Anti-CA15.3 and Anti-CA125 Antibodies and Ovarian Cancer Risk: Results from the EPIC Cohort

In the 1980s, two heavily glycosylated proteins were discovered as potential markers for breast and epithelial ovarian cancer (EOC; 1, 2). mAbs were raised against selected epitopes and assays developed whose names became synonymous with the markers CA15.3 for the breast biomarker and CA125 for the EOC biomarker. Genes for these proteins were eventually cloned and proved to be members of the human mucin family (3, 4)—CA15.3 as human mucin 1 (MUC1) and CA125 as human mucin 16 (MUC16). Both proteins were found to be expressed on mucosal barriers lining the genital, digestive, and respiratory tracts and breast ducts and over expressed in many of the neoplasms originating from these tissues (5, 6). Specifically, for invasive EOC and depending upon the antibody used for histochemical analysis, a majority of all EOC subtypes express CA15.3 (7). Tissue expression of CA125 varies by subtype from 12% (mucinous) to 85% (serous) (8). Higher serum levels of CA15.3 in patients with breast and colorectal cancers were associated with poorer prognosis (9, 10) and thought to reflect an immunosuppressive effect on T-cell proliferation (11). Similarly, in EOC, high serum levels of CA125 predict advanced disease and poorer survival (12). Like CA15.3, this may have an immune basis since CA125 can bind with natural killer (NK) cells and blunt NK response to ovarian tumors cells (13).

Antibodies against CA15.3 and CA125 have also been described. Anti-CA15.3 antibodies were found in some patients with CA15.3-expressing cancers and associated with better prognosis (14). Both CA15.3 and anti-CA15.3 antibodies can also be found in healthy women, including those who are pregnant or breastfeeding (15). This led to the theory that anti-CA15.3 antibodies evoked by these events could explain why they protect against breast cancer (16). CA125 is elevated in pregnancy, but whether anti-CA125 antibodies also form during pregnancy has not been studied. However, events that reduce EOC risk like pelvic surgery, mumps, or puerperal mastitis, may lead to elevated (and presumably protective) levels of both anti-CA125 and anti-CA15.3 antibodies (17–20). Conversely, chronic events that increase EOC risk like repeated damage and repair to the ovary from “incessant ovulation” (ovulations not interrupted by pregnancy or oral contraceptive use) were associated with lower levels of anti-CA15.3 antibodies, suggesting immune tolerance that allows emergence of a CA15.3-expressing cancer (20, 21). These observations form the basis for a general paradigm that risk factors for mucin-expressing cancers may operate because of humoral (and, likely, cellular) immune reactions to mucin-expression invoked by the events themselves. Chronicity of the exposure may determine whether the event raises or lowers risk for the cancer.

In this study, we measured anti-CA15.3 and anti-CA125 antibodies and their corresponding antigens in sera from female participants in the European Prospective Investigation of Nutrition and Cancer (EPIC), including 806 who subsequently developed invasive EOC and 1,927 matched controls. We sought to identify factors associated with levels of anti-CA15.3 and anti-CA125 antibodies by menopausal status and determine whether these antibodies were associated with early detection (elevated levels within three years of diagnosis) or risk (elevated levels three or more years before diagnosis) of EOC overall and by histologic subtypes.

EPIC is an ongoing multicenter prospective cohort study designed to investigate relationships between diet and cancer (22). Briefly, 519,978 participants (366,521 women) were enrolled from 1992 to 2000 in 23 centers in 10 European countries: Denmark, France, Germany, Greece, Italy, Norway, the Netherlands, Spain, Sweden, and the United Kingdom. A total of 385,747 participants (226,673 women) provided a baseline blood sample, kept frozen in long-term storage at ≤−150°C, with the exception of Sweden, where samples are stored at −70°C.

Cancer occurrence was documented through record linkage with cancer registries or active follow-up. Self-report was verified by clinical record review. Vital status was determined by linkage with mortality registries. When this study was initiated, follow-up was complete through 2005 (France) to 2008 (Germany). Among women who had provided a baseline blood sample, a total of 806 incident cases of invasive EOC were identified including ovary (International Classification of Diseases for Oncology O–3code: C569), fallopian tube (C570) or peritoneal cancer (C480, C481, C482, C488). Low malignant potential (borderline) tumors were not included. For the 806 EOC cases, complete information on tumor grade was available for 471 patients (58%) and information on tumor stage was available for 709 patients (88%).

For each case, up to four controls were randomly selected from female cohort members who were alive and cancer-free at diagnosis of the index case using a sampling protocol described previously (23). Case and control participants were matched on study center, age at blood donation, time of the day of blood collection, fasting status, menopausal status, menstrual cycle phase for premenopausal women, current use of oral contraceptives (OC), or hormonal replacement therapy (HRT).

Baseline data were self-reported through questionnaires or interview except for body mass index (BMI) where height and weight were directly measured. Categorized, the variables included: age at blood draw (<41, 41–50, 51–60, 61–70, >70 years), BMI (<18.5, 18.5–24.99, 25–29.99, ≥30 kg/m²), smoking (never, former, current), pack years (≤11, 12–19, 20–31, >31), age at menarche (<12, 12, 13, 14, >14), OC use (never/ever), OC duration (≤2, >2–5, >5–10, >10 years), number (0, 1, 2, 3, 4+) and timing of pregnancies, age at first birth (<20, 20–24, 25–29, ≥30), age at last birth (<25, 25–29, 30–34, ≥35), years since last birth (<21, 21–27, 28–32, >32 years), hysterectomy, oophorectomy, family history of breast cancer, age at menopause (<46, 47–49, 50–51, 52–53, >53), and HRT use (postmenopausal never used HRT, ≤2.5 years, >2.5 years). The number of ovulatory cycles was estimated by calculating years between menarche and current age or age at menopause (if postmenopausal) and subtracting time using oral contraceptives, pregnancy, or breastfeeding. Those missing an exposure of interest were excluded from analyses for that exposure.

Assays were performed at the Brigham and Women's Hospital Laboratory of Genital Tract Biology using a multiplex platform (Meso Scale Discovery). Laboratory personnel were blind to case–control and quality-control (QC) sample status. Methods related to the measurement of CA125 and CA15.3 were previously described (23). Similar batching rules used for the mucin antigens were employed for the mucin antibodies; that is, keep matched sets together and as many from the same center as possible with at least 6 QC samples. Stock assays for Human CA125 (catalog number K151WC) and Human Prototype CA15.3 (catalog number N45ZA-1) were available from MSD in singleplex assays. To measure antibodies against CA15.3 and CA125, antigen-grade CA15.3 and CA125 purified from human breast and ovarian cancer cell lines (Meridian Life Sciences Inc.) were coated on multi-spot plates. The assays included blocking with a blocking buffer for 1 hour followed by wash; 2-hour incubation with samples at multiple dilutions followed by PBS/0.05% Tween-20 wash; detection of human IgG1 bound to the specific protein spots with MSD sulfo-Tag–labeled antibodies (1 μg/mL) for 2 hours; and washing and adding read buffer followed by detection of electro-chemiluminescence (ECL) using an MSD Imager 2400. Split aliquots of this pool were tested at the same dilutions as the test samples on each assay plate and served to assess inter-batch variation. Coefficients of variation were calculated using measurements from blinded aliquots. Average intra-plate CVs were 14% for anti-CA125, 11% for anti-CA15.3, and the inter-plate CVs were 35% and 31%, respectively.

To account for batch-to-batch variation, we recalibrated antibodies to have a comparable distribution with an average batch according to methods described by Rosner and colleagues (24). Other than batch number, no other variables were used in the recalibration models. Anti-CA15.3 and anti-CA125 antibody levels were log-transformed (as were CA15.3 and CA125). Correlations between antigens and antibodies were assessed using Spearman partial correlations adjusted for the study matching factors. We used generalized linear models to estimate mean antibody values separately in controls and cases overall and by menopausal status as risk factors, especially reproductive variables, may differ by menopausal status (25). Results were exponentiated to obtain geometric mean values in the original scale. Linear regression models were adjusted for the study matching factors, BMI, and smoking status. Indicators were used to account for missing data for covariates. Trend tests of continuous variables were based on the medians of each category. We included interaction terms in the linear regression models to test for heterogeneity in antibody levels by menopausal status. To examine the associations between antibodies and ovarian cancer, we classified antibodies into quartiles based on the control distribution and calculated ORs and 95% confidence intervals (CIs) using conditional logistic regression. In the conditional logistic regression analyses, we individually added each potential confounder to the model and examined the change in the OR estimate from the crude OR. The covariates examined were age at menarche, OC use, parity, hysterectomy, oophorectomy, duration of HRT use, BMI, and smoking status (using categories described above). The percent changes in the adjusted ORs from the crude ORs were minimal, ranging from 0% to 3%; therefore, we present crude conditional logistic regression results. We evaluated risk associations separately for cases diagnosed within three years (to reflect early detection) and greater than three years (to reflect risk); these lag-time cutoff points were based on prior studies in the EPIC cohort (23) and by others (20, 26). Finally, using median values in controls as cutoff points, we cross classified participants as having high or low antibodies and antigens and examined the association between these biomarker combinations and ovarian cancer. We used the contrast test method to test whether the association between increasing antibodies and risk, and between antigen–antibody combination and risk, varied by histologic type (27). Analyses were conducted using SAS version 9.4. Statistical tests were two-tailed and significant at P < 0.05.

All participants gave consent for future analyses of their blood samples in their written consent. The principles expressed in the Helsinki Declaration of 1996, conventions of the Council of Europe on Human Rights and Biomedicine, and the UNESCO Declaration on Human Genome have been respected. This study was approved by the ethics committees at the International Agency for Research on Cancer (IARC, Lyon, France) and the University of Heidelberg (Heidelberg, Germany). As the identity of subjects providing specimens was anonymous to Brigham and Women's Hospital (BWH) investigators, the research was declared exempt at BWH.

Table 1 compares the characteristics of EOC cases and controls in this nested case–control study. Cases did not differ from controls in age at enrollment or menopausal status, but were more likely to be heavier, nulliparous, and to have never used oral contraceptives (and used oral contraceptives for a shorter duration). Concentrations of anti-CA15.3 antibodies by epidemiologic variables in all, premenopausal, and postmenopausal controls are shown in Table 2. Among postmenopausal women, we observed higher anti-CA15.3 antibodies with older age at baseline (P_trend = 0.04). Overall, anti-CA15.3 antibodies were lower among women with higher BMI (P_trend = 0.03) and current smoking (P < 0.0001). Among current smokers, more pack-years were associated with lower anti-CA15.3 antibodies (P_trend = 0.002). Among women premenopausal at baseline, ever versus never use of OCs was associated with lower anti-CA15.3 antibodies (P = 0.05); and higher number of ovulatory cycles was associated with a trend for lower anti-CA15.3 antibodies (P_trend = 0.02). Among postmenopausal women, higher anti-CA15.3 antibodies were associated with later age at last birth (P_trend = 0.02), shorter interval since last birth (P_trend = 0.04), and longer duration of OC use (P_trend = 0.01). The difference in the trend in anti-CA15.3 antibody levels by duration and OC use between pre- and postmenopausal women was significant (P_{heterogeneity} = 0.03).

Similar to associations seen for anti-CA15.3 antibodies, current smoking was associated with lower anti-CA125 antibodies (P < 0.0001), and among smokers, anti-CA125 levels declined with increasing pack-years (P_trend = 0.004; Table 3). Among postmenopausal women, later age at last birth (P_trend = 0.002) and shorter interval since last birth (P_trend = 0.005) were associated with higher anti-CA125 antibodies. Among premenopausal women, duration of OC use was associated with a lower level of anti-CA125 antibodies when nonusers were included in the referent nonexposed category.

Among cases, we observed no associations between tumor characteristics and antibody levels (Supplementary Table S1). When we evaluated all cross-sectional associations adjusting for the corresponding antigen concentration, results were unchanged. Other variables that were examined and found not to affect anti-mucin antibodies in either cases or controls included: miscarriage, hysterectomy or unilateral oophorectomy, family history of breast cancer, IUD use, and duration of HRT use. Information on tubal sterilization was not collected at all EPIC sites and data was too limited to evaluate this exposure.

We then evaluated quartiles of anti-CA15.3 and anti-CA125 antibodies and EOC risk in all subjects and subjects stratified by menopausal status at blood collection, histologic type of EOC, and time between blood collection and diagnosis (<3 years and ≥3 years; Table 4). For mucinous EOC arising three or more years after baseline, trends for increased risk were found with both increasing anti-CA15.3 antibodies (P_trend = 0.02) and increasing anti-CA125 antibodies (P_trend = 0.05). Of borderline significance, there was a trend for higher levels of anti-CA15.3 antibodies to be associated with lower risk for serous EOC (P_trend = 0.06) that developed within 3 years of their blood draw. Levels of anti-CA15.3 in the fourth quartile were associated with an OR (and 95% CI) of 0.51 (0.27–0.96) compared with women with levels in the first quartile.

Next, we evaluated risk for EOC by anti-CA15.3 and anti-CA125 antibody levels jointly classified by the corresponding antigen level with both dichotomized as high or low based on median cutoff points in controls (Table 5). Although combining CA15.3 antigen levels with anti-CA15.3 antibody levels was not informative, combining CA125 levels with anti-CA125 antibody yielded several interesting findings. Relative to women with low concentrations of both markers, EOC risk overall was elevated for women with high CA125 levels regardless of anti-CA125 antibody levels, but only for disease arising within 3 years of blood draw. These effects were most apparent in postmenopausal women for risk of serous or endometrioid EOC. Conversely, a reduction in risk for all EOC, both in short and long-term risk, was associated with low CA125 antigen and high anti-CA125 antibody. For women who developed serous EOC, the combination of low CA125 antigen and high anti-CA125 antibody was associated with an OR (95% CI) of 0.60 (0.40–0.89) for disease that developed more than 3 years after blood draw.