Variables Affecting CA15.3 Tumor Antigen Expression and Antibodies against It in Female National Health and Nutritional Survey Participants

The probability plots for the distributions of (log transformed) levels for CA15.3 antigen and the two antibody measures--one adjusted for non-specific antibodies (the anti-CA15.3 antibody index) and one not adjusted (anti-CA15.3 RLU) are shown in supplemental Figure 1 . The distributions for both log transformed CA15.3 antigen and anti-CA15.3 antibody index were close to normal in appearance while the anti-CA15.3 RLU, even log transformed, was negatively skewed. The relationship between the two antibody measures is further examined in supplemental Table 1 . In all women, there was a good correlation between the index and RLU measures of anti-CA15.3 antibodies (r=0.62, p<0.0001). Correlations were also strong and significant within subgroups of women stratified by menopausal status, race, BMI, and smoking status (r ranging from 0.54–0.67, all p-values <0.0001). Based upon the correlation between the index and RLU measures of anti-CA15.3 antibodies and the more normal distribution for the index measure, we focus on the index measure for anti-CA15.3 antibodies in the remaining tables. Table 1 shows key demographic and lifestyle variables associated with CA15.3 antigen and anti-CA15.3 antibody levels. There were no clear associations with age or menopausal status. Non-Hispanic Black women had the highest level of CA15.3 antigen and non-Hispanic White women the highest antibody levels. No associations were seen with height; but, after excluding women who were currently pregnant, both increasing weight and BMI were associated with lower anti-CA15.3 antibody levels. Finally, current smokers had lower anti-CA15.3 antibody levels that further decreased with increasing pack-years. Table 2 examines the associations between reproductive events experienced at the time of blood collection and CA15.3 and anti-CA15.3 levels. Women who were using OCs at the time of sample collection had lower CA15.3 levels, while women who were either pregnant or breastfeeding had substantially higher CA15.3 levels. Women who were breastfeeding also had higher anti-CA15.3 antibody levels. Higher levels of serum CA15.3 were seen in women who reported being in the menstrual phase of their cycle. No significant effect was seen on CA15.3 antigen or antibody among post-menopausal women using hormonal replacement therapy (HRT). The associations of past reproductive events with CA15.3 antigen and anti-CA15.3 antibody levels are shown in Table 3 for all women and separately by menopausal status. Women who had a later age at menarche had higher CA15.3 antigen levels. The trend was significant for all women and similar (but not significant) for pre- and post-menopausal women. Although no effect was seen by ever-use of oral contraceptives (OCs), longer durations of OC use were associated with lower CA15.3 antigen levels in all and in postmenopausal women. All and premenopausal women who had ever given birth had higher CA15.3 antigen levels; but, interestingly, antigen levels in parous premenopausal women were inversely correlated with number of livebirths. No effect of age at first or last livebirth was seen, nor was there any effect of having ever breastfed or with the number of children breastfed. In all women, history of tubal ligation was associated with lower anti-CA15.3 antibodies—a finding of borderline significance. Women who reported a history of endometriosis had higher CA15.3 antigen levels and women who reported a later age at diagnosis of endometriosis had lower anti-CA15.3 antibody levels. No effect was seen by age at natural menopause nor were any effects seen with the history of hysterectomy (without bilateral oophorectomy) or bilateral oophorectomy (with or without hysterectomy). Similarly, age at hysterectomy or bilateral oophorectomy was not significantly associated with CA15.3 antigen or antibody levels, although a later age at hysterectomy was associated with lower antibody levels in postmenopausal women. There was a modest trend for women who had used menopausal hormonal replacement therapy (HRT) to generally have lower antigen levels. Finally, in premenopausal women, an increasing number of ovulatory years was associated with lower CA15.3 levels. In all and postmenopausal women, an increasing number of ovulatory cycles was associated with a decrease in anti-CA15.3 antibody levels.

The NHANES program, within the U.S. Centers for Disease Control and Prevention, began in 1999 with 2-year data collection cycles ( 17 ). In each cycle, adults and children are selected and enrolled using a complex, multistage probability sampling design to create a representative sample of the civilian, non-institutionalized US population. Participants are interviewed once to obtain health-related data via questionnaires and undergo physical exam and laboratory tests ( 18 ). To increase the precision of estimates for certain subpopulations, the NHANES design includes over-sampling of people aged 60 years and older, African Americans, and people of Mexican origin. In this study, we used public de-identified data collected from participants in the 2001 to 2002 cohort. This cohort was chosen as being remote enough to avoid ethical concerns that might require notification of participants about highly elevated tumor markers. In addition, the women’s questionnaire for this cohort asked about genital talc use—a topic of interest to the authors. From the original cohort (n=11,039), we excluded men (n=5,331) and women less than 20 years old (n=2,833). Among the remaining women, we obtained sera from 2,392 to measure biomarkers including CA15.3 (MUC1) and anti-CA15.3 (MUC1) antibodies. We excluded those with outlying CA15.3 and anti-CA15.3 values (n=82), 76 of whom had high CA15.3 antigen outliers and 6 low antibody outliers. There were no missing values. We also excluded women missing reproductive-health questionnaires (n=193) for a final sample of 2117 women. The study was approved by the scientific and ethics panels of NHANES. Because the research involved specimens from women anonymous to us, it was deemed exempt by the Brigham and Women’s Hospital Human Research Committee. Characteristics examined included demographic and lifestyle factors, including age at blood collection, race, marital status, height, weight, body mass index [(weight(kg)/height(m) 2] , and smoking status. Women who reported not having smoked more than 100 cigarettes were classified as never smokers, those reporting smoking at the time of the NHANES interview were classified as current smokers, and all others were classified as former smokers. Reproductive characteristics included: age at menarche; gravidity and parity; ages at first and last birth; breastfeeding; oral contraception (OC) or Depo-Povera use; history of endometriosis, hysterectomy, or oophorectomy, and menopausal hormone use. Women were categorized as premenopausal if they reported regular (or normally irregular) periods in the past year or were less than age 50 and reported periods had stopped because of a hysterectomy without bilateral salpingo-oophorectomy (BSO). Otherwise, they were categorized as postmenopausal. Age at natural menopause was defined as the age at last period for those who either never had a BSO or had a BSO after their periods stopped. Duration of OC use, but not Depo-Provera, was available. We estimated lifetime ovulatory years using the following equation: (age at last period) – (age at menarche) – (years of OC use) – (number of live births x 0.92) – (number of pregnancies resulting in non-live births x 0.23), where livebirths were assigned 11 months (about 0.92 years) of anovulation and non-live births (miscarriages or abortions) were assigned 12 weeks (about 0.23 years). Although breastfeeding delays return to ovulation, it was not included in the model because the total duration of breastfeeding was not available in NHANES. Serum CA15.3 and anti-CA15.3 levels were measured at the Genital Tract Biology Laboratory (Brigham and Women’s Hospital, Boston, MA), using an electrochemiluminescence immunoassay platform (Meso Scale Discovery (MSD) (Gaithersburg, MD, USA). MSD kit number N45ZA-1 was used to measure CA15.3 antigen levels. For the anti-CA15.3 antibody assay, we used a MUC1 protein naturally produced by a breast cancer cell line. We chose this protein over recombinant proteins or peptides to be able to detect more antibodies against a larger number of epitopes than is possible with synthetic or recombinant capture antigens. While the IgG antibodies we are detecting are mostly peptide epitope specific, the variety of peptide epitopes is dependent on different posttranslational modifications. We used human antigen grade CA15.3 derived from human breast cancer BTA cell line supernatant (Meridian, cat # A32000H), which was coated onto MSD plates. A blocking buffer was applied for 1h followed by PBS wash, sample diluent was applied for 30 minutes, followed by the serum samples at multiple dilutions for 2h. The plate was washed with PBS/0.05% Tween-20 followed by detection of human IgG bound to the specific protein spot with an MSD sulfo-Tag-labeled mouse monoclonal detection antibody (10 μg/ml) for 2h; washing and adding read buffer followed by detection of electro-chemiluminescence (ECL) using an MSD Imager S600. The results were reported as relative luminescence units (RLU) and as an index of relative specific autoimmune reactions Further details on the assays used can be found at https://wwwn.cdc.gov/Nchs/Nhanes/search/datapage.aspx?comonent=laboratory&cyclebeginyear=2001 . Levels of MUC1 and anti-MUC1 antibodies were evaluated using antigen preparation marketed under the designation CA15.3; and we use that terminology through the remainder of the manuscript and in figures and tables in place of MUC1. CA15.3 and IgG CA15.3 antibodies were log transformed to achieve approximately normal distributions. Outliers were identified using the extreme studentized deviate many-outlier procedure ( 19 ) and excluded from the analysis (n=82). Pearson correlations were run between CA15.3 index and RLU antibodies, overall and within subgroups. We used SAS survey procedures (SURVEYMEANS, SURVEYFREQ, and SURVEYREG) with the DOMAIN option for analyzing subpopulations and used sampling weights according to NHANES guidelines ( 20 ). To assess associations between CA15.3 antigen and antibodies and demographic and reproductive characteristics, we used multivariable linear regression models to calculate least squares means (and 95% confidence intervals) within each level of the demographic/reproductive characteristic of interest. Means and confidence intervals were then exponentiated back to the original biomarker units. Regressions were run separately with log transformed CA15.3 antigen or CA15.3 antibodies as the dependent variable. Models for demographic characteristics were adjusted for age (continuous), and race/ethnicity (indicator variables for Hispanic [Mexican American or other Hispanic], non-Hispanic Black, non-Hispanic White, and other races). Each model for reproductive characteristics included the reproductive characteristic of interest and potentially confounding variables: age, BMI (continuous), race/ethnicity, and smoking status (never, former, current). The adjusted least-squares means for CA15.3 antigen and antibodies by reproductive events are reported for all women and separately for pre- and postmenopausal women. Trend tests for exposures in ordinal categories were calculated by modeling the median of each category as a continuous term. P-values for categorical and trend demographic and reproductive variables were calculated with Wald’s F test. All statistical analyses were performed using SAS version 9.4 (SAS Institute, Cary, NC). Data on CA15.3 and anti-CA15.3 antibodies measured in this study are publicly available https://wwwn.cdc.gov/Nchs/Nhanes/search/datapage.aspx?comonent=laboratory&cyclebeginyear=2001 .

Our study confirmed some previous observations and identified new factors that may influence levels of CA15.3 and anti-CA15.3 antibodies in healthy women, thereby modulating immunosurveillance of that cancer antigen that may directly impact cancer risk. Race as a demographic factor controlling a cancer antigen expression was previously reported in NHANES participants who were postmenopausal: non-Hispanic Blacks had lower levels of CA125 antigen compared to non-Hispanic Whites ( 21 ). Race was a key demographic factor in our study predicting CA15.3 antigen levels with non-Hispanic Blacks having higher CA15.3 levels compared to non-Hispanic Whites. Our current study adds the new information that non-Hispanic Blacks also have lower levels of anti-CA15.3 antibodies compared to non-Hispanic Whites, which is more relevant as it suggests weaker immunosurveillance. If higher expression of CA15.3 antigen and lower anti-CA15.3 antibodies in African-American women with cancer follows the same pattern as that seen in healthy African-American women, then this could account for the poorer prognosis seen in African-Americans for several MUC1-associated cancers, including breast, uterine, colorectal, and pancreatic ( 22 – 25 ), for which high MUC1 expression and low anti-MUC1 antibody predict worse survival. Several “lifestyle” variables that were also found to be associated with low anti-CA15.3 antibody levels included weight, BMI, and smoking. Heavier women, those with greater BMI, and current smokers were found to have lower anti-CA15.3 antibody levels; and the level of anti-CA15.3 antibody further decreased with accumulated pack-years. Lower anti-CA15.3 antibody levels in heavier women and smokers from NHANES confirm identical observations in a different sample of healthy women from the European Prospective Investigation of Nutrition and Cancer (EPIC) ( 14 ). This study used an assay for anti-CA15.3 antibodies similar to the RLU assay measured (but not analyzed) in the current study. The fact that two different assays for anti-CA15.3 antibodies in two different populations suggest that the associations relating smoking and obesity with lower anti-CA15.3 antibodies increases the validity of both findings. In fact, obesity and/or smoking are associated with lower antibody response to and efficacy of several vaccines, including hepatitis B ( 26 ), influenza ( 27 ), and COVID-19 ( 28 ). A study designed to raise anti-MUC1 antibodies in individuals with colon adenomas to prevent their progression demonstrated that high levels of anti-MUC1 IgG antibodies could be achieved with the MUC1-based vaccine except in individuals with evidence of pre-existing circulating myeloid-derived suppressor cells (MDSC) ( 29 ). This observation may be relevant since both obesity and smoking increase numbers of MDSC ( 30 , 31 ). Women who were using OCs at the time of sample collection had significantly lower CA15.3 antigen levels, while women who were currently pregnant had significantly higher levels ( Table 3 ). We could not identify any studies reporting on CA15.3 levels and OC use, but the observation that CA15.3 antigen levels are elevated in pregnant women has been reported, including the observation that levels increase as pregnancy advances ( 32 , 33 ). The observation that current breastfeeding raised both CA15.3 antigen and anti-CA15.3 antibody levels also has support from prior studies ( 32 , 34 ). Serum CA15.3 antigen levels varied by menstrual cycle phase with highest levels during menstruation and lowest during the peri-ovulatory (mid-cycle) phase. The later observation might be explained by higher MUC1 expression within the endometrium around the time of implantation to increase endometrial receptivity ( 35 ), which is then released into circulation with menstruation, It is plausible that any biomarker whose endometrial expression varies cyclically will have higher serum levels during menstrual disruption, which is also the case for CA125 (MUC16) ( 36 ). However, we should note that the frequency of women who reported being in the menstrual phase of their cycle, 4%, is lower than would be expected by chance. No significant effect of current HRT use on CA15.3 antigen or antibody levels was seen. Past reproductive events also affected CA15.3 antigen and antibody levels including a clear trend for women with a later menarche to have higher antigen levels and longer durations of OC use to be associated with lower CA15.3 antigen levels in all and postmenopausal women. The latter finding was previously reported for the NHANES cohort and appears to pertain to both White (p=0.003) and Black women (p=0.06) ( 21 ). Childbirth was significantly associated with higher CA15.3 levels in all and in premenopausal women; but, in premenopausal women, CA15.3 levels declined with increasing parity. This is a potentially important finding since risk for several MUC1+ cancers are inversely associated with increasing number of children ( 37 , 38 ). Though no clear effects of current HRT use were seen in Table 2 , Table 3 suggests that past use of HRT is associated with lower CA15.3 levels. We observed that women who reported a diagnosis of endometriosis had higher CA15.3 antigen levels. CA125, CA19.9. and CA15.3 have all been mentioned as potential markers for endometriosis with CA15.3 seen, especially, in women with more advance endometriosis ( 32 ). Higher CA15.3 antigen level in women with a history of endometriosis is a potentially important finding, reinforcing the importance of a history of endometriosis as a risk factor for not only ovarian cancer, but other MUC1-positive cancers including colorectal cancer ( 39 ). Inconsistent with our previous findings ( 10 , 13 ), we did not see a greater level of anit-MUC1 antibodies associated with tubal ligation, Our two studies which showed an association between this procedure and anit-MUC1 antibodies used an un-glycosylated peptide repeat as the capture antigen, while this study used CA15.3 from a breast cancer cell line as the capture antigen. Further study will be necessary to clarify whether an immune reaction to MUC1 is the best explanation for the reduced risk for ovarian cancer associated with tubal sterilization or other factor not involving MUC1 are involved (e.g. blocking ascent of endometrial cells or vaginal contaminants ( 40 ) or effects on tubal fimbria cells making them less susceptible to proliferation ( 41 ). Perhaps the most important observation relates to the relationship between an increasing number of estimated lifetime ovulatory years and lower anti-CA15.3 antibody levels. In all and postmenopausal women, anti-CA15.3 antibodies declined with increasing number of ovulatory years. In the Introduction, we noted that we have made similar observations in different populations using different CA15.3 epitopes as the capture antigen ( 13 – 15 ). The importance of this observation stems from the clear relationships between a greater number of estimated ovulatory years and increased risks for breast, endometrial, and ovarian cancers ( 16 ). In that review, we speculated on possible mechanisms for these associations, not only through effects on MUC1 immunity, but also repeated tissue changes within organs most directly affected during ovulatory cycles. It is also possible that repeated ovulations lead to accumulation of cells in the ovary capable of affecting postmenopausal androgen secretion ( 42 ). A new entry to this list may include the immune modulating effect of MDSC mentioned earlier ( 29 ). There are data to suggest that circulating MDSC are elevated during ovulatory cycles ending with menstruation ( 43 ), raising the possibility that the cumulative effect of ovulatory cycles may persistently elevate MDSC and lower anti-CA15.3 antibodies. Observations from this study that appear to be inconsistent with the interpretation that MUC1-immunity mediates the effect of ovulation on cancer risk are that key determinants of ovulatory cycles did not always trend in the direction that might be predicted. Later menarche, more OC use, greater parity, breastfeeding, and earlier menopause would all decrease ovulations and should individually be associated with greater anti-CA15.3 antibodies. Except for age at menopause this was not clearly the case. The fact that the association between estimated ovulatory years and anti-MUC1 antibodies is more apparent for the composite variables than its components suggest greater complexity to our model than we originally envisioned. Factors which may need to be considered include the epitopes reacting with anti-MUC1 antibodies and the degree of its glycosylation and the interaction of antibodies with T-cells. Expression of MUC1 changes from the normal, fully glycosylated and non-immunogenic form (in the absence of pregnancy and breastfeeding) to the hypoglycosylated, immunogenic form (carrying the CA15.3 epitope) during reproductive events. As changes occur with each successive (and different event), they serve as immune boosters for antibody production. However, for a protective effect against cancer or better response to an early cancer, a T cell response is necessary. Thus, a weakness of our study is that we do not have T cell data; only sera was available. To partially address this problem, we chose to measure anti-MUC1 IgG, which is an indirect measure of helper T cell activity. MUC1-specific helper T cells had to have been activated in order to help MUC1-specific B cells switch from IgM to IgG. Besides the fact that results may depend on the assay used to measure anti-MUC1 antibodies, another weakness relates to the fact that there are several different assays for measuring MUC1, making a direct comparison with other studies more difficult. We used a volume efficient assay suitable for multiplexing, but one that did not yield the same normal cutoff commonly used in clinical assay for CA15.3 (generally <30 U/mL). Currently, there are no comparative studies correlating the MSD CA15.3 values with the standard clinical assay. In conclusion, this study identified several lifestyle and reproductive factors associated with MUC1 antigen and anti-MUC1 antibody levels in healthy women, that may shed light on potential immune mechanisms by which these factors may impact risk of MUC1-associated cancers. The large size of this study with detailed data across a racially and ethnically diverse population in the US, increases the generalizability of our findings.

Human mucin 1 (MUC1) is a transmembrane glycoprotein described by a variety of names in the literature, including CA15.3 antigen, DF3 antigen, human milk fat globule (HMFG), epithelial membrane antigen (EMA), polymorphic epithelium mucin (PEM), Episalin, and others ( 1 ). MUC1 is normally expressed at the apical cell surfaces of glandular epithelium of the respiratory, gastrointestinal, and genitourinary tracts ( 1 ). In cancers arising from these organs, MUC1 becomes overexpressed on the entire surfaces of tumor cells. This overexpression favors cancer growth and spread through several mechanisms: dampening immune response; interacting with other transmembrane proteins, like EGFR; and engaging cytoplasmic-signaling proteins such as Src and beta-catenin ( 2 ). Elevated levels of MUC1 are associated with poor prognosis for breast, prostate, ovarian, gastric, bile duct, colon, renal, non-small cell lung, and pancreatic cancers ( 2 ). Anti-MUC1 antibodies have also been identified in some patients with MUC1+ cancers; and their presence at diagnosis has been associated with better prognosis for breast, colorectal, non-small cell lung, and pancreatic cancers ( 3 – 7 ). These observations have prompted interest in immune-based treatments targeting MUC1 including vaccines based on MUC1 coupled with various adjuvants, passive immunization with anti-MUC1 antibodies, and transfer of MUC1-specific cytotoxic T-cells ( 8 , 9 ). Anti-MUC1 antibodies have also been found in healthy individuals and postulated to be a response to elevated levels of MUC1 following events (other than cancer) that led to antibody generation. Anti-MUC1 antibodies generated in the absence of cancer may be promoted by inflammatory events involving tissues that normally produce MUC1, including childhood mumps, puerperal mastitis, and tubal ligation—each shown to raise anti-MUC1 antibodies and associated with lower risk for ovarian cancer in case-control studies ( 10 – 12 ). Importantly, there is prospective data from two studies suggesting that natural antibodies against MUC1 can lower risk for ovarian cancer. Data from the Nurses’ Health Study reported that higher anti-MUC1 antibodies subsequently led to lower risk for ovarian cancer developing before age 65 ( 13 ); and data from the EPIC cohort showed that elevated levels of anti-MUC1 antibodies are associated with lower risk for serous ovarian cancers arising within three years of the blood draw ( 14 ). If the presence of anti-MUC1 antibodies leads to reduced risk for ovarian cancer, the corollary may follow that their absence leads to higher risk for cancer. We have shown that a greater number of ovulatory cycles is associated with lower anti-MUC1 antibody levels in three different populations ( 13 – 15 ). This observation is important because a greater number of ovulatory cycles is associated with increased risk for not only ovarian, but also breast and endometrial cancers ( 16 ). In this study, we looked at data from the National Health and Nutritional Survey (NHANES) to examine the relationships between demographic, lifestyle, and reproductive variables, including ovulatory cycles, and levels of circulating MUC1 and anti-MUC1 antibodies.