Epidemiological and clinical attributes of the study participants
The estimated mean maternal age at conception of women in the MI and MII error groups having DS child were30.15 ± 4.1 (mean±SD) &30.02 ± 3.4 (mean±SD) years, as compared with control group 29.91 ± 3.9 (mean±SD) years (P=0.20 for MI vs Control, P=0.65 for MII vs Control) (Table 1). The mean age of control women with folate regulator wild genotypes was 33.09 ± 3.1 (mean±SD) years which is similar to the estimate 32.90 ± 2.3 (mean±SD) years for the MI women with DS child (P=0.13) and 33.10±3.3 years for the MII womenwithDS child (P=0.96). The mean age ofcontrol womenwith folate regulator mutant genotypeswas 31.91 ± 3.5 (mean±SD), which is again concordant with the estimated mean age 32.01± 4.4 (mean±SD) years for MI mothers (P=0.59) and 31.60 ± 3.8 (mean±SD) years for MII mothers (P=0.18) havingDS child. When we compared the mean age of the women in MI risk variant group with the MII risk variant group, the difference remained insignificant (P=0.13). Moreover, the meiotic outcome groups did not exhibit significant difference in mean paternal age of conception when compared to control group (P=0.24 MI vs control; P= 0.72 for MII vs control). As we dealt with maternal errors and maternal genotypes only, we did not analyze paternal genotypes. The other epidemiological parameters remain concordant between MI and MII case and control groups.
Frequencies of SCT use in Control and Case groups by age categories
A total of 1294 families with child having DS (Case) and 870 families with healthy child (Control) were included in this study. We categorized each of this group as ‘SCT never-user’ (case 0.64 vs control 0.83) and ‘SCT ever-user’ (case 0.36 vs control 0.17) sub-categories according to the declared SCT use status in epidemiological record. We found significant difference in maternal SCT use with odds in favour of the case mothers (OR = 2.772, 95% CI = 2.245 - 3.424, P <0.0001)(Supplementary Table S1).We further stratified the case and the control women according to their age at conception as young (≤28 years), middle (29 to 34 years) and old (≥35 years)groups and tested if there was any significant difference in maternal SCT use pattern across the age. The frequency of SCT use between control and case group differ significantly in the young (OR = 1.996, 95% CI =1.468- 2.634, P<0.0001) and the middle age group (OR = 1.458, 95% CI =1.056-2.015, P = 0.024) but not in the older age group (OR = 1.133, 95% CI = 0.760-1.686, P = 0.608) (Supplementary Table S2).
Frequency of folate regulator polymorphism in case-control and meiotic outcome groups
We calculated frequency of all tested polymorphic variants of folate metabolic regulators and considered the mothers who bear any of four tested polymorphisms together as a group and found ~5 folds increased odds in favour of case mothers over the controls (OR=5.338; CI=4.015 - 7.097; P <0.0001). Out of the 1294 case mothers 956 and 338 were detected as MI and MII error categories, respectively. We have tested all four polymorphic variants, namely MTR A2756G, MTRR A66G, MTHFR C677T & MTHFR A1298C for all MI and MII case samples. The frequency of polymorphic variants in meiotic outcome groups is given in Table 2. All four tested polymorphic variants exhibited strong association with maternal meiosis II NDJ error and this reconfirms the result that we obtained in our previous study 24. However, frequencies of all four polymorphic variants were estimated negligible in the MI NDJ group and so we considered only the MII group as case for further analyses that involve association study (Supplementary Table S3) and logistic regression modelling.
Some of the participating women carried more than one tested folate polymorphic genotype and for them we estimated synergistic effect through gene-gene interactions models for all four tested variants. All the possible genotype combinations of any two given loci at a time were tested taking wild type genotype as reference (Supplementary Table S4).
Further, we have designed hypothetical models to evaluate the additive risk of maternal genotypes in combination with all the tested variants (Supplementary Table S5) and found gradual increase in risk of DS birth with increasing number of risk alleles in the tested loci. When all four loci carry risk alleles together either in homozygous or heterozygous state in the tested models, a ~17 folds increase odds in favour of MII errors was evident (P <0.0001). When three out of four loci carried respective minor allele together, we estimated ~9 folds increased odds in favour of MII error(P <0.0001).
Interactions among Risk factors
We stratified the participating case mothers into three age categories, based on maternal age at the time of conception following our previous definition5: young (≤28 years), middle (29-34 years) and old (≥35 years). For all the analyses, the maternal age of conception was considered as proxy for oocyte age as direct estimation of the later was beyond the scope of the present study. We used binary logistic regression and liner regression models to study a variety of questions regarding interaction among genetic risk factors i.e., amount of recombination, location of recombination, folate regulator gene mutations and epidemiological risk factor i.e., maternal SCT use and their association with maternal age at conception of DS foetus. Our analyses and statistical modelling are designed to address the following principal questions: 1) Does any significant difference exist in the association of smokeless chewing tobacco (SCT) between cases and controls, and does this depend on maternal age and maternal folate regulator mutation or polymorphisms? 2) Considering only cases, is there any difference in SCT use among MI and MII error groups, and does this depend on maternal and folate regulator mutation or polymorphism? 3) Considering MI and MII cases separately, is there any relation between SCT use and the amount of meiotic recombination and does this depend on maternal age and does folate regulator mutation or polymorphisms have any effect on if? 4) Again, considering MI and MII cases separately, is there any relation between SCT use and the location of meiotic recombination and does this depend on maternal age and folate regulator polymorphisms? All these models explored the risk factors separately as well as considered them together to find any interaction among them that predispose women for having DS pregnancy.
Model I: Effects of SCT and Genotype in Cases vs. Controls
In case-control analyses we considered maternal age, maternal genotype and SCT use as predictors variables and DS birth as outcome. The frequencies of occurrence of errors in case and control groups stratified by maternal folate regulator polymorphic genotypes, maternal age at conception and SCT use status is represented in the Table 3.We found maternal age, folate polymorphic genotype and SCT as significant predictors (P=0.00) with young with had more frequent meiotic errors and old group had the least frequent errors (Table 3). In logistic regression analyses we tested various models (Table 3) of interactions among the predictors. We found significant elevated odds in favour of case women in the interaction models, namely young age X SCT ever user X folate polymorphic genotype, middle age X SCT ever user X folate polymorphic genotype and old age X SCT ever user X folate polymorphic genotype. This suggests both the epidemiological and genetic risk factors together increase risk of NDJ error in all the age group. We obtained intriguing result in the model Old× SCT never-use × folate polymorphic genotype which increases odds nearly 6 folds (P=0.00) in favor of case women.
Model II: Effects of SCT and Genotypes in MI vs. MII
This is case only analyses and we considered maternal age, maternal genotypes and maternal SCT use as predictors and meiotic errors as outcome variables. We observed more frequent incidence of DS birth among women with polymorphic genotype and SCT use in both the MI and MII groups (Table 4) in compare to any other combinations of risk factors. Moreover, the DS birth incidence was more frequent among the younger mothers and gradually decreases with age. For example, we recorded frequency of MII error among the women who were SCT user and had polymorphic genotypes as 0.42, 0.32 and 0.26 (Table 4) for young, middle and older groups, respectively. Logistic regression models revealed interaction between SCT use and folate polymorphic genotypes as significant predictors of MII error(Table 4).Again, the interaction proved significant for all the age groups (OR=21.48 for young and middle age, OR= 24.1 for old age group; P=0.00 for all the model). Interestingly, we found significant odds in favour of MII error for the interaction term SCT never use X folate polymorphic genotypes in all the age groups which suggest the maternal folate polymorphic genotypes impose risk of MII NDJ even in absence of SCT.
Model III: Effects of SCT and Genotype on amount of recombination in M-II NDJ group
Table 5 represents the frequency of single observed recombination events among MII women stratified by their SCT use status, folate polymorphic genotypes and age at conception. We observed reduction in double crossover frequency among the polymorphic genotype bearing and SCT user women than any other categories. We scored frequency of double recombinants among young, middle and older women as 0.13, 0.36 and 0.44, respectively for the SCT user mutant genotype women in contrast to 0.48, 0.56 and 0.68 in the respective age categories of wild type SCT never user group. Pair wise comparison among the tested categories was conducted using chi square tests. We found significant difference between ‘young wild type SCT never-user’ and ‘young wild type SCT ever-user’ (P=0.01), ‘young wild type SCT never-user and young folate polymorphic genotype SCT never-user’ (P=0.01),’ young wild type SCT never-user and young folate polymorphic genotype SCT ever-user’ (P=0.0006) pairs. Interestingly, maximum difference in frequency of recombinant events was recorded (P=0.006) for the pairs ‘young wild type SCT never-user’ and ‘young folate polymorphic genotype SCT ever-user’ with only 13% of all observed double recombination in the latter group. No other pair-wise comparisons were proved significant. Another important observation is that the differences were recorded within the young age category, not in other age groups and this suggests that the effects of risk genotypes and SCT use are maternal age independent.
To find out true interactions among the risk factors we performed logistic regression analysis considering maternal age, maternal habits of SCT, maternal folate regulator genotypes as predictors and amount of recombination on nondisjoined Ch21 as outcome variables. In these analyses we used the interaction term ‘young X SCT ever use X wild genotype’ as reference. Significant interactions were recorded only in the young age category with ‘SCT ever use X wild genotype’ (P=0.027), ‘SCT ever use X folate polymorphic genotype’ (P=0.002) and ‘SCT never use X folate polymorphic genotype’ (P=0.021). No other models for other age groups were proved significant (Supplementary Table S6).
Model IV: Effects of SCT and Genotype on spatial distribution of the observed single recombination events in MII NDJ group
Table 6 represents the distribution of single recombinant events along 21q of MII errors group stratified by maternal genotype, SCT use status and age at conception. The single observed recombination events show a change in spatial distribution pattern from the middle of the chromosome arm in the young age group to the centromere proximal position in the older age group in ‘wild type SCT never-user women’ and this observation is consistent with the findings from the previous studies 4,5. We observed more frequent single recombination events in the pericentromeric regions in the SCT ever-user group as well as among the folate polymorphic genotype bearing women. This is a new finding. We scored ~ 18% of all observed single recombination events in the centromere proximal intervals 1 and 2 among the ‘young SCT never-user wild genotype’ women in contrast to ~81% of all single recombination events in ‘young-folate polymorphic genotype SCT ever user group’ (Table 6). Interestingly, the distribution pattern of single observed recombination events across the age groups remained similar among the women who were SCT ever-user as well as had folate polymorphic genotypes. In other words, we did not observe any displacement of single recombinant events towards centromere with age among SCT user folate polymorphic genotype women. This observation is also novel. We compared the spatial distribution of single observed recombinant events among the age groups in pairwise manner through chi square and found significant difference in the younger age group between ‘wild type-SCT never-users’ vs ‘folate polymorphic-SCT ever-user’ (P<0.0001), ‘wild type-SCT ever user’ vs ‘folate polymorphic-SCT ever-users’ (P<0.0001) and ‘folate polymorphic-SCT never-users’ vs ‘folate polymorphic-SCT ever-users’ (P <0.0001). Careful observation revealed both ‘SCT use’ and ‘folate polymorphic genotype’ has an effect on recombination displacement towards centromere (Table 6).
In evaluating the effect of interactions among the predictors on the placement of single recombinant events on the 21q we did linear regression analyses considering younger age group as reference. Unlike logistic regression we considered any two predictors at a time in a given interaction model(Supplementary Table S7).We did this owing to inability of converting the position of single recombinant events into binary variables needed for data entry in logistic regression program. When considered individually, only maternal age at conception revealed as significant predictors of position of single recombinant events. Significant effects were recorded for the models ‘old age X SCT ever-user’ (P=0.00), ‘old age x folate mutant genotype’ (P=0.00) and ‘SCT ever-user X folate mutant genotypes’ (P=0.005). This observation suggests any two risk factors when present together influence effectively the position of recombination events and probably caused more centromere proximal recombinant events on 21q.