Abstract

In oocytes with nondisjoined chromosomes 21 due to a meiosis I (MI) error, recombination is significantly reduced along chromosome 21; several lines of evidence indicate that this contributes to the nondisjunction event. A pilot study found evidence that these oocytes also have reduced recombination genome-wide when compared with controls. This suggests that factors that act globally may be contributing to the reduced recombination on chromosome 21, and hence, the nondisjunction event. To identify the source of these factors, we examined two levels of recombination count regulation in oocytes: (i) regulation at the maternal level that leads to correlation in genome-wide recombination across her oocytes and (ii) regulation at the oocyte level that leads to correlation in recombination count among the chromosomes of an oocyte. We sought to determine whether either of these properties was altered in oocytes with an MI error. As it relates to maternal regulation, we found that both oocytes with an MI error (N = 94) and their siblings (N = 64) had reduced recombination when compared with controls (N = 2723). At the oocyte level, we found that the correlation in recombination count among the chromosomes of an oocyte is reduced in oocytes with MI errors compared with that of their siblings or controls. These results suggest that regulation at the maternal level predisposes MI error oocytes to reduced levels of recombination, but additional oocyte-specific dysregulation contributes to the nondisjunction event.

INTRODUCTION

Nondisjunction is the failure of homologous chromosomes or sister chromatids to segregate to separate daughter cells during cellular division. When this type of error occurs during meiosis, some of the resulting gametes will have too many or too few chromatids compared with the expected haploid number (aneuploidy). After fertilization, these aneuploid embryos will be significantly compromised, leading to early fetal loss or significant defects at birth, including intellectual and developmental disabilities, physical defects and a shortened life span. Aneuploidy is purported to occur in at least 5% of clinically recognized pregnancies (1–4) and accounts for ∼35% of spontaneous abortions (1). Notably, aneuploidy occurs in ∼1–2% of sperm (5) and 20–25% of oocytes (6,7).

Trisomy 21 is the most commonly observed type of aneuploidy in liveborn infants and occurs at a rate of 1 of 700 live births in the United States (8). As with the majority of autosomal trisomies, trisomy 21 is most often the result of errors in the oocyte (referred to as maternal errors), and accounts for ≥ 90% of cases (9,10) Among those meiotic maternal errors, ∼70% are classified as meiosis I (MI errors) and 30% as meiosis II (MII errors). It may be that many of the MII errors are initiated in MI, but end with an oocyte containing sister chromatids (11).

Maternal age has been identified as the strongest risk factor for trisomy 21. The first molecular correlate found to associate with the occurrence of trisomy 21 is altered recombination along 21q. Studies have shown that maternal MI errors are associated with a single distal exchange or absence of any exchange on nondisjoined chromosomes 21 while, MII is associated with a proximal pericentromeric exchange (11–13). The basis for these associated patterns of recombination is unclear. Studies in controls (oocytes with normal meiotic outcomes) (14–18) and oocytes with nondisjoined chromosomes 21(19) have provided evidence for factors that act globally (i.e. trans-acting factors) on the oocyte to influence the variation in recombination counts across the genome.

Studies on recombination location and frequency in controls have used publically available data sets genotyped for genome-wide variants. Considering the importance of recombination in segregating chromosomes during gamete formation, it was expected that the number and placement of recombination genome-wide would be tightly regulated. While each study has found common themes in genome-wide recombination frequency and placement along chromosomes, there is variation in these properties based on individual, sex and ancestry differences (14–17). This suggests that variation may be explained by some inherent genetic characteristic or an environmental factor. For example, genomic loci and genes have been shown to contribute to variation in recombination frequency and placement. Allelic variants in the RNF212 gene were the first to be correlated with recombination count in males and females (15,18). An inversion at genomic region 17q21.31 was also found to be associated with elevated recombination in female carriers versus non-carriers (15,20). PRDM9 is an example of a gene known to be involved in the placement of recombination events and alleles have also been associated with a small but significant change in crossover rate within hotspots (21–23).

As recombination frequency and placement are altered on nondisjoined chromosomes 21, it is of interest to determine whether this anomaly is present among all chromosomes of a disomic oocyte and linked to global dysregulation of recombination. In a previous pilot study, we found that the predicted number of exchanges along chromosome 21 was correlated with the number of genome-wide recombinants for both MI error and control oocytes (19). In addition, the MI error oocytes exhibited a reduction in genome-wide recombination when compared with controls. These findings suggested that there is a factor that influences correlation in recombination counts among the chromosomes of an oocyte (19). These data did not allow us to ask whether the reduced genome-wide recombination of the MI error group was the result of global dysregulation of recombination. By expanding our dataset to include a larger number of MI error oocytes, the siblings of MI error oocytes and MII errors, we were able to address this question along with others.

We examined two levels of recombination count regulation in oocytes with nondisjoined chromsomes 21: (i) regulation at the maternal level that leads to correlation in genome-wide recombination across her oocytes (represented by recombination profiles of probands and their siblings) and (ii) regulation at the oocyte level that leads to correlation in recombination count among the chromosomes of an oocyte. We also asked whether recombination location along chromosome 21 could predict genome-wide recombination patterns among MI or MII error oocytes in order to identify potential global dysregulation of recombination location. Our results have provided new insights into the risk factors for nondisjunction.

RESULTS

Genome-wide recombination count

Meiosis I nondisjunction

We first sought to confirm the results seen in Brown et al. using a larger, extended sample set (19). In the aforementioned study, we obtained genome-wide recombination counts from 15 oocytes with an MI error and no observed recombinants on chromosome 21. We compared the genome-wide recombination counts from these MI errors with controls that were obtained from the CEPH pedigrees and stratified by having no observed recombination on chromosome 21 or having one or more. We found a significant positive correlation between the inferred number of chromosome 21 exchanges (estimate of the actual number of exchanges based on a tetrad analysis) and genome-wide recombination count. It was also shown that the MI error group had reduced recombination when compared with controls.

For our current study, we hypothesized that the reduced genome-wide recombination observed for the MI error group in the previous study is linked to dysregulation of recombination at the oocyte and/or maternal level. First, we tested whether oocytes with MI errors and no detectable chromosome 21 recombinants had reduced genome-wide recombination counts compared with controls with no detectable chromosome 21 recombinants, similar to our previous study. Using a t-test, we found that there was a significant reduction in the mean genome-wide recombination count among all MI errors compared with that of the control group [37.9 (95% CI 36.3–39.4) versus 42.61 (95% CI 42.3–43.0) (P < 0.0001), Table 1]. When we restricted the sample to only those with no observed recombinants on chromosome 21, the difference remained significant (37.6 (95% CI: 35.5–39.8) versus 40.87 (95% CI 40.3; 41.4), P < 0.006, Table 1). Further examination of the mean genome-wide recombination counts of the controls by 0, 1 and >1 chromosome 21 recombinants showed the pattern expected for correlation in recombination number between chromosomes (Fig. 1): the mean genome-wide recombination count increased with increasing number of chromosome 21 recombinant events. Among MI errors, this pattern was perturbed, as there was no difference between the mean values for those with 0 and 1 observed chromosome 21 recombinants, although the mean genome-wide recombination count among those with >1 chromosome 21 recombinants was increased (not statistically significant).

Table 1.

Genome-wide recombination count t-tests

Meiotic outcome group Mean 95% CL lower 95% CL upper Difference from controls P-value 
MI comparisons 
 Controls (N = 2723) 42.61 42.27 42.95 n/a n/a 
 MI (N = 94) 37.89 36.35 39.44 −4.72 <0.0001 
 MI siblings (N = 64) 38.52 36.50 40.53 −4.10 0.0003 
 Controls with 0 recombinants (N = 1044) 40.87 40.34 41.40 n/a n/a 
 MI with 0 recombinants (N = 56) 37.63 35.48 39.77 −3.25 0.0066 
MII comparison 
 Controls >0 recombinants (N = 1679) 42.61 42.27 42.95 n/a n/a 
 MII >0 recom binants (N = 20) 40.45 36.62 44.28 −2.16 0.29 
Meiotic outcome group Mean 95% CL lower 95% CL upper Difference from controls P-value 
MI comparisons 
 Controls (N = 2723) 42.61 42.27 42.95 n/a n/a 
 MI (N = 94) 37.89 36.35 39.44 −4.72 <0.0001 
 MI siblings (N = 64) 38.52 36.50 40.53 −4.10 0.0003 
 Controls with 0 recombinants (N = 1044) 40.87 40.34 41.40 n/a n/a 
 MI with 0 recombinants (N = 56) 37.63 35.48 39.77 −3.25 0.0066 
MII comparison 
 Controls >0 recombinants (N = 1679) 42.61 42.27 42.95 n/a n/a 
 MII >0 recom binants (N = 20) 40.45 36.62 44.28 −2.16 0.29 

T-tests comparing the average genome-wide recombination counts of: all MI errors and controls, all MI siblings and controls, MI errors and controls with 0 observed recombinants on chr21 and MII errors and controls with >0 recombinants on chr21. P-values are from t-tests that compared the genome-wide recombination counts of cases or MI siblings with that of controls.

Figure 1.

Average genome-wide recombination counts for MI errors and controls. Average genome-wide recombination counts of the complimentary autosomes for MI errors and Controls with 0, 1 and >1 observed recombinant on chromosome 21. Error bars are standard errors of the mean.

Figure 1.

Average genome-wide recombination counts for MI errors and controls. Average genome-wide recombination counts of the complimentary autosomes for MI errors and Controls with 0, 1 and >1 observed recombinant on chromosome 21. Error bars are standard errors of the mean.

We next compared the genome-wide recombination counts of the MI error sibling oocytes with that of the control group. As with MI error probands, we found that there was a significant reduction in the mean genome-wide recombination count among MI error siblings compared with that of the control group [(38.5 (95% CI 36.5–40.5) versus 42.6 (95% CI 42.3–43.0) (P < 0.0001)), Table 1]. Upon examination of the mean genome-wide recombination counts stratified by 0, 1 and >1 chromosome 21 recombinations among siblings, we found that all means were reduced when compared with that of the control group (Fig. 2). However, unlike the MI error probands, the mean genome-wide recombination count increased with increasing chromosome 21 recombination, suggesting that oocyte-level regulation was not perturbed for these samples.

Figure 2.

Average genome-wide recombination counts for MI error siblings and controls. Average genome-wide recombination counts of the complimentary autosomes for MI error siblings and Controls with 0, 1 and >1 observed recombinants on chromosome 21. Error bars are standard errors of the mean.

Figure 2.

Average genome-wide recombination counts for MI error siblings and controls. Average genome-wide recombination counts of the complimentary autosomes for MI error siblings and Controls with 0, 1 and >1 observed recombinants on chromosome 21. Error bars are standard errors of the mean.

We next used linear regression to quantify oocyte-level regulation of recombination across all autosomes in the genome and to determine the extent of the dysregulation observed for the MI error group. We used linear regression to determine the relationships between the number of recombinants on an autosome (predictor variable) and the number of recombinants in the rest of the genome (outcome variable) (Table 2). First, we found a highly statistically significant correlation between the genome-wide recombination counts and the observed number of chromosome 21 recombinants among controls (beta coefficient = 2.90, standard error = 0.27, P < 0.001) but not MI errors (beta coefficient = 0.14, standard error = 1.10, P = 0.30), which reflects the pattern of means discussed above. Upon examination of the other autosomes, we found that there was evidence for oocyte-level regulation of recombination count in the control group. As the number of recombinants on each individual chromosome increased by 1, the genome-wide recombination count increased by 1.44 to 2.06 recombinants, as indicated by the beta coefficients (P < 0.001) (Table 2). As for the MI error group, oocyte-level regulation of recombination count was perturbed as indicated by the lower beta coefficients (−0.07 to 2.29 for MI errors versus 1.44–2.06 for controls). The smaller sample size among MI errors led to a larger range in the beta coefficients among the chromosomes (−0.07 to 2.29 for MI errors versus 1.44–2.06 for controls), but does not affect the mean beta (Table 2). The mean beta coefficient which includes the beta coefficient for chromosome 21 was reduced compared with normal outcomes (1.23 versus 1.75, Fig. 3).

Table 2.

Relationship between the number of recombinants on a specified autosome and the sum of all other autosomal recombinants

Chr MI (N = 94) MII (N = 20) MI Siblings (N = 64) Controls (N = 2723) 
Beta coefficient, SE, P-value Beta coefficient, SE, P-value Beta coefficient, SE, P-value Beta coefficient, SE, P-value 
2.29, 0.44, <0.001* 2.32, 0.80, 0.009* 1.06, 0.63, 0.10 1.78, 0.09, <0.0001* 
0.51, 0.53, 0.34 0.11, 1.15, 0.92 1.25, 0.74, 0.10 1.83, 0.10, <0.0001* 
0.65, 0.65, 0.31 0.08 1.07, 0.94 1.16, 0.62, 0.07 1.73, 0.11, <0.0001* 
1.13, 0.53, 0.04* −1.64, 1.60, 0.32 1.35, 0.72, 0.07 1.86, 0.11, <0.0001* 
0.47, 0.70, 0.50 3.26, 1.06, 0.01* 2.48, 0.77, 0.002* 1.88, 0.12, <0.0001* 
1.88, 0.54, <0.001* 2.48, 0.83, 0.01* 2.10, 0.73, 0.006* 1.97, 0.13, <0.0001* 
2.00, 0.54, <0.001* 0.39, 1.59, 0.81 2.03, 0.93, 0.03* 1.85, 0.13, <0.0001* 
1.36, 0.67, 0.05* 1.48, 1.39, 0.30 1.27, 0.93, 0.18 1.57, 0.12, <0.0001* 
1.34, 0.63, 0.04* 1.67, 1.95, 0.40 1.22, 0.79, 0.13 1.88, 0.14, <0.0001* 
10 1.88, 0.63, 0.004* 3.12, 1.77, 0.10 0.87, 0.96, 0.37 1.74, 0.13, <0.0001* 
11 1.74, 0.66, 0.01* 2.12, 1.72, 0.24 2.02, 0.87, 0.02* 1.68, 0.14, <0.0001* 
12 0.43, 0.72, 0.56 −1.34, 1.68, 0.44 1.67, 0.96, 0.09 1.95, 0.14, <0.0001* 
13 0.98, 0.92, 0.29 1.53, 2.44, 0.54 2.35, 1.08, 0.03* 1.44, 0.17, <0.0001* 
14 0.83, 0.89, 0.36 3.41, 1.73, 0.07 1.80, 1.17, 0.13 1.75, 0.17, <0.0001* 
15 0.87, 0.87, 0.32 −0.83, 2.56, 0.75 2.41, 1.02, 0.02* 1.57, 0.17, <0.0001* 
16 1.07, 0.77, 0.17 3.41, 1.74, 0.07 2.79, 1.01, 0.008* 1.62, 0.15, <0.0001* 
17 2.00, 0.71, 0.006* 4.66, 1.11, <0.001* 3.35, 0.88, <0.0001* 1.34, 0.17, <0.0001* 
18 1.32, 0.93, 0.16 2.44, 1.85, 0.20 1.43, 1.15, 0.22 1.84, 0.17, <0.0001* 
19 −0.07, 1.06, 0.95 1.58, 2.02, 0.45 0.58, 1.19, 0.63 1.62, 0.19, <0.0001* 
20 1.89, 0.99, 0.06 −2.63, 2.72, 0.35 2.46, 1.38, 0.08 2.06, 0.18, <0.0001* 
21 1.14, 1.10, 0.30 2.70, 3.32, 0.43 0.58, 1.70, 0.73 1.98, 0.24, <0.0001* 
22 1.40, 1.13, 0.22 3.70, 2.69, 0.19 1.49, 1.46, 0.31 1.58, 0.23, <0.0001* 
Chr MI (N = 94) MII (N = 20) MI Siblings (N = 64) Controls (N = 2723) 
Beta coefficient, SE, P-value Beta coefficient, SE, P-value Beta coefficient, SE, P-value Beta coefficient, SE, P-value 
2.29, 0.44, <0.001* 2.32, 0.80, 0.009* 1.06, 0.63, 0.10 1.78, 0.09, <0.0001* 
0.51, 0.53, 0.34 0.11, 1.15, 0.92 1.25, 0.74, 0.10 1.83, 0.10, <0.0001* 
0.65, 0.65, 0.31 0.08 1.07, 0.94 1.16, 0.62, 0.07 1.73, 0.11, <0.0001* 
1.13, 0.53, 0.04* −1.64, 1.60, 0.32 1.35, 0.72, 0.07 1.86, 0.11, <0.0001* 
0.47, 0.70, 0.50 3.26, 1.06, 0.01* 2.48, 0.77, 0.002* 1.88, 0.12, <0.0001* 
1.88, 0.54, <0.001* 2.48, 0.83, 0.01* 2.10, 0.73, 0.006* 1.97, 0.13, <0.0001* 
2.00, 0.54, <0.001* 0.39, 1.59, 0.81 2.03, 0.93, 0.03* 1.85, 0.13, <0.0001* 
1.36, 0.67, 0.05* 1.48, 1.39, 0.30 1.27, 0.93, 0.18 1.57, 0.12, <0.0001* 
1.34, 0.63, 0.04* 1.67, 1.95, 0.40 1.22, 0.79, 0.13 1.88, 0.14, <0.0001* 
10 1.88, 0.63, 0.004* 3.12, 1.77, 0.10 0.87, 0.96, 0.37 1.74, 0.13, <0.0001* 
11 1.74, 0.66, 0.01* 2.12, 1.72, 0.24 2.02, 0.87, 0.02* 1.68, 0.14, <0.0001* 
12 0.43, 0.72, 0.56 −1.34, 1.68, 0.44 1.67, 0.96, 0.09 1.95, 0.14, <0.0001* 
13 0.98, 0.92, 0.29 1.53, 2.44, 0.54 2.35, 1.08, 0.03* 1.44, 0.17, <0.0001* 
14 0.83, 0.89, 0.36 3.41, 1.73, 0.07 1.80, 1.17, 0.13 1.75, 0.17, <0.0001* 
15 0.87, 0.87, 0.32 −0.83, 2.56, 0.75 2.41, 1.02, 0.02* 1.57, 0.17, <0.0001* 
16 1.07, 0.77, 0.17 3.41, 1.74, 0.07 2.79, 1.01, 0.008* 1.62, 0.15, <0.0001* 
17 2.00, 0.71, 0.006* 4.66, 1.11, <0.001* 3.35, 0.88, <0.0001* 1.34, 0.17, <0.0001* 
18 1.32, 0.93, 0.16 2.44, 1.85, 0.20 1.43, 1.15, 0.22 1.84, 0.17, <0.0001* 
19 −0.07, 1.06, 0.95 1.58, 2.02, 0.45 0.58, 1.19, 0.63 1.62, 0.19, <0.0001* 
20 1.89, 0.99, 0.06 −2.63, 2.72, 0.35 2.46, 1.38, 0.08 2.06, 0.18, <0.0001* 
21 1.14, 1.10, 0.30 2.70, 3.32, 0.43 0.58, 1.70, 0.73 1.98, 0.24, <0.0001* 
22 1.40, 1.13, 0.22 3.70, 2.69, 0.19 1.49, 1.46, 0.31 1.58, 0.23, <0.0001* 

The beta coefficient, standard error and P-values of each autosomal chromosome for MI and MII errors, MI error siblings and controls.

*Significant P-values are marked with an asterisk.

Figure 3.

Average beta coefficients of autosomal chromosomes. Average beta coefficient of autosomal chromosomes for MI (square) and MII errors (diamond), MI siblings (triangle) and Controls (circle). Error bars are standard errors of the mean.

Figure 3.

Average beta coefficients of autosomal chromosomes. Average beta coefficient of autosomal chromosomes for MI (square) and MII errors (diamond), MI siblings (triangle) and Controls (circle). Error bars are standard errors of the mean.

We next assessed siblings of probands with an MI error to determine whether their autosomal beta coefficients were more similar to that of MI error probands or the control group. Despite the small sample size (N = 64), we found that the mean beta coefficient for MI siblings was more similar to that of the control group than to MI error probands: 1.72 versus 1.75 for controls and 1.23 for MI error probands (Fig. 3). However, the range of beta coefficients was large (0.58–3.35, Table 2), most likely due to the small sample size. This shows that our method for quantifying oocyte-level regulation of recombination was successful even when used for a reduced sample size and provides evidence that the effects seen in the MI error probands are not due to low power.

We next assessed the regulation of recombination at the maternal level, or the trait that results in recombinant counts that are more similar among oocytes from the same mother when compared with those from other mothers. We conducted an ANOVA using the mothers with MI errors (N = 38) as the group variable (Table 3). We found no evidence for maternal regulation of genome-wide recombination (P = 0.12, graphically shown in Fig. 4).

Table 3.

Number of MI error siblings.

Number of siblings Number of families 
21 
13 
Number of siblings Number of families 
21 
13 

Number of families with a given sibling count.

Figure 4.

MI error proband and sibling genome-wide recombination counts. The genome-wide recombination counts for MI error probands (diamonds) and their siblings (circles) ordered by increasing GWR counts of probands. The P-value represents the result of the analysis of variance between the GWR counts of offspring.

Figure 4.

MI error proband and sibling genome-wide recombination counts. The genome-wide recombination counts for MI error probands (diamonds) and their siblings (circles) ordered by increasing GWR counts of probands. The P-value represents the result of the analysis of variance between the GWR counts of offspring.

Meiosis II nondisjunction

In the preliminary analysis performed by Brown et al., only MI errors were compared with controls to test for differences in genome-wide recombination rates (19). MII errors have been shown to have an elevated number of recombinants as well as an increased rate of pericentromeric recombinants on chromosome 21q when compared with the MI error group or the control group (24). Here, we asked whether these patterns extended to genome-wide recombination patterns. Although the following studies are limited by the number of MII errors in our sample (n = 20), they provide, for the first time, preliminary results on global regulation of recombination count in oocytes with this type of error.

We first compared the genome-wide recombination rates of the MII error group with that of the normal meiotic error group. As we had to eliminate MII errors with no observed recombinants on chromosome 21 (those that likely represent post- zygotic errors in mitosis), we restricted our comparison with the samples in the control group with at least one observed recombinant on chromosome 21 (N = 1712). The MII group had a mean genome-wide recombination count of 40.45 (95% CI 36.61–44.28) and the control group had a mean genome-wide recombination count of 43.35 (95% CI 42.84–43.85) (Table 1). These values were not significantly different (t-test = 0.11, α > 0.05).

Next, we performed linear regression to quantify oocyte-level regulation of recombinant count using chromosome 21 as the predictor and genome-wide recombination count as the outcome variable. Although we found that the beta coefficient was higher than that of the control group (2.70 versus 1.98), it was not a statistically significant predictor because of the small sample size (standard error = 3.32, P = 0.43) (Table 2). As in our MI cases, we then assessed the beta coefficients of the other autosomes to examine the correlation between individual chromosome recombination counts and the total number of recombinants for the rest of the autosomes. As in the MI error group, the beta coefficients of the chromosomes varied widely because of the small sample size, ranging from −2.63 to 4.65 (Table 2). However, the average beta coefficient (1.54) was closer to what was found in the control group (1.75) than the MI error group (1.23) (Fig. 3). There were not enough siblings of probands to examine maternal regulation of recombination among the MII errors.

Genome-wide recombination location

In order to determine whether there is oocyte-level regulation of recombination location for MI and MII errors, we assessed the relationship between the presence or absence of a proximal or single distal recombination event on chromosome 21 (predictor) and the proportion of genome-wide recombination in the proximal or distal 20% region of chromosome arms (outcome variable). We included genome-wide recombination count as a covariate in the model to correct for the potential effect of interference (25). That is, an increased number of multiple recombinants per arm due to an overall increase in genome-wide recombination counts could lead to recombinants localized nearer to the ends of chromosome arms. Among controls, we found that the presence of a single distal chromosome 21 recombinant or the presence of a proximal chromosome 21 recombinant was significantly correlated with a greater proportion of distal (Table 4) or proximal (Table 5) genome-wide recombination events, respectively. There was no evidence that the location of recombination on chromosome 21 predicts recombination location genome-wide for these nondisjunction error groups (Tables 4 and 5). These findings are inconclusive because we had a reduced sample size for this test compared with other tests performed in this study (we could only include samples that had at least one recombinant).

Table 4.

Relationship between the distal recombination on chr21 and the proportion of genome-wide distal recombinants

Predictor MII (N = 20)
 
MI (N = 41)
 
Normal (N = 1679)
 
Beta coefficient P-value Beta coefficient P-value Beta coefficient P-value 
Genome-wide recombination count −0.000415 0.8139 0.00193 0.2102 0.00041071 0.0018* 
Distal chr 21 recombinant (y or n) −0.01414 0.7692 −0.00132 0.9477 0.03428 <0.0001* 
Predictor MII (N = 20)
 
MI (N = 41)
 
Normal (N = 1679)
 
Beta coefficient P-value Beta coefficient P-value Beta coefficient P-value 
Genome-wide recombination count −0.000415 0.8139 0.00193 0.2102 0.00041071 0.0018* 
Distal chr 21 recombinant (y or n) −0.01414 0.7692 −0.00132 0.9477 0.03428 <0.0001* 

The beta coefficients for the presence or absence of recombination in the distal portion of chromosome 21 (predictor) and the proportion of all other autosomes with a distal recombinant (outcome) for MI and MII errors and controls. Genome-wide recombination count was included as a variable to adjust for its effects on distal recombination.

*Significant P-values are marked with an asterisk.

Table 5.

Relationship between the proximal recombination on chr21 and the proportion of genome-wide proximal recombinants

Predictor MI (N = 41)
 
MII (N = 20)
 
Normal (N = 1679)
 
Beta coefficient P-value Beta coefficient P-value Beta coefficient P-value 
Genome-wide recombination count −0.00049 0.74 0.00079 0.52 0.00031 <0.0001* 
Proximal chr 21 recombinant (y or n) −0.00136 0.96 −0.03544 0.08 0.02612 <0.0001* 
Predictor MI (N = 41)
 
MII (N = 20)
 
Normal (N = 1679)
 
Beta coefficient P-value Beta coefficient P-value Beta coefficient P-value 
Genome-wide recombination count −0.00049 0.74 0.00079 0.52 0.00031 <0.0001* 
Proximal chr 21 recombinant (y or n) −0.00136 0.96 −0.03544 0.08 0.02612 <0.0001* 

The beta coefficient for the presence or absence of recombination in the proximal portion of chromosome 21 (predictor) and the proportion of all other autosomes with a proximal recombinant (outcome) for MI and MII errors and controls. Genome-wide recombination count was included as a variable to adjust for its effects on proximal recombination.

*Significant P-values are marked with an asterisk.

DISCUSSION

Altered recombination patterns have been identified as a risk factor for nondisjunction (11,12,24). As the number and placement of recombination on chromosome 21 are major determinants of proper chromosome segregation, it is important to identify factors that regulate recombination on chromosome 21. Our previous study showed that reduced number of exchanges on chromosome 21 was associated with reduced recombination genome-wide (19). This suggested that variation in factors that act globally (trans-acting factors) may be involved in controlling recombination counts on chromosome 21 and elsewhere.

Here, in our expanded study, we found similar results to Brown et al. (19): MI errors with zero observed chromosome 21 recombinants had significantly reduced genome-wide recombination counts compared with that of controls (Table 1). Upon examination of oocyte-level control of recombination number, we found that the recombination counts among the autosomes of MI error oocytes were less correlated than controls. This analysis showed that as the number of recombinants on each individual chromosome increased by 1, the genome-wide recombination count increased on average by 1.23 recombinants for MI versus 1.75 for controls, as indicated by the beta coefficient (Fig. 3). Thus, this reduction in or lack of oocyte-level regulation of global recombination count indicates that recombination may be dysregulated genome-wide in at least a subset of MI error oocytes.

Siblings of MI error probands provide additional information on the source of the altered recombination patterns observed for MI error oocytes. Siblings of MI error probands showed a similar overall reduction in genome-wide recombination counts as did the MI error probands (Table 1). However, siblings showed a similar correlation among chromosomes as did the control group (1.72 versus 1.75, respectively) (Fig. 3). Thus, they showed evidence for normal oocyte-level regulation of recombination count. Based on these observations, we suggest that a two-step model must be invoked to explain the dysregulation of genome-wide recombination count observed in the MI error oocytes. The first step predisposes a woman's oocytes to reduced genome-wide recombination, perhaps due to a variant in a gene that controls recombination number or to an environment exposure during the woman's fetal life when recombination occurs. This would suggest that recombination regulation at the maternal level contributes to reduced recombination in MI error oocytes. Yet when we performed an ANOVA, we did not detect this maternal-level regulation of recombination among sibships in the MI error group. There are at least two explanations for this lack of evidence. First, the maternal regulation may exist, but our data set was too small to detect it (n = 38 mothers, typically with the proband and one additional offspring). This phenomenon among sibships with normal meiotic outcomes was previously reported among families with large numbers of offspring (14,16). Hence, we cannot rule out maternal-level regulation of recombination count as a contributor to MI error. Alternatively, maternal regulation may not be observed in our dataset due to the inclusion of probands in the analysis, or those oocytes for which we speculate have dysregulated recombination. Nonetheless, reduced recombination is observed for both MI error oocytes and their siblings, yet only MI error oocytes exhibit a lack of oocyte-level recombination count regulation. Hence, we speculate that this may be due to the second step which results from an oocyte-specific factor that leads to dysregulation of recombination in that single oocyte. This factor is unlikely to be environmental as all gametes would be affected. Hence, it is more probable that a de novo mutation has occurred in the MI error oocyte.

With regard to the maternal regulation of recombination, studies in model systems have previously identified genetic variants that result in reduced genome-wide recombination counts across offspring and, in some cases, concomitant nondisjunction of susceptible chromosomes. For instance, the Mei-S282 mutation in female Drosophila melanogaster leads to a decrease in global recombination among gametes and an increased rate of chromosomal nondisjunction during MI (26). Furthermore, nondisjunction events within a gamete were associated with an increase in achiasmate heterologous chromosomes within that same gamete (26). Similar phenotypes have been found in Caenorhabditis elegans that are him-6/him-4 null in which genome-wide recombination was found to be decreased and nondisjunction rates increased in these mutants (27). Taken together, these results indicate that a gene that acts in maternal regulation of recombination count can lead to reduced recombination among sibling-oocytes. Additionally, this gene can lead to nondisjunction in a subset of oocytes, much like what was seen in this study.

Genetic variants that act in maternal regulation of genome-wide recombination counts have also been identified in human genes. RNF212 is an ortholog of the Caenorhabiditis elegan gene Zip-3 and is required for crossover recombination in this organism. Several genome-wide association studies have shown that variants in RNF212 are associated with increased recombination number in the sperm of human males (15,18,28) with two of these studies finding an association in females (18,28). Most recently, a molecular study of RNF212 in mice found that this protein is essential for crossing over and works by stabilizing recombination proteins (29). A heterozygous mutation in this gene leads to reduced recombination (29). Hence, this gene is believed to work in a concentration-dependent manner where mutations that lead to haploinsufficiency cause a reduction in genome-wide recombination. The effects of RNF212 on the crossover number of gametes may be influenced by an interaction between the size of a chromosome and protein concentration (there may be a lower probability of protein binding along the smaller chromosomes when compared with larger chromosomes in low protein concentrations). Also, a common inversion on chromosome 17q21.31 has consistently been shown to be correlated with elevated recombination counts in females (15,20,28). This inversion carries two haplotypes (named H1 and H2) (30). These haplotypes have differing gene expression profiles (31). Our data show that chromosome 17 is a strong predictor of genome-wide recombination count, as it is highly significant even in the MII group (N = 20) (Table 2). Potentially, the presence of the inversion on chromosome 17 effects recombination in two ways: by influencing the number of recombinants on chromosome 17 when acting in cis and by influencing the genome-wide recombination count by acting in trans (haplotype variation causing differential expression of genes that influence recombination).

In summary, we have confirmed that oocytes with MI errors as well as their sibling oocytes have reduced recombination when compared with control oocytes and seems to be the result of maternal regulation of global recombination count. However, dysregulation of recombination at the oocyte-level seems to distinguish MI error oocytes from their siblings. In future studies, it will be important to identify global factors that influence maternal regulation of recombination as this factor may predispose oocytes to reduced global recombination. RNF212 and region 17q21.31 may serve as the first candidate genes for genotyping in mothers of MI error probands. Of greater importance would be the identification of factors that alter oocyte-level regulation of genome-wide recombination as they would provide insight into the recombination-associated mechanism that increases susceptibility for nondisjunction of human chromosomes.

MATERIALS AND METHODS

Ethics statement

All recruitment sites obtained the necessary Institutional Review Board approvals from their institutions.

Sample sets

Trisomy 21 sample Set

Families of probands with standard trisomy 21 were recruited by various sites using a common protocol, with the goal to identify risk factors of nondisjunction. Blood and/or buccal samples were collected from probands and their families for genetic studies. The minimal family unit required for this analysis included the proband with trisomy 21, parents and maternal grandparents (five-membered family). In some cases, we were able to ascertain siblings of the proband. Genome-wide genotyping was performed at the Center for Inherited Disease Research (CIDR) at Johns Hopkins University using the Golden Gate linkage panel. This panel consists of 6056 SNP markers tiled across the genome and spaced on average 0.63 cM apart. CIDR assessed several metrics of data quality in order to exclude low quality data. As a result of their quality control assessment, 358 markers were dropped due to low genotyping rates (<0.98) and atypical intensity plot cluster patterns (forming greater than two clusters). This resulted in a final marker count of 5698 SNPs. The remaining markers had a mean genotyping call rate of 0.9988 (min = 0.9805, std = 0.0021). At the person level, there were 896 samples sent for genotyping and 57 were dropped. These samples either failed genotyping because they were of poor quality or were removed as a result of having genotyping call rates that fell within the 5% area of the lower tail of a normal distribution. Following genotyping quality control measures, families were removed if probands were not found to have an MI or MII error. Nine families were removed due to the following: four were found to be paternal in origin, one due to failure to identify stage or origin and four were likely to have resulted from post-zygotic mitosis. The final sample set contained 114 families of which 94 had probands with MI errors and 20 had probands with MII errors (Table 6).

Table 6.

Study populations

Meiotic outcome group Number of observed recombinants on chromosome 21
 
Total 
>1 
Meiosis I error 56 28 10 94 
MI siblings 23 36 64 
Meiosis II error n/a 15 20 
Controls 1044 1319 360 2723 
Meiotic outcome group Number of observed recombinants on chromosome 21
 
Total 
>1 
Meiosis I error 56 28 10 94 
MI siblings 23 36 64 
Meiosis II error n/a 15 20 
Controls 1044 1319 360 2723 

Study populations stratified by meiotic outcome group and number of recombinants on chromosome 21.

Control sample set

Recombination profiles from controls (normal meiotic outcome group) were obtained from 2762 families that were genotyped through the following three studies: the Framingham Study (FHS), GENEVA Dental Caries Study (GENEVA) or Autism Genetic Resource Exchange (AGRE) study groups. These families were genotyped for at least 500 000 SNPs across the genome and assessed for uniformity in recombination distribution. The AGRE samples were genotyped for 520 017 SNPs tiled across the genome using the Infinium(R) HumanHap550-Duo Bead Chip. However, 11 473 markers were excluded from the analysis due to deviation from Hardy-Weinberg Equilibrium (HWE) (P < 10−7). After completion of quality control measures, the AGRE dataset contained genotype information for 512 207 markers across the genome. The FHS samples were genotyped for 500 568 markers across the genome using the Affymetrix Genome-Wide Human SNP Array 5.0. However, 22 000 markers were excluded from the analysis due to deviation from HWE (P < 10−7). After quality control measures were complete, there was genotype information for 478 658 markers across the genome for the FHS dataset. The GENEVA samples were genotyped for 620 901 markers using the Illumina 610-Quad Array. There were 58 610 markers that were excluded from the analysis due to deviation from HWE (P < 10−5) and an MAF < 0.02. After quality control measures were complete, there was genotype information for 562 291 markers across the genome for the GENEVA population. The final dataset contained 2723 control samples (Table 6). All SNP locations were based on human NCBI Build 36 (hg18).

Determination of the nondisjunction error type

The origin of a nondisjunction event was categorized by both the parent from whom the extra chromosome originated and meiotic stage of origin as described previously (9,10). Briefly, to determine which parent contributed the disomic gamete, we assessed the chromosome 21 allelic contribution from each parent to the child. We included only those cases found to be of maternal origin. The meiotic stage (MI or MII) was inferred using pericentromeric markers. For the most proximal pericentromeric marker heterozygous in the mother, allelic heterozygosity (nonreduction) in the proband led to the classification of an MI error. Allelic homozygosity (reduction) in the proband led to the classification of an MII error. MII errors could result from classical MII errors where sister chromatids fail to separate at MII or to errors initiated in MI followed by abnormal segregation in MII. Additionally, when all markers were found to be reduced to homozygosity, indicating an MII error with no recombination on chromosome 21, the origin of nondisjunction was inferred to be a post-zygotic mitotic error. These cases were excluded from this study. However, this allelic configuration may also have been the result of an MII error in which no recombination had occurred.

Determining recombination phenotypes

Trisomy 21 sample set

For the trisomic dataset (probands), two methods were used to identify recombination location and number. Trisomic chromosomes 21 were analyzed separately because tri-allelic genotype data do not meet the assumptions used in available haplotyping software. For trisomic chromosomes 21, SNP and STR data were combined from our previous studies (32), and used to define the location of recombination events along 21q. STR data were weighted more heavily as this marker type exhibits less genotyping error. The breakpoints of a single recombinant event were defined by a minimum of either one STR or eight consecutive informative SNPs flanking the recombination breakpoint. However, when the most proximal or distal informative markers on 21q indicated a recombinant event, a minimum of either one STR or four consecutive informative SNPs were required to define recombination break points.

For the disomic chromosomes, we used our own method to call recombination events. Our method works on similar principles as that of Coop et al. (33) and Chowdhury et al. (15), except that it is tailored toward three-generation pedigrees also allowing for the use of informative marker density as a quality control measure for calling recombination break points. First, for each SNP the proband's maternal allele is assigned a grandparent-of-origin status (which grand-parent the allele is inherited from). The grandparent-of -origin statuses are then ordered by the SNPs' physical positions to demarcate segments along the chromosome that alternate between the two grandparents of the proband. The region between two consecutive segments, known as a recombination interval, represents the location of a recombination event. In our analysis, each recombination interval that was supported by flanking segments both containing at least two consecutive informative markers was scored as a separate recombination event.

Control sample set

For the AGRE, FHS and GENEVA datasets, genotype data from members of two-generation families with three or more children were used to infer the location and number of recombinants of autosomal chromosomes. The method developed by Coop et al. (33) and Chowdhury et al. (15) was used for these datasets. In this method, markers on the parental chromosomes are assessed for identity by descent allele sharing (IBD) between two siblings, and each sibling is assigned either the same haplotype phase or different haplotype phases depending on the IBD status. In the event where there was a switch from shared to unshared phase, a putative recombination event is noted. The haplotype phases of three or more siblings are then compared to identify which sibling inherited the recombinant chromosome. This method as well as the datasets are described in detail in Chowdhury et al. (15)

Statistical analyses

Genome-wide recombination count

We first performed t-tests to compare the mean number of genome-wide recombination counts between specific meiotic outcome groups defined by meiotic error (MI error, MI-sibling or MII error versus controls) and, in subsequent comparisons, by chromosome 21 recombination count. For the analysis of MI error probands or their siblings versus the controls, the chromosome 21 recombination groups were 0, 1 or >1 recombinant on chromosome 21. For the comparison between MII errors versus the controls, the chromosome recombination groups were 1, 2 or >2 recombinants on chromosome 21. We did not assess MII errors with 0 recombinants because these cases are most likely the result of a post-zygotic mitotic error. A significance level of 0.05 was used for this test as well as for all other tests in this study.

Linear regression models were then used to quantify oocyte-level recombination regulation by determining whether the number of recombinants on chromosome 21 (predictor) was correlated with the total number of autosomal recombinants less those that occurred on chromosome 21 (outcome). We stratified by meiotic outcome group (MI error, MI error siblings, MII error and control) as each group has a different ability to detect observed recombination along chromosome 21 and each is associated with a different chromosome 21 recombination profile. The latter may reflect distinctive mechanisms of recombination regulation. We adjusted the regression model for maternal age at the birth of the proband for the MI and MII datasets, as it is an important covariate for nondisjunction (9,34–36). However, we did not find maternal age to be a significant predictor of genome-wide recombination count and it was therefore omitted from the final model. Maternal age at birth was not available for the control group. However, current data suggest its effect on genome-wide recombination is very small (37). To further assess oocyte-level regulation of recombination count, we conducted the same regression model described above for each of the other autosomes. In these models, we used the number of recombinants on each autosomal chromosome as the predictor variable and the aggregate count of recombinants in the remaining autosomes as the outcome variable.

To assess genome-wide recombination regulation at the maternal-level, we assessed whether oocytes that have had a trisomy 21 nondisjunction event showed similarity in genome-wide recombination count with sibling-oocytes from the same mother. To test this, we performed an ANOVA of the genome-wide recombination counts of siblings from different mothers. We accounted for the effect of random sampling by using a random-effects ANOVA.

Genome-wide recombination location

To determine whether there was evidence for oocyte-level regulation of the location of recombination, each arm of each chromosome was split into three regions: proximal, medial and distal. The proximal and distal regions were defined as the most proximal 20% and distal 20% of a chromosome arm measured in physical distance. The same definition was used by Przeworski et al. (2011) (28). We then performed a linear regression analysis where our outcome measure was the proportion of proximal or distal recombinants genome-wide and our predictor was the presence or absence of a recombination in the proximal or distal region of chromosome 21. Again, we analyzed each meiotic outcome group separately.

FUNDING

This work was supported by the National Institutes of Health [1T32MH087977 (T.O.), R01 HD057029 and R01 HD38979 (S.S., E.F.), R01 HL083300, R01-DE 014899 and U01-DE018903 (M.M.)] and the Center for Inherited Disease Research [HHSN268200782096C (M.M.)] and the Children's Healthcare of Atlanta Cardiac Research Committee.

ACKNOWLEDGEMENTS

We thank our laboratory personnel, recruiters and the families who participated in this study. We thank Dr Mary Marazita (M.M) for contributing data on the location of recombination along properly segregating chromosomes.

Conflict of interest statement. None declared.

REFERENCES

1
Hassold
T.
Hunt
P.
To err (meiotically) is human: the genesis of human aneuploidy
Nat. Rev. Genet.
 , 
2001
, vol. 
2
 (pg. 
280
-
291
)
2
Hassold
T.
Abruzzo
M.
Adkins
K.
Griffin
D.
Merrill
M.
Millie
E.
Saker
D.
Shen
J.
Zaragoza
M.
Human aneuploidy: incidence, origin, and etiology
Environ. Mol. Mutagen.
 , 
1996
, vol. 
28
 (pg. 
167
-
175
)
3
Warburton
D.
Fraser
F.C.
Spontaneous abortion risks in Man: data from reproductive histories collected in a medical genetics unit
Am. J. Hum. Genet.
 , 
1964
, vol. 
16
 (pg. 
1
-
25
)
4
Angell
R.R.
Sandison
A.
Bain
A.D.
Chromosome variation in perinatal mortality: a survey of 500 cases
J. Med. Genet.
 , 
1984
, vol. 
21
 (pg. 
39
-
44
)
5
Hassold
T.J.
Nondisjunction in the human male
Curr. Top. Dev. Biol.
 , 
1998
, vol. 
37
 (pg. 
383
-
406
)
6
Jacobs
P.A.
The chromosome complement of human gametes
Oxf. Rev. Reprod. Biol.
 , 
1992
, vol. 
14
 (pg. 
47
-
72
)
7
Marquez
C.
Cohen
J.
Munne
S.
Chromosome identification in human oocytes and polar bodies by spectral karyotyping
Cytogenet. Cell Genet.
 , 
1998
, vol. 
81
 (pg. 
254
-
258
)
8
Parker
S.E.
Mai
C.T.
Canfield
M.A.
Rickard
R.
Wang
Y.
Meyer
R.E.
Anderson
P.
Mason
C.A.
Collins
J.S.
Kirby
R.S.
, et al.  . 
Updated National Birth Prevalence estimates for selected birth defects in the United States, 2004–2006
Birth Defects Res. A. Clin. Mol. Teratol.
 , 
2010
, vol. 
88
 (pg. 
1008
-
1016
)
9
Allen
E.G.
Freeman
S.B.
Druschel
C.
Hobbs
C.A.
O'Leary
L.A.
Romitti
P.A.
Royle
M.H.
Torfs
C.P.
Sherman
S.L.
Maternal age and risk for trisomy 21 assessed by the origin of chromosome nondisjunction: a report from the Atlanta and National Down Syndrome Projects
Hum. Genet.
 , 
2009
, vol. 
125
 (pg. 
41
-
52
)
10
Freeman
S.B.
Allen
E.G.
Oxford-Wright
C.L.
Tinker
S.W.
Druschel
C.
Hobbs
C.A.
O'Leary
L.A.
Romitti
P.A.
Royle
M.H.
Torfs
C.P.
, et al.  . 
The National Down Syndrome Project: design and implementation
Public Health Rep.
 , 
2007
, vol. 
122
 (pg. 
62
-
72
)
11
Lamb
N.E.
Feingold
E.
Savage
A.
Avramopoulos
D.
Freeman
S.
Gu
Y.
Hallberg
A.
Hersey
J.
Karadima
G.
Pettay
D.
, et al.  . 
Characterization of susceptible chiasma configurations that increase the risk for maternal nondisjunction of chromosome 21
Hum. Mol. Genet.
 , 
1997
, vol. 
6
 (pg. 
1391
-
1399
)
12
Warren
A.C.
Chakravarti
A.
Wong
C.
Slaugenhaupt
S.A.
Halloran
S.L.
Watkins
P.C.
Metaxotou
C.
Antonarakis
S.E.
Evidence for reduced recombination on the nondisjoined chromosomes 21 in Down syndrome
Science
 , 
1987
, vol. 
237
 (pg. 
652
-
654
)
13
Oliver
T.R.
Feingold
E.
Yu
K.
Cheung
V.
Tinker
S.
Yadav-Shah
M.
Masse
N.
Sherman
S.L.
New insights into human nondisjunction of chromosome 21 in oocytes
PLoS Genet.
 , 
2008
, vol. 
4
 pg. 
e1000033
 
14
Cheung
V.G.
Burdick
J.T.
Hirschmann
D.
Morley
M.
Polymorphic variation in human meiotic recombination
Am. J. Hum. Genet.
 , 
2007
, vol. 
80
 (pg. 
526
-
530
)
15
Chowdhury
R.
Bois
P.R.
Feingold
E.
Sherman
S.L.
Cheung
V.G.
Genetic analysis of variation in human meiotic recombination
PLoS Genet.
 , 
2009
, vol. 
5
 pg. 
e1000648
 
16
Kong
A.
Gudbjartsson
D.F.
Sainz
J.
Jonsdottir
G.M.
Gudjonsson
S.A.
Richardsson
B.
Sigurdardottir
S.
Barnard
J.
Hallbeck
B.
Masson
G.
, et al.  . 
A high-resolution recombination map of the human genome
Nat. Genet.
 , 
2002
, vol. 
31
 (pg. 
241
-
247
)
17
Kong
A.
Thorleifsson
G.
Gudbjartsson
D.F.
Masson
G.
Sigurdsson
A.
Jonasdottir
A.
Walters
G.B.
Gylfason
A.
Kristinsson
K.T.
Gudjonsson
S.A.
, et al.  . 
Fine-scale recombination rate differences between sexes, populations and individuals
Nature
 , 
2010
, vol. 
467
 (pg. 
1099
-
1103
)
18
Kong
A.
Thorleifsson
G.
Stefansson
H.
Masson
G.
Helgason
A.
Gudbjartsson
D.F.
Jonsdottir
G.M.
Gudjonsson
S.A.
Sverrisson
S.
Thorlacius
T.
, et al.  . 
Sequence variants in the RNF212 gene associate with genome-wide recombination rate
Science
 , 
2008
, vol. 
319
 (pg. 
1398
-
1401
)
19
Brown
A.S.
Feingold
E.
Broman
K.W.
Sherman
S.L.
Genome-wide variation in recombination in female meiosis: a risk factor for non-disjunction of chromosome 21
Hum. Mol. Genet.
 , 
2000
, vol. 
9
 (pg. 
515
-
523
)
20
Stefansson
H.
Helgason
A.
Thorleifsson
G.
Steinthorsdottir
V.
Masson
G.
Barnard
J.
Baker
A.
Jonasdottir
A.
Ingason
A.
Gudnadottir
V.G.
, et al.  . 
A common inversion under selection in Europeans
Nat. Genet.
 , 
2005
, vol. 
37
 (pg. 
129
-
137
)
21
Baudat
F.
Buard
J.
Grey
C.
Fledel-Alon
A.
Ober
C.
Przeworski
M.
Coop
G.
de Massy
B.
PRDM9 is a major determinant of meiotic recombination hotspots in humans and mice
Science
 , 
2010
, vol. 
327
 (pg. 
836
-
840
)
22
Berg
I.L.
Neumann
R.
Lam
K.W.G.
Sarbajna
S.
Odenthal-Hesse
L.
May
C.A.
Jeffreys
A.J.
PRDM9 Variation strongly influences recombination hot-spot activity and meiotic instability in humans
Nat. Genet.
 , 
2010
, vol. 
42
 pg. 
859
  
+
23
Berg
I.L.
Neumann
R.
Sarbajna
S.
Odenthal-Hesse
L.
Butler
N.J.
Jeffreys
A.J.
Variants of the protein PRDM9 differentially regulate a set of human meiotic recombination hotspots highly active in African populations
Proc. Natl Acad. Sci. USA
 , 
2011
, vol. 
108
 (pg. 
12378
-
12383
)
24
Lamb
N.E.
Freeman
S.B.
Savage-Austin
A.
Pettay
D.
Taft
L.
Hersey
J.
Gu
Y.
Shen
J.
Saker
D.
May
K.M.
, et al.  . 
Susceptible chiasmate configurations of chromosome 21 predispose to non-disjunction in both maternal meiosis I and meiosis II
Nat. Genet.
 , 
1996
, vol. 
14
 (pg. 
400
-
405
)
25
Weber
J.L.
Wang
Z.
Hansen
K.
Stephenson
M.
Kappel
C.
Salzman
S.
Wilkie
P.J.
Keats
B.
Dracopoli
N.C.
Brandriff
B.F.
, et al.  . 
Evidence for human meiotic recombination interference obtained through construction of a short tandem repeat-polymorphism linkage map of chromosome 19
Am. J. Hum. Genet.
 , 
1993
, vol. 
53
 (pg. 
1079
-
1095
)
26
Parry
D.M.
A meiotic mutant affecting recombination in female Drosophila melanogaster
Genetics
 , 
1973
, vol. 
73
 (pg. 
465
-
486
)
27
Zetka
M.C.
Rose
A.M.
Mutant rec-1 eliminates the meiotic pattern of crossing over in Caenorhabditis elegans
Genetics
 , 
1995
, vol. 
141
 (pg. 
1339
-
1349
)
28
Fledel-Alon
A.
Leffler
E.M.
Guan
Y.
Stephens
M.
Coop
G.
Przeworski
M.
Variation in human recombination rates and its genetic determinants
PLoS One
 , 
2011
, vol. 
6
 pg. 
e20321
 
29
Reynolds
A.
Qiao
H.
Yang
Y.
Chen
J.K.
Jackson
N.
Biswas
K.
Holloway
J.K.
Baudat
F.
de Massy
B.
Wang
J.
, et al.  . 
RNF212 Is a dosage-sensitive regulator of crossing-over during mammalian meiosis
Nat. Genet.
 , 
2013
, vol. 
45
 (pg. 
269
-
278
)
30
Pittman
A.M.
Myers
A.J.
Duckworth
J.
Bryden
L.
Hanson
M.
Abou-Sleiman
P.
Wood
N.W.
Hardy
J.
Lees
A.
de Silva
R.
The structure of the tau haplotype in controls and in progressive supranuclear palsy
Hum. Mol. Genet.
 , 
2004
, vol. 
13
 (pg. 
1267
-
1274
)
31
de Jong
S.
Chepelev
I.
Janson
E.
Strengman
E.
van den Berg
L.H.
Veldink
J.H.
Ophoff
R.A.
Common inversion polymorphism at 17q21.31 affects expression of multiple genes in tissue-specific manner
BMC Genomics
 , 
2012
, vol. 
13
 pg. 
458
 
32
Oliver
T.R.
Tinker
S.W.
Allen
E.G.
Hollis
N.
Locke
A.E.
Bean
L.J.
Chowdhury
R.
Begum
F.
Marazita
M.
Cheung
V.
, et al.  . 
Altered patterns of multiple recombinant events are associated with nondisjunction of chromosome 21
Hum. Genet.
 , 
2011
, vol. 
131
 (pg. 
1039
-
1046
)
33
Coop
G.
Wen
X.
Ober
C.
Pritchard
J.K.
Przeworski
M.
High-resolution mapping of crossovers reveals extensive variation in fine-scale recombination patterns among humans
Science
 , 
2008
, vol. 
319
 (pg. 
1395
-
1398
)
34
Yoon
P.W.
Freeman
S.B.
Sherman
S.L.
Taft
L.F.
Gu
Y.
Pettay
D.
Flanders
W.D.
Khoury
M.J.
Hassold
T.J.
Advanced maternal age and the risk of Down syndrome characterized by the meiotic stage of chromosomal error: a population-based study
Am. J. Hum. Genet.
 , 
1996
, vol. 
58
 (pg. 
628
-
633
)
35
Antonarakis
S.E.
Petersen
M.B.
McInnis
M.G.
Adelsberger
P.A.
Schinzel
A.A.
Binkert
F.
Pangalos
C.
Raoul
O.
Slaugenhaupt
S.A.
Hafez
M.
, et al.  . 
The meiotic stage of nondisjunction in trisomy 21: determination by using DNA polymorphisms
Am. J. Hum. Genet.
 , 
1992
, vol. 
50
 (pg. 
544
-
550
)
36
Muller
F.
Rebiffe
M.
Taillandier
A.
Oury
J.F.
Mornet
E.
Parental origin of the extra chromosome in prenatally diagnosed fetal trisomy 21
Hum. Genet.
 , 
2000
, vol. 
106
 (pg. 
340
-
344
)
37
Kong
A.
Barnard
J.
Gudbjartsson
D.F.
Thorleifsson
G.
Jonsdottir
G.
Sigurdardottir
S.
Richardsson
B.
Jonsdottir
J.
Thorgeirsson
T.
Frigge
M.L.
, et al.  . 
Recombination rate and reproductive success in humans
Nat. Genet.
 , 
2004
, vol. 
36
 (pg. 
1203
-
1206
)