCA Analysis
Quality control of DNA was performed using the QC plex kit (Agilent Technolog Santa Clara, CA, USA). The manufacturer's instructions were followed, using the 2 bioanalyzer (Agilent Technologies, Santa Clara, CA, USA). BRCA1 and BRCA2 som mutation analyses were performed with the BRCA MASTR Plus Dx kit (Agilent tech ogies, Santa Clara, CA, USA) according to the manufacturer's instructions, using a V Thermocycler (Thermo Fisher Scientific, Waltham, MA, USA) and the MiSeq (Illum San Diego, CA, USA). The results of the MiSeq run were uploaded to the MASTR repo (Agilent technologies, Santa Clara, CA, USA) for further analysis. Molecular profiles were determined in a stepwise manner. Germline BRCA mutation information was obtained from the NCR database. In all patients without a known germline BRCA mutation, tumor BRCA mutation was determined. In cases without BRCA mutation, BRCA1 promotor methylation status was determined. Patients without a BRCA mutation or BRCA1 promotor methylation were further analyzed with CNV sequencing for non-BRCA mutation (non-BRCAmut) HRD profile and CCNE1 amplification based on low-coverage whole-genome sequencing data. This stepwise manner was used as BRCA1 methylation and CCNE1 amplification are mutually exclusive with BRCA mutation and as BRCA-ness refers to non-BRCA mutated tumors mimicking BRCA loss [3,10,13]. In the remaining patients in which none of the aforementioned molecular profiles were found were categorized as "no specific molecular profile" (NSMP).

Somatic BRCA Analysis
Quality control of DNA was performed using the QC plex kit (Agilent Technologies, Santa Clara, CA, USA). The manufacturer's instructions were followed, using the 2100 bioanalyzer (Agilent Technologies, Santa Clara, CA, USA). BRCA1 and BRCA2 somatic mutation analyses were performed with the BRCA MASTR Plus Dx kit (Agilent technologies, Santa Clara, CA, USA) according to the manufacturer's instructions, using a Veriti Thermocycler (Thermo Fisher Scientific, Waltham, MA, USA) and the MiSeq (Illumina, San Diego, CA, USA). The results of the MiSeq run were uploaded to the MASTR reporter (Agilent technologies, Santa Clara, CA, USA) for further analysis.

BRCA Promotor Methylation
To determine BRCA1 promotor methylation status, Methylation-specific MLPA (MS-MLPA) was performed using a commercial kit (ME0053 kit, MRC Holland, Amsterdam, The Netherlands). The manufacturer's instructions were followed to determine hypermethylation with the use of the Veriti thermocycler (Thermo Fisher Scientific, Waltham, MA, USA). Fragment analysis was performed using the genetic analyzer (ABI-3500, Thermo Fisher Scientific, Waltham, MA, USA). The probes used were designed to contain a HhaI recognition site (GCGC) and thus target one CpG dinucleotide within a CpG island. If the HhaI recognition site is not methylated, HhaI will cut the probe-sample DNA hybrid and no PCR product will be formed. If the target DNA is methylated the fragment will be amplified during subsequent PCR [27]. The total amount of DNA was quantified on the Nanodrop 2000 (Thermo Fisher Scientific, Waltham, MA, USA), and the amount of double-stranded DNA in the genomic DNA samples was quantified with the Qubit dsDNA HS Assay Kit (Invitrogen, Waltham, MA, USA, cat no Q32851). Because of variable DNA extraction efficiencies or sample sizes, varying input amounts of double-stranded DNA were used (from 34 ng to 964 ng). These quantities were Covaris sheared in a standard volume of 130 µL, bead cleaned, and eluted in 50 ul elution buffer that was used to the full extent to start library preparation. Samples were purified using 2× Agencourt AMPure XP PCR Purification beads according to the manufacturer's instructions (Beckman Coulter, Brea, CA, USA, cat no A63881). Sheared samples were quantified and qualified on a BioAnalyzer system using the DNA7500 assay kit (cat no. 5067-1506, Agilent Technologies, Santa Clara, CA, USA). Library preparation for Illumina sequencing was performed with a maximum input of 1 µg sheared DNA using the KAPA Hyper Prep Kit (KK8504, KAPA Biosystems, Wilmington, MA, USA). During library enrichment, 6 PCR cycles were used to obtain enough yield for sequencing. After library preparation, the libraries were cleaned up using 1× AMPure XP beads. Libraries were analyzed with a BioAnalyzer system using DNA7500 chips to define molarity. Three pools were created, two with ninety-three and one with sixty-three uniquely indexed samples were mixed together by equimolar pooling, in a final concentration of 10 nM, and subjected to sequencing on an Illumina HiSeq2500 machine in a total of twenty-two lanes of a single read 65 bp run, according to the manufacturer's instructions.

Non-BRCAmut HRD Classification
A non-BRCAmut HRD copy number profile was determined according to methods described previously by Schouten et al. [28]. In short, reads were aligned to the reference genome GRCh38 using BWA-MEM (version 0.7.17). Reads with a mapping quality of over 15 were counted in 20 kb non-overlapping bins, corrected for CG bias, and corrected for local alignment-bases estimated mappability, resulting in 2 log count ratios. The 20 kb resolution 2 log ratios were mapped to the 1 MB resolution input for the classifier. Subsequently, we corrected the centering and scaling of the data between the sequencing platform of the current study and the oligonucleotide array platform on which the classifier was created. We fitted a linear regression model with Gaussian distribution and identity link function using the R glm function to the sorted location-wise average of the training set and the current dataset. The obtained alpha coefficient to correct the centering and the obtained beta coefficient to correct the scaling of the current data.
All genomic profiles underwent automated and manual quality control. The profiles were subsequently classified using the described shrunken centroids classifier. Profiles with a posterior probability of >0.5 were classified as non-BRCAmut HRD, and profiles with a posterior probability of ≤0.5 were classified as non-BRCA-like.

CCNE1 Classification
A FastQC [29] report was generated for quality control of sequencing reads and further checked in MultiQC [30]. Reads with adapter sequences were trimmed using Trimmomatic [31], and correct trimming was confirmed with a second MultiQC report. Reads were aligned to the hg19 reference genome using BWA mem, and subsequently sorted and indexed using Samtools [32]. Duplicates were marked using Picard MarkDuplicates [33]. Copy Number Profiles were generated using the QDNAseq suite [34]. Reads were binned into 30 kb-sized bins. Blacklisted bins were removed using default filtering. Bins were corrected for GC-content and mappability. Finally, read counts were log2 normalized and outliers were removed. Standard QDNAseq segmentation quality using Circular Binary Segmentation (CBS) resulted in noisy segmentation in some cases. Noisy samples were smoothed (Supplementary Methods S1) while retaining unbiased segmentation in high-quality profiles. Finally, Copy numbers were called using QDNAseq, using its implementation of CGHcall.

Statistical Analyses
Statistical analyses were performed in STATA/SE (version 14.1, STATA CORP, College Station, TX, USA). A p-value < 0.05 was considered significant. The influence of molecular profile was explored by the unique profiles: BRCA mutation (BRCAm) profile (BRCA 1/2 mutation or promotor methylation), non-BRCAmut HRD, CCNE1 gain/amplification, double classifier (non-BRCAmut HRD and CCNE1 gain/amplification), and NSMP. Basic patient characteristics and immune cell densities of these profiles were assessed with Chisquare tests for categorical variables, One-way ANOVA for normally distributed continuous variables, and Kruskal-Wallis for non-normally distributed continuous variables. Ordinal logistic regression was used to investigate the influence of molecular profiles on immune cell densities and the Chi-square test to investigate the relationship between tumor regression and immune cell density groups. Kaplan-Meier survival estimates with the corresponding logrank test and univariable and multivariable Cox regression analyses were used to assess the effect of molecular profile on progression-free survival (PFS) and overall survival (OS). Those found significant in univariate analyses with a p < 0.10, were included in the multivariable regression analyses and assessed using backward selection. PFS was calculated as the time between the start of primary treatment and the date of recurrence or date of death of disease (DoD). OS was calculated as the interval between the start of treatment and the DoD, or if alive, the date of the last check of the municipal population register (31 January 2021). In case no event had occurred (recurrence nor death), patients were right censored at the time of the last follow-up.

Molecular Profiles
We were able to assess the molecular profile in 348 out of 360 patients. Twelve patients were excluded as a result of insufficient DNA quality. The molecular profiles are depicted in Figure 2. In 78 out of the 348 patients (22%), a BRCA mutation was present: 25 and 25 patients with a germline or somatic BRCA1 mutation, respectively, and 18 and 10 patients with a germline or somatic BRCA2 mutation. 27 patients (8%) had a BRCA1 promotor methylation resulting in a total 105 patients (30%) with a BRCAm profile. In 67 patients (19%) a non-BRCAmut HRD profile was found. In 45 patients (13%) increased CCNE1 copy numbers (gain n = 28; amplification n = 17) were detected. 69 patients (20%) depicted both a non-BRCAmut HRD profile and increased CCNE1 copy numbers. 62 patients (18%) had no specific molecular profile.   Table 2 lists the immune cell densities, stratified by molecular profiles. The number of immune cells did not significantly differ between the molecular profiles, except for CD68+ cells, which were highest in BRCAm profile tumors (>100 cells 44.8%, compared to 16-22% in other molecular profiles). Although non-significant, compared to the other profiles BRCAm profile tumors showed higher amounts of CD8+ cells (32.4% > 100 cells),  Table 2 lists the immune cell densities, stratified by molecular profiles. The number of immune cells did not significantly differ between the molecular profiles, except for CD68+ cells, which were highest in BRCAm profile tumors (>100 cells 44.8%, compared to 16-22% in other molecular profiles). Although non-significant, compared to the other profiles BRCAm profile tumors showed higher amounts of CD8+ cells (32.4% > 100 cells), CD103+ cells (27.6% > 100 cells) and CD20+ cells (20.9% > 50 cells). Tumors with a CCNE1 amplification/gain showed the lowest amount of these immune cells (>100 CD8+ cells 20.0%; >100 CD103+ cells 11.1%; >50% CD20+ cells 4.4%).

PFS and OS of Molecular Profiles and Immune Cell Infiltration
The median follow-up of the total cohort was 38.2 months (IQR 22-66). Kaplan-Meier curves showed a significant association between molecular profile and overall survival (OS) (Logrank = 0.0003) ( Figure 3A). Median OS was most favorable in patients with a BRCAm profile (52.5 months), followed by non-BRCAmut HRD (    Kaplan-Meier curves were generated for OS per immune cell type ( Figure 4) and further subgrouped by the two treatment groups: PDS and NACT (Supplementary Figure  S2). Defined as high and low densities, CD8+, CD68+, and CD103+ cells showed a significant association with survival (Logrank; p = 0.0171, p = 0.0151, p = 0.0015, respectively). In all cases, a lower density resulted in poorer survival. Lower densities of CD20+ cells also showed a poorer survival, yet non-significant. Median OS was most favorable in patients with higher immune cell densities (CD8+ cells; 44 (Table 4). Within the treatment types a significant association between immune cell densities and OS was seen in case of PDS in CD8+, CD20+, and CD103+ cells (Logrank; p = 0.0093, p = 0.0148, p = 0.0180, respectively). In NACT there was also a favorable survival in case of higher immune cell densities, yet nonsignificant (Supplementary Figure S2). Furthermore, an association was also seen within the different molecular profiles. Although predominantly non-significant, presumably due to low numbers, higher immune cell densities showed the tendency of a favorable OS within the several molecular profiles, except for CD20+ cells in BRCAm profile patients (Supplementary Table S1). Higher immune cell densities also resulted in a favorable PFS (Supplementary Figure S3, Supplementary Table S2).

Multivariable Analyses
After adjustment for age, FIGO stage, therapy sequence, and completeness of debulking surgery in multivariable cox regression, the associations of molecular profile with OS remained consistent, except for NSMP ( Figure 3B). Compared to the BRCAm profile group, CCNE1 amplification and double classifier profile correlated significantly with poorer OS (HR 1.75; 95%CI 1.16-2.65 and HR 1.53; 95%CI 1.07-2.18, respectively). The same accounted for PFS (Table 3). Compared to the BRCAm profile group, CCNE1 amplification and double classifier profile both correlated significantly with poorer PFS (HR 1.57; 95%CI 1.02-2.43 and HR 1.77; 95%CI 1.22-2.56, respectively). Neither the immune cell densities nor the interaction between immune cell densities and molecular profiles were significant confounders.
Multivariable Cox regression showed that immune cell densities remained associated with OS after adjustment for age, FIGO stage, therapy sequence, and completeness of debulking surgery (Table 4). Higher CD8+, CD20+ and CD103+ densities resulted in a significant favorable OS compared to lower densities (HR 0.72; 95%CI 0.55-0.94, HR 0.71; 95%CI 0.51-0.98 and HR 0.68; 95%CI 0.50-0.93, respectively). CD68+ cells showed a favorable OS for higher densities in the univariate analysis but were non-significant in the multivariate analysis. Notably, in multivariable cox regression, only CD20+ densities remained associated with PFS after adjustment for age, FIGO stage, therapy sequence, and completeness of debulking surgery (Supplementary Table S2).

Discussion
Our study reports an improved OS in patients with a BRCA mutation or promotor methylation and a worse OS in patients with a CCNE1 amplification/gain. This is in line with previous reports describing a correlation between OS and response to platinum-based chemotherapy, which is enhanced in BRCA-mutated tumors and decreased in CCNE1 amplified tumors [17,18,[35][36][37]. Furthermore our study depicted an improved OS in patients non-BRCAmut HRD supporting the hypothesis that non-BRCAmut HRD and possibly BRCA1 promotor methylated EOCs are an important subset of cancers with impaired HRR [10,14].
Our study confirmed that immune cell infiltrates are associated with OS. Higher CD8+, CD20+, and CD103+ cell densities resulted in a more favorable survival compared to lower cell densities. These results are in agreement with previous studies [19][20][21]. This was also seen within the two treatment types, PDS and NACT, even though nonsignificant in case of NACT. The link between tumor infiltration with macrophages and patient survival is more complex with conflicting study results. Tumor-associated macrophages have been described to promote cancer progression [38,39]. In contrast, our cohort and other previous studies demonstrated an association between higher levels of macrophages and improved OS [22] while others show a negative or no influence on OS [23,40] insinuating a complex role of immune cells and other factors influencing OS.
We found that immune cells were most prominent in BRCA mutated or BRCA1 promotor methylation patients. BRCA gene mutations have been correlated with increased immune cell infiltration in HGSOC [9,20,41]. Immune cell infiltration, in turn, has been associated with increased response to immunotherapy [42]. Although immunotherapy is not proven to be beneficial (yet) in the treatment of HGSOC, our results suggest that BRCA mutated HGSOC will be the most eligible candidates for immunotherapy in the future, in contrast to CCNE1 amplified HGSOC.
In EOC patients, non-BRCAmut HRD has been correlated with a favorable response to platinum-based chemotherapy and PARPi [43], similar to the benefit that has been seen from DNA double-strand-break-inducing chemotherapy in breast cancer patients with non-BRCAmut HRD profiles compared to patients without this profile [44][45][46]. In Zhang et al. investigated non-BRCAmut HRD using whole-exome deep sequencing data from the TCGA and showed a similar OS in BRCA mutated and non-BRCAmut HRD ovarian cancer patients, which was significantly better compared to patients without HRD [47]. In the present study, we showed a moderate beneficial effect of non-BRCAmut HRD tumors on OS of approximately 6 months, whereas patients with a BRCAm profile showed a significant increase of 17 months in OS, compared to our NSMP group. The nonsignificance of the effect in patients with non-BRCAmut could be explained by the relatively good performance of the NSMP group, from which patients with a CCNE1 gain/amplification, with a significantly worse OS, were excluded. Presumably, this explains the differences between our study and the results of non-BRCAmut HRD reported in the literature. Furthermore, the survival curve of the double classifier group tends to follow the survival curve of the CCNE1 gain/amplification group and shows a similar median survival rate. This suggests a more dominant influence of CCNE1 gain/amplification compared to the influence of non-BRCAmut HRD.
Remarkably, the immunological response in non-BRCAmut HRD tumors shares similarities with the response in tumors without a BRCA mutation. Hypothetically, non-BRCAmut HRD tumors display lower levels of neo-antigens, which leads to less tumor-cell recognition by T-lymphocytes, compared to BRCA mutated tumors. More clinical trials must confirm to what extent patients with non-BRCAmut HRD tumors are comparable to patients with tumors harboring somatic or germline BRCA mutations in clinical outcome and response to immunotherapy.
Activation of the RB1/CCNE1 pathway is considered to be largely exclusive to BRCA mutations [3,48]. CCNE1 activates transcription of BRCA1 and BRCA2 genes, thereby stimulating the HRR pathway, which gives CCNE1 amplified tumors the ability to better withstand DNA double-strand-break systemic therapy [49]. Our results also suggest that CCNE1 amplified tumors are characterized by relative chemotherapy resistance and therefore might require a different treatment approach. Moreover, we showed that immune cells are less abundant in this molecular profile, suggesting that patients with CCNE1amplified tumors could be less likely to respond to immunotherapy. Other treatment strategies for CCNE1 amplified, BRCA-wildtype tumors have not been investigated in clinical trials yet. However, in both in vitro and in vivo studies, CCNE1 amplified HGSOC showed sensitivity to Cyclin-dependent kinase (CDK)-inhibitor dinaciclib and showed synergism with AKT-inhibitor MK-2206 [50].
Even though immune cell densities differed between the distinct molecular profiles, possibly warranting different treatment approaches, the tendency of a favorable OS for patients with higher immune cell densities was seen within molecular profiles as well. This finding indicates that even within the favorable, or non-favorable, molecular profiles immune cell densities strongly influence survival.
The strength of the present study is that we integrated detailed genomic, immunological, and clinical data from a large cohort of patients with HGSOC. Our data defines distinct prognostic profiles of HGSOC based on molecular and immune profiles. A central pathology rereview of all histological tissue was performed by dedicated pathologists in gynecologic oncology. Lastly, a long follow-up was achieved, and clinical data were complete. This study is not without limitations. A limitation is that immune cells were scored on tissue derived from PDS, but also after NACT. Pre-NACT immune cell levels have shown to be different from post-NACT levels, predominantly with higher levels post-NACT [51,52]. Ideally, immune cells are scored before, during, and after treatment with NACT. Nonetheless, pre-and post-NACT immune cells have been associated with OS and PFS in the same manner, with higher levels being associated with a favorable OS. Including both treatment types allowed us to give an accurate representation of the real-life setting and allowed us to investigate immune cell levels in both groups. Another limitation of our study is that we only determined CCNE1 amplification and non-BRCAmut HRD profile of patients who had no BRCA mutation. In the literature, CCNE1 gain/amplification and BRCA mutation have shown synthetic lethality [3,48]. However, we did identify 69 patients with both a CCNE1 amplification/gain and a non-BRCAmut HRD profile. This is the first study reporting on the presence of CCNE1 amplification and non-BRCAmut HRD simultaneously in a large group of HGSOC (20%). Our results implicate that the tumor microenvironment in this group is similar to non-BRCAmut HRD, however, the OS of this double classifier group is worse than OS in the non-BRCAmut HRD group and shows similarities to the group with CCNE1 amplification. The exact prognostic and clinical relevance, and underlying mechanisms of this finding are yet unclear. Finally, the extent of the influence of immune cell densities within the molecular profiles could not be determined due to inadequate power.

Conclusions
Our results emphasize that HGSOC does not reflect one entity, but comprises different variants based on molecular profiles and tumor microenvironment, which in the future is ideally translated into tailored treatment approaches. Further research is warranted to clarify to what extent molecular profiles are correlated with therapy sequence and response to current targeted therapeutic modalities including PARPi. Additional research is necessary regarding treatment strategies for not only BRCA and non-BRCAmut HRD patients but also for CCNE1 amplified patients. The present study further classified HGSOC into molecular and immunological profiles, which could serve as a basis for future research on new treatment modalities.
Supplementary Materials: The following supporting information can be downloaded at: https:// www.mdpi.com/article/10.3390/cancers14235965/s1, Methods S1: Copy number profile smoothing for CCNE1 classification; Figure S1: Flowchart of patient selection and tissue retrieval; Figure S2: Survival analyses of patients with HGSOC, according to immune cell densities and therapy sequence.; Figure S3: Survival Analyses of Patients with HGSOC, According to immune cell densities.; Table S1: Median overall survival of immune cell densities per molecular subtype; Table S2: Progression-free survival analysis, immune cell composition.