Maternal H-antigen secretor status is an early biomarker for potential preterm delivery

Abstract

Objective

Pre-pregnancy or first trimester biomarkers predicting preterm delivery are lacking. The purpose of this study was to determine whether maternal H-antigen (secretor status) is a potential biomarker for preterm delivery.

Methods

This cohort study examined maternal saliva samples and birth data gathered by the National Children’s Study Vanguard pilot phase (2009–2014) and included 300 women who were ≥18 years old and provided birth data and saliva samples. The maternal secretor status phenotype was determined by quantifying H-antigen in saliva using enzyme-linked immunoassay. Mothers were stratified by secretor status and multivariable analysis estimated adjusted associations with preterm delivery.

Results

Maternal lack of H-antigen production was an independent risk factor for preterm delivery after adjusting for known confounders (aOR 4.53; 95% CI: 1.74, 11.81; P = 0.002).

Conclusions

Maternal H-antigen may be a biomarker identifying women at-risk for preterm delivery. Prospective cohort studies validating these findings are needed.

Introduction

With greater than one million deaths globally each year, preterm birth is the leading cause of mortality in newborns and is the second most common cause of death in children under five [1]. More than 70% of preterm births occur between 34 weeks and 36 weeks, 6 days gestation [2], with perinatal mortality and serious morbidity more pronounced in infants born prior to 34 weeks gestation [3,4,5]. The majority of preterm deliveries are spontaneous, however, the fraction of medically indicated (iatrogenic) preterm deliveries is currently on the rise with 30–40% of singleton deliveries in Canada and the United States being iatrogenic in recent years [3,4,5]. There are many known risk factors for preterm delivery, however, most cannot be detected until late in pregnancy or are not easily amenable to intervention. Biomarkers identifying an increased risk for preterm birth prior to or early in gestation might allow interventions to delay or prevent preterm delivery, however, such biomarkers are lacking. The most commonly used predictor of preterm delivery employs second-trimester sonographic assessment of cervical length, with a shorter length indicating increased risk [6,7,8,9]. While studies have looked at transvaginal cervical measurements as early as 11–14 weeks gestation, the sensitivity prior to the second trimester is low [9], and even high-risk patients who inevitably deliver preterm rarely develop cervical shortening prior to 14 weeks [10]. Most available biomarkers are targeted to the symptomatic mother (rupture of membranes, contractions, and/or cervical changes), and are used to predict delivery within the ensuing 48 h–7 days [11,12,13]. Several studies have identified various second and third-trimester biomarkers purported to predict preterm delivery up to 2 months in advance in asymptomatic women [11,12,13], such as disruptions in the vaginal microbiome [14] and markers of cervical remodeling [15]. However, these studies report conflicting results, and the putative biomarkers they report are invasive, expensive, time-consuming, require further validation, and are not useful prior to or early in pregnancy.

The clinical significance of H-antigen secretor status is an emerging area of research in which secretor positivity appears associated with specific health benefits. Secretor status is determined by the fucosyltransferase 2 (FUT2) gene, which is inherited in an autosomal dominant pattern [16]. Secretor status varies geographically and by race/ethnicity with ~20% of individuals being non-secretors [16, 17]. H-antigen, a glycan, results from the enzymatic addition of fucose by FUT2 to a glycoprotein [16, 17]. In secretors, H-antigen is richly expressed on mucosal surfaces [18] and is present in saliva, vaginal, and other body secretions [19, 20]. Mucosal surface glycans, such as H-antigen, may influence infection susceptibility via host-pathogen interactions [21]. Glycans can act as pathogen receptors or as receptor-decoys [22,23,24], promote commensal microbial composition and resilience [23, 25, 26], and are integral to the physical properties and function of the mucus barrier [27,28,29,30]. Non-secretor individuals are at increased risk for infection from certain pathogenic bacteria [25, 31,32,33,34] and for developing various chronic diseases [35,36,37,38,39,40,41,42,43,44,45]. Non-secretor pregnant women are at increased risk for urinary tract infection presumably from a lack of H-antigen on urothelial cells of the urinary tract mucosa [46]. Furthermore, secretor status has been strongly associated with alterations in the gastrointestinal microbiome during pregnancy with non-secretors exhibiting a preponderance of pro-inflammatory bacteria species [47]. Urinary tract infection [48,49,50], vaginal dysbiosis [51,52,53], and a maternal pro-inflammatory state [54,55,56] have all been directly linked to preterm delivery. Given that non-secretor status is associated with multiple adverse outcomes, and the association of H-antigen with factors known to increase the risk for preterm delivery, the purpose of this study was to determine whether maternal secretor status is a potential early biomarker for preterm delivery. We hypothesized that non-secretor mothers are at increased risk for preterm delivery relative to secretor counterparts.

Methods

The National Children’s Study (NCS) was a prospective longitudinal cohort study intended to measure health and disease factors and to identify drivers of disease with prenatal or developmental origins [57]. The NCS Vanguard pilot phase study enrolled 5608 children and their families from 43 counties across the United States between 2009 and 2014 with the purpose of testing the feasibility of different recruitment strategies and data collection methodologies. The National Institutes of Health (NIH) granted permission for secondary use of de-identified NCS data and biological specimens for our study. We performed no collection of new data or specimens. The Institutional Review Boards at Connecticut Children’s and the University of Texas Health Science Center, San Antonio determined that this study did not meet the definition of human subject research.

Study design and inclusion criteria

The design employed secondary data analysis using biological samples and data previously collected by the NCS Vanguard Study. Participants from the Vanguard Study were eligible for inclusion in our study if they were age ≥18 and provided a saliva sample and corresponding infant birth data. Mothers with incomplete data sets were excluded from the study. The Vanguard Study enrolled 5608 children and their families; however, only 465 mothers provided at least one saliva sample, and of these mothers, only 300 had complete infant birth data sets including all parameters of interest to our study. To prevent the duplication of maternal data, only the first pregnancy for multiparous females during the study period was included, only data for infant “A” was included in the case of multiple gestations, and the higher H-antigen sample concentration was retained for statistical analysis when two saliva samples for the same mother were available. This did not convert any mother from non-secretor to secretor status.

Saliva samples and H-antigen ELISA

Maternal saliva samples were collected by the NCS pre-pregnancy and/or at 36 months post-delivery. Samples were processed, de-identified, coded, and stored at −20 °C by the NCS prior to transfer to our facility. Secretor status was determined by enzyme-linked immunoassay using a monoclonal antibody specific for H-antigen (87-N, Santa Cruz Biotechnology) and methods similar to those previously reported [58,59,60]. Saliva samples were thawed, aliquoted, and boiled for 5 min, diluted 1:200 in 1× PBS, and plated overnight at 4 °C. H-antigen concentrations were calculated using a standard curve (Blood group H-BSA, Carbosynth). Non-secretor status was categorically determined using a cutoff set at the limit of detection for the enzyme-linked immunoassay (8 ng/mL, corresponding to ≤1.6 µg/mL for undiluted samples) with all undiluted samples >1.6 µg/mL categorized as a secretor.

Statistical methods

Maternal demographics were stratified dichotomously by secretor status, including age at delivery, gestation at delivery, and infant gender, as well as known risk factors for preterm delivery including race/ethnicity, multiple gestations, advanced maternal age at delivery (>35 years of age), and proxies for socioeconomic status such as maternal education, employment, household income, and live-in partner. Univariate comparison between secretor statuses applied independent t-test for normally distributed data, Wilcoxon rank sum test for non-normal data, and chi-square test or Fisher’s exact test for categorical data. Diagnostic test parameters (sensitivity, specificity, positive predictive value, and negative predictive value) evaluated the validity of the H-antigen’s usefulness.

Logistical regression models were performed to evaluate the association between secretor status and other known risk factors on preterm delivery (<37 weeks gestation) in unadjusted and multivariable models. Unadjusted single-variable models examined each risk factor variable independently, including secretor status, to determine its impact on preterm delivery. Known risk factors for preterm delivery available in the NCS data sets were determined a priori and included race/ethnicity, proxies of socioeconomic status (maternal education, maternal employment, maternal health insurance, household income, live-in partner); advanced maternal age at delivery (>35 years of age), multiple gestations, and infant sex. Important risk factors, statistically associated with a preterm outcome were hand-selected into the first multivariable model employing backward selection with inclusion and retention thresholds of P ≤ 0.1 and P < 0.05, respectively. A second multivariable model adjusted for potential confounding factors with strong clinical relevance, which were excluded from the first multivariable model. This second model was performed after the initial adjusted model did not retain African American race/ethnicity to ensure race/ethnicity and multiple gestations were appropriately controlled for both as strong clinical predictors for preterm delivery, and since secretor status varies by race/ethnicity. The Akaike Information Criterion, Hosmer–Lemeshow goodness-of-fit test, and receiver operating characteristic (ROC) curves accompanied each model. All model assumptions were tested to ensure the accuracy of the final model. All statistical analyses were conducted using SAS 9.4 (SAS Institute Inc., Cary, NC). A two-tailed P < 0.05 was considered significant.

Results

Salivary H-antigen concentrations ranged from <1.6 to 84 µg/mL (median 44.2; IQR 22.6–59.3) (Fig. 1). Secretor status was determined in 300 maternal saliva samples, of which 81% were secretors (n = 243) and 19% were non-secretors (n = 57). In univariate analyses, no differences in the distribution of secretor status were observed based upon the level of education, maternal employment, health insurance status, household income, live-in partner, median maternal age at delivery, or infant sex (Table 1). Statistically significant differences were detected by secretor status with the non-secretor group showing a higher proportion of Caucasian and African American race/ethnicity and lower proportion of Latina race/ethnicity (P = 0.02), a higher proportion of multiple gestations (P = 0.02), increased rate of premature delivery (P < 0.001) and lower median gestational age at delivery (P = 0.007). Premature delivery was further sub-divided into late-preterm delivery (34 weeks, 0 days to 36 weeks, 6 days gestation) and early to moderate preterm delivery (<34 weeks gestation) as preterm delivery prior to 34 weeks gestation carries greater clinical consequence with increased morbidity and mortality [3,4,5]. A higher proportion of both of these subgroups was detected in the non-secretor mothers with a statistically significant difference for late-preterm delivery (P < 0.001); however, the number of early to moderate preterm deliveries captured was too low to analyze for statistical significance. Diagnostic test parameters (95% CI) include: sensitivity 22.81% (12.74%, 35.84%); specificity 93.83% (90.02%, 96.50%); positive predictive value 46.43% (30.42%, 63.21%); negative predictive value 83.82% (81.76%, 85.69%).

Fig. 1: The distribution of H-antigen concentrations.

Secretor status was determined in 300 maternal saliva samples. H-antigen concentrations ranged from <1.6 to 84 µg/mL (median 44.2; IQR 22.6–59.3), of which 81% were secretors (n = 243) and 19% were non-secretors (n = 57). Non-secretor status was determined using a cutoff set at the limit of detection for the enzyme-linked immunoassay (8 ng/mL, corresponding to ≤1.6 µg/mL for undiluted samples).

Table 1 Maternal characteristics of surveyed participants stratified by secretor status (n = 300).

Unadjusted logistic regression analysis examined each variable independently to determine the impact on preterm delivery. Maternal non-secretor status, maternal African American race/ethnicity, live-in partner, high school completion, and multiple gestations were all found to be independently associated with risk for preterm delivery (Table 2). Further, there was no evidence for a dose–response relationship. Using H-antigen quintiles, secretors were less likely to deliver preterm relative to non-secretors, but among secretors, there was no dose–response relationship between salivary H-antigen concentration and either likelihood or timing of preterm delivery.

Table 2 Unadjusted and adjusted risk for preterm delivery, controlling for known confounders (n = 300).

Using multivariable logistical regression to adjust for potential confounding factors, maternal non-secretor status, live-in partner, and multiple gestations were independent risk factors for preterm delivery (Table 2). The area under the ROC curve was 0.74 (95% CI: 0.64, 0.85) suggesting that the model has good discriminating power. The Hosmer–Lemeshow goodness-of-fit test was not statistically significant (P < 0.89), indicating that these data adequately fit the model.

Our final adjusted model employing the same multivariable logistical regression and controlling for maternal race/ethnicity and multiple gestations, found non-secretor status (aOR 4.53; 95% CI: 1.74, 11.81; P = 0.002) was an independent risk factor for preterm delivery (Table 2). The model also confirmed that maternal African American race/ethnicity (P = 0.01), and multiple gestations (P = 0.03) are significant predictors for preterm delivery (Table 2). The area under the ROC curve was 0.82 (95% CI: 0.74, 0.9), and the Hosmer–Lemeshow goodness-of-fit test was not statistically significant (P < 0.96).

Discussion

Biomarkers that predict a mother’s risk of preterm delivery prior to or early in pregnancy may allow providers to monitor an at-risk pregnancy more closely and potentially intervene to delay or prevent preterm delivery. Using the National Children’s Study data, we report that maternal H-antigen secretor status may be such a marker. Our findings demonstrate that non-secretor mothers were significantly more likely to deliver prior to term after adjusting for known risk factors. Secretor status was more strongly associated with preterm delivery in adjusted analysis than were maternal age, or proxies of socioeconomic status including educational attainment, household income, or presence of a live-in partner. We were unable to locate prior studies examining the relationship of maternal secretor status and the risk for preterm delivery. Further studies in retrospective or prospective cohorts are needed to validate our results.

H-antigen as an early biomarker for preterm delivery would have many clinical advantages. In individuals expressing FUT2, H-antigen production remains stable throughout the gastrointestinal tract during pregnancy [61] and therefore can be measured at any time point. A woman’s secretor status could be established during family planning prior to pregnancy or during the first prenatal visit, allowing early identification of those most at-risk and granting providers ample time to intervene. The majority of risk factors for preterm delivery and available biomarkers are not measurable until well into the second trimester, often too late for intervention especially among extremely preterm fetuses that are the highest risk for morbidity and mortality. Many of these second-trimester biomarkers are solely measured from the amniotic fluid, requiring invasive amniocentesis for acquisition [62,63,64,65]. Furthermore, the clinical utility of existing predictive tests is unclear as study results are conflicting or population dependent [66]. Our findings suggest that H-antigen is easily measured from maternal saliva, and now can be standardized across various laboratories. Using this biomarker within a risk assessment screening could reduce the need for invasive procedures (i.e., amniocentesis) or specially trained practitioners and operators for sample procurement or image acquisition; and furthermore, less risk of operator error as is found with cervical length screening where >25% of images are insufficient [67]. H-antigen measurement can be relatively quick, reproducible, inexpensive, and essentially pain-free.

This study has several limitations. First, this is a secondary analysis using existing NCS data. The Vanguard Study resulted in four different recruitment strategies of varying data set complexity, and was not designed to answer this research question. As a result of the NCS sampling frame and data collection processes, important risk factors for preterm delivery, such as maternal age <18 years, cervical competency, maternal infection (including asymptomatic bacteriuria and chorioamnionitis), preeclampsia, pre-gestational diabetes, chronic hypertension, obesity, smoking, or the need for reproductive technology were either not obtained or not universally available for all mothers. Preterm deliveries captured by the NCS were skewed to late preterm and were not inclusive of extremely preterm births. The total number of early to moderate preterm deliveries (<34 weeks) included in our study was <5, which precluded further analysis of this clinically important subgroup using our models. Furthermore, spontaneous preterm delivery versus medically indicated preterm delivery was not delineated in the available data. Thus, our observed findings may be biased by unmeasured confounding factors. Second, our enzyme-linked immunoassay measured phenotype and not the genotype. It is possible that determining secretor status by H-antigen production (phenotype) is not as sensitive or as specific as genotypic analysis. While our findings identified a potential new risk factor for preterm birth, the sensitivity of this test was low, suggesting that this biomarker is one of many risk factors included in the causal pie (listed above) that can contribute to preterm birth. Additionally, given that secretor status is inherited in an autosomal dominant pattern, we cannot draw conclusions regarding how heterozygosity may influence preterm delivery especially with respect to the wide variation in H-antigen concentrations across all samples. However, the phenotype may arguably be more clinically relevant as the potential for a dose-dependent response in protection from preterm delivery may depend on how much H-antigen is synthesized, which may or may not be influenced by the presence of one versus two functional FUT2 alleles. In a similar manner, an individual that has a secretor genotype but only trace or low H-antigen production may clinically function as a non-secretor. Lastly, this is an associative study. H-antigen was identified as a biomarker indicating increased risk for preterm delivery and does not guarantee pregnancy outcome.

Secretor status varies geographically and by race/ethnicity, with ~20% of individuals being non-secretors [16, 17], which was consistent with our analysis. Pakistan has been identified as having one of the World’s highest preterm birth rates at 15.8 per 100 live births; in contrast, Sweden was reported to have one of the World’s lowest at 5.9 per 100 live births [68]. Interestingly, the frequency of non-secretors in Karachi (Pakistan) [69] was found to be 14.6% higher than a group of Swedish [70] blood donors (35.6% and 21%, respectively). While this comparison of two independent studies is anecdotal, it may provide some indirect epidemiological evidence that secretor status may contribute to preterm delivery.

We could not locate other studies that quantified H-antigen concentrations in human saliva. Prior phenotypic studies have reported the presence or absence of H-antigen using optical density measures [58,59,60]. An H-antigen standard curve, based upon an internationally accepted standard, will facilitate prospective H-antigen studies and allow individual results to be interpreted independent of the investigator.

Our analysis also confirmed previously established risk factors of preterm delivery including multiple gestations and African American ethnicity, further strengthening our model.

Secretor status is strongly associated with alterations in the gastrointestinal microbiome during pregnancy, with non-secretor mothers colonized with a preponderance of pro-inflammatory bacteria species such as Proteobacteria, and fewer species associated with maintaining immunity including Clostridia species, Bifidobacterium species, and Bacteroides [47, 71, 72]. When present on mucosal surfaces, glycans such as H-antigen act as a nutritional source for beneficial bacteria, thereby promoting microbial stability and resistance to pathogen colonization [21, 25, 73]. We speculate, therefore, that maternal secretor status might impact the risk for preterm delivery through host-microbe interactions.

Previously identified second and third-trimester biomarkers for preterm delivery in asymptomatic women include disruptions in the vaginal microbiome [14]. Lactobacillus depletion is associated with poor pregnancy outcomes including late miscarriage and premature delivery [51,52,53]. In the absence of H-antigen, it is possible that vaginal dysbiosis could occur. Furthermore, recent evidence describes the potential role of dysbiosis at other mucosal surfaces, resulting in a maternal “pro-inflammatory state”, which has been directly linked to preterm delivery [54,55,56]. Thus, a deficiency of H-antigen may predispose to dysbiosis and overall inflammation during the pregnant state, potentially further exacerbated by the finding that FUT2 gene expression during pregnancy is upregulated in the endocervix of secretors [61]. Glycans also can influence the physical properties and function of the cervical mucus barrier, preventing pathogen invasion and host inflammatory responses [27,28,29]. Non-secretor status during pregnancy increases the occurrence of urinary tract infection [46], a well-established risk factor for preterm delivery [48,49,50], presumably due to a lack of H-antigen on urothelial mucosal cells.

Maternal secretor status in conjunction with good prenatal care and evaluation for other known risk factors for preterm delivery may provide families and caregivers additional information to help develop a risk assessment, and evaluate possible interventions with a goal of reducing preterm delivery. Early identification and stratification by secretor status could help guide appropriate prenatal care by allowing providers ample time to intervene with those most at-risk, thus allowing for a multidisciplinary approach with interventions early and throughout pregnancy.

Despite the clear association between secretor status and susceptibility to infection and chronic disease [25, 30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45, 58, 73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90,91,92] the exact mechanisms remain to be delineated. Our finding of preterm delivery as an adverse outcome associated with non-secretor status warrants additional confirmatory observational and translational mechanistic studies, and may justify randomized controlled trials testing specific antimicrobial protocols or probiotic therapies to mitigate the increased risk for preterm delivery associated with maternal non-secretor status.

In summary, non-secretor mothers have a significantly higher risk for preterm delivery relative to their secretor counterparts. These data suggest that non-secretor status may be a predictive biomarker for preterm delivery, however further research is needed. The role of secretor status in predicting preterm deliveries inclusive of extremely preterm deliveries, identifying the mechanism through which secretor status may influence pregnancy outcome, as well as protective factors associated with non-secretor mother’s delivering at term are all unknown. A particular focus should be placed on the relationship between secretor status and reason for preterm delivery to further elucidate the role non-secretor status may have on preeclampsia, preterm labor, premature rupture of membranes, and infection during pregnancy. Studies comparing both genotype and phenotype could further clarify if a dose-dependent response exists, and whether a dose–response association is proportional to one versus two functional FUT2 alleles.