Maternal Telomere Length and Risk of Down Syndrome: Epidemiological Impact of Smokeless Chewing Tobacco and Oral Contraceptive on Segregation of Chromosome 21

Nearly a century has passed since advanced maternal age was identified as a risk factor for giving birth to a child with Down syndrome (DS) [1, 2]; however, the molecular basis of this relationship remains incomprehensible. DS is predominantly caused by nondisjunction (NDJ) of chromosome 21 (Ch21), and the effect of advanced maternal age is restricted only to NDJ that occurred in maternal gametogenesis and not to paternal or inferred postzygotic errors. Both maternal meiosis I (MI) and meiosis II (MII) errors show an association with advanced maternal age of conception [3, 4].

To explain the relationship between advanced maternal age and the high incidence of NDJ of Ch21, several hypotheses have been put forward that revolve round the general aspect of ovarian functions, such as maternal age-related change associated with oocyte pool size or hormone function [5, 6, 7]. These hypotheses inspired researchers to explore the underlying mechanisms by assessing specific subcellular components for their vulnerability due to maternal aging; the list includes mitochondrial components, spindle apparatus and sister cohesion protein complexes [8, 9, 10, 11, 12]. Other researchers have investigated indicators of oocyte pool size, such as hormone titers or antral follicle count, to determine if women who have had an NDJ event have smaller oocyte reserves than the controls [13, 14, 15]. Warburton [15] proposed the ‘biological aging' theory to explain the maternal age effect on the higher incidence of DS, which states that biological aging differs among women of the same chronological age and that the frequency of trisomic conceptions depends upon the biological age of women rather than on their chronological age. According to this hypothesis, women bearing a child with DS are biologically older than women with a healthy baby. To test whether ‘biological aging' in DS has a molecular basis, we have estimated maternal telomere length (TL) in our previous work [16]. TL has been argued by some to be an authentic molecular marker of biological aging [17]. We found that mothers aged 34 years or older having a child with DS had shorter TL on average than the controls, with the shortest telomere among women with maternal MII errors.

In the current study, we wanted to know whether shorter TL of mothers bearing a child with DS might be related to behavioral/environmental factors that are associated with the risk of conception of a child with DS. In other words, we were curious to unravel the interactions among maternal chronological age, maternal TL, and various maternal periconceptional behavioral or environmental factors. We recruited families with children with DS and control families from Kolkata metropolitan and adjoining suburban areas and assayed TL as well as risk factors. We chose two maternal periconceptional risk factors: use of smokeless chewing tobacco (SCT) and oral contraceptive (OC) use. The practice of periconceptional cigarette smoking and OC use are known to associate with DS, as reported in a previous study [18]. However, the epidemiology of risk exposure of our Indian cohort is somewhat different from that in reported Western populations. Specifically, in our Indian cohort, smoking is unusual in women; rather, they use SCT, often from early adolescence. In addition, a considerable number of women in this society start irregular OC use without physician consultation immediately after commencement of their sexual activity and continue this practice irregularly even after they have conceived. Because of the heavy exposure to these two important environmental risk factors, we do believe that this population is highly suitable to study risk factor interaction with maternal chronological and molecular age.

The study was designed following the Declaration of Helsinki and subsequently reviewed by the Institutional Ethics Committee constituted by the University of Calcutta, Kolkata, India; the committee specifically approved this study. Written informed consent was given by the participating women as well as by their husbands in the preprinted format. The records have been kept secret, using laboratory codes. The Ethics Committee has also approved this consent procedure.

The present work is an extension and continuation of our previous study [19], whose outcome suggested that the maternal practice of SCT and OC use imparts the risk of Ch21 NDJ in maternal chronological age in an age-independent and age-dependent manner, respectively. The aims of the present study are to refute our previous findings [19] and to find out whether these two risk factors impart their effect on chromosome segregation through the route of maternal molecular aging.

A total of 206 families, each with an infant with suspected DS, were enrolled for the study. They were referred to our laboratory after initial phenotypic screening by a birth defect surveillance group consisting of pediatricians from different hospitals and medical colleges in the Kolkata metropolitan area and suburbs. These families were reported between 2003 and 2010. The sample included those who participated in our previous study [19]. Eligibility for enrolment included the availability of a complete set of DNA samples from the father, the mother and the infant with DS, free trisomy 21 and a live-born child with DS as determined by classical karyotyping at our laboratory, as well as the completion of a lifestyle questionnaire, including information on periconceptional SCT and OC use. The study subjects are mothers with a child with DS. The interviews with these women were performed privately in person after consent from both the participants and their husbands. An extensive printed questionnaire was used to collect detailed family history, information about lifestyle, birth control preference and other relevant epidemiological details. All the subjects have used either SCT or OC or both periconceptionally, and the frequency of use was at least 3 times per day (≥25 mg nicotine/day). The enrolled cases consisted chiefly of Bengali-speaking families from West Bengal, the majority of whom were Hindus and Muslims.

A control cohort of 195 families, each with a euploid infant, was enrolled at our laboratory. The controls were identified and selected randomly among healthy newborn euploid babies without any birth defect from the enrolled patient databases and birth registers of the hospitals that provided the cases. We chose those hospitals for control selection to ensure maximum possible similarity in demographics between cases and controls (table 1). Control mothers were stringently age matched with the case mothers. The minimum requirement for the enrollment of control families was the completion of the maternal questionnaire and the availability of DNA samples. The families were unrelated between and within the groups.

To include only the free trisomy 21 cases and true euploid control mothers, and to exclude women with hidden mosaicism, conventional karyotyping was performed. At least 40 metaphase plates were analyzed for each to confirm the chromosomal profiles of cases and controls.

Each participating family was genotyped with a battery of highly polymorphic microsatellite markers spanning from the pericentromeric region to the telomere of 21q. The order of markers was centromere - D21S369, D21S215, D21S258, D21S120, D21S1432, D21S11, D21S1437, D21S210, D21S1270, D21S167, D21S1412, D21S2055, D21S1260, D21S1411, D21S1446 - qter. The maternal origin of NDJ was determined by establishing the contribution of the maternal alleles to the child with DS. The decision was considered confirmed when at least 5 markers were informative and the allelic status of the rest of the markers was consistent with that inference. The first 4 markers were used to interpret the stage of meiotic NDJ, i.e., MI or MII error. We inferred an MI error when maternal heterozygosity of these markers was retained in the trisomic child, i.e., the marker was ‘non-reduced'. If maternal heterozygosity was ‘reduced' to homozygosity in the trisomic child, we concluded that the case was of MII origin. Some proportions of so-called MII errors actually originate at MI when homologous partners fail to segregate properly, followed by an obvious error at MII in which sister chromatids become tangled and do not separate precisely. Despite this fact, we chose the conventional approach and treated apparent ‘MI' and ‘MII' errors separately in many of our analyses. The determination of MI or MII was done blinded to the risk factors and other epidemiological variables. Genotyping was also done for controls with the same set of markers to eliminate cryptic mosaicism. At least 5 informative markers were used as inclusion criterion for further analyses.

We estimated TL as proxy of maternal molecular age and oocyte age. The logic behind this approach is that telomere shortening in dividing peripheral lymphocytes is an indication of molecular aging and thus of DS risk [16], though the telomere of germinal tissue does not exhibit shortening. Blood samples were collected within a week of the birth of the child for both cases and controls. This ensured that the maternal age at which TL is estimated is the same as the maternal age at the time of giving birth. Due to either insufficient DNA samples or noise signals from the gel matrix, we successfully recorded TL of 148 cases and 183 controls. TL estimation was blinded to the case/control status, risk factors and other variables. We followed the restriction digestion/Southern blot method for TL estimation using the TeloTAGGG Telomere Length Assay kit from Roche chemicals.In brief, the DNA samples (10 μg) were digested with HinfI and RsaI (20 units/μl each), following the protocol described in the kit. The digested DNA was resolved in 0.8% agarose gel. The gel was dried, denatured and neutralized. The DNA fragments in the gel were then transferred to the membrane by Southern blot technique. UV radiation was used to fix the fragments on the membrane. The membrane was then hybridized with the labeled telomere-specific probe provided in the kit. An autoradiogram was obtained after exposing the membrane to X-ray film. The signal intensity, i.e., density, was recorded and calculated using the freely available NIH software Image. The signal was optimized and subdivided into 1-kbp intervals. The telomere restriction fragment length (L) was estimated using the formula L = Σ(OD_i)/Σ(OD_i)/(L_i), where OD_i is the signal intensity and L_i is the length of the terminal restriction fragment at the midpoint of position i. The calculation took into account the higher signal intensity from larger terminal restriction fragments because of multiple hybridization of the telomere-specific probe.

All statistical calculations were performed using the SCT and OC data of 148 cases and 183 controls, for whom TL was estimated unambiguously. The cases and controls were divided into three groups on the basis of their age at the time of conception, following our previous definition [20]: ≤28 years (young), 29-34 years (middle) and >34 years (old). All age groups were further stratified into ‘non-user', i.e., has never used either SCT or OC, and ‘user', i.e., has used SCT and/or OC, around the time of conception with a frequency of at least 3 times per day. Since the goal of the study was to look whether the TL of the mothers is associated with maternal periconceptional behaviors, linear regression methods were used to estimate telomere shortening as outcome variable; predictors were SCT and OC use, as well as age group and meiotic outcome, i.e., MI and MII. Simultaneously, interaction terms in the logistic regression models were employed, which allowed us to test for differences in risk factor effects among the groups. The regression models were implemented and analyzed in the software package STATA.

We have categorized the participating women into two groups, namely SCT and OC users and non-users, and stratified them further by maternal age at birth of the child with DS and the meiotic state of origin of the error (table 2). Both SCT and OC use are significantly associated with giving birth to a child with DS in a case-control analysis (p = 0.003 for SCT use and p < 0.001 for OC use by Pearson's χ² test). The distribution of tobacco users exhibits a decreasing trend with the old age groups (for MI, 0.42 in the young age group, 0.23 in the middle age group and 0.35 in the old age group; for MII, 0.5 in the young age group, 0.13 in the middle age group and 0.19 in the old age group); on the contrary, OC user frequency exhibits a gradual increase with aging (for MI, 0.19 in the young age group, 0.30 in the middle age group and 0.51 in the old age group; for MII, 0.13 in the young age group, 0.27 in the middle age group and 0.6 in the old age group). These trends are consistent with and confirm our previous results [19]. In MI versus MII analysis, SCT use is marginally increased in MII (p = 0.083), and OC use does not show a statistically significant difference (p = 0.43).

Tables 3 and 4 show the mean TL of SCT and OC users/non-users; table 5 exhibits the TL of women who used both SCT and OC. Statistical associations among these factors were tested in linear regression models that predicted TL as an indicator of the age group, the meiotic outcome group, SCT and OC use (fig. 1). As previously shown [16], TL decreases with age and is lower in the MII group than in the MI group or controls.

The logistic regression models with all predictors and interactions showed that SCT use was a significant predictor of TL (p = 0.001). There was a significant interaction between MI/MII NDJ and SCT use (p = 0.04), indicating that the effect of SCT use on TL varies in the different meiotic outcome groups. In particular, as seen in table 3, it appears that MI SCT non-users have similar TL to controls, while MI SCT users and all MII mothers have reduced TL. All of these results are consistent across the age groups. A faster rate of telomere shortening with increasing age was evident among the SCT users bearing a child with DS compared to non-users (fig. 1), using linear regression models (β = -0.143, R² = 0.491 for MI SCT users; β = -0.207, R² = 0.464 for MII SCT users; β = -0.159, R² = 0.463 for combined MI and MII SCT users; β = -0.109, R² = 0.158 for combined MI and MII SCT non-users), confirming our previous finding [16]. Among controls, there is a suggestion that SCT use is associated with decreased TL, although this is not statistically significant.

The model did not find a statistically significant association between OC use and TL (p = 0.18), although TL is slightly shorter in OC users. There was no evidence of an interaction between OC use and the meiotic outcome groups, and thus, the data are represented in the combined MI and MII group (table 4). In the model that explores the relationship between TL and coexposure of SCT and OC use, we found a significant effect (p = 0.03) in the older women with a MII NDJ event (table 5).

The purpose of the present study is to ask whether SCT or OC use imparts a risk of DS by its effect on maternal molecular age or not. The decreasing frequency of SCT users and the increasing frequency of OC users with advancing maternal age (table 2) is strongly concordant with our previous findings [19] and confirms our earlier notion that SCT use is a maternal age-independent and OC use a maternal age-dependent risk factor for DS [19]. In addition, the outcome of the present analyses exhibited a significant association with telomere shortening in mothers with MI error; the effect is consistent across age groups (table 3).

Combining these data with previous observations that SCT use is a risk factor for giving birth to a child with DS, it seems that SCT use might induce telomere shortening and imperil the chromosome segregation apparatus at an early phase of oocyte development, leading to missegregation of Ch21 at MI. If so, it is a matter of interest whether SCT use separately affects the telomere or chromosome disjunction or if it distorts the activity of some common factors that act as a molecular link between the telomere maintenance system and the chromosome segregation apparatus. Intuitively, the later possibility could gain indirect support from observations in a mouse model [21] in which the BubR1 mutant exhibits chromosomal aneuploidy as well as accelerated senescence. Support for this prediction is also seen in the Chinese hamster [22]. Recently, TL estimation in human oocytes has revealed that eroded telomeres are associated with random embryonic aneuploidy due to genetic instability [23].

Consistent with our previous observation [16], the MII NDJ group of our case sample exhibited the shortest telomere, but without any significant association with SCT use (table 3). Thus, MII mothers have shorter telomeres than controls regardless of SCT use. This result is intriguing, because SCT use is a probable risk factor for MII NDJ, as observed in our previous epidemiological study [19]. This suggests that the effect of SCT use on MII NDJ is not imparted through the molecular pathway to which the TL maintenance system is linked, but rather through some other etiological routes. The apparent similarity of TL between the MI SCT and the MII SCT users suggests the tentative timing in oogenesis at which SCT use exerts its adverse effect on the chromosome segregation system. Probably, it is the early phase of meiosis which is common to both the MI and MII NDJ categories.

Our OC data do not show a statistically significant relationship with TL (table 4), which suggests that OC use is not related to maternal molecular age. However, the model that evaluated an association of coexposure of SCT and OC use (table 5) with TL exhibits significantly shorter TL among older mothers with MII NDJ. This finding is intriguing and needs careful justification. At this point, it is difficult to fit these two apparently contradictory facts. Our previous finding [19] and the present data have confirmed that OC use is a maternal age-dependent risk factor. The possible explanation is that OC use alone does not have, but the interaction with SCT use does have, an effect on telomere shortening. Another possibility is that OC use alone associates with the subcellular aging process which does not include the telomere shortening system. On the contrary, the most likely possibility is that the relationship of OC use with TL would benefit from being re-examined with a larger sample size in future, as we estimated the p value of 0.18 in a logistic regression model.

In summary, our results clearly demonstrate, for the first time ever, that an environmental genotoxic agent like chewing tobacco interacts with the genetic component of the oocyte that causes molecular aging, as is evident from shorter TL among user women, irrespective of their chronological age, and this imperils the proper segregation of Ch21 in the MI phase of oogenesis. All of our results would be strengthened by replication in other populations, and while we cannot make any claims to understand the causality or mechanism, it appears that the effect of advanced maternal chronological age on Ch21 NDJ may be related to maternal molecular age, which probably is affected by SCT use.

The authors are thankful to the participating families who have donated their tissue samples. Our work was supported by the University Grant Commission (UGC), New Delhi, India (grant sanction No. F-30-32 BSR; 2014).

The authors have no conflicts of interest with any person or institution.