Maternal Meiosis I Non-Disjunction of Chromosome 15: Dependence of the Maternal Age Effect on Level of Recombination

Download (0)

Full text


1998 Oxford University Press Human Molecular Genetics, 1998, Vol. 7, No. 6 1011–1019

Maternal meiosis I non-disjunction of chromosome

15: dependence of the maternal age effect on level of


W. P. Robinson


*, B. D. Kuchinka


, F. Bernasconi


, M. B. Petersen


, A. Schulze



K. Brøndum-Nielsen


, S. L. Christian


, D. H. Ledbetter


, A. A. Schinzel



B. Horsthemke


, S. Schuffenhauer


, R. C. Michaelis


, S. Langlois


and T. J. Hassold


1Department of Medical Genetics, University of British Columbia, Vancouver, Canada, 2Department of Medical

Genetics, The John F. Kennedy Institute, Glostrup, Denmark, 3Department of Human Genetics, University of Chicago, Chicago, IL, USA, 4Institut für Medizinische Genetik der Universität Zürich, Zürich, Switzerland, 5Institüt für

Humangenetik, Universitätsklinikum, Essen, Germany, 6Kinderpoliklinik der Universität Abteilung für Pädiatrische

Genetik, München, Germany, 7Greenwood Genetic Center, Greenwood, SC, USA and 8Department of Genetics and The Center for Human Genetics, Case Western Reserve University School of Medicine, Cleveland, OH, USA

Received January 12, 1998; Revised and Accepted March 9, 1998

Non-disjoined chromosomes 15 from 115 cases of uni-parental disomy (ascertained through Prader–Willi syndrome) and 13 cases of trisomy of maternal origin were densely typed for microsatellite loci spanning chromosome 15q. Of these 128 cases a total of 97 mei-osis I (MI) errors, 19 meimei-osis II (MII) errors and 12 mito-tic errors were identified. The genemito-tic length of a map created from the MI errors was 101 cM, as compared with a maternal length of 137 cM based on CEPH con-trols. No significant differences were detected in the distribution of recombination events along the chromosome arm and a reduction was seen for most of the chromosome 15 intervals examined. It was esti-mated that 21% of tetrads leading to MI non-disjunc-tion were achiasmate, which may account for most or all of the reduction in recombination noted. The mean age of mothers of cases involving MI errors which showed no transitions from heterodisomy to iso-disomy was significantly lower (32.7) than cases show-ing one or more observable transitions (36.3) (P < 0.003, t-test). However, even among chiasmate pairs the highest mean maternal age was seen for multiple exchange tetrads. Chromosome-specific differences in maternal age effects may be related to the normal distribution of exchanges (and their individual sus-ceptibilities) for each chromosome. However, they may also reflect the presence of multiple factors which act to ensure normal segregation, each affected by

ma-ternal age in a different way and varying in importance for each chromosome.


It is estimated that 15–20% of clinically recognized pregnancies are spontaneously aborted, with cytogenetic abnormalities ac-counting for a large proportion of these (1,2). The cause of this extraordinarily high rate of aneuploidy in the human species is not well understood. However, two factors which have been clearly associated with aneuploidy are advanced maternal age and altered meiotic recombination. Although generally increased, the effect of maternal age on the rate of non-disjunction is strongly chromosome dependent (1). Chromosome-specific differences are also apparent in studies of origin of the non-disjunction event. The majority of non-mosaic trisomies so far studied appear to be due to maternal meiosis I (MI) errors (for reviews, see refs 3,4). In contrast, trisomy 18 is mostly associated with maternal meiosis II (MII) errors (5) and paternal errors account for about half the cases of trisomy 2 (T.J. Hassold and W.P. Robinson, unpublished data), as well as XXY (6). Apparent post-zygotic duplication of one chromosome may also occur, but is much more commonly observed in mosaic than non-mosaic aneuploidy (7–10).

It is widely accepted that chromosome pairing and recombina-tion is important to ensure segregarecombina-tion of homologous chromo-somes to opposite poles at the end of MI (reviewed in ref. 11) and there is ample evidence in yeast, Caenorhabditis elegans and

Drosophila to show that chromosome gain or loss may be

intimately associated with failure of chromosomes to pair and recombine. In addition, not just absent but altered levels and positioning of recombinational events have been observed in MI

*To whom correspondence should be addressed at: BC Research Institute for Children’s and Women’s Health, Room 170, 950 West 28th Avenue, Vancouver, BC V5Z 4H4, Canada. Tel: +1 604 875 3229; Fax: +1 604 875 2496; Email:


non-disjunction in female Drosophila (12,13). A reduction in recombination associated with maternal MI non-disjunction in humans has been observed for trisomies 16, 18, 21 and XXX and XXY (5,6,14,15). For trisomies 16 and 21 recombination is primarily reduced near the centromere, with increased recombina-tion observed more distally along the chromosome arms. In contrast, increased recombination is observed in chromosome 21 MII non-disjunction, implying that what we observe as an ‘MII’ non-disjunction may be triggered by an event occurring at MI (16). Chromosome non-disjunction may also lead to uniparental disomy (UPD), whereby the chromosome number is normal but both homologs of a chromosome pair have originated from a single parent. The non-disjunction event leading to UPD15 is also predominantly due to a maternal MI segregation error and shows a maternal age effect similar to that observed for other trisomies involving acrocentric chromosomes (17). In a preliminary study of 24 cases of maternal MI-derived UPD15 we concluded that a reduction in recombination was associated with non-disjunction of chromosome 15, but that this could not be attributed simply to a proportion of cases in which recombination had not occurred (18). However, those analyses were limited by the small sample size and we were unable to assess the possible effect of maternal age on recombination frequencies. We have now extended our analyses to include detailed marker typing on 116 cases of UPD15 or trisomy 15, all attributed to a maternal meiotic non-disjunction (97 MI and 19 MII cases). The present report summarizes these studies and provides evidence of an association between maternal age and recombination levels in the non-disjunctional meioses.

RESULTS Interval analysis

The tetratype frequency and map distance (cM) for each interval is estimated from the number of observed exchanges (transitions from non-reduced to reduced markers or vice versa). Summing the estimated length for each interval, the genetic map from MI non-disjunction events is estimated as 101 cM, which is 26% shorter than the estimate of 137 cM from the CEPH controls (Fig. 1). For most intervals the MI genetic distance was shorter than for the CEPH controls, with the notable exception of the interval from D15S123 to D15S117. Interestingly, of eight transitions occurring between these two markers which could be further localized, five occurred between D15S123 and CYP19, a distance estimated as only 3 cM from the CEPH control haplotypes (CYP19 is listed as lying 1 cM proximal of D15S123 in the CEPH/Généthon on-line data, which is inconsistent with the present analysis of data from both UPD15/T15 cases and CEPH controls). Although there appears to be increased recombination in the most distal interval, most cases were not typed for D15S642 and this result is based on the observation of a single transition from only 21 informative meioses. Only the χ2 value for the interval from FES to D15S100 was significant between the UPD15/T15 and control groups (P < 0.05). However, this value is not considered significant when correcting for number of comparisons, nor was the overall χ2 significant.

Figure 1. A comparison of the genetic maps constructed from non-disjoined

chromosomes 15 of maternal MI errors (n = 97) and CEPH maternal haplotypes (n = 92).

Figure 2. Comparison of estimated distribution of exchange classes from

maternal MI errors and CEPH maternal haplotypes.

Observed and inferred meiotic exchanges

Another method to examine the effect of recombination on non-disjunction is to determine the distribution of the total number of exchange events per case (rather than averaging the amount of recombination among all cases). The distribution of maternal transitions identified in the UPD15/T15 families is given in Table 1. In 91 of the 97 MI cases markers were fully informative throughout the entire chromosome (i.e. informative for all or most of the 11 marker clusters indicated in Table 2, including the most distal one, and with no uninformative gaps >30 cM). Of these, 35 showed heterodisomy throughout the entire chromosome with no evidence of recombination. However, this does not mean that there was no meiotic recombination in these cases, only that the two recovered products from the original tetrad were not informative to show a recombination event. To take this effect of sampling into account we calculated the likely original distribution of meiotic configurations using the tables given previously (18). Standard errors were calculated using properties of a multinomial distribu-tion as exemplified elsewhere (19).



Nucleic Acids Research, 1994, Vol. 22, No. 1

Human Molecular Genetics, 1998, Vol. 7, No. 6 1013

Table 1. Observed distribution of transition classes in meiotic errors and the most likely underlying distribution of exchange

Observed exchange Inferred distance of Observed exchange Inferred distance of

MI (n = 91) chiasmata MI (SE) MII (n = 19) chiasmata MII (SE)

0 35 0.21 (0.08) 1 16 0.22 (0.12) 7 0.11 (0.18) 2 25 0.00 (0.19) 5 0.31 (0.27) 3 13 0.48 (0.16) 5 0.31 (0.32) 4 2 0.09 (0.06) 1 0 (0.30) 5 1 0.25 (0.21) Map length 101 cM 145 cM

Table 2. Female genetic distances

Marker cluster Marker Robinson and CEPH/Généthon CEPH reference

Lalande, 1995 (29) on-line haplotypes

I D15S541a 0 NA 0 D15S542 0 NA 0 D15S543 0 NA 0 D15S11 0.3 3.4 0 II D15S97 6.2 8.5 6.5 GABRB3a 6.2 3.2 6.5 III D15S1048 19.7 D15S165a 12.7 19.9 15.2 D15S1031 21.9 D15S1043 21.9 IV D15S231 23.3 D15S144 23.9 20.1 ACTCa 28.0 23.9 V D15S1006 46.2 D15S1028 46.2 D15S123a 46.3 39.3 CYP19 45.4 42.3 D15S982 46.6 VI D15S117a 56.1 49.2 D15S155 58.4 D15S98 61.4 VII D15S993 71.7 D15S108a 72.9 62.4 D15S1020 75.0 VIII D15S977 91.3 D15S131a 92.3 82.6 D15S204 IX D15S116 109.6 FESa 111.8 102.8 IPM15 115.3 X D15S130 130.3 D15S100a 135.0 120.6 XI D15S120 147.5 131.6 D15S203 147.5 D15S87a 148.0 136.1 D15S966 149.4 D15S86 149.4 D15S642 149.4 137.5


Table 3. Observed distribution of maternal crossover classes in CEPH families and the most likely underlying distribution of exchange estimated by (i) the ‘Weinstein

method’; (ii) constraining all frequencies to ≥0; or (iii) constraining all frequencies to ≥0 and setting the zero exchange class as equal to 0 Class Observed CEPH Model

exchange Weinstein (unconstrained) Classes constrained ≥0 Classes constrained ≥0, 0 class = 0 Estimated chiasma frequency (SE) Estimated chiasma frequency Expected exchanges Estimated chiasma frequency Expected exchanges 0 20 0.13 (0.12) 0.08 20.2 0.00 17.7 1 32 –0.22 (0.39) 0.00 33.2 0.13 36.8 2 31 0.57 (0.54) 0.27 28.6 0.22 27.5 3 8 0.35 (0.37) 0.51 9.1 0.50 9.2 4 1 0.17 (0.17) 0.14 0.8 0.15 0.9 Total cM 132.6 132.0 133.5 G statistic 0.42 (NS) 1.57 (NS)

The estimated frequency of achiasmate meioses leading to MI non-disjunction is 21%. Although the standard error of this estimate is reasonably large, this value is significantly different from zero. The overall map lengths are calculated using the relationship that one exchange is equivalent to 50 cM. Our estimate of an MI non-disjunction map length of 101 cM is identical to that estimated from summing over the distances estimated for each interval (Fig. 1). There are too few data to make any conclusions from the MII errors. However, the estimated mean map length of 145 cM lies between the two estimates from the CEPH haplotypes and the on-line Généthon map and is not significantly different from either one.

In order to determine the normal distribution of exchange these patterns were estimated from the CEPH haplotype data (Table 3, Fig. 2). In this case only one of the four products of meiosis is obtained and the accuracy in estimating the original pattern of exchange is reduced as compared with UPD15/T15 data. The inferred pattern of exchange was estimated in two ways: (i) the ‘Weinstein’ method; and (ii) the best-fit ‘constrained’ method. Under the Weinstein method it is possible to obtain estimates <0 (20). Under a ‘best fit’ scenario the deviation from observed values (measured by the G statistic) is minimized while constraining all frequencies to be ≥0. The disadvantage of this approach is in the difficulty of obtaining error estimates on these values. The advantage is that the estimated values are more realistic. In either case the overall genetic map length was estimated to be ∼133 cM, which is slightly lower than estimated from summing over intervals. This is probably because not all haplotypes were informative for the most telomeric markers (D15S87, D15S86 and D15S642) and thus recombination events in that region may be missed. Although the estimate of the zero chiasma class is 13% (±24) under the Weinstein method, the confidence intervals, estimated as twice the standard error, include a frequency of 0%. Furthermore, although our best fit ‘constrained’ model indicated an estimate of 8% for the zero crossover class, a model fixing the zero chiasma class at 0% was also completely consistent with the observed data (Table 3).

Maternal age and recombination

The mean maternal age for each transition class was calculated for MI errors to determine if there are age differences between

Figure 3. The relative frequency of observed transition classes among maternal

MI errors in each maternal age group. Among mothers of UPD15 or trisomy 15 cases who were <30 years of age two-thirds of cases showed no evidence of recombination and the three or more transition class was completely absent from this group. Yet this latter class made up almost half the cases from women >40 years of age.

non-disjunction in recombinant and non-recombinant tetrads (Table 4). Maternal ages were available for 71 cases and the mean among non-recombinant cases (those with no observable ex-change) was significantly less than among those cases showing one or more transition (32.7 versus 36.3 years) (P < 0.003, Student’s

t-test). The mean maternal age for control random births from a

Zurich population was 28.0 and similar figures were estimated from both Swiss and Canadian (British Columbian) population databases (17). The overall variability in maternal age between groups was not statistically significant when excluding the zero recombination class (ANOVA, P = 0.06). However, the mean age of the three or four transition class was significantly higher when compared with the one and two transition classes considered together (P < 0.01, Student’s t-test). Specifically, the three/four transition class is rare among mothers <35 years of age, but becomes increasingly important in the older age groups (Fig. 3).



Nucleic Acids Research, 1994, Vol. 22, No. 1

Human Molecular Genetics, 1998, Vol. 7, No. 6 1015

Table 4. Mean maternal age within each class of observed transitions among

MI errors

Observed transitions Number observed (n = 71) Mean maternal age

0 27 32.7

1 13 35.2

2 19 35.1

3 or 4 12 39.3

The two cases for which four transitions were observed had maternal ages of 37 and 38, but are lumped together with the three transition class due to sample size considerations.


Level of recombination and non-disjunction

A reduction in recombination among MI errors can result from a general tendency for tetrads with fewer exchanges to be at increased risk of non-disjunction or may be due to increased susceptibility of tetrads which have failed to undergo recombina-tion altogether. In the case of chromosome 15 a 26% reducrecombina-tion in genetic length was observed when comparing MI errors (101 cM, a mean of 2.0 exchanges) with controls (137 cM, a mean of 2.7 exchanges). From the observed distribution of transitions among UPD15/T15 cases it was estimated that 21% of non-disjoined chromosomes were associated with complete absence of recom-bination between the two homologs at MI. Thus if absence of pairing and recombination is rare in normal meioses, then a high rate of non-disjunction in this class of tetrads can explain most or all of the observed reduction in recombination. This contrasts with the conclusions of our earlier paper, in which we suggested that a reduction in recombination associated with chromosome 15 non-disjunction could not be completely explained by an increased susceptibility of the non-recombining class (18). The main reason for the differing conclusions is shortening of the ‘control’ map over time. Previously the CEPH female genetic map from D15S11 to D15S87 was estimated as 181 cM. The present estimate is only 137 cM based on 92 maternal haplotype transmissions from the eight CEPH reference families. In addition, there are now considerably more data than in our initial study, both in numbers of cases and in numbers of markers typed per individual.

Estimating the population frequency of achiasmate meioses for chromosome 15 is difficult due to large standard errors in these estimates (as indicated in Table 3), although it is certainly lower in normal meioses than in tetrads undergoing non-disjunction. Lamb et al. (19) estimated that the achiasmate class occurs with zero or near-zero frequency for chromosome 21 in normal meioses and suggested that there is probably not a ‘back-up’ system in humans to allow proper segregation of non-recombi-nant homologous pairs. A genetic map constructed from MI errors leading to trisomy 21 was only 33 cM, compared with 72 cM in controls (54% reduction) (21), with an estimate of the achiasmate class being 45% (20). Thus the majority of reduced recombinations for both chromosomes can be accounted for by a high susceptibility of achiasmate pairs to missegregation, with the greater reduction in recombination in trisomy 21 due to a higher contribution of achiasmate non-disjunction. Nonetheless, the frequency of the achiasmate class was estimated as 8% for

meioses involving chromosome 15. Although this value was not significantly different from zero, some low background rate of non-exchange meioses must exist for them to be sampled so frequently among non-disjoined chromosomes.

Distribution of recombination and non-disjunction

Despite the ability of non-disjunction from achiasmate pairings to explain the overall reduction in recombination for chromosome 15, the majority of segregation errors have occurred in tetrads with normal levels of exchange (e.g. two–four chiasmata). It is likely that at least two mechanisms leading to non-disjunction are occurring: one that involves segregation in achiasmate tetrads and one resulting in abnormal segregation of chromosome pairs with a ‘normal’ level of recombination. There are several lines of evidence to suggest that the distribution of recombination among the chiasmate pairings, and not just presence/absence of recom-bination, is important in susceptibility to non-disjunction. For example, a 25% reduction in recombination has been reported for chromosome 16 MI non-disjunction, but in this case there is no evidence that any cases are associated with an achiasmate meiosis (14). In addition, data from both chromosomes 16 and 21 suggest that a chiasma located near the centromere is of particular importance in protecting against non-disjunction (14,20,21). Specifically, an ∼50 cM region around the centromere was reduced to 0 cM in the trisomy 16 map and the most proximal 20 cM was reduced to 9 cM in the trisomy 21 map. A 50% reduction (from 15.3 to 7.7 cM) is also noted in the most centromeric section of chromosome 15, which is consistent with the previous data.

An increase in recombination in the distal region of the chromosome arm is seen in the trisomy 16 and 21 MI data, but is not apparent in the UPD15/T15 MI error data. If an exchange occurring proximally is protective against non-disjunction, then those single exchange tetrads which non-disjoin will consequen-tially be those with distally located chiasmata. In the case of chromosome 16 and 21, which commonly have only one exchange per chromosome arm, this increase should be more apparent than for larger chromosome arms. In the case of chromosome 15 any apparent ‘increase’ in recombination (due to removing tetrads with proximal exchange from the pool of susceptible configurations) may be spread over the entire remainder of the chromosome. In addition, the compensatory increase may be most pronounced in the ‘next’ interval, as might be suggested by the increase in recombination observed in the interval between D15S123 and D15S117.

Maternal age and non-disjunction

The most novel finding in the present data is an association of maternal age with level of recombination. One hypothesis is that ‘achiasmate’ pairs may be subject to aberrant segregation regardless of maternal age. The observed zero transition class represents 38% of the total sample, of which ∼50% are estimated to be true zero exchanges (21% of the total sample). Thus the age distribution of the observed zero transition class should fall between that of the recombinant classes and normal controls. This appears to be the case, except that there were no cases within the UPD15/T15 0 transition class attributed to mothers <25 years of age, despite the fact that 25% of normal births are expected to fall into this maternal age class (based on Swiss and BC controls; 17). This would suggest that achiasmate tetrads can segregate properly in very young women, but are less resistant to the effects


of maternal age than the recombinant classes. A back-up system for ensuring normal segregation in the absence of recombination has been shown to exist in other organisms (22). In addition, as the highest mean age was observed for the three or four transition class there remains the possibility that there is overall a positive association between maternal age and level of recombination, rather than simply a difference between the exchange/non-exchange classes. As there is no evidence of a difference in recombination levels at different maternal ages in control populations (A. Chakravarti and A. Lynn, personal communica-tion), this would potentially reflect a difference in susceptibility of different chiasmate configurations to the effects of maternal age.

A dilemma, however, is how to reconcile these data with those from trisomy 16 and 21, for which no evidence of an association between age and recombination has been observed (20; T.J. Hassold, unpublished data). In particular, non-disjunction of chromosome 21 shows many similarities to that for chromosome 15 and would seem likely to occur by a similar mechanism. Both predominantly involve a maternal MI event and the effect of maternal age on frequency of trisomy 21 is roughly similar to that for trisomy 15 (1) or UPD15 (17).

One possibility is that an interaction between maternal age and recombination does exist for chromosome 21, but that it is obscured by other factors. There is much less recombination for chromosome 21 than for chromosome 15. For trisomy 21 multiple transitions were observed in only 7% of cases (from Table 1 in ref. 20), as compared with 44% of the chromosome 15 cases. Thus for chromosome 21 an effect may be harder to detect if the increased maternal age in recombinant cases is predomi-nantly due to the multiple transition classes. If an association between maternal age and recombination does exist, then the mean age of the ‘two or more’ transition class for trisomy 21 should be increased over the other classes when analyzed separately. If the ‘older’ maternal age classes are relatively rare within the population studied an effect of maternal age would also be less pronounced. The mean age of 35 years from the UPD15/T15 data is considerably higher than the mean of 31 years reported for trisomy 21 MI errors (16). This difference may be due to the higher mean maternal age in European populations (from which most of the UPD15 cases with ages available were ascertained) as compared with US populations (from which many of the trisomy 21 cases were ascertained) (23). There has also been observed to be an increased rate of trisomy 21 in very young (<20-year-old) women as well as in the older age classes (1). The explanation for this phenomenon is not known, but if non-dis-junction of chromosome 21 in young mothers involves mostly recombinant tetrads, this could also obscure a relative increase in chiasmate non-disjunction in older women. No mothers in the present UPD15/T15 data were younger than 20 years of age.

Another possibility is that there is an additional back-up system to ensure normal segregation which acts on large (e.g. 15) but does not (or at least to a lesser extent) act on small (e.g. 21) chromosomes. Thus the fundamental system of segregation and susceptibility to non-disjunction may be similar for chromosomes 15 and 21, but the secondary forces which help stabilize multiple chiasmate tetrads may be more resistant to the effects of maternal age and do not break down except in women within the highest range of maternal ages. An additional system for ensuring normal segregation which acts to a greater extent on chromosome 15 than on 21 might also explain some of the differences between

chromosomes 15 and 21 in terms of presence/absence of an altered distribution of recombination in association with non-dis-junction. It would also be consistant with the observation that the risk of UPD15 with maternal age rises more rapidly in women >35 than for trisomy 21 (17).

In summary, a reduction in recombination among the MI errors is confirmed and may be explained by an increased frequency of achiasmate tetrads. However, most MI non-disjunction of chromosome 15 occurs in tetrads with two or more exchanges, indicating that there are at least two distinct mechanisms of non-disjunction occurring, one involving achiasmate pairs and one involving normally paired and recombining tetrads. Al-though there was not a dramatic alteration in the distribution of recombination along the chromosome arm among the recombi-nant cases, we cannot exclude the possibility that local position-ing of exchanges within each interval plays a role in susceptibility. Non-disjunction in achiasmate pairs appears to be less influenced by the effect of maternal age than in recombinant tetrads. Therefore, the degree to which recombination is altered in association with non-disjunction may be dependent on the age structure of the population being sampled. Chromosome-specific differences in maternal age effects may be related to the normal distribution of exchanges (and their individual susceptibilities) for each chromosome; however, they may also reflect the presence of multiple factors which act to ensure normal segregation, each affected by maternal age in a different way and varying in importance for each chromosome.

MATERIALS AND METHODS Patient ascertainment

A total of 115 cases of maternal UPD15 were ascertained through investigation of Prader–Willi syndrome patients. Of these, 12 were probable mitotic errors and were excluded from analysis (see below), leaving 103 cases attributed to a maternal meiotic non-disjunction event. Partial marker data on 24 of the 103 UPD cases have been previously published (18). Trisomy 15 (T15) cases were ascertained through cytogenetic analysis of sponta-neous abortions. All 13 maternal trisomy 15 cases have been analyzed previously for parental and meiotic stage of origin (24). However, in two of these cases typing of additional proximal chromosome 15 markers indicated the presence of a proximal transition. This caused a change in assignment of trisomy 15 case S287 from an MI to an MII event and case S827 from an MII to an MI event.

DNA studies

Microsatellite polymorphisms were detected by PCR amplifica-tion and, usually, visualized on a 6% polyacrylamide denaturing gel by silver staining. For 11 of the cases polymorphisms were detected by radioactive end-labeling of primers and visualization by autoradiography (25). Most primer pairs were obtained from Research Genetics (Huntsville, AL) and information on each is available through the Genome Database. For some of the patients RFLP data were also used and hybridization of probes to digested DNA from probands and parents was done as described previously (26). In all, 72 loci spanning 15q were used in the analyses (see Fig. 1 for the more commonly used markers). Typically 20–30 of these markers were typed in any one patient.



Nucleic Acids Research, 1994, Vol. 22, No. 1

Human Molecular Genetics, 1998, Vol. 7, No. 6 1017

Determination of meiotic/mitotic stage of origin. If the parent

transmitting two chromosomes 15 is heterozygous and the UPD15 child has inherited both alleles (i.e. is also heterozygous) this is referred to as non-reduction of parent of origin hetero-zygosity (or heterodisomy). Non-reduction at the centromere is indicative of an MI error. Reduction to homozygosity (or isodisomy) at the centromere is classified as an MII error if other markers elsewhere along the chromosome are heterozygous or a somatic error if all markers throughout the chromosome show reduction to homozygosity. D15S541, D15S542 and D15S1035, the most proximal markers to the centromere, were predominant-ly used to determine meiotic stage of origin and were informative to classify 85 cases as MI errors and 22 as MII or mitotic errors. If these markers were all uninformative origin was assigned based on inheritance at D15S543 (three MI and four MII/mitotic cases), D15S11 (seven MI and three MII/mitotic cases), D15S210 (one mitotic case), D15S13 (one mitotic case), D15S113 (one MI case) or D15S97 (one MI case). Because the markers used are near, but not at, the centromere there may be a small error involved in assignment (for a discussion, see ref. 18). Of the total of 128 cases studied 12 were reduced to homozygosity at a proximal marker, as well as for all additional markers typed spanning 15q, and were therefore classified as somatic errors and excluded from further analysis. Therefore, 116 meiotic cases remained for analysis, of which 97 were classified as MI and 19 as MII errors.

Two or more copies of a duplication of part of the GABRA5 locus map proximal to D15S541 and D15S542 (M. Lalande and R. Ritchie, personal communication; see also ref. 27). A primer pair (738CA) which amplifies a microsatellite locus within both the native and duplicated GABRA5 copies was used to confirm (but not to determine) meiotic stage of origin in some cases, including two (one MI and one MII error) in which a transition had occurred between D15S541 and D15S543 and 10 for which D15S541 or D15S542 were uninformative and origin had been determined using more distal markers. Because of overlapping bands, it is impossible to distinguish alleles from each of the duplicated sequences. However, only two types of patterns can occur in the case of maternal UPD: either the child shows the identical banding pattern and dosage as the mother or only some bands of the mother will be present in the child. In the first case an MI error is most likely (especially if the mother presents more than three distinct bands) and in the second case one or more copies of the duplicated region are presumed to be reduced to homozygosity. This is somewhat similar to interpretations of meiotic origin using the pBAMX7 X-centromeric probe (28). In all 13 cases classified as MI errors which were tested the 738CA pattern was the same in mother and patient.

Analysis of recombination between non-disjoined chromosomes.

In this paper the term ‘transition’ refers to an observed change in marker state between heterodisomy and isodisomy in a sample of two chromosomes from a disomic gamete and the term ‘cross-over’ refers to an observed change in marker state between grandmaternal and grandpaternal inherited markers on a single chromosome (i.e. as in normal controls). Only a portion of the total exchanges which have occurred in the meiotic tetrad at meiosis will be observable as crossovers or transitions. The number of transitions observed in a sample of non-disjoined chromosome pairs will also be dependent on the number and informativeness of the markers used, as transitions may be missed if markers are widely spaced. In general enough informative

markers are needed to divide the chromosome into linked intervals (i.e. intervals showing <50% recombination) and 20 cM intervals should be sufficient to observe the vast majority of transition events (18). Previously we estimated from theoretical expectations that typing five markers for a 175 cM chromosome (∼45 cM intervals) would be sufficient to identify the majority of recombinant versus non-recombinant chromosomes (18). The main effect of typing 10 markers versus five was the ability to detect larger numbers of transitions (i.e. >3), while the expected frequency of the zero transition class was only slightly reduced. Therefore, clusters of markers were defined to divide the chromosome into 10 intervals of an average of ∼14 cM. An attempt was made to make each case informative for at least one marker in each cluster. All recombination events were confirmed using additional markers. Further typing of our previously published 21 MI UPD15 cases for these intervals (and additional markers) resulted in the identification of three double transitions which previously had been missed due to having had large gaps between the informative markers.

In order to subdivide the chromosome into intervals and choose clusters of markers normal female map distances between markers were estimated from the latest on-line CEPH/Généthon mapping data (, with the exception of markers from D15S541 to D15S165 (15q11–q13), for which our own previous estimates were used (29; Table 2). This was done because there were known discrepancies in the order of these markers in the CEPH/Généthon map as compared with well-established maps of this region (30).

The original intention was to use each cluster of markers chosen to define a ‘megalocus’ within which there was little to no recombination. However, because the CEPH/Généthon map distances have changed since the clusters of markers were first chosen, several of the clusters now span 5 cM (for example FES and IP15M9 were formerly mapped to the same genetic location on chromosome 15, whereas the most recent map indicates that they are 3 cM apart). Therefore, rather than define megaloci, analysis of intervals was performed as follows. One marker was chosen from each cluster as the reference marker. When possible, for individuals which were uninformative or not typed at the reference marker marker state (non-reduced or reduced) was inferred in one of two ways: (i) if a marker 1 cM or less away from the reference marker was informative then the state of that marker was used; or (ii) if the reference marker lay within an interval of

≤15 cM between two other informative markers then it was assumed that a double transition would be extremely unlikely (no double crossovers were observed in the CEPH haplotypes within this distance; see below) and that the reference marker could be assumed to be of the same state as the two bounding markers. If there happened to be a transition between the two bounding markers, then the two possible states of the reference marker were given probabilities based on its relative distance between the reduced and non-reduced bounding markers. These strict criteria resulted in a loss of data, as many cases were informative for each interval but not always informative for the ‘reference marker’.

Control female genetic map. The current CEPH/Généthon

estimate of chromosome 15 map length is 149 cM, which is considerably shorter than the 181 cM (based on an earlier CEPH/Généthon map) which we used in our previous study. Other genetic maps, all based on the same CEPH data and published within the last 3 years, estimate the female genetic


length as anywhere between 135 (31) and 216 cM (32). The presence of multiple published lengths for chromosome 15 and known errors in the most current map make it difficult to compare data on recombination in non-disjoined chromosomes with any published data. Therefore, to try to control for discrepancies in map length due to incomplete error checking and to method of analysis the eight CEPH reference pedigrees (92 maternal haplotypes) were haplotyped using the same approach as for the UPD15/T15 cases. Informative markers were identified from the CEPH genotype database and from previous typing of these families for proximal 15 markers (29) for the same intervals used in the analysis of UPD15/T15 cases. Only crossovers which could be confirmed using nearby markers were considered valid (no maternal crossovers were excluded for this reason, but two paternal ones were unable to be confirmed). Our estimate of the female genetic map length from these eight families (92 haplotypes) is 137 cM using the Rao mapping function (with a mapping parameter of 0.35) to convert observed crossovers to map distance. This difference from the 149 cM cited above may be the result of the smaller sample (the CEPH/Généthon maps are based on data from additional families) or use of a different mapping function. Additionally, it may reflect errors in the on-line map; in this regard it is noteworthy that GABRB3 is listed as 5 cM proximal to D15S97 and CYP19 is listed as lying 1 cM proximal to D15S117 in the on-line data, but these locations are inconsistent with our analysis of both UPD15/T15 cases and CEPH controls. Of 50 identified double crossovers on the maternal haplotypes only four potentially occurred within an interval of <30 cM (see W.P.R. abstract for SCW15:1997 for details, available at citation/ConferencePaper). Thus such events are relatively rare and would only be missed if the intervening interval did not happen to span any typed marker.

Analysis of recombination. Recombination for sequential

pair-wise intervals was compared in UPD/T15 cases with control data following Shahar and Morton (33) and as done previously (18). Equating observed recombination in non-disjoined chromosomes to map distance must take into account the fact that exchanges are twice as likely to be observable as a transition if the marker on the centromeric side of the exchange event is reduced to homozygos-ity than if it is not reduced. For each interval the number of recombinants in CEPH controls was compared with the number observed in the UPD15/T15 data using a χ2 test of a 2 × 2 contingency table. In the UPD15/T15 data a correction must be made for the fact that when the proximal marker is reduced all chiasmata which have occurred are observable, whereas in control data and for non-disjoined chromosomes where the proximal marker is heterozygous half of the exchanges which have occurred will be missed. Therefore, the number of transitions for each interval was estimated as (TN–R) + 0.5 ×

(TR–N), where TN–R represents transitions from non-reduction to

reduction and TR–N represents transitions from reduction to



This work was funded by Medical Research Council of Canada grant no. MA-13694 (W.P.R.). We would also like to thank Drs Marc Lalande and Rachael Ritchie for the GABRA5-dup primers.


MI, meiosis I; MII, meiosis II; T15, trisomy 15; UPD, uniparental disomy.


1. Hassold, T. and Jacobs, P. (1984) Trisomy in man. Annu. Rev. Genet., 18, 69–97.

2. Warburton, D., Byrne, J. and Canki, N. (1991) Chromosome Anomalies and

Prenatal Development: An Atlas. Oxford University Press, Oxford, UK, pp.


3. Koehler, K.E., Hawley, R.S., Sherman, S. and Hassold, T. (1996) Recombina-tion and nondisjuncRecombina-tion in humans and flies. Hum. Mol. Genet., 5, 1495–1504.

4. Hassold, T. et al. (1996) Human aneuploidy: incidence, origin and etiology.

Environ. Mol. Mutat., 28, 167–175.

5. Fisher, J., Harvey, J., Morton, N. and Jacobs, P. (1995) Trisomy 18: studies of the parent and cell division of origin and effect of aberrant recombination on nondisjunction. Am. J. Hum. Genet., 56, 669–675.

6. MacDonald, M., Hassold, T., Harvey, J., Wang, L., Morton, N. and Jacobs, P. (1994) The origin of 47;XXY and 47;XXX aneuploidy: heterogeneous mechanisms and role of aberrant recombination. Hum. Mol. Genet., 3, 1365–1371.

7. Robinson, W.P., Binkert, F., Bernasconi, F., Lorda-Sanchez, I., Werder, E.A. and Schinzel, A.A. (1995) Molecular studies of chromosomal mosaicism: relative frequency of chromosome gain or loss and possible role of cell selection. Am. J. Hum. Genet., 56, 444–451.

8. Robinson, W.P. et al. (1997) A meiotic origin of trisomy in confined placental mosaicism is correlated with presence of fetal uniparental disomy, high levels of trisomy in trophoblast and increased risk of fetal IUGR. Am. J. Hum.

Genet., 60, 917–927.

9. James, R.S. and Jacobs, P.A. (1996) Molecular studies of the aetiology of trisomy 8 in spontaneous abortions and the liveborn population. Hum. Genet.,

97, 283–286.

10. Pangalos, C. et al. (1994) Understanding the mechanism(s) of mosaic trisomy 21 by using DNA polymorphism analysis. Am. J. Hum. Genet., 54, 473–481. 11. Carpenter, A.T.C. (1994) Chiasma function. Cell, 77, 959–962.

12. Merriam, J.R. and Frost, J.N. (1964) Exchange and non-disjunction in female

Drosophila melanogaster. Genetics, 49, 109–122.

13. Koehler, K. et al. (1996) Spontaneous X chromosome MI and MII nondisjunction events in Drosophila melanogaster oocytes have different recombinational histories. Nature Genet., 14, 407–413.

14. Hassold, T., Merrill, M., Adkins, K., Freeeman, S. and Sherman, S. (1995) Recombination and maternal age-dependent nondisjunction: molecular studies of trisomy 16. Am. J. Hum. Genet., 57, 867–874.

15. Warren, A., Chakravarti, A., Wong, C., Slaugenhaupt, S., Halloran, S. and Watkins, P. (1987) Evidence for reduced recombination on the nondisjoined chromosomes 21 in Down syndrome. Science, 237, 652–654.

16. Lamb, N.E. et al. (1996) Susceptible chiasmate configurations of chromo-some 21 predispose to non-disjunction in both maternal meiosis I and meiosis II. Nature Genet., 14, 400–405.

17. Robinson, W.P., Langlois, S., Schuffenhauer, S., Horsthemke, B., Michaelis, R.C., Christian, S., Ledbetter, D.H. and Schinzel, A. (1996) Cytogenetic and age-dependent risk factors associated with uniparental disomy 15. Prenatal

Diagnosis, 16, 837–844.

18. Robinson, W.P., Bernasconi, F., Mutirangura, A., Ledbetter, D.H., Langlois, S., Malcolm, S., Morris, M. and Schinzel, A.A. (1993) Nondisjunction of chromosome 15: origin and recombination. Am. J. Hum. Genet., 53, 740–751. 19. Lamb, N.E., Feingold, E. and Sherman, S.L. (1997) Estimating meiotic exchange patterns from recombination data: an application to humans.

Genetics, 146, 1011–1017.

20. Lamb, N.E. et al. (1997) Characterization of susceptible chaisma configur-ations that increase the risk for maternal nondisjunction of chromosome 21.

Hum. Mol. Genet., 6, 1391–1399.

21. Sherman, S.L. et al. (1994) Non-disjunction of chromosome 21 in maternal meiosis I: evidence for a maternal age-dependent mechanism involving reduced recombination. Hum. Mol. Genet., 3, 1529–1535.

22. Hawley, R.S. and Theurkauf, W.E. (1993) Requiem for distributive segregation: achiasmate segregation in Drosophila females. Trends Genet., 9, 310–317.



Nucleic Acids Research, 1994, Vol. 22, No. 1

Human Molecular Genetics, 1998, Vol. 7, No. 6 1019

23. Robinson, W.P., Lorda-Sanchez, I., Malcolm, S., Langlois, S. and Schinzel, A.A. (1993) Increased parental ages and uniparental disomy 15: a paternal age effect? Eur. J. Hum. Genet., 1, 280–286.

24. Zaragoza, M., Jacobs, P., Rogan, P., Sherman, S. and Hassold, T. (1994) Non-disjunction of human acrocentric chromosomes. Hum. Genet., 94, 411–417.

25. Petersen, M.B. et al. (1992) Comparative study of microsatellite and cytogenetic markers in detecting the origin of the nondisjoined chromosome 21 in Down syndrome. Am. J. Hum. Genet., 51, 516–525.

26. Robinson, W.P., Bottani, A., Yagang, X., Balakrishnan, J., Binkert, F., Mächler, M., Prader, A. and Schinzel, A. (1991) Molecular, cytogenetic, and clinical investigations of Prader–Willi syndrome patients. Am. J. Hum.

Genet., 49, 1219–1234.

27. Robinson, W.P. et al. (1997) Report of the third international workshop on chromosome 15 mapping 1996. Cytogenet. Cell Genet., 76, 1–13.

28. Morton, N., Keats, B., Jacobs, P., Hassold, T., Pettay, D., Harvey, J. and Andrews, V. (1990) A centromere map of the X-chromosome from trisomies of maternal origin. Annls Hum. Genet., 54, 39–47.

29. Robinson, W.P. and Lalande, M. (1995) Sex-specific meiotic recombination in the Prader–Willi/Angelman syndrome imprinted region. Hum. Mol.

Genet., 4, 801–806.

30. Christian, S.L. et al. (1998) Integrated YAC contig map of the Prader–Willi/ Angelman region on chromosome 15q11–q13 with average STS spacing of 35kb. Genome Res., in press.

31. Dib, C. et al. (1996) A comprehensive genetic map of the human genome based on 5, 264 microsatellites. Nature, 380, 152–154.

32. Spurr, N.K. et al. (1994) European Gene Mapping Project (EUROGEM): genetic maps based on the CEPH reference families. Eur. J. Hum. Genet., 2, 193–203.

33. Shahar, S. and Morton, N. (1986) Origin of teratomas and twins. Hum. Genet.,


Figure 1. A comparison of the genetic maps constructed from non-disjoined chromosomes 15 of maternal MI errors (n = 97) and CEPH maternal haplotypes (n = 92).

Figure 1.

A comparison of the genetic maps constructed from non-disjoined chromosomes 15 of maternal MI errors (n = 97) and CEPH maternal haplotypes (n = 92). p.2
Figure 2. Comparison of estimated distribution of exchange classes from maternal MI errors and CEPH maternal haplotypes.

Figure 2.

Comparison of estimated distribution of exchange classes from maternal MI errors and CEPH maternal haplotypes. p.2
Table 1. Observed distribution of transition classes in meiotic errors and the most likely underlying distribution of exchange

Table 1.

Observed distribution of transition classes in meiotic errors and the most likely underlying distribution of exchange p.3
Table 2. Female genetic distances

Table 2.

Female genetic distances p.3
Figure 3. The relative frequency of observed transition classes among maternal MI errors in each maternal age group

Figure 3.

The relative frequency of observed transition classes among maternal MI errors in each maternal age group p.4
Table 3. Observed distribution of maternal crossover classes in CEPH families and the most likely underlying distribution of exchange estimated by (i) the ‘Weinstein method’; (ii) constraining all frequencies to ≥0; or (iii) constraining all frequencies to

Table 3.

Observed distribution of maternal crossover classes in CEPH families and the most likely underlying distribution of exchange estimated by (i) the ‘Weinstein method’; (ii) constraining all frequencies to ≥0; or (iii) constraining all frequencies to p.4
Table 4. Mean maternal age within each class of observed transitions among MI errors

Table 4.

Mean maternal age within each class of observed transitions among MI errors p.5