A Proteogenomic View of Synchronous Endometrioid Endometrial and Ovarian Cancer.

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact , Ernst Lengyel: [email protected] This study did not generate new materials. The mass spectrometry proteomics data have been submitted to the ProteomeXchange Consortium ( http://proteomecentral.proteomexchange.org ) via the PRIDE partner (RRID:SCR_003411) with the dataset identifier PXD059172. The lead contact will share all data reported in this paper upon reasonable request. Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request. The study was approved by the Institutional Review Board of the University of Chicago, the University of British Columbia Office of Research Ethics, and the University of Tübingen Ethics Committee and was in accordance with the Declaration of Helsinki. Written informed consent was obtained from each patient. Targeted genomics was performed on FFPE samples. First, nucleic acids were extracted using the QIAamp DNA Tissue Kit (Qiagen) according to the manufacturer’s instructions. 100ng of total DNA per sample was then sheared to an average DNA length of 200bp using a Covaris ultrasonicator. Libraries were then generated using the QIAseq Targeted DNA Human Comprehensive Cancer Panel (Qiagen) according to the manufacturer’s instructions. Sequencing of final libraries was performed by Novogene (Novogene Co., Ltd) on a HiSeq 4000. Targeted DNA sequencing data from QIAseq were processed, and somatic variants were called using smCounter2 ( 15 ), a UMI-based variant caller for target sequencing data with default parameters. The reference genome used was human hg19. The called variants were then annotated with ANNOVAR (RRID:SCR_012821, https://annovar.openbioinformatics.org/en/latest/ ) ( 16 ) and R maftools (RRID:SCR_024519) ( 17 ) was applied to visualize the results and perform downstream analysis, which included driver gene detection and analysis of pathways. Phylogenetic trees based on the variants were reconstructed using LICHeE ( 18 ). We assessed the clonal relationship of the ovarian- and endometrial carcinomas in each SEOC patient to investigate recent findings that the carcinomas are clonally related. This assessment was performed by the clonality index calculation described in ( 19 ). Where P k is the proportion of cases (23 samples of single ovarian- or endometrial carcinomas from this study, 512 samples of ovarian- or endometrial carcinomas from TCGA) harboring a given mutation from the sequencing panel of this study, and "n" is the number of mutations shared between two carcinomas. TCGA data were retrieved from the UCSC Xena Browser (RRID:SCR_018938) ( 20 ). For mutations not identified in the TCGA cohort, we conservatively set Pk as 1/(23+512) instead of zero. A pair of SEOCs was deemed clonal if the clonality index was above 0.995. Mutation data from 2843 COSMIC carcinoma samples (2405 uterine, 438 ovarian) was retrieved from: https://cancer.sanger.ac.uk/cosmic/download ( 21 ). Mutation data from 512 TCGA carcinoma samples (446 uterine, 62 ovarian) were retrieved from the UCSC Xena browser as described in the clonality index section above. Mutation rates in the genes of the cancer panel were compared to the samples sequenced in this study (27 synchronous cancers, 20 uterine cancers and 8 ovarian cancers). For spatial proteomics and laser microdissection (LMD), 10 μm FFPE tissue sections were mounted on polyethylene naphthalate (PEN) membrane slides (ThermoFisher). Prior to LMD, each individual slide was hematoxylin and eosin (H&E) stained. Using a Leica LMD6 microscope, areas of interest (tumor or stromal compartment) were chosen for sampling based on H&E-stained tissue morphology. A total volume of 50 nl tissue was collected per sample (5E6 μm 2 , 10 μm thick) into adhesive caps for proteomics analysis. Collected tissue was transferred in 100 μl lysis buffer (300 mM Tris/HCl pH8, 50% 2,2,2-trifluoroethanol) into 0.2 ml PCR tubes for further processing. After controlled heating (to avoid cap opening from overpressure) for 90 min at 90°C, samples were sonicated in a Bioruptor (15 cycles, duty cycle 50%) and vacuum-dried for approximately 1 h at 60°C until 20 μl remained. Dithiothreitol was added for a final concentration of 5 mM and incubated for 20 min at room temperature. 20 mM chloroacetic acid was added for alkylation and samples were incubated for another 20 min at room temperature. Liquid chromatography-grade water was added to adjust the sample volume to 100 μl and tryptic digestion started by the addition of LysC and Trypsin at an enzyme to protein ratio of 1:100. Samples were digested overnight at 37°C at 1400 rpm. The next day, trifluoracetic acid was added for a final concentration of 1% (v/v) to stop digestion. Peptide clean-up was carried out using StageTip. Details of the entire microproteomics workflow can be found in our publication ( 22 ). Liquid chromatography-mass spectrometry (LC-MS) analysis was performed with an EASY-nLC-1200 system (Thermo Fisher Scientific) connected to a trapped ion mobility spectrometry quadrupole time-of-flight mass spectrometer (timsTOF Pro, Bruker Daltonik GmbH, Germany) with a nano-electrospray ion source (Captive spray, Bruker Daltonik GmbH). Peptides were loaded on a 50 cm in-house packed HPLC-column (75 μm inner diameter packed with 1.9 μm ReproSilPur C18-AQ silica beads, Dr. Maisch GmbH, Germany). Peptides were separated using a linear gradient from 5–30% buffer B (0.1% formic acid, 80% ACN in LC-MS grade H2O) in 86 min followed by an increase to 60% buffer B for 14 min, then to 95% buffer B for 10 min and back to 5% buffer B in the final 10 min at 300nl/min. The total gradient length was 120 min. Buffer A consisted of 0.1% formic acid in LC-MS grade water. We used an in-house made column oven to keep the column temperature constant at 60°C. Mass spectrometric analysis was performed essentially as described in Meier et al. ( 23 ) in data-dependent (ddaPASEF) mode. 1 MS1 survey TIMS-MS and 10 PASEF MS/MS scans were acquired per acquisition cycle. Ion accumulation and ramp time in the dual TIMS analyzer were set to 100 ms each, and we analyzed the ion mobility range from 1/K0 = 1.6 Vs cm-2 to 0.6 Vs cm-2. Precursor ions for MS/MS analysis were isolated with a 2 Th window for m/z 700 in a total m/z range of 100–1.700 by synchronizing quadrupole switching events with the precursor elution profile from the TIMS device. The collision energy was lowered linearly as a function of increasing mobility, starting from 59 eV at 1/K0 = 1.6 VS cm-2 to 20 eV at 1/K0 = 0.6 Vs cm-2. Singly charged precursor ions were excluded with a polygon filter (otof control software, Bruker Daltonik GmbH). Precursors for MS/MS were picked at an intensity threshold of 1.000 arbitrary units (a.u.) and resequenced until reaching a ‘target value’ of 20.000 a.u., taking into account a dynamic exclusion of 40 s elution. Mass spectrometric raw files acquired in ddaPASEF mode were analyzed with MaxQuant ( 24 ) (version 1.6.7.0, RRID:SCR_014485). The Uniprot database (RRID:SCR_002380, 2021 release, UP000005640_9606) was searched with a peptide spectral match (PSM) and protein level FDR of 1%. A minimum of seven amino acids was required, including N-terminal acetylation and methionine oxidation as variable modifications and cysteine carbamidomethylation as a fixed modification. Enzyme specificity was set to trypsin with a maximum of two allowed missed cleavages. First and main search mass tolerance was set to 70 ppm and 20 ppm, respectively. Peptide identifications by MS/MS were transferred by matching four-dimensional isotope patterns between the runs (MBR) with a 0.4-min retention-time match window and a 0.05 1/K0 ion mobility window. Label-free quantification was performed with the MaxLFQ algorithm ( 25 ) and a minimum ratio count of one. The reported MaxLFQ label-free quantification values were further filtered as described below. Proteomics data analysis was performed with Perseus (version 1.6.15.0, RRID:SCR_015753) ( 26 ), within the R environment (RRID:SCR_001905, https://www.r-project.org/ ) and Python (RRID:SCR_008394). MaxQuant output tables were filtered for ‘Reverse’, ‘Only identified by site modification’, and ‘Potential contaminants’ before data analysis. In addition, only unique protein groups with a minimum number of two razor or unique peptides were kept for analysis, resulting in 8,253 unique protein groups. Due to a low identification rate, one sample was excluded, resulting in 152 quantified proteomes. Replicates of adjacent tissue regions were acquired for three dual cancer samples, and resulting proteome measurements were averaged prior to analysis. The final dataset comprised 144 proteome samples, 88 from the tumor and 56 from the stromal compartment. Missing values were imputed after stringent data filtering (maximum 30% missing values per protein group) based on a normal distribution (width = 0.3; downshift = 1.8). For pairwise proteomic comparisons (two-sided unpaired t-test), we applied a permutation-based FDR of 5% to correct for multiple hypothesis testing including an s 0 value ( 27 ) of 0.1. Multiple group comparisons were performed based on ANOVA analysis with multiple hypothesis correction performed with permutation-based FDR of 5%. For protein feature ranking, normalized protein measurements were subjected to feature importance analysis using the recently introduced open-source tool, OmicLearn ( 28 ). The ExtraTrees method was used for feature selection and AdaBoost was selected as the ML algorithm.

We performed a systematic proteogenomic analysis of three patient groups (64 patients) diagnosed after their surgery with endometrioid tumor histology in either the ovary, the uterus, or in both organs ( Fig. 1A ). The group with single-site endometrioid ovarian cancer (EOC) contained 9 patients and the group with single-site endometrial endometrioid cancer (EEC) contained 26 patients. Twenty-nine patients with synchronous cancers/ SEOC, an endometrioid tumor in both the uterus and the ovary, made up the third group. Patients received treatment at one of three international gynecologic oncology centers (Chicago, Vancouver, Tübingen). ( Fig. 1A - B , Supplementary Table S1 ). All synchronous cancers were reviewed by a gynecologic pathologist (RL) and met the Scully criteria for synchronous cancer ( 1 ). We first analyzed the genomic data from SEOCs, which revealed that missense mutations were the most dominant mutation type (> 90%), followed by nonsense and splice site mutations ( Supplementary Fig. 1A ) C>T single nucleotide variations (SNV) were the most frequent SNV class (>90%, Supplementary Fig. 1B ). Known drivers previously described as mutated in SEOC ( 6 , 7 ), such as PTEN , PI3KC A and ARID1A , were among the top ten altered genes ( Fig. 1C , Supplementary Table S2 ). In agreement with previous studies ( 5 – 7 ), 25 of the 27 SEOCs were clonally related and showed high variability in the number of shared core mutations between cancer sites (1 – 88, Fig. 1D - E , Supplementary Table S3 ). We re-confirmed a clonal relationship for eight SEOCs (clonality indexes > 0.995, Supplementary Fig. 1C , D ), which we had previously characterized using a different gene panel ( 6 ), but were not able to confirm the clonal relationship for two Vancouver cases (VAN _04 and VAN_22). This could be partly explained by the fact that our studies were performed on two independent samples from potentially heterogenous tumors with a varying mutation burden as well as the differences in the gene panels; our data only covered one of the genes, where shared mutations were previously found, for each case. The inclusion of single-site EOC and EEC patients, in addition to dual cancer cases, allowed us to investigate genomic and proteomic features unique to SEOCs despite their similar histologic appearance ( Fig. 2A ). Chromatin remodeling and transcription factor related genes, including the KMT gene family ( 29 ), EP300 , ARID1A , and KAT6A were mutated in all three patient groups, as well as CTNNB1 (β-catenin), LRP1B , NOTCH2 and PIK3CA , pointing to common genetic drivers of endometrioid histology, independent of the anatomical site ( Fig. 1C , Fig. 2B , Supplementary Fig. 2A ). A comparison to the TCGA pan-cancer study ( 30 ) revealed that the top mutated genes in our entire cohort overlapped to a much greater extent with the genetic profile of uterine endometrioid carcinoma than with epithelial ovarian cancers of non-endometrioid histology ( Fig. 2C ). Mutational signatures found in SEOCs shared similarities with EEC and differed from EOC ( Fig. 2D ). This was particularly apparent in the disease drivers PTEN and ARID1A , and in many other cancer-related genes (e.g., CREBBP , NOTCH2 , ERBB4 ) ( Fig. 2D , E ). We investigated whether endometrial and ovarian SEOCs showed differences in allele frequency of the shared mutations, which could point to the directionality of metastasis. Endometrial tumors had a mean ratio of 0.43 (std = 0.32), and ovarian tumors 0.51 (std = 0.28), which was significantly different (p=8.12e-05, Supplementary Fig. 2D ). This supports the view that the SEOCs are likely to have an endometrial origin and progress from the endometrium (subclonal) to the ovary (clonal) ( 31 ) . Genomic profiling established that SEOC are clonally related and more similar to single EEC than single EOC. To comprehensively characterize SEOCs, we performed laser microdissection followed by MS-based proteomics ( 14 ) on 146 samples from 64 patients to resolve the tumor and stromal compartments ( Fig. 1A , Supplementary Table S1 ). Samples were analyzed on a trapped-ion mobility mass spectrometer and label-free proteome quantification ( 24 ) was performed, resulting in a final set of 88 tumor and 56 stroma samples. We quantified a median of 6,378 protein groups per tumor and 5,400 protein groups per stromal sample with a similar protein level variability across groups ( Supplementary Fig. 3A , B ). Proteomes showed strong compartment specificity with enriched epithelial and stromal markers in the respective sample groups ( Fig. 3A - C ) while the sample collection site (Vancouver, Tübingen, and Chicago) had a negligible effect on the number of differentially expressed proteins ( Fig. 3D ). Generally, intra-patient dual cancer proteomes (uterus versus ovary) were more correlated to each other than inter-patient proteomes of either dual or single cancers ( Fig. 3E ). The anatomic site of the resected tumors was a strong confounder on the proteome, which was more pronounced in the stroma ( Fig. 3D ). Nevertheless, the comparison of all single site and SEOCs identified 400 and 335 uniquely regulated proteins in the tumor and stroma compartment, respectively ( Fig. 3D , F ). The identification of significantly expressed proteins in the tumor and stromal compartments of SEOCs compared to single cancers prompted us to further investigate these signatures. The stromal proteomes differentiated better than the tumor proteomes between dual and single cancers, particularly when we accounted for organ related proteome differences ( Fig. 4A - D , Supplementary Fig. 4A , B ). Similar to our genomic analysis, we next asked whether the SEOC tumor proteomes were more similar to single EEC or single EOC, to shed light onto their putative site of origin. We first incorporated healthy uterine and ovarian proteome signatures, consisting of 47 proteins, obtained from the Human Protein Atlas’s tissue enrichment profiles ( 32 ) and, second, used an unbiased approach to filter our data for the 1,000 most variably expressed proteins. In both approaches, the averaged tumor proteome of SEOC clustered closer with EEC than with single EOC ( Fig. 4E ). Collectively, our genomic ( Fig. 2 ) and proteomic data revealed that the average synchronous cancer is more similar to single uterine cancers, suggesting that the majority of low-grade synchronous tumors originate in the uterus. The histopathologic criteria currently used for diagnosing SEOC are not very stringent ( 1 ) Yet, distinguishing them from single EEC or EOC is of clinical importance, since patients with SEOC generally have a more favorable prognosis, and are usually treated solely with surgery without adjuvant treatment ( 3 ). We, therefore, leveraged our compartment-resolved quantitative proteome data to prioritize proteins for the discrimination of SEOC and single cancers. We performed a feature importance analysis using machine learning (ML) based on over 6,000 unique proteins quantified across all three groups. Normalized protein measurements were subjected to feature importance analysis based on the ExtraTrees method for feature selection and the AdaBoost ML algorithm within the open-source tool OmicLearn ( 28 ). Here, data was split into training (80%), and test (20%) data with five cycles of cross-validation ( Fig. 5A ). Receiver operator characteristics (ROC) curves resulted in high area under the curve (AUC) values (0.92 for stroma and 0.88 for tumor), clearly differentiating the different cancer entities ( Fig. 5B - E ). The top-ranked protein features for the stroma predominantly overexpressed in single cancers compared to synchronous cancers included CFAP20, TMEM115, AKTIP, GALNT3 and ARAP1 ( Fig. 5C , Supplementary Fig. 5A ). The top-ranked tumor proteins were EPRS, PIGS, AKR1B1, FAM3C, ATP13A1, and COMMD1 ( Fig. 5E ) Supplementary Fig. 5B ). Notably, protein fold changes of the top-ranked features were small to moderate, emphasizing the power of quantitative MS-based proteomics to identify molecular markers for disease profiling. Because there is currently no other proteomics data available that could be used to further validate these findings, we analyzed the TCGA pan-cancer study to demonstrate marker specificity. Since most top-ranked tumor proteins were overexpressed in the single cancers, we reasoned that these markers would also be overexpressed in the TCGA endometrial endometrioid cancer study. Indeed, uterine endometrioid carcinoma and endometrial cancer showed the highest mRNA up-regulation compared to all other cancer entities, much more than ovarian cancer, clearly reflecting specificity for uterine histology ( Supplementary Fig. 5C ).

In the not-so-distant past, a patient with a low-grade endometrioid uterine and a low-grade endometrioid ovarian cancer was considered to have synchronous tumors that developed independently. However, recent genomics-only-based studies suggest that they are primary and metastatic disease ( 6 – 8 ) as opposed to independent primary cancers. Our proteogenomic study of SEOC confirmed that most sporadic synchronous cancers are clonally related, share several mutations, and have similar proteomic profiles. Our data point to common genetic drivers of endometrioid-type histology, independent of anatomical site. Chromatin remodeling and transcription factor-related genes, including EP300 , ARID1A , and KAT6 A, were mutated in all four groups (EEC, EOC, SEOC-Ut, SEOC-Ov), as were CTNNB1 (β-catenin), NOTCH2 , and PIK3CA . We also found mutations in several histone lysine methyltransferases family genes, known as ‘writers’ methylating histones and proteins, ( 29 ) such as KMT2C and KMT2D , which both play a role in the transcriptional regulation of estrogen response genes ( 33 ) and are among the most frequently mutated genes in endometrial cancer ( 34 ). Given that unopposed estrogen is associated with low-grade endometrioid ovarian ( 35 ) and endometrial cancer, these findings align with current concepts explaining the tumorigenesis of endometrioid tumors ( 36 ). The identified genes probably contribute to the de-differentiation of normal epithelial cells and their glandular transformation into endometrioid-type cancers in the female reproductive tract. Independent of endometrioid histology, the patient group with single EOC had several gene alterations unique to this subtype, including ERBB4 , BRCA2 , and RNF43 mutations and a low frequency of PTEN mutations ( 37 ). The unique genomic profile of single EOC as compared to single EEC and synchronous cancers was also reflected in their global proteomes. Our compartment-resolved proteomics data, quantifying more than 6,000 unique proteins from laser-microdissected micro-regions of tumor and stromal areas, delineated, for the first time, the dual cancer-specific proteome. Moreover, our data also revealed a distinct stromal microenvironment for SEOC. Unsupervised clustering not only showed proteomic similarities for the majority of SEOC stromal proteomes, but also that they were more related to single endometrial tumors. Several lines of evidence make us believe that SEOC originates in the uterus: (i) comparison of the mutational signatures across the single and synchronous cancers showed that single ovarian cancers clustered separately ( Fig. 2D , E ), and (iii) the ovarian cancer proteome from synchronous cancers resembled uterine-derived cancers much more than single EOC. Additionally, the quantitative proteome of single EOC differs substantially from the dual or single endometrioid cancers of the uterus and ovary ( Fig. 4E ). It would have been helpful to identify a single marker or a few markers that are uniquely expressed in low-grade endometrioid tumors metastasizing to the ovary (“synchronous cancers”). We stained for several top markers (e.g., FAM3C and PGK2) from our list of differentially regulated proteins ( Table 5 ), but the results did not show discernable differences ( Supplementary data ). This is likely due to the relatively subtle protein fold changes quantified by state-of-the-art MS-based proteomics. Nevertheless, our extracted protein panels are based on a 20-protein signature from tumor and stromal regions, which unambiguously distinguishing dual from single cancers. While there are currently no other dual cancer-specific proteome (or even transcriptome) data available to further validate and refine our candidate marker panel, we found that our dual cancer-specific protein signature showed a strong enrichment for uterine cancer related signatures (TCGA pan-cancer analysis). Consistent with this, our genetic- and proteome profile clustering analyses also showed a closer relationship between SEOC samples and EEC than SEOC samples and EOC, suggesting that the SEOC could have a uterine origin. While more omics-level data is needed to substantiate this hypothesis, a recent study supports this idea: Based on clonal composition analyses, Moukarzel et al. concluded that a minor subclone within an endometrioid uterine tumor gives rise to the ovarian tumor, supporting a uterine origin for SEOC ( 31 ). We also compared the frequency of shared mutations of SEOC patients and found that the frequency is lower in the endometrial samples compared to ovarian samples ( Supplementary Fig. 2D ), suggesting a shift from subclonal to clonal which supports the findings of Moukarzel et al. Our proteogenomic study therefore indicates that when patients present with synchronous well differentiated endometrioid cancer in the uterus and ovary, the tumor most probably originated in the uterus and metastasized to the ovary. However, additional analyses are needed to confirm the directionality of metastasis, considering confounding factors such as differences in tumor cellularity in endometrial and ovarian tumors. We believe that clinically useful diagnostic tests for synchronous cancers will require either whole genome (WGS) or exome sequencing to evaluate the similarity between the ovarian and uterine cancers and further comparison to already published WGS data. Alternatively, targeted proteomic assays quantifying proteins like the ones we prioritized here, could offer a promising complement to sequencing-based approaches to aid clinical decision making. Our cohort was restricted to low-grade single uterine and ovarian tumors and SEOC, therefore limiting our conclusions to patients who present with low-grade endometrioid disease. Reijnen et al ( 8 ) showed that when the endometrial cancer in a diagnosed SEOC is of low grade endometroid histology and the ovarian carcinoma has a high grade non-endometrioid histology, the two cancers share no mutations, which makes it very probable that these tumors developed independently. Also, when the histology of apparent synchronous cancers is discordant (e.g. serous histology in the uterine tumor and endometrioid histology in the ovarian tumor), these tumors have also developed independently, as is the case with synchronous tumors in patients with Lynch syndrome ( 31 ). An additional limitation of this study is that most EOC cases were grade 2 and above (5 out of 9), while most EEC cases were grade 1 (18 out of 26). Furthermore, we only had limited access to patient data, such as survival / outcome data and history of endometriosis, which might further explain why mutations associated with endometriosis-related neoplasms (ARID1A and PI3K) were in the top ten genes mutated in synchronous cancers. Another caveat is that there was no significant difference in age across the cohorts, though synchronous cancers are often found in younger pre-menopausal women and EEC and EOC are most often found in postmenopausal women. We would have liked to report the MMR promoter methylation status in our molecular subtyping analysis, as this is the most common cause of MMR deficiency, but did not have this information available, so we used the protein level of the MMR complex proteins as a marker for low MMR protein levels. It should also be noted that since we started this study, WHO and FIGO changed their SEOC staging to slightly different categories.

Synchronous endometrial and ovarian cancers (SEOC) are usually endometrial endometrioid cancer (EEC) and endometrioid ovarian cancer (EOC) that co-occur in the same patient. In the last century, Scully established histopathologic criteria to differentiate SEOC, which were thought to arise independently in both organs simultaneously, from metastatic disease to either organ ( 1 ). These criteria include a shallow invasion pattern within the uterus or ovary without surface involvement and similarity in histology and grade between the two sites. About 5% of all EEC and 10% of all EOC present concurrently ( 2 , 3 ). Synchronous endometrial and ovarian cancers are often found in younger premenopausal women and most often display low-grade endometrioid histology in both the ovarian and the uterine tumors. Surprisingly, women who present with SEOC have a favorable prognosis despite having tumors in two anatomic sites ( 4 ). Patients are usually cured by surgery alone, rarely requiring additional adjuvant treatment. The advent of molecular analysis made it possible to accurately determine if SEOC are truly synchronous cancers. In 1998, Lin et al. analyzed PTEN mutations and loss of heterozygosity at both sites and found that many SEOC are clonally related and originate from the same cancer ( 5 ). Two landmark studies ( 6 , 7 ) solidified this finding. Using targeted sequencing of SEOC from 18 patients, Anglesio et al. ( 6 ) revealed mutations unique to the ovarian or endometrial cancer site, although most SEOC shared at least one mutation at each site. Schultheis et al. and Reijnen et al. independently confirmed this ( 7 , 8 ), showing that all low-grade SEOC shared nonsynonymous somatic cancer-driving mutations found in EEC or EOC as defined by the TCGA analysis ( 9 , 10 ). The most frequently mutated genes in SEOC were ARID1A , PTEN , PIK3CA , and KRAS , suggesting they are clonally related and did not arise independently in the uterus and ovary. Based on these observations, we aimed to investigate whether SEOC demonstrates proteogenomic profiles more aligned with metastatic EEC, metastatic EOC or if proteogenomic signatures can offer new insights into the molecular underpinning of synchronous cancers. We analyzed how known genomic alterations and clonality shape the proteome of the cancer and its adjacent stroma. The use of mass spectrometry (MS) based proteomics on laser microdissected formalin-fixed and paraffin-embedded (FFPE) tissue sections ( 11 – 14 ) enabled the quantification of thousands of proteins in cancer and stroma. Our unbiased and comprehensive approach allowed us to analyze the underlying dual cancer-specific proteome and to characterize the proteogenomic landscape of synchronous cancers. This research sheds light on the directionality of metastasis, supporting that low-grade endometrioid synchronous cancers generally originate in the uterus and metastasize to the ovary.

Table 1: Clinico-pathologic patient characteristics Table 2: Top 30 mutated genes in single (ovary, endometrium) and dual (SEOC) cancers Table 3: Overlap of patients and shared mutations between Anglesio et al. and this study Table 4: LC-MS based proteomics data of all 144 samples Table 5: Significantly up – and downregulated proteins in the tumor and stroma of synchronous cancers Table 6: Top 20 discriminating proteins between dual and single cancers