Preneoplastic Lesions Fimbria Early Diagnosis Markers Underlying Timeline Mechanisms at The Origin of Ovarian Cancer in BRAC1/2 Patients: Case Reports Based on Proteogenomic Study

Background : BRCA 1 or 2 mutations have been known to be drivers of high-grade serous ovarian carcinomas (HGSC). As no ecient screening is currently available, a bilateral prophylactic salpingo-oophorectomy is the recommended procedure to avoid this often lethal carcinomas. However, a proportion of BRCA mutation carriers refuse to undergo this procedure due to signicant impacts on quality and quantity of life, especially if a hormone replacement therapy is contra-indicated. Such decision expose them to developing a pelvic serous carcinoma. A better undersanding of high-grade serous cancer pathogenesis, rather stemming from fallopian tubes than from ovary itself, makes interesting a two-step prophylactic strategy in which tubes are removed as soon as possible, followed by oophorectomies at the time of menopause. Thus women are protected against HGSC with no immediate need of an HRT. Following this idea and considering the concept of junctional “hot spots “ (called mesothelo-müllerian junctions) for the development of HGSC, we suggested as rst step to perform a bilateral prophylactic radical mbriectomy rather than salpingectomy with a delayed oophorectomy (RF/DO). This operation was tested in a national phase 2 trial from 2011-2014 (NCT01608074). Radical Fimbriectomy (RF) consists of the surgical removal of both fallopian tubes along with a tiny part of the adherent ovary to suppress the tubal source of possible dysplastic cells from which can stem a high-grade invasive tumor, through stepwise cellular and genomic altereations ranging from p53 signatures to serous tubal intraepithelial carcoinoma (STIC), just before genuine invasive HGSC, through intermediate aspects (STIL) and (TILT). Methods : In this context, we carried out an in-depth proteomics analysis of these “pre-invasive ” epithelial lesions based on spatially resolved proteomic/ and bioinformatics systems biology platforms platform guided by immunohistochemistry (IHC) technique. Results : Specic markers related to each preneoplastic lesions and in particular from normal to p53 signatures were evaluated. We identied EIF3B, MOB1B and Emilin2, CAVIN1 as bad prognosis markers of P53 signature whereas FBLN5 as good prognosis markers. Among the novel markers mutated proteins as hidden proteins translated from alternative Proteome been sequenced and characterized. specic markers for p53 signature, the protein RNA-binding protein 10 (RBM10)- 3 specic proteins in STIC lesions -i.e., Glucosamine (N-acetyl)-6-sulfatase (GNS), Upstream binding transcription factor, RNA polymerase I (UBTF) and ATPase family, AAA domain containing 3 (ATAD3A/B). We identied specic mutated from each lesion stages i.e. mutated Histone H4, Histone H2B type 1-C/E/F/G/I) in p53 signature, mutated 60S ribosomal protein L14 in STIL, and mutated (Na(+)/H(+) exchange regulatory cofactor NHERF1in STIC.


Abstract
Background : BRCA 1 or 2 mutations have been known to be drivers of high-grade serous ovarian carcinomas (HGSC). As no e cient screening is currently available, a bilateral prophylactic salpingo-oophorectomy is the recommended procedure to avoid this often lethal carcinomas. However, a proportion of BRCA mutation carriers refuse to undergo this procedure due to signi cant impacts on quality and quantity of life, especially if a hormone replacement therapy is contra-indicated. Such decision expose them to developing a pelvic serous carcinoma. A better undersanding of high-grade serous cancer pathogenesis, rather stemming from fallopian tubes than from ovary itself, makes interesting a two-step prophylactic strategy in which tubes are removed as soon as possible, followed by oophorectomies at the time of menopause. Thus women are protected against HGSC with no immediate need of an HRT. Following this idea and considering the concept of junctional "hot spots " (called mesothelo-müllerian junctions) for the development of HGSC, we suggested as rst step to perform a bilateral prophylactic radical mbriectomy rather than salpingectomy with a delayed oophorectomy (RF/DO). This operation was tested in a national phase 2 trial from 2011. Radical Fimbriectomy (RF) consists of the surgical removal of both fallopian tubes along with a tiny part of the adherent ovary to suppress the tubal source of possible dysplastic cells from which can stem a high-grade invasive tumor, through stepwise cellular and genomic altereations ranging from p53 signatures to serous tubal intraepithelial carcoinoma (STIC), just before genuine invasive HGSC, through intermediate aspects (STIL) and (TILT).
Methods : In this context, we carried out an in-depth proteomics analysis of these "pre-invasive " epithelial lesions based on spatially resolved proteomic/ and bioinformatics systems biology platforms platform guided by immunohistochemistry (IHC) technique.
Results : Speci c markers related to each preneoplastic lesions and in particular from normal to p53 signatures were evaluated. We identi ed EIF3B, MOB1B and Emilin2, CAVIN1 as bad prognosis markers of P53 signature whereas CAVIN2, SPTANi, FBLN5 as good prognosis markers. Among the novel markers identi ed, mutated proteins as well as hidden proteins translated from alternative ORF Proteome have been sequenced and characterized.
Conlusion : In summary, our results represent a novel pioneer approach to identify the unique under investigated markers that can characterize the molecular mechanism occurring in the preneoplastic lesion in mbria at the origin of the HGSC.
Whereas ovarian cancer (OC) is ranked at the 8th position among cancers, it is by far the most lethal cancer; with an approximate of 30-40% survival rate in a period of ve years. Amongst the different classes of varian cancers, high-grade serous carcinoma (HGSC) is the most represented (> 75%) having ~ 15% of the patients with mutations in the BRCA1/2 gene (Braicu et al., 2011;Pal et al., 2005;Seidman et al., 2004). These patients exhibit a 20-40 percent increased risk of developing HGSC between 30 and 35 years old (versus 1.4% in the global population).
During the last 10 years, special attention has been paid towards the pathological examination of operative specimens of risk-reducing salpingo-oophorectomy in women carrying a familial risk of OC. Most of the ovarian carcinomas related to BRCA 1 or 2 mutations are of invadsive high grade serous pathological subtype. They seem stem from the fallopian tube and particularly from its distal part called the mbria.. Unfortunately as they are rapidly evolutive HGSC cannot be effectively screened. In addition if hormonal contraception seems e cient at preventing them, this treatment is sometimes not well tolerated or even contraindicated in the case of breast cancer history.
Risk-reducing surgery is therefore suggested to these high-risk ladies when child bearing project has benn completed. Bilateral risk-reducing salpingo-oophorectomy (RRSO) surgery is the gold standard prophylactic procedure worlwide.
Some BRCA-mutated women (or at-risk of ovarian cancer but without identi ed mutations) are reluctant to undergo this procedure considering the numerous adverse effects on the body and quality of life, especially when hormonal replacement is proposed. These woman are at higher rsk to develop serous pelvic carcinoma. The goal of the bilateral laparoscopic radical mbriectomy is to suppress the tubal source of possible dysplastic cells from which high-grade tumor originates, while preserving a natural ovarian hormonal secretion. This procedure is offered only in case of rejection of BSO and it is aimed to salvage these high-risk women and keep them under adapted surveillance of their gonads. The reduction in bene ts of castration over breast cancer risk, should be evaluated and assessed by the patinets. Complementary castration is recommended at 50 years old cohorts. This new prophylactic operation, based on ovarian carcinogenesis emergece, preserves femininity and possibly fertility as well. It seems an interesting option for selected women who are reluctant to undergo prophylactic BSO. Our work will evaluate Müllerian cells in the mbria end of the fallopian tube to nd a relationship between the pre-neoplastic lesions and (HGSC), a type II ovarian carcinoma. Based on Sectioning and Extensively Examining the Fimbriated End Protocol (SEE-FIM) (Mingels et al., 2014), a systematic serial examination of the fallopian tubes will be performed, coupled with IHC evaluation of the p53 and Ki-67protein expression. In this group of patients, an unusual rate of certain cancer or at least cellular abnormalities are observed in the fallopian tube epithelium (FTE), especially at its terminal end, the mbria (Crum et al., 2007;Medeiros et al., 2006b). Three main lesions are de ned according to the IHC pattern for the two markers and the architectural alterations of the cells. The p53 signature is identi ed by a small number of epithelial cells (10-20) with a proliferation activity like the adjacent normal epithelium, with a p53 staining pattern corresponding to a missense TP53 mutation. On the other hand, (serous tubal intraepithelial carcinoma (STIC) involves many cells with architectural and nuclear alterations, TP53 mutations and a high proliferative activity. While STILs (serous tubal intraepithelial lesions (STILS) are characterized by a lower level of abnormalities compared to the STIC along with a normal proliferative activity Visvanathan et al., 2011a). accordingly, morphological changes concomitant with multi-step accumulation of molecular and genetic alterations on the pre-neoplastic lesions of FTE suggest that HGSCs derive their origin, not in the ovary, but from the mbria part of the fallopian tube (Auersperg, 2013;Kobayashi et al., 2017;Roh et al., 2010). In fact PAX8, a Mullerian marker has been found to be expressed in most of the HGSC but not calretinin (mesothelial marker) (Bowtell, 2010;Folkins et al., 2009;Mingels et al., 2014). This view has not been universally accepted, primarily as it con icts with traditional theories of the OC origins coupled with variation in the detection of tubal lesions in association with HGSC, which may, be due to differences in sampling or di culties in diagnostic interpretation (Bachert et al., 2020;Singh et al., 2015). Recently, advanced studies have provided a stepwise progression of FTE to precursor lesions to carcinoma, with the aid of -p53 signature -STIL -STIC HGSC sequence's model (Kobayashi et al., 2017;Wu et al., 2019;Zeppernick et al., 2015).
In this context, we performed a pan-proteomic study based on the state-of-the-art of the spatially resolved proteomics that have been previously applied on different kinds of cancers (Delcourt et al., 2017;Le Rhun et al., 2015;Wisztorski et al., 2016;Wisztorski et al., 2013;Wisztorski et al., 2017). This spatially resolved proteomic study is guided by IHC analysis performed on formalin-xed para n-embedded (FFPE) tissue section of patients   presented with p53 signature lesion, STIL, and STIC lesions compared to non-pathological fallopian tissues and HGSCs. Here, we investigate the different pre-neoplastic lesions, their proteome pro le, their protein mutations landscape as well as assessing alternative ghost proteome translated from alternative ORF, and coupled with a bioinformatics sytems biology analysis to construct proteome molecular pathways indicative of the pathological underlying mechanisms (Cardon et al., 2020;Cardon et al., 2019;Delcourt et al., 2018;Vergara et al., 2020). By applying our approach, we identi ed the several early-stage markers such as EIF3B, MOB1B and Emilin2, CAVIN1 as bad prognosis markers and CAVIN2, SPTANi, FBLN5 as good prognosis markers. In addition, we have surveyed the molecular mechanisms occurring across pre-neoplastic lesions and con rm the bene t of the radical frimbioectomy in BRCA1/2 mutated patients .

Experimental Procedures
Experimental Design and Statistical Rationale.
Shotgun proteomics experiments were conducted on formalin-xed para n-embedded (FFPE) tissue section of patients presented with p53 signature lesion, STIL lesions and STIC lesions which were compared to nonpathological fallopian tissues and HGSC samples; all experiments were conducted in biological triplicate. Statistical analysis: For the proteomics statistical analysis of extracted proteins or secreted media, only proteins presenting as signi cantly determined by the ANOVA test were used with an FDR of 5%. Normalization was achieved using a Zscore with a matrix access by rows. Obtained data from Western blot were reported as mean ± SEM. Mean values among different experimental groups were statistically compared by one-way ANOVA tests using Graph pad PRISM software or by student t-test.

Reagents and Chemicals
For the different experiments, we used high purity chemicals from various suppliers: HPLC grade methanol (MeOH), ethanol (EtOH), Chloroform (CHCl3), acetonitrile (ACN), water and tri uoroacetic acid (TFA) were obtained from Thermo Fisher Scienti c (Courtaboeuf, France). Tri uoracetic acid (TFA, 99%), and ammonium bicarbonate (NH4HCO3) were purchased from Sigma-Aldrich (Saint-Quentin Fallavier, France). Xylene and formic acid (FA, ≥ 96%) were purchased from Biosolve (Dieuze, France). Sequencing grade modi ed porcine trypsin was obtained from Promega (Charbonnieres, France). Surgery Procedure BRCA1/2 carriers or any women with a documented familial risk of breast/ovarian cancer were rst counseled to perform a classical laparoscopic RRSO. If they denied, they were offered to enter the RF/DO trial. All specimens were submitted to the Sectioning and Extensively Examining-the FIMbriated end (SEE-FIM) pathological protocol.
Pathological data along with all intra-and 30-day and beyond post-operative adverse events were prospectively recorded. Follow-up visit consisted of an annual clinical breast and gynecological examination, with tumor markers and hormonal status assessment. Primary endpoint was the rate of pelvic serous carcinoma. Secondary endpoints were procedure morbidity, rates of tubal abnormalities, breast cancer, secondary oophorectomy. An IRB Protocol NCT-01608074 was approved by the national and institutional review boards (IRB) and accepted by our local ethics committee. An interim analysis by a French DSMB was performed in January 2014. It was decided to complete the study until its end, without protocol modi cations. This national phase 2 prospective study, started in 2012, ended in October 2014 after accrual completion. Eleven French centers actively participated in patients' recruitment.
Case selection and sample processing Tissue samples were obtained from patients of the Centre Oscar Lambret (Lille, France. All experiments were approved by the local Ethics Committee (CPP Nord Ouest IV 12/10) in accordance with the French and European legislation. Prior to the experiments, patients signed an informed consent and authorization form describing the experimental protocol. No personal information was used in these experiments, and a random number was assigned to each sample. In our study, all samples were FFPE tissue from prophylactic adnexectomies of women with BRCA mutations. Our study was performed on patients presenting with pre-neoplastic lesions without any concomitant ovarian high-grade serous carcinoma. All samples were examined following SEE-FIM (Sectioning and Extensively Examining the FIMbria) protocol (Medeiros et al., 2006a). On these samples, tissues sections of 7 µm thickness were cut using a microtome and were deposited on a glass slide, HPS (Haematoxylin Phloxine Saffron) and immunostaining against P53 and Ki-67 (Dako, Japan) were performed and examined by a pathologist to nd preneoplastic lesions (p53/STIL/STIC/HGSC) (Visvanathan et al., 2011b). A cohort of eight patients was selected.
Four patients presented p53 signature lesion, 2 patients with STIL lesions, 2 patients with STIC lesions, 1 patient with HGSC (see Table 1). The normal tissue was analyzed from the normal part (presenting no abnormalities at IHC) of the tissue section from patients with p53 signature lesion. On-tissue spatially resolved proteomics A complete work ow for sample preparation is illustrated in Fig. 1A Sample preparation The same slides as the one used and annotated by the pathologist were unmounted; resin was removed by soaking them overnight in xylene and rinsing them with xylene and ethanol baths. The tissues were rehydrated using 5' each successive bath of decreasing ethanol degree (2x95°, 1x30°) and two baths of 10mM NH4HCO3 buffer. Then, an antigen retrieval step was performed to increase the trypsin access to biomolecules. For this, the slides were dipped in 90°C pH9 20mM Tris for 30 minutes, rinsed twice with NH4HCO3, and dried under vacuum at room temperature.

In situtrypsin digestion
On localized pre-neoplastic lesions tryptic digestion was performed using a Chemical Inkjet Printer (CHIP-1000, Shimadzu, Kyoto, Japan). The trypsin solution (40µg/mL, 50mM NH4HCO3 buffer) was deposited on a region de ned to 1mm² during 2h. During this time, the trypsin was changed every half-hour. With 350 cycles and 450 pL per spot, a total of 6.3µg of trypsin was deposited. To stop digestion, 0.1% TFA was spotted during 25 cycles.

Liquid Extraction
After microdigestion, the spot content was gathered by liquid microjunction using the TriVersa Nanomate (Advion Biosciences Inc., Ithaca, NY, USA) using Liquid Extraction and Surface Analysis (LESA) settings. With 3 different solvents mixture composed of 0.1% TFA, ACN/0.1% TFA (8:2, v/v), and MeOH/0.1% TFA (7:3, v/v). A complete LESA sequence run 2 cycles for each mixture composed of an aspiration (2µL), a mixing onto the tissue, and a dispensing into low-binding tubes. For each interesting spot, 2 sequences are pooled into the same vial.
The separation, prior to MS, used online reversed-phase chromatography realized with a Proxeon Easy-nLC-1000 system (Thermo Scienti c) equipped with an Acclaim PepMap trap column (75 µm ID x 2 cm, Thermo Scienti c) and C18 packed tip Acclaim PepMap RSLC column (75 µm ID x 50 cm, Thermo Scienti c). Peptides were separated using an increasing amount of acetonitrile (5%-40% over 140 minutes) with a ow rate of 300 nL/min. The LC eluent was electrosprayed directly from the analytical column and a voltage of 2 kV was applied via the liquid junction of the nanospray source.
The chromatography system was coupled to a Thermo Scienti c Q-Exactive mass spectrometer. The mass spectrometer was programmed to acquire in a data-dependent mode for the 10 most intense peaks. The survey scans were acquired in the Orbitrap mass analyzer operated at 70,000 (FWHM) resolving power. A mass range of 200 to 2000 m/z and a target of 3E6 ions were used for the survey scans. The MSMS analysis was performed using HCD with a normalized collision energy of 30 eV, a mass range between 200 to 2000, an AGC of 5e4 ions, a maximum injection time of 60ms and a resolution set at 17,500 FWHM. The method was set to analyze the top 10 most intense ions from the survey scan and dynamic exclusion was enabled for 20 s.
Peptides identi cation were obtained according to target decoy search against Homo sapiens Uniprot database (version of 2017_02, 20,172 sequences) and database containing 262 commonly detected contaminants. The human uniport database was used as the forward database and a reverse one for the decoy search was automatically generated in MaxQuant Mass tolerances were set to 10 ppm and 20 ppm respectively for parent and fragments measurement. Enzyme speci city was set to "trypsin" with a maximum of 2 missed cleavages allowed.
Methionine oxidation and acetylation of protein N-terminal were set as variable modi cations. FDR < 1% was set for peptides and proteins identi cation. Two peptides with one unique were necessary to assess protein identi cation.
For label-free quanti cation, the MaxLFQ algorithm was used (Cox et al., 2014). The option "Match Between Runs" was enabled to maximize the number of quanti cation events across samples. This option allowed the quanti cation of high-resolution MS1 features not identi ed in every single measurement. Data generated by MaxQuant were analyzed using Perseus (version 1.6.2.3, (Tyanova and Cox, 2018). LFQ values were used and proteins were removed if found in the category only identi ed by site modi cations, in the decoy reverse database, or identify in the contaminant database. We removed proteins that were not presented in at least three of four replicates. We took the average expression per group to perform a comparison and visualized using a Venn diagram. Individual LFQ values were used to perform a multi scatter plot and calculate a Pearson Correlation between samples. Missing values were imputed based on normal distribution (width = 0.3, down-shift = 1.8).
Principal component analysis (PCA) was done to compare the protein content of each sample. An ANOVA Multisample test was performed and consolidated by a Permutation-based FDR (FDR < 0.05, 250 randomizations). A speci c comparison between normal and p53 signature samples was performed using a student's T-test.
Proteins with signi cant differences were ltered out and values were z-scored. The samples were then clustered according to a Euclidean average as a distance measure for column and row clustering.
Up-regulated and down-regulated proteins in the different groups were used to perform an annotation analysis of gene ontology terms by using Funrich (v3.1.3) (Pathan et al., 2015). A hypergeometric test was performed against all annotated gene/protein list by comparing the multiple datasets. Enrichment for biological process, transcription factor and cellular component were present as a bar chart. PANTHER Classi cation System (v14.1) was also used.
PANTHER Overrepresentation test (Released 20190701) was performed using each list of up-or down-regulated proteins as "analyzed list" and Homo sapiens as "reference list". Fisher's Exact test with false discovery rate correction was used.
Subnetwork Enrichment Analysis (SNEA) from Elsevier's PathwayStudio version 10.0 (Elsevier) (Bonnet et al., 2009;Yuryev et al., 2009) was used to extract statistically signi cant altered biological and functional pathways in the different clusters of proteins.
The immunohistochemical (IHC) data of different proteins of interest were investigated from the Human Protein Atlas (HPA) database (http://www.proteinatlas.org, (Uhlen et al., 2017)). Evaluation of the prognostic effects of these proteins on the overall survival (OS) was also extracted from HPA. Comparisons are performed with the 20 genes of highest signi cance associated with unfavorable prognosis for ovarian cancer, cervix cancer, endometrial cancer, and breast cancer.

Mutation identi cation:
The MS data were also processed to search for potential mutated peptides using the XMAn database (Yang and Lazar, 2014). This database contains information on mutations observed in cancers and diseases. Proteome Discoverer 2.1 was used to query the data using MS Amanda as a search node against the XMAn database, the human Uniprot database, and a database containing potential contaminants. Peptides identi ed only in the XMAn database at a high level of con dence were selected, and the MSMS spectra were manually inspected to con rm the presence of the mutation.  Table 1). We used immunostained tissue sections obtained from the SEE-FIM protocol (Fig. 1A). This protocol consisting of systematic examination of the totality of the mbria is di cult to perform and will not allow observation of a lot of pre-cancerous lesions. Indeed, the number of samples is then critically limited allowing minimal slide examination. Moreover, the pre-cancerous lesions presented an extremely limited number of cells on the epithelial layer with no speci c morphological features (Fig. 1B). The presence of p53 signature, STIL, and STIC lesions were; thus, con rmed-thanks to a double IHC against p53 and Ki-67 markers (Fig. 1B). In order to process these samples for spatially resolved proteomics, we developed a new protocol directly using the same IHC slide used by the pathologist which has served to the detection of the lesions. The cover slide and the mounting medium were removed by extensive bathing of xylene and ethanol. A visual inspection was carried out to check whether additional washing is required to remove all the mounting medium from the tissue section. One critical point was the antigen retrieval procedure that needed to be performed after tissue rehydration. Next, the region of interest corresponding to a lesion is covered with a droplet of trypsin solution. The resulting tryptic peptides are extracted from the tissue by a droplet-based liquid microextraction derived from the LESA method Quanico et al., 2013;Wisztorski et al., 2016;Wisztorski et al., 2013;Wisztorski et al., 2010;Wisztorski et al., 2017) and analyzed by MS-based proteomics.
Considering all the pre-neoplastic lesions and non-pathological tissues, 10375 unique peptide sequences corresponding to 1617 distinct protein groups have been identi ed using our work ow. MaxLFQ algorithm was used to perform label-free quanti cation of proteins and resulted in a total of 1571 protein quanti cations. After ltering one according to a minimum number of values (2/3 of valid values) in at least one group of the four de ned groups, 1046 proteins were obtained. This result is like those obtained in our previous experiments on FFPE tissue. Furthermore, using immunostained tissues, after removing the coverslip glass and the mounting medium, did not appear to affect not affect the microproteomic analysis.
To con rm these ndingss, we performed a correlation analysis based on all quanti ed proteins. Proteomes inside samples from the same group were similar (mean Pearson correlation 0.92) compared with inter-group variation.
The most differences were observed between STIC and STIL lesions (mean of 0.84) ( Fig. 2A). This is in accordance with what is generally observed considering the inter-patient variability, and the protocol we used did not appear to alter the proteomic content of the tissue sections. These results allowed us to perform a comparative analysis to obtain a speci c proteomic signature of the different lesions.
Spatially resolved proteomic study of p53 signature, STIL, STIC, and normal Fimbria Comparing the content of each group by averaging the replicates and using a Venn diagram representation, 853 common proteins (81.5% of overlap) were observed between normal tissues and pre-neoplastic lesions ( and a hierarchical clustering analysis were performed to visualize the correlation between samples. PCA based on whole proteome levels allow visualization of the difference among each group. (Fig. 2D). Components 1 and 2 account for 42.5% of the total data variation. Samples from normal tissue and p53 signature are closed and could not be easily differentiated. Samples from STIL and STIC lesions differed from normal and p53 signature samples in component 2. STIL lesions were closer to normal/p53 lesions in component 1 than STIC. Adjunction of the HCSG data con rmed the separation of the nature of the proteome between p53 signature, STIL, STIC and HGSC (Fig. 2D).
Hierarchical clustering showed a signi cant difference in expression between the 5 groups as shown in the heatmap (Fig. 2E). Two main branches separate HGSC from the other lesions. The second branch is subdivided again into two subbranches i.e., one separating the STIC to p53 signature/Normal/STIL (Fig. 2E). The second subbranches were then separated between STIL to p53 signature/Normal. The last group (p53/Normal) was di cult to be differentiated. Nevertheless, the difference in cell composition between the ovary and the fallopian tube did not allow a direct conclusion to be drawn on the relationship between HGSC and the pre-neoplastic lesions observed in the fallopian tube. However, we could observe that the proteomic content of the STIC lesions is closer to HGSC than to the other lesions and that each lesion seemed to present a speci c proteomic signature.

Proteomic analyses along the tumoral process
For a deeper analysis of the pre-neoplastic lesions of the FTE, ANOVA testing was performed without the HGSC data. A total of 197 proteins showed a variation of expression on the four groups (Supp. Data 3). A hierarchical cluster based on the expression of these proteins resulted in three groups and two mains clusters (Fig. 3A, Supp. Data 3). As observed previously in PCA, normal and p53 signature could not be resolved and formed one group.
This group and another group composed of the STIL lesions samples clustered together. The proteomic signature of STIC differed from the two previous groups and samples from STIC lesions formed the second cluster suggesting that proteomics pro les differed a lot from the two other groups. The heatmap presents ve main clusters of proteins. Proteins that were enriched in STIL and STIC regions were represented in cluster 1. Those that were more abundant in both normal tissue and p53 signature lesions were represented in cluster 2, proteins presented in normal, p53 signature lesion and STIL but not in the STIC were represented in cluster 3. Proteins that were speci cally overexpressed in the STIC were represented in cluster 4 and those speci c to normal tissue, p53 signature lesion, and STIC were represented in cluster 5. Cluster 1 (STIL/STIC) was composed of only 5 proteins i.e., MARCKS-related protein (MARCKSL1), Myosin-10 (MYH10), Protein SET (SET), Double-stranded RNA-binding staufen homolog-1 (STAU1), and D-3-phosphoglycerate dehydrogenase (PHGDH). This SET of proteins participates in numerous cellular functions including DNA repair, transcription, cell survival and proliferation. This protein is also involved in many cancer processes such as metastasis, the development of therapeutic drug resistance and plays a key role in tumorigenesis (Bayarkhangai et al., 2018). MARCKSL1 is known to promote the progression of lung adenocarcinoma by regulating Epithelial-mesenchymal transition (EMT) (Liang et al., 2020). STAU1 has been previously identi ed in colorectal cancer (Zhu et al., 2005), whereas PHGDH protein has been identi ed in ovarian cancer (Francavilla et al., 2017;Zhang et al., 2018). In cluster 2 (normal-p53), 15 proteins showed an overexpression for normal and p53 signature lesion. Among these proteins, CD166 antigen (ALCAM/CD166) is shown to be a potential cancer stem cell marker (Shimamura et al., 2014). It has been found as a poor prognosis marker in pancreatic cancer (Kahlert et al., 2009) or gastric cancer (Levin et al., 2010).This protein is known to contribute to local invasion and tumor progression by acting on the detachment of tumor cells (Lunter et al., 2005;Ofori-Acquah and King, 2008). Moreover, TNPO1, ACLY, NME1, FLOT1, KLC4 are proteins already know to be involved in epithelial ovarian cancer (Creekmore et al., 2011;Kalra and Bapat, 2013;Li et al., 2018). Gene Ontology (GO) analyses con rmed that these proteins are involved in cell proliferation, adhesion, and growth (Fig. 3Ba). Cluster 3 (normal- (Tubulin beta-4B chain)). This cluster showed an overrepresentation of proteins involved in neoplasia and cancer transitions (Fig. 3Bd). Functional enrichment analyses point out differences between clusters. Different biological processes have been identi ed i.e. immune response, regulation of immune response, cell growth and/or maintenance, metabolism, energy pathways, signal transduction and protein metabolism (Fig. 4A). Enriched transcription factors (TFs) that regulate the overexpressed proteins in the different clusters were also obtained (Fig. 4B). The cluster containing proteins overexpressed in normal-p53 signature showed a high abundance of proteins involved in metabolism and energy pathways as well as an enrichment of SP1, SP4 and TEAD1 transcription factors that targeted the most genes of this cluster. For the cluster normal-p53/STIL, proteins were mostly involved in cell growth and/or maintenance and protein metabolism. Enrichment of the TFs NFIC and EGR1 was observed. Concerning Normal-p53/STIC, identi ed proteins were implicated in cell growth and /or maintenance, metabolism, energy pathways and protein metabolism while KLF7 protein was identi ed as a TF. For STIL/STIC cluster, enrichment of the TFs ZEB1 and ETS1 was observed. STIC cluster presented a strong enrichment of proteins involved in protein metabolism and regulation of immune response and SP1, SP4 and KLF7 as TFs. We also observed a diminution of the proteins involved in immune response from normal-p53 to STIC. Cellular components analyses re ected that all proteins of the 5 clusters were in cytoplasm or exosomes (Fig. 4C) but with some differences. In fact, in STIL-STIC and STIC, we observed a strong enrichment for exosomes; whereas, p53 signature and STIC involved cytoplasmic proteins. We also observed a high enrichment of cytoskeletal proteins in STIC lesions. Altogether, these analyses re ectedsome clear transition occurring between Nomal-p53 signature, p53 to STIL and STIL to STIC with a modulation of proteins involved in cell growth and/or maintenance and the different metabolism processes. To con rm these observations, further analyses were carried out, this time grouping together all proteins overexpressed in each lesion, i.e., normal-p53 signature, STIL, and STIC (Fig. 4D, E and F and Supp. Data 3). In STIL, an enrichment of proteins involved in cell growth and/or maintenance (Fig. 4D), in integrin cell surface interactions (Fig. 4E), in extracellular matrix (ECM) structural constituent and cell adhesion molecule activity (Fig. 4F) were observed and the ones in various metabolism and energy pathways ( Fig. 4D and E), catalytic activity (Fig. 4F) were lower than for the other cellular transition. We also investigated whether the proteins involved in the Warburg effect (Scatena et al., 2010) are activated to compensate for the overall decrease in the classical metabolism processes observed ( Figure S1). These proteins are not altered overall with a decreasing rather than an increasing trend, which means that no Warburg effect seemed to occur. In STIC, proteins involved in metabolism, energy pathways, cell communication, signal transduction ( Fig. 4D and E), calcium ion binding, catalytic and structural molecule activities (Fig. 4F) were higher than the other transition. In the same time, immune response ( Fig. 4D), integrin interactions (Fig. 4E), cell adhesion molecule activity and ECM structural constituent ( Fig. 4F) were highly repressed. Statistical overrepresentation test using PANTHER classi cation system analyses (Supp. Data 4) con rmed such transition.
Nevertheless, there are still di culties in distinguishing between p53 signature and normal tissue. Using the Student's T-test, 34 proteins have been shown to be highly discriminant between Normal and p53 lesions with 26 overexpressed in Normal and 8 in p53 signature (Fig. 5,Supp. Data 5). Table 2 presents the immunoreactivity in normal FTE and prognostic effect on the overall survival (OS) in OC for these 8 proteins extracted from the HPA.
PTRF, also named CAVIN1, is an unfavorable prognosis marker for ovarian, urothelial, and colorectal cancers, whereas SDPR (known as CAVIN2) is a renal cancer favorable prognosis marker and unfavorable in stomach cancer. The IHC data indicated that these proteins are not detected in normal FTE, showed a moderate positivity in OC for CAVIN1 and a strong positivity in a rare case of endometrioid carcinoma of ovary for CAVIN2. SPTAN1, FBLN5 are a favorable prognostic marker in renal cancer but unfavorable in OC (considering a p < 0.05). EIF3B, MOB1B and Emilin2 are already known to bad prognostic markers for liver cancer. Moreover, EMILIN-2 and EIF3B are also considered as bad prognosis markers for renal cancer and for head neck cancer.

Protein Mutation
The lineage relationship between HGSC and the different lesions is suggested by the presence of same genomics changes and mutations . In order to identify speci c modi cations that may have occurred during the different potential pre-cancerous lesions, we explored the presence of protein mutations per each lesion. For that purpose, we used the human database combined with the XMAn database (Flores and Lazar, 2020). This database contains information concerning mutated peptides that could be found in some cancers extracted from the COSMIC database. This database integration resulted in the identi cation of 83 peptide sequences containing possible mutations (Supp. Data 6). For four of them, it was possible to determine the amino acid modi cation directly via the tandem mass spectra (Fig. 6A). These peptides were derived from four proteins i.e., the vitamin D binding protein (GC), the polyubiquitin-C (UBC), the Histone H2B (H2B1C), and Histone H3.1 (H31). The mutations found in the protein GC (1296T > G p.D432E), lead the sequence LPEATPTELAK in the protein and has been identi ed in stomach cancer studies according to the COSMIC database (Alfaro et al., 2017;Dopazo and Erten, 2017;Gnad et al., 2013;Marquard et al., 2015). It was observed in data issued from two patients in a normal tissue and a p53 signature lesion. Interestingly, this protein was found to be signi cantly under-represented in the STIC lesion (Supp. Data 3). Similarly, the mutated peptide of the UBC was found in two normal tissues, two p53 signature lesions, and one STIL. The mutation (c.368G > C p.G123A), which leads to the new peptide the LIFAAKQLEDGR was previously identi ed in breast cancer. The mutated peptide (QVHPDTGISTK) from Histone H2B (mutation c.169T > A p.S57T) was identi ed only in one normal tissue while the protein was not dysregulated (Supp. Data 3). Interestingly, this mutation has been identi ed before in the endometrium. The peptide of the histone H3.1 (VTIMPRDIQLAR) was observed in two normal tissue, one p53 signature, and one STIL lesion. This has been found to be downregulated in the STIC lesion (Supp. Data 3). This mutation (c.368A > G p.K123R) has also been identi ed in the endometrium.
Considering the 83 mutations identi ed (Table 3, Sup. Data 6), some have been detected in normal to all lesion stages including the proteins: vimentin or collagen, but some are speci c to one particular type e;g. mutations on Histone H4, Histone H2B type 1-C/E/F/G/I in p53 signature, mutation on 60S ribosomal protein L14 in STIL, and mutation on (Na(+)/H(+) exchange regulatory cofactor NHE-RF1 in STIC. Interestingly, in the transition from Normal to p53 signature, 7 mutated peptides have been identi ed. For p53 signature/STIL and Normal-p53/STIL, only one mutation has been observed respectively on Isoform 2 of Tropomyosin beta chain and Histone H1.4 proteins. From normal-p53/STIL and STIC, 3 mutated peptides have been identi ed corresponding to Laminin subunit gamma 1, Vimentin, and Histone H3.1.It is note worthy that the last two mutations have been detected in the endometrium. Ghost Proteome Using our proteomic data against the OpenProt database, 59 AltProts have been identi ed as ghost proteome. More than 64% of these proteins are associated with sequences from non-coding RNA (ncRNA), and from the mRNA coding for RefProt 23% are from 5'UTR, 33% from the 3'UTR and 42% from a shift in the CDS (Supp. Data 7). Among these proteins (Table 4 (Table 4). Using the ANOVA test, 12 AltProts present a signi cant variation in the different lesion stages. Nine AltProt are derived from ncRNA, one from 3'UTR, and two from a shift in the CDS. The hierarchical clustering and heatmap representation (Fig. 6Ba) revealed a clear separation between STIC and the other lesion. The second branch separates STIL from the normal-p53 signature. The pseudogene KRT8P11 (IP_563986) is overexpressed in STIC, whereas LOC105376924 and, COL18A1 (IP_290397) pseudogene were overexpressed in the STIL samples. The other AltProts are differentially expressed in p53 lesions and Normal tissue, but all are underexpressed in the STIC or non-expressed in STIL (Fig. 6Ba). When data from HGSC were added, the separation of STIC from other samples is still observed, carcinoma samples composed the second branch, and a third branch is divided into subbranches. These subdivisions group included three of the fourth STIL samples and three of the fourth Normal samples (Fig. 6Bb). With regards to the results, KRTP8P11 (IP_563986) is a speci c marker of STIC, B4GALN (IP_191334) is speci c of HGSC, and KRT8P32 (IP_602534) is overexpressed in two p53 signature samples and one of the normal tissue sample but also highly decreased in HGSC. Finaly, COL18A1 (IP_290397) is speci c to STIL. The other AltProt were observed between Normal-p53 signatures with most of them under-expressed in STIC (Fig. 6Bb). AltProts identi ed. 14 AltProt have been deleted because they are identi ed but without a su cient abundance to be quanti ed. The AltProt identi cation accession ID is according to the OpenProt Database. Venn diagram representation is based on the abundance obtain after analysis, function to the triplicate samples type (normal, p53, STIL, STIC, Carcinoma).

Altprot
Normal P53 STIL STIC Carcinoma IP_794359 X X X X X

Discussion
The most recent theories regarding the origin of HGSC propose an involvement of cells in the region of the tubalperitoneal junction (Leblanc et al., 2011;Mingels, 2015;Seidman et al., 2011). This junction corresponds to the region of the peritoneum covering the serosal surface of the fallopian tube and meets the specialized epithelium of the tubal mbriae, with a transition between the Müllerian and the mesothelial cells (Gan et al., 2017). Such junctional sites between different types of epithelia represent cancer hot spots in which neoplastic transformation occurs. This is con rmed by evidence that cells derived from this transition region demonstrate a cancer-prone stem cell phenotype (Gan et al., 2017;Leblanc et al., 2011;Mingels, 2015;Seidman et al., 2011). Subsequent studies have led to the designation of these changes as STIC recognition of their preferential localization in the epithelium of the mbriae end of the fallopian tube, and the recommendation of standardized and detailed sampling of tubal specimens with the SEE-FIM protocol. However, the diagnosis of STIC based on morphology alone is reported to have low reproducibility. With the emergence of detailed histological examination of the fallopian tubes, with and without IHC, a variety of confounding pre-neoplastic lesions is now being encountered which had not been previously characterized included p53 signature and STIL lesions (Carlson et al., 2010), but the temporal relationship between each lesion and HGSCs is still debated.
The aim of our study was to investigate and the molecular mechanisms associated with the development of the pre-neoplastic lesions found in the FTE. To the best of our knowledge, no similar proteomics/bioinformatics studies have been condiucted yetinvestigating alterabtive and ghost proteomes, due to the the robust uniqueness of the protocol established for sampling the tissue sections. We rstly developed a novel strategy based on pathology routine protocol. Several tissue sections with pre-cancerous lesions are drastically limited. In fact, these types of lesions could only be observed using a SEE-FIM protocol as discussed above. It allows localization and differentiation of each lesion type but on a limited number of slides. For example, p53 signatures are observable only on one or two tissue sections. In this context, the direct use of the IHC slides was applied to perform other experiments has many advantages. In our study, to a better understanding of the different transitions of the lesions, i.e. from Normal to p53 signature, then from p53 signature to STIL and nally from STIL to STIC, IHC-guided spatially resolved proteomic analyses were realized on each identi ed lesion. We rst demonstrated that the work ow we developed is compatible with proteomic analysis. Using targeted digestion of each region highlight by the IHC, we can preserve spatial information and perform analysis of different regions in the same slide. The tissue section surface recovery protocol does not alter the proteomic content and allows an accurate comparison of quanti ed proteins between each lesion.
If we compare the proteomic content of HGSC and the lesion found in the fallopian tube, we observed interesting groupings that may correspond to what different studies propose for the chronology of HGSC development from pre-neoplastic lesions. The normal tissue and the p53 signature cluster together and are close to the STIL. The STIC lesion presents a more different signature and is closer to HGSC than the two other lesions. Due to the difference in composition between the fallopian tube tissue and the ovary, we have focused our analysis to deepen our understanding of the biological processes involved in the different pre-neoplastic lesions.
The p53 signature lesion is only characterized by a limited number of epithelial cells (~ 10) presenting a p53 overexpression. In our analyses, the p53 signature samples are close in their protein content to the normal tissue. But some proteins speci c to these lesions could be obtained. A signature emerged and demonstrated that among the 8 proteins overrepresented in p53 signature, some of them are unfavorable prognosis markers of OC (CAVIN1, SPTAN1 and FBLN5) and some for renal, head and neck cancers (EIF3B) or liver cancer (MOB1B,EMILIN-2, EIF3B), by contrast, 3 are favorable prognosis markers for renal cancer (SPTAN1, SPTBN1, CAVIN2). We also dissected the molecular mechanisms that occur and the whole proteome analysis showed that the transition from normal to p53 signature was characterized by a residual immune response and translation activation, regulation of tra cking through cavins proteins, and ECM modi cations through emilins.
From p53 signature to STIL: upregulated proteinswere involved in the adhesion and anchoring of cells in the ECM, allowing stabilization of the cells, along with an increase in maintenance activity instead of cell proliferation. Immune response and in ammation processes are still maintained, but an overall decrease of all metabolic processes, especially in TCA cycle is observed and with no Warburg effect occurs. These observations are in line with a recent study suggesting that STIL could be considered as "Dormant STICs" and take a prolonged time of more than one decade to develop into STIC (Wu et al., 2019). We observed that mostly pathway concerning metabolism and cell growth and maintenance is downregulated in STIL that could explain this "dormant" characteristic.
From STIL to STIC: Contrary to the dormant pro le of STIL lesion, observations done on the different function dysregulation for STIC lesion, suggest a more aggressive pro le with a decrease of proteins involved in the cell adhesion and could re ect the start of an invasion step. Proteins that are upregulated in the STIC show various functions inside the cells and appear to be involved in different steps of cancer processes. An increase of proteins involved in the metabolism process and energy pathway are observed. If we compare the molecular functions involved in the three different groups, calcium-dependent protein binding is overrepresented. Also, heat shock protein activity is highly represented in this dataset. Proteins involved in telomere activities and centrosome maturation are speci c to STICs compared to other lesions.
Interestingly, these two processes have been shown to be linked with early tumorigenesis (Kuhn et al., 2010) and that STIC precedes the development of many HGSCs (Kuhn et al., 2016). A decrease of expression of proteins involved in the extracellular structure organization and cell adhesion molecule activity was observed in STIC lesion, suggesting a breaching of the basement membrane and an increase of cell motility that could be conducted to an escape of these pre-cancerous cells to a distant organ. Finally, we observed a high level of proteins involved in extracellular vesicles including exosomes production which could consequently modify the microenvironment and promote cancerization processes.
From lesions to HGSC: One protein is found common in the pre-neoplastic lesions and HGSC, the Fibrillarin (FBL) protein. FBL is one of the core proteins of box C/D small nucleolar ribinucleoprotein complexes (snoRNP). This complex is involved in the rst steps of pr-rRNA processing (Newton et al., 2003). An overexpression of FBL contributes to tumorigenesis and confers cellular resistance to chemo drugs (El Hassouni et al., 2020). It has also been shown that dysregulation of ribosome biogenesis plays key roles in oncogenesis (Truitt and Ruggero, 2016). Overactivation in cancer cells of ribosome biogenesis could be due to a loss of function of RNA polymerase repressors such as p53 (Zhai et al., 2000). Particularly FBL expression has been demonstrated to be correlated with p53 activity (Marcel et al., 2013). High levels of FBL protein were associated with the expression of mutant p53 and contributed to tumorigenesis by altering translational control of key cancer genes. UBTF is only present in STIC and HGSC. This protein is known to be a transcription factor involved in the regulation of the RNA polymerase I (Pol I), implicated in the regulation of cell cycle checkpoint and DNA damage response (Sanij et al., 2015)and regulated by MYC (Rossetti et al., 2018). different studies have shown a link between Pol I initiation factor and chemoresistance of ovarian cancer (Cornelison et al., 2017) Similarities between these pre-neoplastic lesions and HGSC have been found especially concerning TP53 mutations and other genomic changes, but without excessive cell proliferation (Yamamoto et al., 2016). In that context, we decided to explore the presence of mutations at the protein level in the different steps of tubal tumor development. Using a speci c database for protein mutation, XMAn, resulted in the identi cation of 83 mutated peptides sequences, of which 4 were observed directly in the tandem mass spectra. Taken together, it is interesting to observe that most of the protein mutated in the p53 signature have been initially identi ed in endometrium cancer and are related to Histones and ribosomes suggesting a link with the epigenetic and translation.
Polyubiquitin C is related to proteasome and self-antigen presentation. Synaptopodin-2 and the Isoform 2 of Drebrin-like protein are actin-binding proteins known to be respectively an invasive cancer biomarker (Kai et al., 2015) and a potential marker for breast, lung, and colorectal cancers (Chen et al., 2016). Finally, the Actin, alpha cardiac muscle 1 is involved in cisplatin ovarian cancer cell resistance (Pan et al., 2009) and the synaptic vesicle membrane protein VAT-1 homolog is a marker of epithelial cells (Koch et al., 2003) and an oncogene in gastric cancer (Mottaghi-Dastjerdi et al., 2016). None of these mutated proteins have been found in STIL and STIC except the ones occurring on histones. Most of the mutated peptides observed in STIL, STIC, p53/STIL, or in Normal/p53 signature/STIL/STIC are linked to the cytoskeleton and migration. These mutations could be key drivers of the cancer process from the epigenetic modi cation step, and then on cytoskeleton modi cations for cell migration and cancer development.
In our present study, some of the identi ed AltProt or transcripts have been described in other cancer studies. The pseudogene KRT8P32 has been identi ed in breast cancer (Hardman et al., 2019), the pseudogene ACTBP11 is observed in GM12878 (B cells) cell line according to Diana-LncBase V3 (Karagkouni et al., 2020). Concerning the ANKRD28, this AltProt has been recently demonstrated to act as a novel BRCA1-interacting protein in breast and ovarian cancer (Vincent et al., 2016). RP11-13K12.2 is described to be overexpressed in pediatric ovarian brosarcoma (Melendez-Zajgla et al., 2018). COL18A1 and B4GALNT4 are overexpressed in endometrial and ovarian cancers according to the HPA and are unfavorable markers for overall survival for both cancers. Some of these are speci c to the type of lesion -i.e. LOC105376924 and COL18A1 in STILL, KRT8P11 in STIC and ANKRD28 in STILL-STIC. These alternative proteins seem to be clinically relevant, and further exploration will lead to a better understanding of the disease and new targets to improve diagnosis.
Finally, the Identi cation of the AltProt and the high number of mutations affecting histones and cytoskeleton proteins observed between normal to p53 signature compared to the other lesions, are in line with the hypothesis of epigenetic reprogramming towards carcinogenesis and cell transformation (Bartlett et al., 2016). A recent study established that epigenetic reprogramming occurs speci cally in the proximal end of the fallopian tubes in BRCA mutation carriers. This epigenetic reprogramming event is driven by aberrantly high AICDA (also named AID, activation-induced cytosine deaminase) expression and is an integral early pre-malignant event in HGSC development. In this context, our results would provide an interesting starting point for further studies, in respect to the potential link to endometrium and exfoliation of endometrial carcinoma cells due to epigenetic reprogramming to carcinogenesis (Mingels et al., 2014) or that STIL represents exfoliated precursor cells that eventually undergo malignant transformation within the peritoneal cavity .

Conclusion
Overall, the present study identi ed several processes involved in the various neoplastic lesions that are observed in the mbriated end of the fallopian tube. Besides, we have demonstrated the association across the different lesions, allowing us to identify the potential timeline and underly mechanisms of origin of ovarian cancer. Several of the identi ed protein hits represent potential markers for differentiating several classes of ovarian related cancers which require further validation in bigger cohorts of clinical samples. Declarations Ethics approval and consent to participate : All experiments were approved by the local Ethics Committee (CPP Nord Ouest IV 12/10) in accordance with the French and European legislation. Prior to the experiments, patients signed an informed consent and authorization form describing the experimental protocol. No personal information was used in these experiments, and a random number was assigned to each sample. Consent for publication : All patient consent for publication at least 12 cells without morphological abnormalities and low Ki-67 index; STIL: same accumulation of p53 in more than 20 cells with some morphological abnormalities and an higher Ki-67 proliferating index (10-40%); STIC: high p53 and Ki-67 index and cells atypical morphology (carcinoma-like). Proteomic analysis of the pre-neoplastic lesions. A) Hierarchical clustering of the most variable proteins between normal tissue, p53 signature, STIL and STIC (n=4 for each category, ANOVA with permutation-based FDR < 0.05); B) Subnetwork Enrichment Analysis was done to highlight altered biological and functional pathways in the different clusters of proteins.

Figure 4
Annotation analysis of gene ontology terms. A) Biological process, B) Transcription factor and C) Cellular component for the clusters of proteins. D) Biological process, E) Biological pathway and F) Molecular function of the proteins overexpressed in each lesion. (hypergeometric test against all annotated gene/protein list of Funrich database, p-value is represented by stars: *** p < 0.001, ** p < 0.01, * p < 0.05 no star for p > 0.05).

Figure 5
Comparison between p53 signature and normal tissue. A) Visualization of a t-test in form of a volcano plot comparing normal to p53 lesion (proteins with p < 0.05 in red.) B) Hierarchical clustering of the signi cant variable proteins between normal tissue and p53 signature (t-test with p < 0.05).

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.