Identification of Tumor Antigens in Ovarian Cancers Using Local and Circulating Tumor-Specific Antibodies

Ovarian cancers include several disease subtypes and patients often present with advanced metastatic disease and a poor prognosis. New biomarkers for early diagnosis and targeted therapy are, therefore, urgently required. This study uses antibodies produced locally in tumor-draining lymph nodes (ASC probes) of individual ovarian cancer patients to screen two separate protein microarray platforms and identify cognate tumor antigens. The resulting antigen profiles were unique for each individual cancer patient and were used to generate a 50-antigen custom microarray. Serum from a separate cohort of ovarian cancer patients encompassing four disease subtypes was screened on the custom array and we identified 28.8% of all ovarian cancers, with a higher sensitivity for mucinous (50.0%) and serous (40.0%) subtypes. Combining local and circulating antibodies with high-density protein microarrays can identify novel, patient-specific tumor-associated antigens that may have diagnostic, prognostic or therapeutic uses in ovarian cancer.


Introduction
Ovarian tumors are the most lethal of all female reproductive cancers, with over 70% mortality within five years of diagnosis [1]. Tumor progression is characteristically accompanied by few or ambiguous symptoms, resulting in patients often presenting with advanced metastatic disease and a poor prognosis. By contrast, patients diagnosed with early disease exhibit survival rates above 90%, with disease often cured by surgery alone [2]. For this reason, early detection is widely believed to be the best strategy to improve patient outcomes, but no screening tests are currently available [2,3].
Ovarian cancer is now recognized to be a collective term for several disease subtypes (including serous, mucinous, clear cell and endometrioid carcinomas) with different tissue origins, hereditary susceptibilities and gene expression profiles [4][5][6]. New biomarkers and more targeted treatment options that take into account the high degree of heterogeneity associated with ovarian cancer are therefore required [7,8].
Tumor antigens are expressed by malignant cells and can arise through several mechanisms, including mutations, post-translational modifications and aberrant tissue expression [9,10]. Irrespective of their initiation, tumor antigens can induce highly specific antibodies that, in some cases, can be detected several years prior to the manifestation of clinical symptoms [11][12][13], making them attractive targets for early diagnosis. Antibodies enable the identification of patient-specific tumor antigens [9,10], which are targets for immunotherapies such as CAR-T cell therapy and vaccines [14]. Antibody profiles have also been shown to predict immune-related adverse events during treatment with immune checkpoint inhibitors in melanoma [15,16].
Antibody generation requires the induction of antibody-secreting B cells (ASCs) in tissue-draining lymph nodes, with the subsequent release of antibodies into the blood. Previous studies have shown that ASCs recovered from tissue-draining lymph nodes are highly enriched for disease-related antigens compared to those in the blood [17,18] and can be used to identify novel tumor antigens [19,20]. To identify the cognate antigens recognized by tumor antigen-induced antibodies, high-density protein microarrays have emerged as useful tools, although challenges with cross-reactivity and high background are common [21].
In this study, we hypothesized that tumor antigen-enriched antibodies harvested from ASCs present in the lymph nodes (ASC probes) of ovarian cancer patients can be used to identify patient-specific tumor antigens. We screened these ASC probes using two commercial high-density protein microarrays and identified 50 candidate antigens. A custom 50-antigen array was constructed and used to screen sera from a separate patient cohort to investigate the diagnostic potential.

Antibody Reactivity Profiles of ASC probes against Ovarian Cancer Cell Line
ASC probes were generated from one to three separate pelvic lymph nodes (LN1, LN2, LN3) of 11 ovarian cancer patients (A-K). The IgG concentrations of ASC probes ranged from 0.04 µg/mL to 1.05 µg/mL, excluding an outlier (C-LN1, 15.00 µg/mL) ( Table 1). Western blots of ASC probes with the highest IgG content were performed against soluble (1) or insoluble (2) extracts of an ovarian cancer cell line (OVCAR3), and showed unique and shared patterns of band reactivity with higher reactivity seen in soluble fractions (Figure 1), as well as instances of no reactivity (A-LN2, H-LN1).

Candidate Antigen Selection Using High-Density Protein Microarrays
This study utilized two commercial high-density protein microarrays-the CDI HuProt TM array (16,086 human proteins and protein isoforms) and the Sengenics CTA protein array (257 cancer-testis and -associated antigens)-to screen ASC probes and matched sera from cohort 1 (n = 11) to identify candidate antigens.
Using the CDI HuProt TM array, pooled healthy (n = 2) and benign (n = 2, adenomas) serum controls showed detectable autoantibodies (Z-score > 3) against 122 and 177 antigens (133 unique to benign disease), respectively, which were used to eliminate non-specific hits and determine cancer-specificity of candidate antigens detected by ASC probes. A total of 45 cancer-specific antigens showed high ASC probe reactivity (RFU > 6000) and no reaction in healthy and benign sera ( Table 2). Overlapping cancer-specific antigen specificities were seen for a subset of antigens in all cases with matched ASC probes and sera. Table 2. Antigen reactivities identified with ovarian cancer cohort 1 using CDI HuProt TM arrays. Ovarian cancer ASC probes (n = 11, A-K) and selected matched sera (n = 5) were screened using CDI HuProt TM arrays.

Shared ASC Probe and Serum Cancer-Specific Hits
Using the Sengenics CTA protein array, pooled benign controls (n = 2, adenomas) showed detectable autoantibodies (Z-score > 3) against five antigens (AURKA, CTAG1A, CYP450 3A4, MAGEA4 and RAF1). A total of 25 cancer-specific antigens showed high ASC-probe reactivity (Z-score > 3) and no reaction in benign serum (Table 3). Overlapping cancer-specific antigen specificities were seen for a subset of antigens in two out of five cases with matched ASC probes and sera. Table 3. Antigen reactivities identified with ovarian cancer cohort 1 using Sengenics CTA protein arrays. Ovarian cancer ASC probes (n = 11, A-K) and selected matched sera (n = 5) were screened using Sengenics CTA protein arrays.

Diagnostic Screen Using a Custom Protein Microarray
Upon candidate antigen identification using the above described commercially available high-density arrays, a 50-antigen custom protein array was generated via CDI (Table S1). This custom array consisted of 41 cancer-specific antigens selected from the highest-reacting antigens for each ASC probe and nine benign disease-specific antigens (AURKA, CROCCP2, CTAG1A, MAGEA4, MAGEA9, PALM2, PMEPA1, SMYD5_frag and USP14) selected from the highest-reacting antigens in benign serum (and no reaction in healthy serum) on the high-density protein microarrays. This custom protein array was used to screen serum from cohort 2 (n = 80) to determine its diagnostic potential.

Discussion
Identifying novel diagnostic biomarkers or therapeutic targets in ovarian cancer, particularly those that prevail over the heterogeneity seen across disease subtypes, has proven challenging to date and presents an area of unmet clinical need. Antibodies against tumor antigens have been suggested as promising biomarkers for ovarian cancer [22]. In this study, we aimed to use ASC probes (antibodies harvested from ASCs recovered from tissue-draining lymph nodes) as a source of antibodies enriched against tumor antigens. Here, we showed that ASC probes can be screened on high-density protein microarrays to identify candidate cognate tumor antigens, yielding a proportion of overlapping antigen specificities when compared to matched serum. Moreover, this enabled the identification of novel tumor-associated antigens that may be used to detect ovarian cancers or targeted therapeutically.
In this study, most ASC probes isolated from a small cohort of ovarian cancer patients (n = 9/11) reacted to extracts of a high-grade serous ovarian cancer cell line (OVCAR3) with varying intensity. While antigen reactivity was lower than previously detected in breast cancer, both shared and unique patterns were also seen [19]. Interestingly, higher antigen reactivities were seen in soluble cell fractions containing cytoplasmic antigens, when compared to insoluble fractions mostly including membranous antigens. This is in agreement with previous studies, and has been associated with immunosuppression [23,24]. Given that the utility and origin of ovarian cancer cell lines have previously been questioned [25], reactivity against cell line extracts from additional ovarian cancer subtypes was not investigated in this study.
Screening the ASC probes on two high-density protein microarrays, the CDI HuProt TM array and the Sengenics CTA protein array, led to the identification of 41 candidate cancerspecific antigens. Whilst shared antigen reactivities were seen across patients, each patient exhibited a unique antibody profile, as was previously observed in breast cancer [19]. Matched sera for a subset of patients revealed instances of shared antigen reactivities in all patients between the ASC probe and serum pairs, as per previous studies [19,26]. The CDI HuProt TM array showed broader antigen reactivity in serum when compared to matched ASC probes, most likely due to the more targeted, temporal lymph node response in the latter. A subset of antigens was present on both protein microarrays, and detectable antibodies against one antigen (TP53) were simultaneously seen in ASC probes from patient J (J-LN1).
A custom array was constructed using the above identified 41 cancer-specific and 9 benign candidate antigens and used to screen a separate ovarian cancer cohort (n = 80), along with age and gender-matched healthy (n = 35) and benign (n = 12) controls. This diagnostic screen identified 28.8% of patients across disease subtypes and led to the identification of 11 tumor-associated antigens (CTAG2, MAGEB4, SHARPIN, MAGEA12, MAGEA4, MRFAP1L1, PNMA2, AURKA, SERPINB1, XAGE3 and BAGE4) with detectable antibodies exclusive to cancer patients, including several cancer-testis antigens. These data are comparable to previous antibody profiling studies in serous ovarian cancers, which reported sensitivities ranging from of 23 to 45% using multi-antigen signatures [27,28]. As a means of validating our findings, corresponding gene expression was investigated in a separate cohort (n = 307) accessible via TCGA, confirming expression at the transcript level ranging from 15.6% to 100% of serous ovarian cancer samples. While antibody profiling of these candidate antigens was not performed in other cancers, it is unlikely that these antigens are ovarian cancer-specific given the abundant expression reported across cancer tissues (TCGA).
Cancer-testis antigen expression has previously been reported in ovarian cancer, with specific associations to tumor progression, prognosis and therapeutic resistance, thereby supporting their potential uses as novel diagnostic biomarkers or immunotherapeutic targets [29]. Previous studies that investigated MAGEA4 in ovarian cancer reported tissue expression in 57% of serous carcinomas [30] and 47% of epithelial ovarian cancer patients [31], serum levels in 22% of primary ovarian cancer patients [32] and autologous antibodies in 9% of epithelial ovarian cancer patients [31]. Of note, all studies reported correlation between MAGEA4 and poor patient survival. Similarly, CTAG2 (alias LAGE-1, sharing 94% homology with NY-ESO-1 [32]) was expressed in 21% of epithelial ovarian cancer tissues, 30% of which also had detectable autologous antibodies [33].
Combining the use of ASC probes with sensitive, high-density protein microarrays is an attractive method to identify novel, patient-specific tumor-associated antigens that may have diagnostic, prognostic or therapeutic uses in ovarian cancer. Clinical evaluation is warranted to investigate the further applicability of these findings.

Patient Cohorts
Specimens from two separate cohorts of ovarian cancer patients were used in this study. Cohort 1 (n = 11) included multiple surgically resected pelvic lymph nodes and matched serum samples, while cohort 2 (n = 80) included serum samples from clear cell, endometrioid, mucinous and serous carcinoma subtypes (Table 4). In addition, serum from healthy females (n = 35) and those with benign gynecological conditions (n = 12) was also obtained ( Table 5). This study was conducted with ethical approval from the La Trobe University Human Ethics Committee (HEC17106) and the Research Ethics Committee of Tokai University Hospital (IRB registration number: 09R-082, 15R-253), including the use of clinical information and patient samples.

Generation of ASC Probes and IgG Quantification
Single cell suspensions were prepared from separate lymph node tissues, cultured for five days, and the supernatants collected (ASC probes) and used as previously described [18,19]. The total IgG content of ASC probes was measured using a human IgG quantification kit (FastElisa; R&D Biotech, Besancon, France).

Western Blotting
Western Blotting of ASC probes against soluble and insoluble extracts of a high-grade serous ovarian cancer cell line (OVCAR3) was performed as previously described [19].

Antibody Profiling Using Protein Microarrays
ASC probes and matched sera from ovarian cancer cohort 1 (n = 11), as well as serum from healthy (n = 2) and benign (n = 2) controls, were screened using the CDI Human Proteome (HuProt TM ) array (CDI Laboratories Inc., Mayaguez, PR, USA) and the Sengenics Cancer-Testis Antigens (CTA) protein array (Sengenics Corporation, Singapore), as per the manufacturers' instructions. Both arrays were used to ensure the maximum coverage of the human proteome and tumor antigens relevant to cancer. The CDI HuProt TM array included 16,064 human proteins and protein isoforms and was used to screen ASC probes at a 1:2 dilution and serum at a 1:500 dilution. The Sengenics CTA protein array included 256 cancer-testis and associated antigens and was used to screen ASC probes at a 1:2 dilution and serum at a 1:800 dilution. While some protein content was shared between the HuProt TM and CTA arrays (189 antigens), the protein cloning and expression methods used by each array platform differed (N-terminal GST-His-tagged proteins produced using the pEGH-A expression vector in a Saccharomyces cerevisiae strain versus BCCP-tagged fusion proteins produced in SF9 insect cells and biotinylated in vivo, respectively) [21]. These data were used to identify candidate antigens relevant to ovarian cancer, and a custom protein array including 50 candidate antigens (Table S1) was generated via CDI (CDI Laboratories Inc.). Serum samples from ovarian cancer cohort 2 (n = 80), as well as all healthy (n = 35) and benign (n = 12) controls, were screened at a 1:500 dilution using the CDI custom protein array, as per the manufacturer's instructions (Figure 4). . Study design diagram. Ovarian cancer cohort 1 and selected healthy and benign controls were screened using the CDI HuProt TM array and the Sengenics CTA protein array, which led to the identification of 50 candidate antigens. A custom array was developed and used to screen ovarian cancer cohort 2, as well as healthy and benign controls.

Gene Expression Profiling Using the Cancer Genome Atlas (TCGA)
Gene expression profiling using the identified tumor-associated antigens was investigated using publicly available ovarian cancer datasets accessible via TCGA research network (http://cancergenome.nih.gov/ (accessed on 3 August 2021)) and analyzed using the cBioPortal [34,35]. The TCGA dataset used was that of ovarian serous cystadenocarcinoma (TCGA, Firehose legacy), consisting of 606 samples (594 patients). Absolute mRNA transcript values were used to determine the levels of expression across TCGA patient samples, with any transcript level above zero considered "positive" for expression. The mRNA expression levels of the respective genes are shown as Z-scores (RNA Seq V2 RSEM) and represented in a log2 scale.  Data Availability Statement: Publicly available datasets were analyzed in this study. These data can be found here: http://www.cbioportal.org/study/summary?id=ov_tcga (accessed on 3 August 2021). The remaining data presented in this study are available within the article or on request.