All patient samples were obtained deidentified, according to approved institutional IRB protocol, and after patient consent. Sample processing occurred within 2-3 hours from collection time. Fresh tumor tissue was washed with DPBS (Corning), cut into 1 to 2 mm³ pieces and put into a T75 flask (Corning) with 15 ml complete RPMI-1640 (RPMI supplemented with 10% fetal bovine serum (FBS, Gibco), 10,000 U/L penicillin (Sigma), 10,000 U/L streptomycin (Sigma), and 2 mmol/L L-glutamine (Gibco). Cells were maintained at 37°C, 5% CO₂, and fresh culture medium was added every 3-5 days. Cells were trypsinized at confluency, using TrypLE™ Express (Gibco), and the tissue organoids and 1/5 of the cells were replated. For B lymphoblastoid cell lines, after about six weeks, the lymphocyte-like cells were collected by pipetting and moving to a new flask and cultured with complete RPMI. For human ovarian cancer cell lines, attached stromal cells were removed by short period trypsinization (2-3 mins) while the epithelial tumor cells were still attached to the flask, followed by trypsinization for another 5 to 10 minutes to recover the tumor cells, 1/5 of which were replated. Cells were propagated for 15 to 20 passages to form cell lines.

Ascites fluid was centrifuged at 400 g, 10 minutes. Pelleted cells were placed in a T75 flask in 20 ml complete RPMI-1640 and cultured long term.

Peripheral blood mononuclear cells (PBMCs) from female healthy donors were purchased from STEMCELL technologies and cultured in the same conditions.

All EOC cell lines and the human EBV-related Burkitt lymphoma cell line Raji were obtained from ATCC, except for UP04, UP13 and UP14 cell lines, which were derived in house from primary tissue. All cancer cells were cultured in complete RPMI-1640.

The B cells derived from ovarian cancer (BOC cell lines) were transferred to serum-free AIM-V medium (Gibco) for a minimum of 5 days (and up to three weeks) prior to antibody purification, so only the BOC- secreted immunoglobulins would be pulled down by the protein A/G beads (34). Cell free culture supernatants were co-incubated with protein A/G Plus agarose (Pierce) or PureProteome™ Kappa Ig Binder Magnetic Beads (Millipore) at room temperature (RT) for 60 minutes. After three washes in PBS, antibodies were eluted with 0.2 M glycine-HCl, pH2.2 followed by neutralizing 1M Tris-HCl solution, pH8.9. A fraction of isolated IgG (5μg) was loaded onto 4-20% SDS-PAGE to check for purity. Purified antibodies were conjugated with PE (Ab36, Ab73) or PerCP-Cy5.5 (Ab89) using PE/R-Phycoerythrin or PerCP-Cy5.5 Conjugation Kit - Lightning-Link^® (Abcam). Isotypes of purified antibodies were determined using Pro-Detect Rapid Antibody Isotyping Assay Kit-Human (ThermoScientific) according to the provided protocol. Serum IgG fraction was purified using caprylic acid (35) and sodium sulfate precipitation, as previously described (36).

Cells were stained with fluorescent labeled primary antibodies or isotype control in 1% bovine serum albumin (BSA, Fisher Bioreagents) in DPBS, on ice, for 30-40 min, and the data were acquired with LSR II and analyzed with FACSDiva or FlowJo (all from BD). The following anti-human antibodies were used: FITC-CD19 1:5, FITC-CD20 1:5, APC-CD27 1:5, BV421-anti-human IgD 1:100, PE-Cy7-CD28 1:100, PE-anti-human IgG 1:5 (all from BD), PE-CD138 1:100 and PerCP-Cy5.5-CD45 1:100 (both from BioLegend).

Cells (1x10⁶) suspended in 100 μl DPBS were loaded onto cell chips (ZellSafe, Canopy Biosciences) and allowed to settle for 5-10 minutes at RT. The cell chips were washed with 1 ml DPBS, fixed with fixation buffer (Canopy Biosciences) for 45 minutes at 4°C, and washed with 1 ml DPBS followed by 1 ml ZELLKRAFTWERK storage buffer (Canopy Biosciences). Cells were stained with PE-anti-human Ig, kappa 1:100 BV421-anti-human Ig, lambda 1:100, PE-anti-human IgG1 1:100, BUV395-anti-human IgG3 1:100 (all from BD), PE-Ab36 1:100. All antibodies were titrated in 300 μl of ZELLKRAFTWERK storage buffer and incubated at RT for 5 minutes. After washing the cell chips with 5 ml DPBS two times at a 2-minute interval, images were acquired with ZellScanner ONE or CellScape™ and analyzed with ZKWapp DataWizard (Canopy Biosciences). Tissue chips were mounted with 4 μm sections from formalin fixed and paraffin embedded (FFPE) tissue blocks following the protocol provided by Canopy Biosciences and stained with AF488-anti-CA125 1:100 (R&D systems), PE-anti-Vimentin 1:500 (BD), PerCP-Anti-CD3 1:100, PerCP-anti-EpCAM 1:100 (Canopy Biosciences Spatial Immune Profiling Kit), PE-Ab36 1:100, PE-Ab73 1:100 and PerCP-Cy5.5-Ab89 1:300 (labeled in house with PE, or PerCP-Cy5.5 according to manufacturer’s instructions and as described above), at room temperature, for one hour.

Sectioning of formalin-fixed, paraffin embedded tumor blocks into four-micron sections, and hematoxylin/eosin (H&E) staining were performed at the Magee-Womens Research Institute Pathology Core. Histopathology examination was performed by a trained pathologist (EE). Blank slides were used in house for IHC. Slides were left at 56°C overnight and rehydrated with xylene, 100% ethanol, 95% ethanol, 70% ethanol and water. Peroxidase blocking with 3% H₂O₂/methanol (Fisher Chemical) and antigen retrieval was performed by boiling the slides for 20 minutes in pH 9 Tris-EDTA buffer. The slides were blocked with 1% BSA in DPBS, 30 min at RT. The positive signal was detected using the ImmPACT DAB EqV Substrate Kit (Vector Laboratories, Burlingame, CA) and the slides were counterstained with hematoxylin. To ensure specificity of staining, control sections were stained with HRP-conjugated human IgG isotype control.

Single stain for PNAd (BioLegend, clone MECA-79, 1:50) was performed using Tris-EDTA (pH 9, 20 min). Biotin mouse anti-rat IgM Antibody (BioLegend, clone MRM-47, 1:100) was used as secondary antibody, followed by Pierce™ High Sensitivity Streptavidin-HRP (Thermo Scientific, 1:500). Agilent Liquid DAB+ Substrate Chromogen System was used as chromogen. For CD3/CD20 dual IHC, the protocol provided with Vector ImmPRESS Duet Double Staining Polymer Kit was used to carry out the staining. Mouse anti-human CD3 (Agilent, Clone F7.2.38, 1:100) and rabbit anti-human CD20 (Abcam, Clone EP459Y, 1:200) were used as the primary antibodies. Images were captured with Axiocam 105 color camera on ZEISS Light Microscope.

Cell pellets were washed with cold DPBS and lysed with RIPA (Pierce, Rockford, IL, USA) supplemented with Halt protease inhibitor cocktail, Halt phosphatase inhibitor cocktail and 0.05 M ethylenediaminetetraacetic acid (Thermo Scientific, Rockford, IL, USA). Protein lysates (10–20 μg) were loaded onto 4–20% Mini-Protean precast gel (Bio-Rad Laboratories) and run at 150V for 1h and transferred at 120 mA for 90 minutes to nitro-cellulose membranes (Bio-Rad Laboratories). Membranes were incubated in blocking buffer -2% BSA (Fisher Scientific), at RT for 1h and then blotted with primary antibodies in blocking buffer with 0.05% Tween-20 (Fisher Bioreagents), at 4°C overnight. Primary antibodies to the following antigens were used: EBV latent membrane protein 1 (Abcam, [CS 1-4], ab78113, 1:2000), β-actin (Sigma, A1978, 1:5000). CCDC155 (polyclonal antibody (Sigma-Aldrich, HPA019940, 200 mg/ml, 1:1000), human Ig kappa light chain (R&D Systems, Cat. # MAB10050-100, 1:2000), human Ig, lambda light chain (clone 1000040, R&D Systems™, 1:2000), mouse-anti-GST (Abcam, ab92, 3G10/1B3, 1:5000).

Goat-anti-mouse (170–6516) or goat-anti-rabbit (170-6515) IgG-HRP (both from Bio-Rad), or rabbit-anti-human IgG-HRP (Dako, 1.34 g/L) were diluted 1:6000 in blocking buffer. SuperSignal West Femto kit (Thermo Scientific) was used to develop the membranes. Images were taken with Chemidoc XRS darkroom system (Bio-Rad).

Primers (IDT) used for all PCR protocols are listed in Table 1 . Two EBV genes - LMP1 (37) and EBNA3C (38)- were detected using two sets of primers, respectively. Wild type LMP1 shows up as a band of 316 bp, while LMP1-Del30 has a band of 286 bp. Type I virus shows a EBV3C band of 153 bp and type II of 246 bp. PCR conditions: 95°C, 3 min; (95°C, 30s; 58°C, 30s; 72°C, 1 min) x 40; 72°C, 5 min; 12°C hold. 1.5% TAE gel, 120v, 40 min.

PCR conditions for CCDC155: ~90ng DNA templates; 94°C 3min; (94°C, 30s;58°C,30s;72°C, 1min) x40; 72°C, 5min; 12°C. 1.0% TAE gel, 120v, 40 min. Wild type gene appears as a single band. Upon knock-out, multiple bands or smears will appear due to the DNA repair attempts.

A total of six pair of primers ( Table 1 ) were used for antibody epitope mapping. BamHI and NotI restriction enzymes cut sites ( Table 1 , italic) were inserted into forward and reverse primers, respectively, for cloning purposes.

Templates for CCDC155 F1/R1, F2/R2 and F3/R3: 1ng pcDNA3.1-CCDC155 (GeneScript, OHu32184), 94°C 3min; (94°C, 30s;65°C,30s;72°C, 1min) x35; 72°C, 5min; 12°C. 1.0% TAE gel, 120v, 33 min. 10 μl loaded.

Templates for CCDC155 F4/R4 and F5/R5: 10ng pcDNA3.1-CCDC155 in 200μl total reaction volume. 94°C 3min; (94°C, 30s;67°C,30s;72°C, 1min) x 37; 72°C, 5min; 12°C. 1.5% TAE gel, 120v, 32min, 8 μl loaded.

Genomic DNA was extracted using Qiagen AllPrep DNA/RNA Micro Kit and sent to Adaptive Biotechnologies for BCR sequencing. Sequencing data were analyzed using immunoSEQ Analyzer 3.0.

Purified antibodies were shipped to CDI LABS, to be analyzed with the High-Spec^® cross-reactivity assay on the HuProt™ human proteome array, which contains >21,000 unique, full‐length, individually purified human proteins on a single microscope slide. According to the manufacturer’s description, the full‐length recombinant proteins are expressed in the yeast S. cerevisiae, purified, and printed on glass slides in duplicate, along with a set of control proteins (GST, BSA, histones, IgG, etc.). The HuProt™ expression library was created by inserting full-length human open reading frames (ORFs) into a yeast high-copy expression vector that produces GST-His6 fusion proteins when induced in yeast (39, 40). The company-provided protocol describes that the arrays were incubated with blocking buffer (1xTBS/0.1% Tween 20/5% BSA) at room temperature for 1 hr, with gentle shaking. Antibodies were diluted to 1 μg/mL in blocking buffer and probed on the arrays for 1 hour. After probing, the arrays were washed three times with TBST (1xTBS/0.1% Tween 20) for 10 min, and then exposed to Alexa647-anti-mouse IgG Fc secondary antibody (0.25 μg/mL) for 1 hour in a light-proof box with gentle shaking, followed by 3 washes with TBST for 10 min each and 3 rinses with ddH₂O. The arrays were then dried with an air duster and scanned using a GenePix 4000B scanner for data collection. Non-specific hits (due to binding to the secondary antibody) were eliminated from the analysis. CDI software was used to quantify binding to specific proteins on the array, based on Z Scores, as described in the statistical analysis section.

Six pair of primers were designed to amplify CCDC155 cDNA fragments (further detailed in the “PCR” section). The full CCDC155 protein contains 562 amino acids (aa). The first three pairs of primers were used to amplify fragments F1, F2, and F3, encoding aa 1 to 204, aa 182 to 383, and aa 358-562, respectively, with a 22 aa overlap between F1 and F2, 25 aa overlap between F2 and F3. The amplified fragments were digested with BamHI and Not I (Thermo Fisher), at RT for 10 minutes, and linked into BamHI and Not I digested pGEX-4T-1 (Cytiva, Product No. 28954549) vector, respectively, using the Rapid DNA Ligation Kit (Thermo Fisher). The constructed vectors (5 μl) were transformed into 50 μl BL21 (DE3) competent cells (Thermo Fisher) at 42°C for 30 seconds, followed by incubation on ice, for 2 minutes. After adding 250 μl of S.O.C. Medium (Thermo Fisher), the transformed E. coli BL21 (DE3) cells were placed on a shaker, at 200 rpm, 37°C for 1 hour. The E. coli cells from each transformation reaction were spread on separate LB plates containing 100 μg/ml ampicillin (Thermo Fisher) and incubated overnight at 37°C. Colony PCR was performed using the same primers and conditions for amplification of the corresponding fragments, as described in the PCR section. Three positive colonies from each transformation were picked, placed in 3 ml LB medium with 100 ug/ml ampicillin at 200 rpm, 37°C overnight and 1:100 diluted in 3 ml fresh LB medium with 100 ug/ml ampicillin and 0.5M IPTG (Thermo Fisher), 200 rpm, 37°C for 4 hours. Cell pellets from 1.5 ml were resuspended in 500 ul 1x SDS-PAGE loading buffer (Bio-Rad), ultrasonic treatment (Sonics GEX130 Ultrasonic Processor) for 30 seconds and heated at 95°C for 5 minutes. SDS-PAGE and Western blot were used to confirm the expression of the protein fragments.

Immulon 4 HBX ELISA plates (Thermo Fisher) were coated with one of the following proteins: CCDC155-GST, CPNE1-GST, PDP2-GST, NO66-GST, GMPS-GST, PTPN2-GST (all from CDI). Recombinant S. japonicum GST protein (Abcam) was used as control. All proteins were diluted in DPBS at 4 μg/ml, 100 μl/well in duplicates, incubated at 4°C overnight. Excess protein was washed four times with PBS-T (with 0.05% Tween-20) and blocking buffer (1% BSA in PBS) was added for one hour at RT. Purified antibody was added at 2 μg/ml, 100 μl/well. Human serum/plasma were 1:50 diluted with blocking buffer and added at 100 μl/well. Both were incubated for one hour at RT. HRP-rabbit-anti-human IgG, 1:5000 (Dako), HRP-mouse anti-human IgG3 (hinge) 1:1000, HRP-goat anti-human IgG1 (heavy chain) 1:1000, HRP- goat anti-human lambda light chain (1:2000), HRP- goat anti-human kappa light chain (1:2000), (all four from Invitrogen), or AP-goat-anti-human IgG (1:5000, Sigma) were added at 100 μl/well for one hour at RT. Reaction was developed using TMB (BD) at RT for 12 to 15 minutes or p-Nitrophenyl phosphate (Sigma) at RT for 20 minutes, and plates were read at 450 nm or 405 nm, respectively with Infinite 200 Pro ELISA plate reader (Tecan). The reported values represent optical density units at 450nm (OD450) and were calculated after subtraction of the anti-GST background, as follows: CCDC155 (OD450) = CCDC155-GST (OD450) - GST (OD450). The same formula was applied to samples read at 405nm wavelength.

Ovarian cancer TOV21G cells (4 x 10⁵ cells in 3 ml cRPMI/well) were seeded in 6-well tissue culture plate, without antibiotics. UltraCruz™ Transfection Reagent (Santa Cruz Biotechnology, sc-395739, 10 µl) was diluted in 140 µl Plasmid Transfection Medium (Santa Cruz Biotechnology, sc-108062). CCDC155 double nickase plasmids (2 µg) (Santa Cruz Biotechnology, sc-410329-NIC) or control plasmids (2 µg) were added into Plasmid Transfection Medium (Santa Cruz Biotechnology) to final volume of 150 µl, mixed with the transfection solution and added dropwise to the well containing TOV21G cells in 2ml fresh antibiotic-free growth medium. After culturing for 24 hours, cells were trypsinized with 1 ml/well Gibco™ TrypLE™ Express, washed and analyzed via flow cytometry to check the transfection efficiency. GFP sorting and puromycin selection were used sequentially to achieve high yield (~90%) of CCDC155 gene knockout. GFP positive cells were sorted, resuspended in 10 ml cRPMI without antibiotic and divided into four wells (3x10⁴/well/2.5ml) in a 12-well plate and cultured overnight, followed by adding puromycin to 400, 500, 600, 700 ng/ml cRPMI, respectively, for a 7-day selection. PCR and Western blot were used to confirm the loss of CCDC155.

OVCA432 cells were seeded in a well of a 6-well plate, 1x10⁶ cells/well in cRPMI and cultured overnight at 37°C, 5% CO₂. Cells were transfected with either CCDC155_OHu32184D_pcDNA3.1+/C-(k)-DYK (3 µg) or control pcDNA3.1+ (3 µg) (GenScript) using UltraCruz^® Transfection Reagent (Santa Cruz Biotechnology). The transfected cells were cultured for 48 hours and CCDC155 expression was confirmed by Western blot.

Z Score is the average Z Score of the duplicate spots of a given protein (each protein was printed in duplicate on the HuProt™ array). The Z Score of each spot on a given array is calculated with the formula Z= [F635 – F635(avg)]/F635(std).

F635(avg) and F635(std) are the average and standard deviation of the F635 values of all spots on the array, respectively. F635 is the average foreground signal intensity of 2 replicate spots of a given protein in the detection channel (IgG = 635 nm). The complete statistical methodology has been previously reported (41).

We used clinical specimens from four different EOC cases, with disease characteristics detailed in Table 2 . All patients had high grade serous EOC, FIGO stage IIIc (3 patients) or IIIb (one patient). Two of the samples were chemo-naïve and were obtained at the time of primary surgical debulking, while the other two were collected at the time of interval surgery, 27 and 16 days post completion of neoadjuvant chemotherapy, respectively, according to standard of care.

All samples (three fresh tumor tissue and one malignant ascites, Table 2 ) were handled as described in Materials and Methods, and as outlined in the study diagram ( Supplementary Figure 1 ). Following tissue processing, all cell populations exhibited the expected morphological heterogeneity, with cancer cells, fibroblasts and immune cells being visible in the cell culture flask ( Supplementary Figure 2A ). After several weeks in culture, outgrowth of a single cell phenotype was observed. The rapid expansion indicated that cells have entered the stage of exponential proliferation, and the growth pattern showing cell clumping and rosette morphology (42) was suggestive of a proliferating lymphocytic population ( Figure 1A ). Cells were continuously propagated in vitro for an average of 15 weeks, until stably maintained. Multi-parameter flow cytometry and chip cytometry data ( Figures 1B, C ) revealed that the established cells are homogeneously positive for the pan-leukocyte marker CD45, negative for the CD3 T cell marker, and positive for CD19 and CD20, indicating that all four lines are lymphoblastoid B cells ( Figure 1B and Supplementary Figure 2B ), exhibiting the expected size, morphology, and cell surface marker distribution ( Figure 1C ). Importantly, all cells are positive for the CD27 memory marker (43) and express high levels of plasma cell markers CD138 (44) and CD28 (45). Three of the four cell populations (BOC-36, BOC-73, and BOC-89) are IgG positive ( Figure 1B ). The fourth cell line (BOC-78) is IgM positive, with most IgM+ cells co-expressing IgD, as shown by flow ( Figure 1D ) and chip cytometry ( Figure 1E ). Altogether, these phenotypic markers point to an antigen-experienced, IgG or IgM memory plasma cell phenotype.

Histopathologic examination of the originating tumor tissue reveals an immune inflammatory tumor profile, with presence of tumor infiltrating, small mononuclear cells visible on hematoxylin-eosin (HE) staining ( Figure 2A ). The cells are positive for T and B cell markers on dual stain IHC or tissue chip cytometry ( Figure 2B ). The T and B cell infiltrates resemble lymphoid aggregates ( Figure 2B ), the hallmark of inflamed tumor phenotypes (46). The starting material for cell line BOC-78 was ovarian cancer ascites fluid (sample 22-78-Ascites, Table 2 ), consisting of mostly CD45+ inflammatory immune cells, including CD3+ T and CD19+ B lymphocytes, CD14+ monocyte/macrophages and CD56 NK cells ( Figure 2C ). Importantly, this patient’s corresponding tumor tissue (sample 22-78-Tumor) also shows presence of infiltrating CD20+ B cells, mostly as T-B cell conglomerates ( Figure 2B ). Of the four samples, staining for peripheral lymph node addressin (PNAd), important for lymphocyte homing, was positive in tumor samples 19-36 (the source of BOC-36, Table 2 ) and 21-73 (the source of BOC-73, Table 2 ), suggesting possible TLS formation in these two cases ( Figure 2D ). For the other two PNAd negative cases (22-89 Tumor and 22-78 Tumor), we classify the T and B cell infiltrates as lymphoid aggregates. Together, these results point to the presence of a B TIL-rich environment and presence of intratumor T/B lymphoid conglomerates (either as aggregates or TLS) (11, 47) in all four EOC cases.

To identify the potential cause for the observed spontaneous immortalization, we interrogated all four B lines for the presence of EBV, an oncogenic virus with tropism for and latency in memory B cells. PCR analyses revealed the presence of LMP1 and EBV3C viral genes in all four cell lines ( Figure 3A ). The LMP1 PCR product from samples BOC-78 and BOC-89 had a size consistent with the wild-type sequence (316 base pairs, bp), also found in the well characterized, EBV positive RAJI B lymphoma cell line (48), included as positive control. The smaller size PCR products detected in BOC-36 and BOC-73 suggest that these patients had been infected with the LMP1 30-bp deletion variant (49). The presence of EBNA3C 153 bp product points to type-A EBV in all cell lines ( Figure 3A ) (49). Western blot results revealed viral LMP1 protein expression in all four cell lines ( Figure 3B ), confirming viral activity. The observed size difference may be caused by genetic variations in LMP1, such as deletions and/or different numbers of tandem repeats, as previously described (50).

To check whether these spontaneously immortalized cell lines, which express CD138 plasma cell marker, are antibody-secreting, we tested the cell culture supernatant for the presence of human immunoglobulins. Two bands, corresponding to immunoglobulin heavy (IGH) and light (IGL) chains were observed in the supernatants of all samples, confirming that all four cell lines are actively secreting antibodies ( Figure 4A ). A single immunoglobulin light chain lambda (IGLλ) was detected in samples BOC-36 (Ab36) and BOC-89 (Ab89), whereas sample BOC-73 (Ab73) revealed two light chain bands. Western blot using IGL-specific antibodies confirmed that BOC-73 sample was positive for both IGL κ and λ ( Figure 4B ).

In line with the flow data in Figure 1D , the IGH in sample BOC-78 shows a single, larger band of approximately 75kDa, typical of reduced IgM (51) ( Figure 4A ). Further analyses using isotype detection strips, identified the antibody isotypes secreted by the four cell lines as being IgG1, λ (Ab36), mixture of IgG1,κ and IgG1,λ (Ab73), IgM, κ (Ab78), and a mixture of IgG1,κ and IgG3,κ (Ab89) ( Figure 4C ). The presence of both light chains κ and λ in BOC-73 cells was confirmed by flow cytometry ( Figure 4D ) and chip cytometry ( Figure 4E ), verifying that most cells are light chain κ positive and that κ and λ are separately expressed. FACS analysis of BOC-89 shows prevalence of IgG3, while also showing a dual expressing, IgG1+IgG3+ population. ( Figure 4F ). The presence of double positive cells was also captured via chip cytometry ( Figure 4G ).

BCR sequencing of the variable region (CDR3) of the heavy chain, an approach typically used to determine clonality (52), revealed that all four B lymphoblastoid lines are highly clonal, with one dominant sequence showing productive frequencies of 93.6%, 93.2%, 99.9% and 98.6% in BOC-36, BOC-73, BOC-78, and BOC-89, respectively ( Figure 4H ). The top amino-acid sequence of each BCR is listed in Figure 4I , while remainder sequences and their frequencies are listed in Supplementary Table 1 . In samples BOC-36, BOC-78, and BOC-89, except for the dominant (top listed) sequence, all others occurred with small frequencies (0.5% or less Supplementary Table 1 ), consistent with sequencing artifacts. To further address the clonality of sample BOC-73, which shows mutual expression of immunoglobulin light chain kappa (IGLκ) and lambda (IGLλ), we performed BCR sequencing on the original (mixed, unsorted) population as well as on cells sorted using antibodies specific to either IGLκ or IGLλ. The top sequence (CARGKGQSLGQTSGYKRPFDVW) was enriched from 93.2% in unsorted cells to 99.4% in cells sorted on kappa ( Supplementary Table 1 ), revealing that this is the IGH sequence mostly likely pairing with IGLκ. In cells sorted on IGLλ, two other sequences (CARGGSYHSGTGGYYKPPTFDYW and CAREFGTVAWFDPW) showed significant (100 fold) increases in productive frequency, compared to unsorted cells, suggesting that these may represent the two other clonal candidates for lambda ( Figure 4J and Supplementary Table 1 ). Interestingly, the top sequence in cells sorted on IGLκ (99.4% frequency) is also the top sequence in cells sorted on IGLλ, albeit at a lower (49.6%) productive frequency ( Figure 4J and Supplementary Table 1 ). This suggests a potential contamination during sorting on IGLλ, with more rapidly proliferating IGLκ positive cells. While co-expression of kappa and lambda light chain in the same cell has been reported before and remains a possibility ( Figure 4D , half of λ positive cells also co-express κ by flow cytometry), co-localization by fluorescent microscopy could not be observed in these cells ( Figure 4E , separate expression). Together, the results indicate that BOC-73 (κ) are monoclonal while BOC-73 (λ) may represent a bi-/oligo-clonal population.

For a detailed analysis of the target antigen, we focused on BOC-36 and the antibody released by this cell line, (Ab36 IgG1,λ Figures 4A, B ). BCR sequencing of this cell line shows a high frequency of the top sequence (94%, Figure 4H ), with the rest being distributed among “satellite”, likely artifactual clones, with a low production frequency (0.4% or less, Supplementary Table 1 ). Based on this, we believe BOC-36 cells are very likely monoclonal.

Asking if Ab36 can recognize antigens present in EOC tumor cells, we probed Ab36 purified from cell culture supernatant against cell lysates from 17 different human ovarian cancer cell lines. Western blot results show binding in 8 of the 17 human EOC lines, to a target slightly larger than 50kDa ( Figure 5A ). Chip cytometry imaging shows that Ab36 binds to an intracellular protein and confirms that cell lines TOV21G, UP14-TL and OVCA429 express target antigen with high, intermediate, and low expression levels, respectively, in line with Western blot data ( Figure 5B ). When tested against the patient’s own tumor, the staining pattern also showed that Ab36 binds to a target with perinuclear/cytoplasmic distribution. The staining was confined to larger, EpCAM positive, epithelial tumor cells, and was negative in non-tumor cells (such as CD3 T cells, for example) ( Figure 5B ).

For target identification, we queried the purified Ab36 on the HuProt™ human proteome arrays (CDI Labs), comprising more than 21,000 proteins. Intensity of antigen-specific binding was reported using Z score histograms and protein candidates with highest scores were further interrogated. Of the top four proteins ( Figure 5C ), we chose to further test the coiled-coil domain containing 155 protein (CCDC155, also known as KASH5 or Nesprin-5), which has been previously reported to be a high grade serous EOC antigen (53) and which, unlike the other three candidates, has a molecular weight (of 62kDa) that fits the western blot predicted size ( Figure 5A ). Using full length CCDC155 recombinant protein as target (and 6 additional proteins, as controls) we developed an ELISA protocol that confirmed that mAb36 specifically binds to CCDC155 ( Figure 5D ). Importantly, the patient’s serum also showed CCDC155 reactivity, suggesting a systemic humoral response in this patient ( Figure 5E ). In validation of these results, we further demonstrated that binding of Ab36 is lost in TOV21G cells in which we deleted CCDC155 ( Figures 5F, G ). Additionally, western blot using cell lysate from OVC432 cells (which naturally express undetectable levels) transfected with a plasmid encoding the full length CCDC155 shows increased Ab36 binding, compared to control transfected cells ( Figure 5H ).

To map the binding epitope more exactly, we amplified three DNA fragments, each encoding for approximately one third of the full-length CCDC155 protein. Fragment F1 encodes for aa 1 to 204, F2 for aa 182 to 383, and F3 for aa 358 to 562, with a 22-aa overlap between F1 and F2, and a 25-aa overlap between F2 and F3 ( Figure 6A , step 1). The three PCR fragments were inserted into pGEX4T-1 expression vectors ( Supplementary Figure 3 ) and E coli BL21(DE3) were transformed with vectors containing F1, F2, F3 fragments, respectively. Bacterial cell lysates show Ab36 binds exclusively to F2 while the commercially available antibody (rabbit polyclonal anti-human CCDC155), included as control, binds to F1 ( Figure 6B ). Fragment 2 was further divided into three smaller domains (F4, F5 and F6, with small overlaps ( Figure 6A step 2) and Ab36 was found to bind to F4 ( Figure 6C ). This stepwise approach allowed us to narrow down the binding domain to a stretch of 32 amino-acids ( Figure 6A , step 3- underlined sequence).

To verify the extent to which anti-CCDC155 circulating IgG antibodies are detectable in EOC patient sera, we used a cohort of 20 advanced stage EOC patients, of which 18 had recurrent disease. ELISA results show a detectable response in 9 (45%) of patients ( Figure 6D ). A pairwise comparison in an additional cohort of seven patients with advanced stage, high grade serous EOC shows similar antibody levels in primary and recurrent disease setting ( Figure 6E ), suggesting that patients are largely maintaining the upfront level of antibodies to this antigen. Together, these results further demonstrate that CCDC155 is immunogenic and antigenic in late stage, primary and recurrent EOC, and that the antigen-specific monoclonal antibody Ab36 IgG1, λ, secreted by the spontaneously immortalized BOC-36 cells, captures this phenomenon at a single clone level.

To interrogate the antigenic specificity of the other three clones, we applied the same approach, probing the purified secreted antibody fraction from each cell line, on the HuProt™ human proteome arrays. The Z score distribution of Ab73, purified from BOC-73 unsorted cells, shows three potential targets with high scores: copine 1-CPNE1, pyruvate dehydrogenase phosphatase 2- PDP2, and nucleolar protein NO66 ( Figure 7A ). To further identify which of these candidates is the specific antigen, we developed an ELISA protocol using the three recombinant proteins (and four additional proteins as controls). The results showed that Ab73 (secreted by unsorted cells) specifically binds to PDP2 ( Figure 7B ). Given that the immunoglobulin fraction secreted by unsorted cells comprises two different light chains (IGLκ and IGLλ) ( Figure 4 ), we asked which light chain is responsible for binding to PDP2, using IGLκ− and IGLλ - specific secondary antibodies. Interestingly, both IgG1,κ and IgG1,λ bind to PDP2 and the amplitude of the response was equally high, suggesting that the less abundant IgG1,λ may have higher avidity for PDP2 than the prevalent IgG,κ ( Figure 7C ).

Protein array analysis of Ab89 revealed a single candidate -growth factor receptor-bound protein 2, GRB2- as the potential antigenic target ( Figure 7D ). GRB2 is a signaling adapter protein involved in cell signaling pathways, linking receptor tyrosine kinases to MAPK pathways (54). ELISA testing confirmed that Ab89 specifically binds to GRB2 and not to any of the additional 7 proteins, included as non-specific controls ( Figure 7E ). As expected, only Ab89 but not Ab36 or Ab73 (used as negative controls) specifically binds to GRB2 ( Figure 7F ). The serum from the same patient (Serum 89) also shows a positive signal, suggesting a systemic response ( Figure 7F ).

BOC-89 is a highly clonal population, with the top BCR sequence showing a productive frequency of almost 99% ( Figure 4H ). However, two IgG subclasses were detected in BOC-89 supernatant (IgG1 and IgG3) ( Figure 4A ), and some cells seemed to co-express them, with IgG3 being the more abundant fraction ( Figures 4F, G ). To verify which of the two handles binding to GRB2, we used sub-class specific secondary antibodies. The ELISA results revealed that IgG1, the less abundant secreted immunoglobulin, is the one responsible for binding to GRB2, with a much weaker signal coming from IgG3 ( Figure 7G ).

Using a healthy donor PBMC fraction, we similarly generated a fifth B cell line (BHD-1, Supplementary Figures 4A, B ), which is type I EBV positive ( Supplementary Figures 4C, D ) and is antibody secreting (AbHD1, Supplementary Figure 4E ). ELISA testing using the purified antibody fraction showed that AbHD1 has no significant binding to CCDC155, PDP2 or GRB2 protein ( Supplementary Figure 4E ), further demonstrating the specific binding of Ab36, Ab78 and Ab89 to their cognate antigens, respectively.

Compared to the other three antibodies described, the Z scores for Ab78 are much lower, and the score distribution shows no clear candidate ( Figure 7H ), raising the possibility that this antibody (of IgM isotype- Figure 4 ), might actually bind to non-peptide epitopes, such as glycans (55).

As with Ab36 ( Figure 5B ), Ab73 and Ab89 also recognize their respective tumor samples ( Figure 7I ). Importantly, their target antigens seem to be expressed by the epithelial tumor cells, highlighted by either EpCAM or CA125 epithelial cell markers, and not by the vimentin-expressing tumor stroma. Together, these results provide a platform for antigenic target identification using tumor-isolated B cells and protein arrays, and reveal CCDC155, PDP2 and GRB2 as the tumor-expressed target antigens recognized by Ab36, Ab73 and Ab89, respectively.