Aneuploidy is the most common chromosome abnormality in humans, and is the leading genetic cause of miscarriage and congenital birth defects. Since the identification of the first human aneuploid conditions nearly a half-century ago, a great deal of information has accrued on its origin and etiology. We know that most aneuploidy derives from errors in maternal meiosis I, that maternal age is a risk factor for most, if not all, human trisomies, and that alterations in recombination are an important contributor to meiotic non-disjunction. In this review, we summarize some of the data that have led to these conclusions, and discuss some of the approaches now being used to address the underlying causes of meiotic non-disjunction in humans.
It has now been approximately 50 years since the identification of the first human aneuploid conditions ( 1 , 2 ). In the intervening half-century, research on the genesis of aneuploidy has addressed three basic questions: How frequent (and clinically important) is aneuploidy? What is the parental and meiotic/mitotic source of the extra or missing chromosome? What are the underlying non-disjunctional mechanisms that yield aneuploidy? The first of these questions has now been answered: aneuploidy is astonishingly common and extremely important clinically in our species. No fewer than 5% of all clinically recognized pregnancies are trisomic or monsomic. Most of these terminate in utero , making aneuploidy the leading known cause of miscarriage but some (e.g. trisomy 21) are compatible with livebirth, making aneuploidy the leading cause of congenital birth defects and mental retardation ( 3 ).
In this brief review, we summarize recent work on the remaining two questions.
WHERE DOES THE EXTRA CHROMOSOME COME FROM?
Beginning in the 1980s, several groups initiated DNA polymorphism studies to determine the origin of human aneuploidy. Because monosomies almost always spontaneously abort, these investigations focussed on trisomies, especially those compatible with livebirth (e.g. autosomal trisomies 18 and 21 and the 47,XXY and 47,XXX conditions). Based on these analyses, three general ‘rules’ of human non-disjunction emerged: first, regardless of the specific chromosome, most trisomies originate during oogenesis; second, for most chromosomes, maternal meiosis I (MI) errors are more common than maternal meiosis II (MII) errors; and third, the proportion of cases of maternal origin increases with maternal age ( 3 ). However, against this background, chromosome-specific differences in non-disjunction have become apparent. For example, maternal MI errors predominate in trisomy 21, while trisomy 18 typically involves maternal MII errors, and the 47,XXY condition is as likely to be paternal as maternal in origin.
Recently, information on origin has become available for other, previously under-studied, trisomies, including trisomies 13, 16 and 22 ( 4–6 ). Combined with the data from previously analyzed trisomies, these results provide a more complete picture of the way in which different human trisomies originate (Table 1 ). Perhaps most importantly, it now seems likely that there are at least three different types of non-disjunctional mechanisms: those that affect all chromosomes, those that affect groups of chromosomes, and those that affect individual chromosomes. For example, with the exception of the 47,XYY condition, maternal MI errors are likely an important contributor to all human trisomies. This is not particularly surprising since, in oogenesis, the first meiotic division begins in the fetal ovary and is not completed until the time of ovulation, 10–50 years later. However, there also appear to be mechanisms that differentially affect subsets of chromosomes. For example, non-disjunctional patterns are similar for trisomies 13, 14, 15, 21 and 22, suggesting that mechanisms of non-disjunction are shared among the acrocentric chromosomes. Further, some patterns of non-disjunction appear to be chromosome-specific. Most notably, virtually all cases of trisomy 16 are linked to errors at maternal MI, while MII errors are surprisingly common in trisomy 18. Presumably, these patterns reflect differences in the genomic architecture of individual chromosomes. The basis for these differences is unclear, and is one of the major unanswered questions in human trisomy research.
|MI (%)||MII (%)||MI (%)||MII (%)|
|MI (%)||MII (%)||MI (%)||MII (%)|
Meiotic recombination and human trisomies
One of the most important spin-offs of the studies of the origin of trisomy has been the ability to study recombination in non-disjunctional meioses, and to ask whether alterations in recombination are associated with human trisomy. The approach is straightforward: by studying the inheritance of polymorphic alleles in trisomic conceptuses, we can identify the location and number of crossovers that occurred between the non-disjoining chromosomes, and compare this information with data from normal meioses (e.g. with CEPH linkage data). The results have been remarkable. For all trisomies examined, significant alterations in recombination have been identified ( 7 ). Indeed, other than increasing maternal age, altered recombination is the most important known etiological factor associated with human trisomy. Most often, the effect is attributable to homologs that fail to crossover; this would be expected to produce random segregation at metaphase I and, consequently, a 50% chance of non-disjunction. However, for certain chromosomes the situation is more complicated, with unusual locations of crossovers being correlated with non-disjunction. The most compelling evidence for this is from studies of trisomy 21 conducted by Stephanie Sherman and her colleagues. Recently, her group has published a series of important papers that make it clear that—at least for trisomy 21—the association between recombination and non-disjunction is complex ( 8 , 9 ), and changes with the age of the mother. For example, among maternal MI-derived cases, telomeric exchanges are an important contributor to trisomy among younger women, but are less important among older women. From this one might conclude that, with increasing age, factors other than aberrant recombination become more important. However, a second type of ‘susceptible’ crossover configuration—pericentromeric exchanges in MII trisomies—is more common in older women, and a third abnormal recombination situation (‘achiasmate’ homologs, in which the two chromosomes 21 failed to crossover) accounts for ∼50% of maternal MI errors in both the youngest and oldest maternal age groups. How can we interpret these apparently contradictory observations? Probably the simplest explanation is that non-disjunction is not simple, even for individual chromosomes. For trisomy 21—and presumably other chromosomes as well—there are almost certainly different non-disjunctional mechanisms. Some are likely associated with failure to crossover, others with crossovers that occur too close to (or too far away from) the centromere, and others have nothing to do with recombination, but are attributable to abnormalities in other meiotic processes (e.g. loss of sister chromatid cohesion or defects in spindle assembly/disassembly). Further, maternal age probably affects each mechanism differently, being irrelevant for some, increasing the frequency of others, and being totally responsible for some. Thus, when we reach the point where we begin thinking about methods of preventing meiotic non-disjunction, it will be important to recognize that likely we will need not just one silver bullet, but scores of them.
WHERE DO WE GO FROM HERE: HOW DO WE UNCOVER THE NON-DISJUNCTIONAL MECHANISMS RESPONSIBLE FOR HUMAN TRISOMIES?
Studies of the origin of trisomy have been extremely instructive: we now know when and in which gamete the extra chromosome originates, and we have identified the first molecular correlate of human non-disjunction, altered recombination. Will this approach continue to provide new and important insights? Probably not. The commonly occurring trisomies have been thoroughly studied, providing a general understanding of the where and when of human non-disjunction. This is not to say that additional studies of origin are without merit, just that they will most likely provide incremental, and not novel, advances in the field.
So where do we go now? Clearly, there are a number of possible approaches, but recently three—direct studies of human meiosis, the generation of mammalian models of non-disjunction, and analyses of environmental influences—have attracted considerable attention.
Going to the source of the errors: direct studies of human meiosis
Starting about 15 years ago, investigators studying yeast and flies began to characterize proteins involved in meiotic pairing, synapsis, recombination and segregation. Because the basics of meiosis are extraordinarily conserved, these studies have provided a general outline of the important players in humans as well. Recently, several groups have taken advantage of this to study the mechanics of meiosis in humans, and to ask whether specific meiotic defects are associated with infertility or aneuploidy.
The majority of this work has involved males, where all stages of meiosis are available in the post-pubertal testis. One of the most intriguing aspects of the studies has been the use of immunofluorescence methodology to analyze pairing, synapsis and recombination ‘as it happens’ during meiotic prophase. Importantly, it has become apparent that certain proteins (e.g. the DNA mismatch repair proteins MLH1 and MLH3) localize on synaptonemal complexes in pachytene cells at the sites where chiasmata will form. Several groups have exploited this association to investigate basic aspects of human male recombination. For example, from these studies we know that the overall number of exchanges per spermatocyte varies from about 45 to 55 in control men, with each chromosome arm (except short arms of acrocentric chromosomes) having at least one exchange ( 10 , 11 ). In contrast, in men with idiopathic infertility the number of exchanges is frequently reduced, suggesting an association between recombination and male infertility ( 12 , 13 ). The relevance of these observations to the production of aneuploid sperm has not been directly demonstrated. Nevertheless, such an effect likely exists since unrelated molecular studies of aneuploid sperm and paternally derived trisomies indicate an association between aneuploidy and reduced recombination (reviewed in 14 ).
While analyses of human males are instructive, they are largely irrelevant to the study of human aneuploidy, since the majority of aneuploid conceptions are attributable to errors during oogenesis. Thus, recent attempts to extend direct studies of synapsis and recombination to human females take on particular importance. These analyses are technically challenging because they require collection of human fetal ovarian tissue. Further, given the many years of separation between prophase and the segregation of chromosomes at the first meiotic division, all stages of meiosis cannot be analyzed (unlike the situation in males). Nevertheless, studies of fetal oocytes provide a window into the critical early stages of meiosis, allowing us to ask whether the chromosomes are ‘set-up’ to mal-segregate years later in the adult female. Initial reports provide evidence that this is the case. For example, Lenzi et al . ( 15 ) reported remarkable variation in recombination levels (measured as the number of MLH1 foci) among different oocytes. Further, MLH1 foci were identified much earlier in prophase than in human males or male and female mice, suggesting that the temporal controls on recombination are different in human females. In general, these observations have been supported by other groups ( 16 ), who have also identified a high incidence of synaptic defects in human prophase oocytes (Fig. 1 ). Thus, the small amount of available data suggests that, for reasons that remain unclear, the early events of human female meiosis are comparatively ‘sloppy’. Although it is tempting to conclude that these prophase defects contribute to the high level of aneuploidy in our species, a significant limitation of these direct studies is the inability to distinguish between those oocytes that will survive and those that will be eliminated in the wave of perinatal atresia that dramatically reduces oocyte numbers. Thus, to assess the contribution of prenatal events to human non-disjunction it will be important to determine if chromosomes known to non-disjoin at high frequencies (e.g. chromosomes 16, 21 and 22) also exhibit an increased frequency of synaptic and/or recombinational defects.
Making mouse chromosomes behave badly
One of the major challenges in human aneuploidy research is the absence of appropriate model systems: reliable in vitro approaches to mammalian meiosis have not yet been developed, systematic analyses of chromosome abnormalities in non-human primates are prohibitively expensive and have not been attempted, and other mammalian species (including mice) have rates of chromosome abnormality an order of magnitude lower than those observed in humans.
However, against this bleak background, there have been recent successes in utilizing the mouse to model aspects of human non-disjunction. In one approach, Koehler et al . ( 17 ) attempted to recreate the recombinational deficiencies identified in human non-disjunctional meioses. They mated two closely related species of mouse, and analyzed the oocytes of the female offspring for meiotic abnormalities. The expectation was straightforward: by generating homologous pairs of chromosomes with substantial sequence divergence, recombination might be disturbed and, consequently, meiotic non-disjunction increased. Indeed, both outcomes were observed. Most importantly, the levels of aneuploidy approached those identified in humans and increased with maternal age, from ∼10% in young females to over 20% in older females. This is a level of spontaneous aneuploidy previously unreported in control mice, and provides optimism that the mouse can be manipulated to yield a useful model of spontaneous non-disjunction.
Other groups have used a more conventional approach; i.e, mutating meiotic genes and assessing the phenotypic consequences in mice. To date, several hundred engineered mutations affecting reproduction have been reported ( 18 ), and many of these disturb meiosis. Indeed, just this year there have been characterizations of mutations in genes involved in meiotic synapsis ( Syce2 , 19 ), recombination ( Mre11 , 20 ; Rad51c , 21 ; Dmc1 , 22 ) and the RNAi pathway ( Dicer , 23 ), among others. These studies are invaluable in elucidating the temporal and functional relationships of meiotic players. However, they have impacted our understanding of aneuploidy in a limited way, for a simple reason: most mutations are nulls that cause meiotic arrest during prophase, so it has not been possible to assess their effect on meiotic chromosome segregation.
A few notable exceptions bear mentioning. In an analysis of a knockout for the meiosis-specific cohesin gene Smc1β , Hodges et al . ( 24 ) were able to visualize oocytes throughout pachytene and up to diakinesis/metaphase I. By comparing the placement of MLH1 foci (at pachytene) with the location of chiasmata (at diakinesis/metaphase I), they demonstrated ‘slippage’ of crossovers in the absence of SMC1β. That is, while MLH1 foci were positioned normally along synaptonemal complexes on condensed chromosomes, chiasmata were located nearer the telomeres. Indeed, connections between homologous chromosomes were frequently lost by diakinesis. Importantly, the effect was exacerbated by maternal age: in the oldest females, the chromosome connections were simply gone, and the chromosomes were present as individual sister chromatids. These observations make it clear that cohesion is crucial to meiotic chromosome segregation, and suggest that age-related degradation of cohesion may contribute to the maternal age effect in humans.
Similarly, mutations in synapsis and recombination pathway genes have also been linked to mal-segregation. In an early study, Yuan et al . ( 25 ) observed increases in aneuploidy in oocytes of females homozygous for a null mutation in the synaptonemal complex gene Sycp3 . Further, there was an increase in embryonic death in offspring of older Sycp3 −/− females, consistent with an age-related increase in non-disjunction. More recently, Kuznetsov et al . ( 21 ) generated a hypomorphic mutation for the recombination gene Rad51c , overcoming the embryonic lethality associated with the Rad51c −/− genotype. Intriguingly, following hormonal stimulation oocytes of females heterozygous for null and hypomorphic alleles (i.e. Rad51c −/neo ) progressed to MII and a few embryos were generated. However, as most of the embryos were developmentally abnormal, the authors suspected that they might be chromosomally abnormal and investigated chromosome dynamics at MI and MII. Indeed, profound disturbances were identified. Most strikingly, at MII most of the oocytes contained chromosomes that had lost sister chromatid cohesion, similar to the phenotype observed in Smc1β mutant (discussed earlier). While it seems unlikely that RAD51c is a component of the cohesin complex, these results make it clear that disturbances in recombination can impact the way in which chromosomes are held together.
Taken together, these mutational analyses demonstrate that abnormalities in any of the three major events of meiotic prophase—cohesion of sister chromatids, pairing and synapsis of homologs and recombination between homologs—can yield errors in chromosome segregation. This also provides many avenues for future aneuploidy research, chief among them are questions regarding the maintenance and turnover of proteins that are laid down during fetal development in the female.
Are there predisposing factors to human non-disjunction?
Since the identification of the first human aneuploidies, investigators have been interested in identifying predisposing factors. This has led to tests of a wide variety of possible associations, including occupational exposures (e.g. pesticide applicators), medical treatments (e.g. radiation therapy), reproduction-related products (e.g. oral contraceptives), habituating agents (e.g. cigarettes), miscellaneous environmental exposures (e.g. toxic waste sites) and intrinsic factors (e.g. polymorphisms in folate pathway genes). However, these efforts have been remarkably unsuccessful, and maternal age, aberrant recombination and the occurrence of a previous trisomy ( 26 ) remain the only three factors incontrovertibly linked to human aneuploidy.
Does this mean that environmental factors are irrelevant to human aneuploidy? Possibly, but it seems more likely that exogenous factors that influence the fidelity of the meiotic process exist, but are difficult to identify. There may be several reasons for this. First, the magnitude of the maternal age effect likely interferes with our ability to detect other factors. Secondly, the fact that the female meiotic process is initiated during fetal development but chromosome segregation events occur in the adult female makes ascribing cause and effect difficult. Lastly, most epidemiological studies consider the origin of aneuploidy to be homogeneous, while it is clear that there are different non-disjunctional mechanisms, and each may respond differently to environmental stimuli.
Sherman and colleagues have attempted to overcome the last difficulty by combining molecular analyses of the origin of trisomy 21 with epidemiological analyses of putative etiological agents ( 27 , 28 ). This approach is not for the fainthearted, as it has taken over a decade to accumulate ∼800 cases that are informative for parent and meiotic/mitotic stage of origin ( 28 ). Nevertheless, several positive correlations have been reported along the way (e.g. cigarette smoking and maternal MII errors, low socio-economic status and maternal MII errors), and this approach likely provides the only way to directly identify environmental effects in the human population.
Hunt and colleagues have taken a completely different approach, using the mouse to test the aneugenic properties of bisphenol A (BPA), a chemical widely used in the manufacture of plastics and resins. Their route to these studies was serendipitous, involving the accidental damaging of plastic animal caging materials and the concomitant 10-fold increase in aneuploidy in oocytes of exposed females ( 29 ). Subsequently, they intentionally exposed females and males for defined periods during different parts of their reproductive life cycles and recently reported a remarkable effect after exposing pregnant females during the time when oocytes of their female fetuses would be entering meiosis. Specifically, synaptic defects increased, recombination rates were perturbed and—most importantly—when exposed female fetuses reached adulthood and their MII oocytes were examined, there was a dramatic increase in the incidence of aneuploidy. This indicates a ‘grandmaternal’ effect; i.e. by exposing pregnant females to BPA, the chromosome constitutions of their daughters’ eggs were affected. These are very worrying observations, since BPA is one of a group of compounds known as endocrine disruptors that are ubiquitous in modern society, and have been shown to have a number of adverse effects on reproduction. Designing studies to test the effect of BPA on human meiosis will be challenging but, given the implications to human reproduction, clearly necessary.
A half-century of aneuploidy research has yielded a wealth of information on the incidence, origin and etiology of human aneuploidy. This, however, has been the easy part. Now comes the real work—identifying the molecular basis of meiotic non-disjunction, understanding the mechanisms responsible for the age-related increases and, ultimately, developing therapies to reduce or eliminate non-disjunction. In vitro systems would be an invaluable tool for meiotic studies and, during the past four years, several groups have reported apparent progress toward the goal of generating gametes from embryonic stem cells ( 30–32 ). It may well be possible to achieve this in males, since a few liveborn offspring from embryonic stem cell derived mouse ‘sperm’ have been reported ( 30 ). This result, however, remains to be replicated and, similar success has not yet been reported for females. Importantly, meiosis has not been the focus of any in vitro efforts. Given the complexity of the meiotic process, this is a bit like building a house without giving thought to the foundation. Coupled with the difficulties women experience in vivo , this makes it unlikely that advances directly relevant to the human female will happen anytime soon. Thus, an important consideration—and a major challenge—in the development of in vitro approaches to mammalian gametogenesis is the genetic quality of gametes. Clearly, an alliance of stem cell biologists and aneuploidy researchers would be mutually advantageous: by focussing on both areas simultaneously, it may be possible to develop in vitro systems that generate genetically normal and functional eggs while at the same time uncovering the reasons why meiosis in the human female so frequently goes awry.
Conflict of Interest statement . None declared.
National Institutes of Health (HD21341 and HD42720 to T.H., HD37502 and ES13527 to P.H.).