Proteomic Analysis of Pre-Invasive Serous Lesions of the Endometrium and Fallopian Tube Reveals Their Metastatic Potential

Serous endometrial cancer (SEC) and high grade serous ovarian cancer (HGSOC) are aggressive gynecological malignancies with high rates of metastasis and poor prognosis. Endometrial intraepithelial carcinoma (EIC), the precursor for SEC, and serous tubal intraepithelial carcinoma (STIC), believed to be the precursor lesion for HGSOC, can also be associated with intraabdominal spread. To provide insight into the etiology of these precancerous lesions and to explore the potential molecular mechanisms underlying their metastatic behavior, we performed a proteomic mass spectrometry analysis in a patient with synchronous EIC and STIC. Through histological and molecular identification of precancerous lesions followed by laser capture microdissection, we were able to identify over 450 proteins within the precancerous lesions and adjacent healthy tissue. The proteomic analysis of STIC and EIC showed remarkable overlap in the proteomic patterns, reflecting early neoplastic changes in proliferation, loss of polarity and attachment. Our proteomic analysis showed that both EIC and STIC, despite being regarded as premalignant lesions, have metastatic potential, which correlates with the common presentation of invasive serous gynecological malignancies at advanced stage.


INTRODUCTION
Endometrial cancer (EC) is the 6 th most common cancer in women worldwide and is the most common gynecological malignancy (1). Despite significant advances in early detection (2), molecular subtyping (3,4), and new or improved treatment regimens (5), the relative survival of patients with EC has declined in recent times (6,7). Serous endometrial carcinoma (SEC) is a highly aggressive malignancy (8). It represents only 10% of EC cases (9, 10) but is responsible for 39% of EC related deaths (11) and is frequently diagnosed at late stage when prognosis is poor (9,10). The current model of SEC development suggests that it evolves from a pre-neoplastic lesion in atrophic endometrium called endometrial glandular dysplasia (EmGD) (12,13). This lesion then progresses further into endometrial intraepithelial carcinoma (EIC) and, finally, into SEC (14).
EmGD is characterized by loss of cell polarity, nuclear atypia, and nuclear hyperchromasia (12). It has a marked loss of heterozygosity in TP53 and chromosome p1; however, this is to a lesser extent than that seen in EIC and SEC (12). Although quite difficult to identify, nucleomegaly and staining for p53, MIB-1 as well as IMP3 are characteristics of EmGD (14). The connection between EmGD and SEC has been confirmed through identical mutations observed in EmGD and subsequent SEC (15).
EIC was first described in the 1990s (16) and seen to arise almost exclusively in atrophic endometrium and in the context of SEC in a majority of cases (17). EIC exhibit similar features as EmGD but with further nucleomegaly, nuclear irregularity and hyperchromasia (14). However, cases of EIC associated with extrauterine metastasis suggest that EIC may more closely resemble SEC (18,19). This early peritoneal spread is in stark contrast to non-serous endometrial cancers which usually do not show early peritoneal spread but preferentially invade the myometrium and spread to the lymph nodes (18,19).
SEC closely resembles other serous cancers of the female genital tract, such as high grade serous ovarian cancer (HGSOC) (4). Both diseases share similar molecular features and clinical properties (20) and, consequently, are treated in a similar way (21,22).
Results from studies using mouse models have established a connection between STIC and subsequent HGSOC (34,35). Inactivation of PTEN, p53 and BRCA1/2 in the fallopian tubes of mice resulted in STIC and concurrent HGSOC with ovarian and peritoneal spread (34,35). However, in the absence of BRCA1/2 inactivation, STIC developed but did not progress to metastatic HGSOC in this mouse model (34).
Cases of HGSOC arising in the absence of STIC (35)(36)(37) have also been reported suggesting that other, as yet unidentified, precursor lesions might exist (38). "Early Serous Tubal Proliferations" (ESTP) have been identified as a potential HGSOC precursor. They are found in the fimbria (30), demonstrate DNA damage (30), and are found in non-ciliated cells, which also give rise to STIC (39). A physical and lineage continuity has been demonstrated between ESTP and STIC suggesting that some ESTP give rise to STIC and subsequently to HGSOC (30,32).
The understanding that HGSOC arises from the fallopian tube in many cases has changed the understanding of serous ovarian cancer. Now serous cancers of the fallopian tube, peritoneum and ovary are thought to share a common origin in the fallopian tube (20). While it is well established that many HGSOC arise from the fallopian tube, it has not been excluded that some serous cancers of the endometrium and ovary may share common origins. For example, Roelofsen et al. (40) suggested that some serous ovarian cancers (SOC) may arise from EIC by showing that they shared TP53 mutations, similar expression of p53, Ki67, estrogen, and progesterone receptors (40). Additionally, Tolcher et al. (20) analyzed 38 patients with SEC and investigated their fallopian tubes. They found STIC, without evidence of tubal metastasis, in 2 of these cases (20).
To better understand the potential link between serous preinvasive lesions of the female genital tract and serous gynaecological cancer, molecular investigations of the precursor lesions of the endometrium and fallopian tube are required. Here, we present the first proteomic analysis of synchronous precancerous lesions of the endometrium and fallopian tubes in a patient without invasive malignancy, by means of mass spectrometry. This precludes the possibility of these premalignant lesions representing metastases from established primary tumors. The analysis of EIC and STIC in this context provides insight into the temporal and mechanistic features of their development and dissemination.

Sample
Archived formalin-fixed paraffin-embedded (FFPE) fallopian tube and endometrial tissues from a 67-year-old female who had undergone a total abdominal hysterectomy and bilateral salpingo-oophorectomy for endometrial hyperplasia was retrieved for analysis with approval of the Research Ethics Committee of the Royal Adelaide Hospital. The fallopian tube and endometrial tissues were processed using standard procedures, stained with hematoxylin and eosin, and annotated by a pathologist. P53 and MIB1 immunostaining was also performed to confirm the location of precancerous lesions.

Laser Microdissection and Sample Preparation
FFPE tissues were sectioned at 4-µm thickness, water bath mounted onto PEN membrane slides (Micro-Dissect, Herborn, Germany), and deparaffinized by submersion in xylene for 5 min, following by two 2-min incubations in 100% ethanol, and two 5min incubations in water. Areas of STIC, EIC and adjacent healthy epithelium were dissected using a Leica AS LCM microscope (Leica Microsystems, Wetzlar, Germany) into 20 ml of 10 mM citric acid buffer (pH = 6) and subjected to heat induced antigen retrieval by incubation at 100°C for 90 min.

Nanoflow Liquid Chromatography Tandem Mass Spectrometry
Nanoflow liquid chromatography tandem mass spectrometry (Nano-LC-MS/MS) was performed on each sample in duplicate using an Ultimate 3000 RSLC system (Thermo-Fisher Scientific, Waltham, USA) coupled to an Impact HD ™ QTOF mass spectrometer (Bruker Daltonics, Bremen, Germany) via an Advance CaptiveSpray source (Bruker Daltonics). Peptide samples were pre-concentrated onto a C18 trapping column (Acclaim PepMap100 C18 75 mm × 20 mm, Thermo-Fisher Scientific) at a flow rate of 5 ml/min in 2% (v/v) ACN 0.1% (v/v) TFA for 10 min. Peptide separation was performed using a 75 mm ID C18 column (Acclaim PepMap100 C18 75 mm × 50 cm, Thermo-Fisher Scientific) at a flow rate of 0.2 ml/min using a linear gradient from 5 to 45% B (A: over 130 min, followed by a 20min wash with 90% B, and a 20-min equilibration with 5% A. MS scans were acquired in the mass range of 300 to 2,200 m/z in a datadependent fashion using Bruker's Shotgun Instant Expertise ™ method. This method uses IDAS (intensity dependent acquisition speed) to adapt the speed of acquisition depending on the intensity of precursor ions (fixed cycle time), and RT 2 (RealTime Re-Think) to exclude previously selected precursor ions from undergoing refragmentation unless the chromatographic peak intensity of the ion has increased by a factor of 5. Singly charged precursor ions were excluded from acquisition. Collision energy ranged from 23% to 65% as determined by the m/z of the precursor ion.

Data Analysis
Spectra were analyzed using the MaxQuant software (version 1.5.2.8) with the Andromeda search engine (43) against the UniProt non-redundant human database. The standard Bruker QTOF settings in MaxQuant were used with a mass error tolerance of 40 ppm. The variable modifications of oxidation of methionine and the fixed modification of carbamidomethyl of cysteines were specified, with the digestion enzyme specified as trypsin. The protein false discovery rate (FDR) and peptide spectrum match FDRs were both set to 1% using a target decoy approach, with a minimum peptide length of 7 amino acids (43). Only unique and razor peptides were used when reporting protein identifications.

Gene Expression Analysis
In order to assess the gene expression levels of the corresponding proteins of interest in early stage I ovarian carcinoma tissues, the dataset of Yoshihara et al. (44)

Histological Analysis Identifying STIC and EIC
Upon analysis of the H&E stained tissue sections, atypical intraepithelial proliferation involving a small population of cells in the endometrium were identified. These changes were consistent with early stages of EIC (Table 1). Similarly, atypical changes involving a small population of cells in the fimbriated tube ( Table 1), consistent with early stages of STIC, were detected and both EIC and STIC are represented in Figure 1. Immunoperoxidase staining for p53 and MIB1 revealed atypical intraepithelial epithelial proliferations in both the fimbriated tube and endometrial lining, confirming the presence of STIC and EIC (data not shown).

Proteomic Comparison of Healthy Epithelia to STIC and EIC
Regions of STIC, EIC, and adjacent healthy epithelium were laser microdissected (LMD) from sectioned tubal and endometrial specimens and analyzed by Nano-LC-MS/MS. In total, 453 proteins were detected across the 4 tissue types (369 proteins in the STIC, 110 proteins in healthy tubal epithelium, 428 proteins in EIC, and 162 proteins in healthy endometrial epithelium (Supplementary Table 1).
When comparing the numbers of identified proteins, the greatest overlap occurred between STIC and EIC with 348 identical proteins identified in both samples. Across all tissue types 73 proteins were detectable, with 96 identical proteins identified in both the STIC and healthy tubal tissue, 157 identical proteins detected in both the EIC and healthy endometrial tissue, and 85 identical proteins were detected in both healthy tissues ( Figure 2).

Proteins Relevant to STIC, EIC, and the Development of Gynecological Cancer
In analyzing potential links between STIC and EIC cells and their respective invasive carcinomas, two of the identified proteins were of particular interest based on their involvement in gynecological cancer development: Epithelial cell adhesion molecule (EPCAM) and Calcyphosin (CAPS) (47)(48)(49). EPCAM was identified in both the STIC and EIC tissue samples but was not detected in either of the healthy tissues while CAPS was only detected in the healthy tube and in STIC.  To evaluate the expression of EPCAM and CAPS in early stage I serous ovarian carcinoma tissues, oligonucleotide microarray data was considered from the GEO data set GSE12470 (44). We chose to investigate the expression of these genes in early stage ovarian cancer as this is expected to be the stage following STIC in the development of serous ovarian cancer. STIC lesions were not deemed an appropriate comparison samples as they are often taken in the context of metastatic disease and potentially represent metastatic implants, and therefore more developed cancer, rather than true precursor lesions. The gene expression levels were analyzed in 10 healthy peritoneal control tissues compared to 8 Stage I serous EOC tissues. The median expression levels of both CAPS and EPCAM were found to be significantly increased in the Stage I serous OC tissues compared to healthy peritoneal controls (p = 0.00014 and p = 3.3 × 10 −7 , respectively) ( Figure 3).
Healthy OSE is infrequently available for research purposes given healthy ovaries are rarely removed during any type of medical procedure. Healthy peritoneum is an effective control because the lining of the ovaries is comprised of a single-cell mesothelial layer of poorly differentiated epithelium derived from the coelomic epithelium and extended to the serosa peritoneal cavity (50). In addition, it has been reported that peritoneal mesothelium and OSE are structurally very similar (51) and are both negative for EPCAM and CAPS expression (52,53).

Expression of EPCAM and CAPS in Early Stage Endometrial Carcinoma Compared to Healthy Endometrial Tissue
To assess the expression of EPCAM and CAPS in early stage endometrial cancer, oligonucleotide microarray data was considered from the GEO data set GDS4589. The gene expression levels were analyzed from 79 endometrioid and 12 serous papillary endometrial carcinomas as well as 12 healthy controls. The median expression of CAPS was increased significantly in EEC (p = 0.00066), while it was significantly decreased in SEC (p = 0.023) compared to healthy endometrium. The median expression of EPCAM was significantly increased in EEC (p = 0.02) and further increased in SEC (p = 0.0007) compared to healthy controls ( Figure 4).

Expression of Proteins Identified Exclusively in the STIC or EIC Across Ovarian Cancer Subtypes
Proteins detected exclusively in either the STIC or EIC tissues were analyzed to determine if their expression is specific to  Table 2. The remaining proteins appeared to be more homogenously expressed and hence were not analyzed further. The gene expression levels corresponding to the proteins listed in Table 2 were compared using oligonucleotide microarray data from the GEO data set GSE6008 ( Figure 5).  For the gene TPPP3 there was reduced expression in ovarian clear cell carcinoma compared to healthy tissue and other ovarian cancer subtypes. SORD showed significantly lower expression levels in the normal control tissues, and to a smaller extent, in serous ovarian carcinoma compared to the other subtypes. The high expression of SORD, whose related protein was detected only in EIC in our data set, in non-SOC was unexpected as the protein atlas reports low expression of SORD in ovarian cancer. CSPG2/VCAN expression, which was detected at the protein level in EIC only, was increased in CCOC when  compared to the normal tissue and other EOC subtypes ( Figure 5).

Expression of Proteins Identified Exclusively in STIC or EIC Across Endometrial Cancer Subtypes
By investigating the gene expression corresponding to proteins in our data set which are enriched in specific gynecological tissues, we aimed to investigate the connection between preneoplastic lesions of the endometrium and tube with subtypes of endometrial cancer as defined by the TCGA. The expression levels of the genes corresponding to the proteins listed in Table 1 were compared using oligonucleotide microarray data from Kandolth et al. (4) (https://gdc.cancer.gov/node/875) ( Figure 6). They were selected as their related proteins were found exclusively in a precancerous lesion and their abundance  (according to protein atlas) was unique to specific gynecological tissue. However, most did not show different gene expression between the TCGA defined subtypes, except for TPPP3 and SPATA18, which both showed decreased expression in the CN high (serous) subgroup compared to other subtypes ( Figure 6). These proteins were detected exclusively in STIC in our analysis and their expression was seen to be decreased in SEC compared to healthy controls (Supplementary Figure 2).

Proteins Associated With Metastasis Identified in STIC or EIC
The 34 identified proteins expressed exclusively in STIC and/or EIC which are known to be involved in metastasis or migration were listed in Supplementary Table 2. Most of these proteins are implicated in the promotion of metastasis, with the exceptions of Galectin-9 (54, 55), Mimecan (56,57), and Catenin alpha-1 (58,59) which have been implicated in the inhibition of metastasis.

DISCUSSION
The application of proteomic techniques, particularly that of mass spectrometry, hold the potential to provide a temporal snapshot of the molecular features within a given sample. Here, we provide what is, to the best of our knowledge, the first proteomic analyses of synchronous precursor lesions of serous endometrial and high-grade ovarian cancer. Through histological and molecular identification of precancerous lesions, followed by laser-capture microdissection and mass spectrometry analysis, we were able to identify over 450 proteins within the precancerous lesions and adjacent healthy tissue. The proteomic profiles of the precancerous lesions showed striking similarity ( Figure 1) and shared a molecular profile indicative of metastatic transformation (Supplementary Table 2). To investigate the connection between these precancerous lesions and serous gynecological malignancies we investigated the gene expression of several proteins of interest in serous endometrial and ovarian cancer data sets (Figures 3-6). EPCAM and CAPS were selected for further investigation based on their implication in the development of ovarian and endometrial cancer (47)(48)(49). Genomic analysis of their expression in Stage I SOC compared to peritoneal control tissue revealed significantly increased levels in the cancerous specimens ( Figure 3). It is well recognized that EPCAM expression has a complex relationship to SOC development (60); however, it has not previously been identified in STIC or other gynecological precancerous lesions. This transmembrane adhesion molecule plays a role in migration and proliferation in wound healing (61) as well as the maintenance of pluripotency in stem cells (62).
The exact function of CAPS is unknown, but it is a suggested target of cAMP-dependent protein kinase and has been implicated in the cAMP and calcium-phosphatidylinositol signaling cascades (63). According to protein atlas, CAPS has high expression in the fallopian tube but is not expressed in healthy OSE (http://www.proteinatlas.org/ENSG00000152611-CAPSL/tissue) which is in agreement with our identification of this protein in both the healthy fallopian tissue and STIC.
Interestingly, CAPS gene expression was seen to be increased in SOC compared to control tissue ( Figure 3). Furthermore, EPCAM is not expressed in healthy ovarian surface epithelium (OSE) (53) but is frequently up regulated in ovarian cancer (60). The single case study presented here is insufficient to draw broader conclusions about potential markers of tissues of origin, but we believe that expression of EPCAM and CAPS merit further investigation in a larger cohort.
To further investigate proteins which may act as markers of tissue of origin in SEC or SOC, we investigated proteins identified exclusively in either the STIC or EIC which were unique to specific gynecological tissues according to protein atlas. CSGP/VCAN was identified exclusively in EIC, was seen to have increased expression in COCC compared to other ovarian cancer cell types ( Figure 5) and had increased expression in SEC compared to healthy control tissue (Supplementary Figure 2). This protein has been previously seen to be increased in COCC (64,65) and represents a potential link between EIC and COCC. COCC has been suggested to arise from endometriosis lesions (66) potentially representing a pathway for endometrial origin of COCC.
A previous mass spectrometry-based analysis of ovarian cancer precursor lesions was performed by Levenon et al. (67) and investigated ex vivo culture derived cells from fallopian tube fimbria (67). Their proteomic analysis identified 11 different ovarian cancer biomarkers present in this ex vivo model. A larger study by M. Eckart et al. (68) investigated both tumor and stroma tissue from STIC, invasive fallopian tube lesions, invasive ovarian lesions and omental metastasis (68). In addition to identifying Nmethyltransferase (NNMT) as a metabolic regulator of cancer associated fibroblasts, they also showed that the molecular profiles of primary cancers and metastatic implants were remarkably similar within the same patient while the microenvironment showed site specific differences. The STIC lesion they investigated showed lower expression compared to normal tube in 4 of our proteins of interest (CAPS, ERO1A, TPPP3, and SPATA18) and similar expression in 2 (EPCAM and SORD). Only VCAN showed marginally increased expression in STIC compared to normal fallopian tube epithelium in this analysis. The authors' focus on stromal tissue, a lack of normal epithelial ovarian or omental controls and the potential that STIC represents metastatic lesions from a primary ovarian tumor limits further comparison to this data set.
The traditional understanding of cancer development is that it acquires metastatic capacities over time within the primary lesion (69). However, the identification of STIC as the origin site of HGSOC implicates that these pre-invasive lesions can leave their site of origin and establish themselves in distant locations. In addition, EIC is often identified with extrauterine spread (18,19) which can take the form of EIC like growths on the ovaries, peritoneum and fallopian tube in the absence of obvious disease in these locations (16,18). Together, this suggests that these premalignant lesions possess some migratory or metastatic ability facilitating their translocation to distant sites within the gynecological system. Here, we identified numerous metastasis and migration related proteins in precancerous lesions in the absence of malignant disease (Supplementary Table 2).
A major limitation of this study is that the data is derived from a single patient case study which makes generalized interpretation of biological implication of this data difficult. Furthermore, only a modest number of proteins was identified from these small tissue regions. As highly abundant proteins are identified preferentially in mass spectrometry analysis (70), the low number of proteins identified potentially masks differences in lower abundance proteins. The analysis of a larger cohort, coupled with the utilization of advanced sample preparation and mass spectrometry techniques to improve proteome coverage, holds the potential to build upon the data presented here and paint a clearer picture of the molecular landscape of precancerous lesions of serous cancers of the ovary and endometrium.

CONCLUSION
Here, we present the first proteomic investigation of precancerous lesions of the gynecological tract in a patient without advanced gynecological malignancy. Interpretation of the data is limited by the single case study and the modest number of proteins identified; however, we provide a foundation for further analysis of the molecular links between precancerous lesions and subsequent caner. In addition, we identified several metastasis-related proteins in precancerous tissues. The understanding that precancerous lesions of the female genital tract potentially possess metastatic potential raises many questions about when, where, and how these cancers develop. Though the early steps are not well understood, further proteomic analyses of gynecological precancerous lesions hold the potential to unravel the early temporal and molecular events underlying the development of these malignancies which, in turn, holds the potential to improve detection and treatment.

DATA AVAILABILITY STATEMENT
Publicly available datasets were analyzed in this study. These can be found in the NCBI Gene Expression Omnibus. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository (71) with the data set identifier PXD018538.

ETHICS STATEMENT
The studies involving human participants were reviewed and approved by Research Ethics Committee of the Royal Adelaide Hospital. The patients/participants provided their written informed consent to participate in this study.

AUTHOR CONTRIBUTIONS
Design of the work: MO, PH, GA, and MA. Acquisition, analysis, and interpretation of data: all authors. Drafting the work or revising it critically: all authors. All authors contributed to the article and approved the submitted version.

FUNDING
PH gratefully acknowledges the support of the National Collaborative Research Infrastructure Strategy (NCRIS) Bioplatforms Australia Node for Tissue Imaging Mass Spectrometry.