Murine oviductal epithelial (MOE WT, equivalent of human fallopian tube epithelial cells) cells were donated by Dr. Barbara Vanderhyden from the University of Ottawa. SKOV3 cells were purchased from ATCC (HTC-77). MOE SCR^shRNA and MOE PTEN^shRNA cells are from the Burdette Lab and have been validated (Dean et al., 2019).

MOE SCR^shRNA and MOE PTEN^shRNA cells were maintained in ɑMEM with Earle’s salts, ribonucleosides, deoxyribonucleosides, and l-glutamine (Thomas Sci 10–022-CV) supplemented (in 500 mL bottles) with 11 µL gentamicin (50 mg/mL stock, Cellgro 30–005-CR), 5 mL l-glutamine (GIBCO 25030–081), 275 µL pen/strep (10,000 U/mL or 100X stock, Thermo Fischer 15140–122), 10 µL EGF (0.1 mg/mL stock, Roche 855731), 550 µL ITS (1000X stock, Roche 1074547), 10 µL β-estradiol (1 mg/mL in 100% EtOH, Sigma Aldrich E2257), and 10% fetal bovine serum (FBS). SKOV3 cells were maintained in McCoy’s 5A (modified) medium (GIBCO 1660108) supplemented with 1.1 g sodium bicarbonate and 1% pen/strep (Thermo Fisher 15140–122), and 10% FBS. All cells were maintained in T-75 flasks, incubated at 37°C and 5% CO₂, and passaged every 3–4 days.

MOE PTEN^shRNA cells were plated in 6 well plates at a density of 75,000 cells per well 24 h prior to transfection. A total of 400 ng/mL of SPARC endoribonuclease small interfering RNA (siRNA) (Sigma-Aldrich EMU088951) was transfected into MOE PTEN^shRNA cells using Mirus TransIT X2 (Mirus Bio LLC MIR6000) according to manufacturer’s protocol. Media was changed after 5 h, and cells were split and re-seeded at 48 h post-transfection into a 10 cm plate. Cells and conditioned media were then incubated for 48 h and collected for protein analysis. A plasmid that expresses mCherry was used as a positive control. A universal negative control siRNA was used as a negative control (Sigma-Aldrich SIC001).

CD-1 mice were obtained from in-house breeding. Animals were housed in a temperature and light (12L:12D) controlled environment. Water and food were provided ad libitum. All animals were treated in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals. On day 16–18 after birth of pups, the ovaries were removed dissected free of the uterus, fallopian tube, and bursa using a dissecting microscope (Leica MZ6).

Conditioned media for Figure 1 and Figure 2 were collected by allowing cells to incubate in normal cell culture media (see above) for 4 days at 37°C and 5% CO₂ in T-75 flasks. Many of the subsequent analyses (Figure 4 and Figure 5) for the identification of proteins required BCA assays to determine protein concentration. Since the conventional media used for MOE cells is rich in proteins and phenol red interferes with the colorimetric readout of BCA analysis, these experiments utilized conditioned media that were serum-free and phenol-red free.

For these later analyses, MOE SCR^shRNA and MOE PTEN^shRNA cells were maintained in phenol red- and serum-free minimum essential medium, alpha (ɑMEM) with nucleosides (GIBCO 41061037) supplemented (in 500 mL bottles) with 11 µL gentamicin (50 mg/mL stock, Cellgro 30–005-CR), 5 ml l-glutamine (GIBCO 25030–081), 275 µL pen/strep (10,000 U/mL or 100X stock, Thermo Fischer 15140–122). SKOV3 cells were maintained in phenol red- and serum-free RPMI 1640 medium (GIBCO 1185055) supplemented with 1.1 g sodium bicarbonate and 1% pen/strep (Thermo Fischer 15140–122). All cells were maintained in T-75 flasks, incubated at 37°C and 5% CO₂ for 4 days. To ensure biological consistency, IMS analyses of biomolecular fractions and protein fractions were repeated using serum-free and phenol red-free conditioned media, yielding the same results as with conventional media (data not shown).

Conditioned media from T-75 flasks (10 mL) were concentrated to approximately 4 mL using a centrifugal evaporator (Labconco CentriVap Concentrator (7810016) and CentriVap Cold Trap (7385020)) at 35°C to accommodate volumes of spin column filters (Millipore Sigma UFC8003). To generate the protein fraction, conditioned media was spun in a 3 kDa spin column at 4000 rpm and 25°C for 40 min in a centrifuge using a swing rotor (Eppendorf Centrifuge 5810 R). Material over 3 kDa was collected from the filter into a pre-weighed microcentrifuge tube, and the filter was rinsed and collected twice with 100 µL DI water, which was combined with the protein material. The flow through media was collected and further partitioned into lipid and polar fractions using a separatory funnel. Chloroform (3 × 1 mL) was added to the media and gently shaken. The lipid fraction was collected and the aqueous flow through was transferred to pre-weighed microcentrifuge tubes and dried in vacuo.

The material retained by the 3 kDa spin column was further fractioned by successive centrifugation through spin columns with molecular weight cut-off filters in the following order: 100 kDa (Millipore Sigma UFC8100), 50 kDa (Millipore Sigma UFC8050), 30 kDa (Millipore Sigma UFC8030), spun at 4000 rpm (Eppendorf Centrifuge 5810 R). This generated four fractions of molecular weight ranges: >100 kDa, 50–100 kDa, 30–50 kDa, and 3–30 kDa (Supplementary Figure S1). Each concentrate was transferred to a pre-weighed microcentrifuge tube and the filter was rinsed twice with 100 µL of DI water which was transferred to the tube. All fractions were dried in vacuo.

General setup followed the protocol established in Zink et. al., except that conditioned media was used in place of a cellular mixture for co-culture with halved murine ovaries (Zink et al., 2018). Low-melting agarose (2%, Sigma A9414) was liquified from a solid at 70°C. All conditioned media fractions were normalized to the fraction with the lowest weight, generating the highest normalized concentration possible into 150 µL of media. Then, conditioned media fractions were mixed 1:1 with the liquified 2% low-melting agarose. Halved ovaries were placed in the center of wells in an 8-well chamber (Lab-Tek 177445) adhered to an ITO-coated glass slide (Bruker 8237001). Each condition (300 µL) was plated into wells, avoiding air bubbles, and ensuring the ovarian tissue remained in the center of the well. In conditions with addition of exogenous SPARC, the compound was resuspended in DI water and used at 50 ng/150 µL (Mus musculus, R&D Systems 942SP050). Slides were placed in the humidified incubator at 37°C and 5% CO₂ and left to incubate for 4 days. After 4 days, the agarose plugs were cut away from the 8-well chamber and the chamber was detached from the slide. The agarose plugs were dried on the ITO-coated glass slide in an oven at 37°C for 2–4 h, monitored closely for wrinkling of the agarose (Lusk et al., 2022).

Matrix application protocol was used as previously described, aside from an increase to the concentration to 10 mg/ml (Zink et al., 2018).

Due to differing access to instrumentation, data were collected in two different ways. In each figure legend, the method is noted to indicate how the IMS data were generated. All RAW and processed data are available at the MassIVE (massive.ucsd.edu) accession number MSV000089866.

Method A: Prior to IMS analysis, slides were scanned at 1200 dpi, and resulting images were used to guide irradiation. IMS data were acquired using flexControl v 3.4 at 50 μm spatial resolution on an Autoflex Speed LRF instrument (Bruker Daltonics) over the mass range 100–2000 Da. In positive reflectron mode, laser power was set to 40%, laser width to 2 (small), and reflector gain to 2.0 ×. For each raster point, 500 laser shots at 2000 Hz were shot in a random walk method. Data were subsequently analyzed in flexImaging v 4.1 × 64 (Bruker Daltonics). All spectra were normalized to the total ion count (TIC). The instrument was calibrated manually using phosphorus red.

Method B: Prior to IMS analysis, slides were scanned using Tissue Scout (Bruker Daltonics), and resulting images were used to guide irradiation. IMS data were acquired using timsControl v2.0.51.0_9669_1571 and flexImaging 5.1 software at 100 μm spatial resolution on a timsTOF fleX instrument (Bruker Daltonics). The data were collected using the mass range 50–1500 Da in positive mode with laser power set to 90.006%, and laser width to 100 µm imaging. For each raster point, 1,000 laser shots at 1,000 Hz were shot. Data were subsequently analyzed in SCiLS™ Lab version 2021b core (Bruker Daltonics). The instrument was calibrated manually using phosphorus red.

Protein concentrations for concentrated conditioned media samples were determined via BCA assay and 50 µg of protein was enzymatically digested prior to mass spectrometry analysis. Ammonium bicarbonate (ABC, 50 mM in DI H₂O) was added to create a uniform sample volume. Dithiothreitol (DTT) was added to the digest at final concentration of 10 mM and incubated for 15 min in a 55°C water bath. After cooling to room temperature, iodoacetamide was added to a final concentration of 30 mM and then incubated for 20 min in a 37°C water bath. Trypsin was added (1 ng trypsin/20 ng protein) and samples incubated overnight in a 37°C water bath. After overnight incubation, the resulting tryptic peptides were enriched and desalted via C18 ZipTips using the manufacturer’s suggested protocol and eluant dried in vacuo. Solution was resuspended in 0.1% formic acid to 1 μg/μL prior to LC-MS/MS analysis.

The peptide digests were analyzed as described previously by Pergande et al. Using a Q Exactive Hybrid Quadrupole-Orbitrap mass spectrometer (Thermo Scientific) which was interfaced to an Agilent 1260 nano/capillary high-performance liquid chromatography (LC) system (Pergande et al., 2021). Mobile phase A was 0.1% formic acid in water and mobile phase B was 0.1% formic acid in acetonitrile. First, peptides were loaded onto an Zorbax 300SB-C18 trap cartridge (Agilent Technologies) and washed with 0.1% formic acid at a flow rate of 2 μL/min and 10 min. Next, the peptides were chromatographically resolved on a Zorbax 300SB-C18 column (3.5 µm i.d. × 150 mm, particle size 5µm, pore size 100Å, Agilent Technologies) using a 5–60% B 60-min linear gradient for at a flow rate of 250 nL/min. The mass spectrometer was operated in data-dependent acquisition mode. The instrument conditions were set to acquire data by data-dependent parameters, utilizing the top 10 peptides per duty cycle, switching between MS and MS/MS. The scan mode for data acquisition in MS1 was set to perform full scan (m/z 375–1600) at a resolution of 70,000 using automatic gain control and with a target of 1 × 10⁶ ions. For MS/MS data acquisition the resolution was set to 17,500 with a target of 1 × 10⁵ ions. The isolation window set was m/z 1.5, utilizing a collision energy of 27.0 eV and dynamic exclusion at 20.0 s. The source ionization had a spray voltage of 1.9 kV with a capillary temperature set to 280°C, s-lens RF, 50.0. MOE PTEN^shRNA and MOE SCR^shRNA samples (N = 5) were analyzed in duplicate (n = 2). SKOV3 samples (N = 5) were analyzed in triplicate (n = 3). All raw and mzML mass spectrometry data is publicly available at MassIVE (massive.ucsd.edu) under the accession number MSV000089866. Protein ID’s were filtered with the following parameters: abundance ratio (MOE PTEN^shRNA/MOE SCR^shRNA) > 1.0, molecular weight (MW) 25–40 kDa as we suspected the protein of interest was approximately 30 kDa in size, and sequence coverage >8%. For the MOE PTEN^shRNA and MOE SCR^shRNA conditioned media, we filtered for the Mus musculus species; for SKOV3 conditioned media, we filtered for Homo sapiens species.

Raw data files from LC-MS/MS analysis of the trypsin digested samples were imported into Proteome Discoverer (version 2.2, Thermo Fisher) and analyzed using a label-free, relative quantitation method. Protein identifications for MOE PTEN^shRNA and MOE SCR^shRNA cohorts were obtained by searching against both the Mus musculus (25230 sequences) and contaminants (298 sequences) databases. Similarly, protein identifications for SKOV3 samples were obtained by searching against the Homo sapiens (42368 sequences) and contaminants database employing a 1% false discovery rate and the Perculator algorithm. Here, trypsin was set as the protease with two missed cleavages and searches were performed with precursor and fragment mass error tolerances set to 10 ppm and 0.02 Da, respectively, where only peptides precursors of +2, +3 and +4 were considered. Peptide variable modifications allowed during the search were oxidation (M) and deamination (NQ), whereas carbamidomethyl (C) was set as fixed modifications. Then, a label-free relative quantitation analysis was performed for MOE PTEN^shRNA relative to MOE SCR^shRNA significance determined by applying an unpaired t-test (p ≤ 0.05).

We hypothesized that a secreted factor was produced from tumorigenic FTE and that it was responsible for inducing the release of ovarian NE. The induction of ovarian NE release occurred when ovaries were cultured with tumorigenic MOE PTEN^shRNA cells but not with non-tumorigenic MOE SCR^shRNA cells. The only engineered difference between the MOE PTEN^shRNA and MOE SCR^shRNA models is the shRNA directed against the tumor suppressor gene PTEN, which enables the MOE PTEN^shRNA line to form rapid cancer from intraperitoneal injection in murine models. As such, we continued to use the MOE SCR^shRNA as a non-tumorigenic control while investigating the tumorigenic MOE PTEN^shRNA, its conditioned media, and potential secreted factors.

As shown in our previously described IMS co-culture model, the typical distance between the cells and the ovary in the divided chambers was 1–2 mm and the matrix supporting the culture was comprised of agarose (Zink et al., 2018). Therefore, we predicted that the FTE factor responsible is likely both secreted and readily diffusible in order to reach and interact with the ovary. To test whether secreted factors from the cells were sufficient to induce ovarian NE release, we opted to alter our co-culture model to utilize conditioned media rather than cells themselves. Woznica et al. described an innovative bioassay-guided fractionation scheme to determine the protein responsible for inducing mating in a choanoflagellate species (Woznica et al., 2017). Inspired by this approach, we aimed to fractionate conditioned media from MOE PTEN^shRNA cells to determine whether a specific class of biomolecule was responsible for inducing the release of NE. Supplemental Figure S1 outlines the steps taken to separate cell-free conditioned media into three biomolecular classes: lipids, polar small molecules, and proteins.

After 4 days of incubation with half of a murine ovary, all the cell free conditioned media fractions cultured with an ovary were analyzed using MALDI-TOF IMS. NE was detected from the ovary that was cultured with the protein fraction (>3 kDa) (N = 3) (Figure 1A), indicating that the FTE factor influencing ovarian NE release was proteinaceous. Further fractionation was then applied to the protein fraction to generate four protein fractions with unique mass ranges: 3–30 kDa, 30–50 kDa, 50–100 kDa, and >100 kDa. Co-incubation of ovarian tissue with these protein fractions, normalized by dry weight, resulted in the ionization of NE from both the 3–30 and 30–50 kDa protein fractions (N = 3) (Figure 1B).

The release of ovarian NE was detected when incubated with MOE PTEN^shRNA secreted proteins fractionated to 3–30 kDa but not with those from MOE SCR^shRNA (Supplemental Figure S2). We used the mass range 3–50 kDa because the NE signal was seen in both the 3–30 kDa and 30–50 kDa co-cultures. Additionally, SDS-PAGE of the serum-free MOE PTEN^shRNA protein fraction in the 3–50 kDa size range generated a prominent band just below 35 kDa, which was not present in the same protein fraction from MOE SCR^shRNA as identified by Coomassie Blue and Zinc staining (Supplemental Figure S3). To identify the protein factor, the area surrounding this band from both MOE PTEN^shRNA and MOE SCR^shRNA fractions were digested for bottom-up proteomics. With MOE SCR^shRNA used as the control, digested peptides (N = 5, n = 2) were analyzed using LC-MS/MS. All identifications across runs were tallied, duplicates removed, and proteins were compared between MOE PTEN^shRNA and MOE SCR^shRNA. The identification workflows for all biological replicates cumulatively identified 45 proteins in MOE PTEN^shRNA and 48 in MOE SCR^shRNA. Of these proteins, 27 were shared between the cell lines, 21 were unique to MOE SCR^shRNA, and 18 were unique to MOE PTEN^shRNA (Supplemental Table S1). The relative protein abundances from each cohort were used to generate a list of proteins which was filtered to select for increased fold change in the MOE PTEN^shRNA samples. Ultimately, three proteins were observed to have an increased abundance in the MOE PTEN^shRNA samples relative to the MOE SCR^shRNA samples. SPARC (secreted protein acidic and rich in cysteine), also known as osteonectin, had the highest sequence coverage and number of unique peptides (Table 1).

To explore possible cross species confirmation, we investigated whether a human ovarian cancer cell line previously reported to display enhanced migration and peritoneal spread in response to NE was also capable of inducing NE production from the ovary. The ability of SKOV3 to respond to NE signaling is well-established (Sood et al., 2006; Sood et al., 2010). For instance, invasion assays indicated that SKOV3 cells increased invasion in response to NE at least 2-fold over vehicle control, and blockage of NE signaling with propranolol decreased tumor spread in vivo (Sood et al., 2006; Zink et al., 2018). From this data suggesting that SKOV3 cells are responsive to NE, we hypothesized that the ovary may also release NE in the presence of SKOV3 cells. Increased ovarian NE release when cultured with SKOV3 conditioned media would confirm the presence of a conserved secreted protein and provide an orthogonal cross-species cell model for validating the IMS analysis and prioritizing the bottom-up proteomics.

The IMS mass-guided assay was repeated with SKOV3 cells. When co-cultured with SKOV3 cells, ovarian NE release was detected (Figure 2A). We subsequently collected SKOV3 conditioned media and fractionated it to isolate the 3–30 and 30–50 kDa protein fractions. Ovaries incubated with the SKOV3 conditioned media protein fractions indicated that a factor secreted from the human ovarian cancer cell line in the 3–50 kDa range was also capable of inducing ovarian NE release (Figure 2B). These data suggest that both a murine fallopian tube cell line and a human ovarian cancer cell line produce a secreted factor that causes the ovary to release NE.

Since our IMS assay validated that a human ovarian cancer cell line also secreted a protein that results in the production and ionization of ovarian NE in the 30–50 kDa fraction, the SKOV3 protein fractions were prepared for bottom-up proteomics analysis as described above. While the evaluation and identification of proteins in the MOE PTEN^shRNA fractions were assessed for relative quantitation against peptide abundances in the MOE SCR^shRNA collection, in the case of SKOV3, there is no equivalent engineered cell line to represent non-cancerous cells. As such, the identification of all proteins in this fraction was determined and then compared against the MOE PTEN^shRNA proteins. SKOV3 samples were also analyzed to identify proteins in the 3–50 kDa fraction for consistency. 51 proteins were identified from all SKOV3 samples (N = 5, n = 3), 14 of which were also identified in MOE PTEN^shRNA, and their comparison to MOE PTEN^shRNA protein identifications are presented in Supplemental Table S2. Of the 14 proteins in common between SKOV3 and MOE PTEN^shRNA, only 5 fit the filter parameters, of which SPARC had the highest number of unique peptides and second highest sequence coverage (Table 2). When we compared all the proteomic data, only three proteins were shared between SKOV3 and MOE PTEN^shRNA that were not shared with MOE SCR^shRNA: galectin-1, hemopexin, and SPARC (Figure 3A). As SPARC was the only one of these 3 proteins that was within the filter parameters, SPARC was prioritized as the leading candidate protein of interest. We then confirmed with RNA-sequencing that MOE PTEN^shRNA cells had higher levels of SPARC expression than MOE SCR^shRNA (Figure 3B), and that both MOE PTEN^shRNA and SKOV3 conditioned media contained more SPARC than MOE SCR^shRNA via western blot (Figure 3C).

To determine if SPARC was involved in the induction of ovarian NE release, we sought to increase endogenous expression of SPARC in MOE WT cells whose conditioned media was not previously able to induce ovarian NE release. Therefore, we transfected a SPARC overexpression plasmid in MOE WT cells and confirmed expression via western blot (Figure 4A). The resulting IMS images demonstrated that increasing SPARC expression in MOE WT cell conditioned media enabled the induction of ovarian NE release (Figure 4B). We also sought to determine whether it was possible for MOE PTEN^shRNA to maintain induction of ovarian NE release with SPARC depleted. We transfected MOE PTEN^shRNA cells with SPARC siRNA or control siRNA and collected conditioned media from these cells. Western blotting confirmed the knockdown of the SPARC protein had occurred in the conditioned media (Figure 4C). Next, conditioned media from these transfected cells were co-cultured with murine ovaries, and IMS was performed on these co-cultures. Analysis demonstrated that the MOE PTEN^shRNA SPARC^siRNA conditioned media induced a less intense ovarian NE signal than the conditioned media from MOE PTEN^shRNA control^siRNA (Figure 4D). These data suggest that SPARC is necessary to induce ovarian NE release.

We next examined whether overexpression of SPARC in MOE PTEN^shRNA cells would stimulate more release of NE from the ovary than previously visualized. However, while transfecting MOE PTEN^shRNA with the same SPARC construct increased the amount of SPARC in the resultant conditioned media (Figure 5A), it was unable to significantly alter the NE release from an ovarian explant in co-culture compared with control transfected MOE PTEN^shRNA conditioned media (Figure 5B). To see if SPARC alone could induce the release of ovarian NE, we treated a murine ovary with recombinant SPARC. The addition of recombinant SPARC (10 μ g/150 μ L) did not result in the release of ovarian NE as seen with IMS (Figure 5C). Additionally, when MOE PTEN^shRNA conditioned media is combined with recombinant SPARC (50 ng/150 μ L), the additional SPARC is not seen to increase ovarian NE release compared to MOE PTEN^shRNA conditioned media (Figure 5D). Hence, we posit that the SPARC secreted into conditioned media of MOE PTEN^shRNA cells expressed a key post-translational modification that was missing on the recombinant protein or that SPARC only elicits NE induction when combined with another factor missing in the recombinant only condition.