1996 Oxford University Press Human Molecular Genetics, 1996, Vol. 5 Review 1495–1504 Recombination and nondisjunction in humans and flies Kara E. Koehler1, R. Scott Hawley1, Stephanie Sherman2,3 and Terry Hassold3,* 1Department of Genetics, Section of Molecular and Cellular Biology, University of California at Davis, Davis, CA 95616, USA, 2Department of Genetics and Molecular Medicine, Emory University School of Medicine, Atlanta, GA 30322, USA and 3Department of Genetics and The Center for Human Genetics, Case Western Reserve University and University Hospitals of Cleveland, Cleveland, OH 44106, USA Received June 25, 1996 Recent studies of Drosophila and humans indicate that aberrant genetic recombination is an important component of nondisjunction in both species. In both, a proportion of nondisjunction is associated with failure to pair and/or recombine and in both, exchanges which are either too distal or too proximal increase the likelihood of malsegregation. In this review we provide two perspectives on these observations: first, a review of exchange and chromosome segregation in model organisms, focusing on Drosophila, and secondly an overview of nondisjunction in humans. This format allows us to describe the paradigms developed from studies of model organisms and to ask whether these paradigms apply to the human situation. INTRODUCTION Over the past decade, studies of model organisms—in particular, studies of yeast and Drosophila—have yielded remarkable advances in our understanding of how meiotic chromosome pairing, synapsis and segregation occur in lower eukaryotes (1,2). Further, these studies have provided compelling evidence that abnormalities in genetic recombination perturb normal meiotic chromosome segregation. Meiotic mutants which reduce the level of recombination invariably increase the frequency of nondisjunction and, in some situations, the location of exchanges on the bivalent also affects the likelihood of nondisjunction. Our understanding of the human meiotic process, and of factors which affect meiotic chromosome segregation, is much less advanced. Nevertheless, recent studies of human trisomies suggest similarities between nondisjunction in our species and nondisjunction in model organisms. Specifically, reductions in recombination are a feature of all human trisomies thus far studied and, in several trisomies, the location of the exchanges is different than those observed in normal meioses. In this review we provide a summary of nondisjunction from two perspectives: first, a review of exchange and chromosome segregation in model organisms, focusing on Drosophila, and secondly an overview of nondisjunction in humans. This format allows us to describe the paradigms developed from studies of model organisms and to ask whether these paradigms apply to the human situation. EXCHANGE AND CHROMOSOME SEGREGATION IN MODEL ORGANISMS Genetic recombination, or crossing-over, is the primary means by which the disjunction of homologous chromosomes at meiosis I *To whom correspondence should be addressed (MI) is ensured. Recombination events themselves do not commit chromosomes to segregate; rather, disjunction is mediated by specific structures, known as chiasmata, that are formed at the sites of exchange. We and others have recently shown that, in addition to ensuring the segregation of homologous chromosomes, chiasmata play a crucial part in the control of the meiotic cell cycle (3–6). How chiasmata work Chiasmata serve to link homologous chromosomes together by taking advantage of the strong sister chromatid cohesion maintained along both of the homologs. The resulting pair of interlocked chromosomes, known as a bivalent, then attaches to the spindle such that one kinetochore is oriented towards each pole. The process by which that orientation is achieved—indeed, the process that ensures disjunction—is explained by the simple physical model described below. In centriolar meiotic systems, and thus in the male meiotic systems of most animals, during prometaphase each kinetochore of the bivalent attaches to the spindle and begins to move toward one of the two poles. Most bivalents achieve a bipolar orientation immediately (7). The high frequency of initial proper orientation is due both to the fact that chromosomal spindle fibers connect to the pole to which a given centromere most nearly points and to the observation that at the start of prometaphase, the two homologous centromeres are usually oriented in opposite directions, such that if one centromere is pointed at one pole the other centromere is pointed at the opposite pole [Ostergren (8), as modified by Nicklas (7)]. To quote Nicklas (9), ‘while the upper half-bivalent’s kinetochores most nearly face the upper pole, those of its partner most nearly face the lower pole, simply because bivalents are so 1496 Human Molecular Genetics, 1996, Vol. 5, Review constructed’. Once the two centromeres have oriented towards opposite poles, the progression of the two centromeres towards the poles is halted at the metaphase plate by the chiasmata. This represents a stable position in which the bivalent will remain until anaphase I. Although the mechanism described above is sufficient to explain bivalent attachment in many meiotic systems, it should be remembered that in some oocytes the spindle appears to be acentriolar and is organized by the chromosomes themselves. A well-studied example of this phenomenon occurs in the oocytes of the fruit fly Drosophila melanogaster. Prior to metaphase, the chromosomes of Drosophila oocytes are condensed into a single mass known as the karyosome. Meiotic spindle formation begins with the establishment of an array of microtubules, lacking a defined pole, that emanate from the major chromosomes. As prometaphase continues, these bundles of microtubules are sculpted together on each side of the metaphase plate to form a bipolar spindle. It is often possible to observe that the early spindle is comprised of four sets of microtubule bundles, presumably corresponding to each of the four pairs of homologous chromosomes. The spindle then lengthens and tapers. The ability of chromosomes to organize a functional spindle is not limited to Drosophila oocytes. Chromosomes have been shown to organize the spindle in the absence of centrosomes in several meiotic systems. At least in Drosophila, and probably in other similar meiotic systems as well, the co-orientation of homologous chromosomes appears to reflect maintenance of heterochromatin pairing in the vicinity of the centromeres (10,11). Because these paired centromeres are both involved in organizing the spindle, their first poleward movements along the developing spindles will serve to lock them in opposite orientations. Thus, the ability of bivalents to orient their centromeres towards opposite poles is achieved by balancing the tension between the bivalent and the two spindle poles. This balancing of tension requires chiasmata to hold the two homologs together. This rather simple mechanism of chiasma function requires that each bivalent must be capable of orienting independently of other bivalents in the cell. The functional significance of chiasma position The number and position of recombination events is also tightly controlled. In metazoans, exchange only occurs in the euchromatin, and the amount of exchange is not proportional to physical distance (12). For each of the five major chromosome arms of Drosophila melanogaster, the frequency of exchange is extremely low near the base and tip of the euchromatin and reaches its highest levels in an interval beginning approximately 30% of the distance from the tip to the base of the euchromatin and ending some 50–60% of that distance. This pattern is not unique to Drosophila females but is a general feature of chiasma distribution in a large number of organisms (13). The distribution of meiotic recombinational events results from the combined action of three types of genetic control: (i) trans-acting regulators of exchange position, which appear to act at the level of entire chromosome arms; (ii) local cis-acting regulators of exchange; and (iii) chromosomal elements such as centromeres and telomeres that can suppress exchange in a polar fashion over long chromosomal distances (2). All three of these levels of regulation appear to act towards one simple purpose: to keep exchanges, and hence chiasmata, a substantial distance away from the centromeres and telomeres. As virtually all exchanges are sufficient to ensure segregation, it seems paradoxical that the organism goes to such enormous efforts to position precisely its exchanges. The positioning of exchanges reflects a compromise between the problems inherent in resolving exchanges at anaphase I, which will preclude more proximal exchanges, and the fact that very distal exchanges occasionally do not ensure segregation. Recall that the resolution of chiasmata occurs not by a terminalization process, but rather by a precise release of sister chromatid cohesion between the chiasma and the telomere (13). Moreover, sister chromatid cohesion must be maintained proximal to that chiasma and most essentially in the centromeric regions to ensure reductional separation at MI. Resolution of extremely proximal exchanges would require the release of sister chromatid cohesion in regions too close to the centromere to allow normal centromere function at MI. By limiting most exchanges to regions in the more distal euchromatin, the cell also limits the extent of sister chromatid release required for proper separation and chiasma resolution at the onset of anaphase, thus keeping its centromeres out of jeopardy. Evidence in both flies and yeast argues that distal exchanges are not always by themselves sufficient to ensure segregation (reviewed in 2). In an elegant series of studies of mini-chromosome recombination in yeast, Dawson and his collaborators (14) have shown that the ‘ability of an exchange to enhance segregation fidelity appears to be proportional to its distance from the telomere’. Similarly, when the two back-up achiasmate segregational mechanisms in Drosophila melanogaster are mutationally ablated, bivalents with very distal chiasmata often nondisjoin at very high frequencies (15,16), suggesting that the ability of a very distal chiasma to guarantee segregation may often depend on additional functions, such as achiasmate back-up systems. We are fond of the idea that very distal exchanges are simply locked in by far too little chromosomal material and that the resolution of chiasmata at the very start of sister chromatid release may well result in a precocious separation of homologous centromeres. Indeed, as noted below, the ord and mei-S332 mutations, which cause precocious sister chromatid separation at anaphase I, also cause high levels of nondisjunction. In summary, the precise genetic control of exchange position reflects a delicate balance between sites of exchange that allow chiasma function and those sites which can permit the sister chromatid release required for chiasma resolution. Exchanges must be proximal enough to provide sufficient sister chromatid adhesion distal to the cross-over in order to ensure that the bivalents stay locked together. This requirement for more proximal exchanges is balanced against the need to prevent sister chromatid separation anywhere even remotely in the vicinity of the centromere. Cases where even properly positioned chiasmata do not ensure segregation There are two important categories into which the nondisjunction of exchange bivalents fall: so-called ‘spontaneous’ nondisjunction and nondisjunction due to meiotic mutations (a third category, radiation-induced nondisjunction, will not be considered here). Although each type of event is discussed in detail below, it may be said in summary that in each case it is a failure 1497 Human Molecular Review Nucleic AcidsGenetics, Research,1996, 1994,Vol. Vol.5,22, No. 1 1497 of some other component of the meiotic process, such as the non-repair of DNA damage or the construction of a faulty spindle, that induces the nondisjunction. Chiasmata cannot be blamed for these segregational failures any more than automobiles can be blamed for the effects of adverse road conditions. Spontaneous nondisjunction in Drosophila. In humans, considerable information is available on the the incidence and origin of spontaneous nondisjunction, as described later in this article. In contrast, in yeast and many other model organisms intolerance to aneuploidy has prevented extensive studies of spontaneous nondisjunction; thus, relatively little is known about its occurrence in most model systems. It is best characterized in Drosophila melanogaster. Instead of trisomic conceptuses or disomic gametes, nondisjunctional events in Drosophila can be recovered as ‘exceptional’ females who have received both of their X chromosomes from their mothers and no sex chromosomes from their fathers. The single published study of spontaneous X chromosome nondisjunction in Drosophila (17) was completed over 30 years ago and raised more questions than it answered; however, it has provided us with the only previous insights into the mechanisms responsible for the spontaneous failure of chromosomes to segregate from each other during meiosis. In many ways the data of Merriam and Frost (17) parallel the basic observations in humans. They suggested, albeit without the use of a centromere marker for verification, that the majority of spontaneous X chromosome nondisjunction in flies occurred at MI; as demonstrated in Table 1 and discussed later in this paper, this is also the case for most human trisomies. Additionally, they noted that exchange precedes spontaneous nondisjunction in as many as 75% of cases, although there was also an excess of achiasmate bivalents among nondisjunctional X chromosomes. As discussed later, the majority of cases of human nondisjunction are preceded by exchange but—especially for the sex chromsomes—a proportion are associated with achiasmate bivalents. Thus, it is reasonable to expect that the origin of nondisjunction in flies will be useful and relevant in elucidating similar mechanisms in humans. This classic study in Drosophila (17) also reported a number of unusual phenomena that remained unaddressed until recently (18). One such observation was the possible occurrence of pericentromeric exchange, a phenomenon recently reported for human X chromosome nondisjunction (19). Another unusual result was the appearance of new lethal mutations at a rate several orders of magnitude greater than background levels on nondisjunctional X chromosomes. Although this idea has not been addressed experimentally in humans, at least one case of trisomy 21 has been noted to carry a new cytogenetic heteromorphism not attributable to either parent (20). These findings in flies and humans led Hawley et al. (18) to propose that at least some spontaneous nondisjunction—in both species—is the result of the repair of double-stranded breaks (DSBs) generated by the movement of transposable elements during meiotic prophase (Fig. 1). Such repair events will form Holliday junctions and thus strongly resemble normal meiotic recombination intermediates. Hawley et al. (18) further suggested that if such repair events occur at the wrong time or wrong place during meiosis, either outside of the narrow temporal window allowed for normal meiotic exchange or in a location in which recombination is normally prohibited, such as the centromeric heterochromatin, the resulting physical connection between the Figure 1. A possible transposon-based model for spontaneous nondisjunction (A) Hawley et al. (18) have proposed a model in which at least some spontaneous nondisjunction is the result of the repair of DSBs generated by the movement of transposable elements during meiotic prophase. DSBs in the DNA resulting from transposon excision (1) are repaired off of the homolog of the damaged chromosome (2), and the resulting repair events are thought to form recombination intermediates (3) similar or identical to those produced by normal meiotic recombination. (B) This model further suggests that such a physical connection between the two homologous chromosomes could, if it occurred at the wrong time or place, actually impair the proper segregation of homologs during anaphase I. For example, an exchange of meiotic origin in the central euchromatin (4) ensures regular disjunction, while a proximal DSB-repair event in the heterochromatin (5) may cause nondisjunction, either by interlocking the homologs at anaphase I or by compromising the integrity of the centromere. The latter event could result in precocious sister chromatid separation at MI. homologs could interfere with rather than ensure their proper segregation at anaphase I. An experimental reexamination of these ideas, using Drosophila, is currently underway and has uncovered two lines of evidence supporting the idea that spontaneous nondisjunction is correlated with the movement of transposable elements. First, a reversion rate of at least 1% due to the excision of a transposable element has been observed (Koehler and Hawley, unpublished observations) at a locus whose normal reversion rate is 1.5×10–5 (21). Second, we have recovered new lethal mutations at a rate of roughly 2% from a population of spontaneously nondisjoining Drosophila X chromosomes (Koehler and Hawley, unpublished observations). The transposon-based model for spontaneous nondisjunction postulated by Hawley et al. (18) predicts that the molecular lesions associated with the elevated mutation rate originally observed by Merriam and Frost (17) will be defects due to the insertion of a transposable element into a vital gene. Regardless of whether this is verified by molecular studies, the origin of these mutations will provide crucial insights into the mechanism of spontaneous nondisjunction. Meiotic mutants that allow the nondisjunction of exchange bivalents. There are six Drosophila meiotic mutations that cause chiasmate bivalents (exchange bivalents) to nondisjoin at high frequency, two of which also impair sister chromatid cohesion. These mutations are ald, mei-38, ord, mei-S332, nodDTW (when homozygous), and ncd. The functions of the first two genes, ald and mei-38, remain unclear. Other than noting that these two mutations apparently define some function necessary for chiasma function, these mutants will not be considered further. Both nod DTW and ncd define functions required for spindle assembly and/or maintenance at prometaphase, while ord and mei-S332 define functions required for the control of sister chromatid cohesion. Each of these functions is considered in detail below. 1498 Human Molecular Genetics, 1996, Vol. 5, Review Chiasma failure under conditions where spindle structure and/or assembly is perturbed. The nod gene in Drosophila melanogaster encodes a kinesin-like protein required for normal meiotic segregation (15,22,23). Although recessive loss-of-function nod alleles affect only the segregation of achiasmate bivalents, a dominant allele (nod DTW) impairs the segregation of both chiasmate and achiasmate bivalents at MI (16,23). Interestingly, the number and frequency of exchanges in nondisjoining bivalents of nod DTW-bearing oocytes parallel those observed for nondisjoining chromosomes 21 in humans (18). Both are enriched not only for achiasmate chromosomes but for bivalents with distally located exchange events; indeed, as discussed later, in trisomy 21-generating meioses, recombination in distal 21q is actually increased over that observed in normal female meioses (Fig. 2). However, it is important to note that the defect in nod DTW is not in recombination per se; e.g. the mutation does not affect the frequency of recombination. Instead, the chromosomes in mutant oocytes are defective in their ability to maintain contact with both the homolog comprising the other half of the bivalent and with the forming meiotic spindle, resulting in precocious disassociation of one or more bivalents or univalent chromosomes from the defective spindle. Although proximal exchanges are still sufficient to ensure proper segregation in this situation, bivalents with very distal exchanges or without any chiasmata at all are much more likely to nondisjoin (16,24). Such chromosomes, ejected from the developing spindle, might frequently either be lost or reform a connection with one of the two poles at random. We understand these observations and ideas in very simple mechanical terms: on a normal spindle there are several mechanisms, including tight centromere apposition as a result of a proximal chiasma and/or proper NOD function, to hold homologs together and to hold bivalents on the spindle. We have previously proposed that age impairs the ability of human females to form a normal meiotic spindle (18), perhaps through the time-dependent decay of an important component on which spindle function is dependent. Observations similar to those made for nod DTW have been made for mutations at the ncd locus in Drosophila melanogaster (Paul Szauter, personal communication). It is possible that there are multiple elements whose functional deterioration could potentially have such an impact in human females as well. Like the nod gene, ncd encodes a kinesin that is required for proper meiotic spindle assembly (25). It has been recently demonstrated that the NCD protein is critical to the stable assembly of a bipolar spindle in living Drosophila oocytes (26). with random segregation at both divisions in both sexes. Like the original allele, these new alleles also decrease the frequency of exchange in females by 10–20-fold and ablate the ability of those exchanges that do occur to ensure segregation. Thus, the pleiotropic effects of the original allele on sister chromatid adhesion, exchange and chiasma function must be due to a single biochemical defect. Presumably, that defect is in sister chromatid adhesion, and the effects on exchange and chiasma function are consequences of that defect. The ord gene has been recently cloned and found to encode a novel protein; mutational analysis of a number of ord alleles suggests that the C-terminal portion of this protein is required for proper sister chromatid adhesion and plays a part in moderating protein binding (29). The mei-S332 mutation elevates equational nondisjunction in both sexes. Like ord, it has been shown cytologically to exhibit defects in sister chromatid cohesion in spermatocytes, although these do not appear until anaphase I, later than the first manifestation of the ord defect. However, unlike ord, mei-S332 does not alter recombination location or frequency. Because of this and the other phenotypic differences between these two mutations, mei-S332 has been postulated to define a function important for maintaining sister chromatid cohesion in the centromeric region (27,30). Consistent with this idea, MEI-S332 has recently been localized specifically to meiotic centromeres, where it remains until dissociation or degradation at anaphase II, concomitant with sister centromere separation (30). The cytology of mei-S332 mutants suggests that the majority of nondisjunction occurs through the following pathway: sister chromatids separate precociously by anaphase I, but complete the reductional division in this separated state. However at meiosis II (MII) the two sister chromatids segregate randomly. The observations on ord and mei-S332 are intriguing, in light of recent cytogenetic studies of nondisjunction in human oocytes. In analyses of unfertilized MII oocytes obtained from patients attending fertility clinics, Angell et al. (31) found a high proportion of oocytes contain either missing or extra single chromatids—not missing or extra intact chromosomes. While these observations are yet to be confirmed, they suggest that errors in segregation at MI may result in the premature segregation of sister chromatids prior to MII. As the precocious equational segregation of chromosomes that nondisjoin at MI has been documented in a number of experimental systems (for review see 13,32), this may represent an important potential mechanism for nondisjunction in many organisms. The ord and mei-S332 mutations and sister-chromatid cohesion. In both males and females ord causes a dramatic increase in both reductional and equational nondisjunction as detected genetically. In cytological studies of male meiosis, Goldstein (27) observed abnormal sister chromatid association during prophase as well as precocious sister chromatid separation during anaphase I. In fact, the majority of chromosome misbehavior in ord mutants appears to be precocious sister chromatid separation at the first rather than the second meiotic division (27). Recently, five additional alleles of ord have been characterized (28); a decrease in the frequency of sister chromatid cohesion is observed cytologically in the male germline for all five of these alleles (female meiosis has not been examined). Indeed, for the strongest allele, both genetic and cytological studies are consistent Control of the metaphase-to-anaphase transition is a central component of cell cycle regulation, and arrest at metaphase I is a common component of oogenesis in both flies and mammals. In many male meiotic systems, the failure to recombine will result in the production of two univalent chromosomes, neither of which can be balanced at the metaphase plate. Several lines of evidence indicate that the absence of kinetochore tension in these organisms results in the production of a chemical signal that triggers a terminal metaphase arrest. For example, Li and Nicklas (33) recently reported that in praying mantid spermatocytes, the absence of tension on even a single kinetochore is sufficient to signal a meiotic error and trigger a prolongation of metaphase which results in cell death. Restoration of tension on that kinetochore by micromanipulation relieves that error signal and Chiasma release and the cell cycle 1499 Human Molecular Review Nucleic AcidsGenetics, Research,1996, 1994,Vol. Vol.5,22, No. 1 allows the cell to proceed into anaphase I. The triggering of a terminal metaphase I arrest by unaligned chromosomes acts to prevent these spermatocytes from producing mature, but potentially aneuploid, sperm (33). Thus, in these males the function of an arrest at metaphase is to abort an aberrant meiotic process. We have recently shown that metaphase I arrest in Drosophila is triggered by the presence of chiasmata (34). Although the presence of even a single chiasmate bivalent is sufficient to trigger metaphase arrest, in the absence of chiasmata metaphase arrest does not occur. While it is possible that the arrest reflects nothing more than the physical inability of homologs to separate until all of the chiasmata are resolved, we favor a model in which the presence of chiasmata induces one of a number of cell cycle regulators to cause arrest. According to either model, the release of this metaphase arrest is concomitant with chiasma resolution. In both the mantid spermatocyte and the Drosophila oocyte, kinetochore tension is being used to halt the cell cycle at metaphase, albeit towards rather different ends. This dramatic sexual asymmetry reflects the fact that evolution has allowed tension to be used in two very different ways to regulate the meiotic process in males and females. The absence of a chromosome misalignment checkpoint during Drosophila female meiosis (and many others, reviewed in 35) is striking. Meiosis proceeds unarrested with any number of univalent chromosomes or spindle disorganization, demonstrating that there is no checkpoint for mono-oriented or misaligned chromosomes in Drosophila female meiosis. A similar case can be made for oocytes from XO female mice. In this case, oocytes 1499 enter anaphase with a single unpaired X chromosome and without signaling a metaphase I checkpoint. Furthermore, the univalent segregates either equationally or reductionally at the first division, showing that the nature of the spindle attachment (bipolar or monopolar) does not effect the metaphase-to-anaphase transition (35). It is conceivable that most or all oocyte systems lack a method to detect mono-oriented chromosomes on the meiotic spindle. This may be related to the fact that, unlike spermatocytes and yeast cells, most oocytes naturally arrest at some point during MI or II. The lack of a metaphase checkpoint in human oocytes may cause the very high frequency of meiotic errors in human oocytes as compared with sperm (35). MOLECULAR STUDIES OF HUMAN NONDISJUNCTION As is obvious from the preceding discussion, analysis of meiotic mutants has guided our thinking about the relationship of exchange and chromosome segregation in lower organisms. In contrast, meiotic mutants which affect human chromosome segregation have not yet been identified; thus, this approach has not been available. Instead, investigations of human meiotic nondisjunction have typically involved studying the incidence or origin of the additional chromosome in trisomic conceptuses. The occurrence of such abnormalities is a major concern to human reproductive health. Trisomy is the most common class of human chromosome abnormality, occurring in at least 0.3% of all newborns and in approximately 25% of spontaneous abortions. Considered as a class, it is the most common known cause of mental retardation and the leading cause of pregnancy wastage. Figure 2. Genetic maps of chromosomes 16 and 21, generated from normal female meioses (from CEPH family data) and trisomies of maternal origin. (A) For trisomy 16, there is an overall reduction in map length, attributable to a reduction of recombination in proximal 16p and 16q. (B) For trisomy 21, both MI and MII maps are associated with aberrant recombination; in the former case, the map is shorter, especially in proximal 21q and in the latter, the map is longer, especially in proximal 21q. 1500 Human Molecular Genetics, 1996, Vol. 5, Review Table 1. Molecular studies of parental and meiotic stage of origin in autosomal and sex chromosome trisomies (adapted from reference 54 and Hassold and Sherman, unpublished results) Trisomy No informative Paternal cases I 2–12 16 13–15 54 16 62 18 101 21* 776 22 Maternal II 133 XXX 46 I Percentage II I or II 3 1 4 2 9 17 27 19 XXY I or II 2 58 2 maternal 13 81 12 8 27 87 51 1 10 100 16 35 41 90 556 176 6 94 11 89 40 13 22 56 24 10 10 94 *For trisomy 21, we have presented only those cases having information on both parent and meiotic stage of origin of trisomy. Despite the high frequency of human trisomy and its obvious clinical importance, until recently little was known about the origin or causes of human nondisjunction. However, beginning in the late 1980s, several laboratories initiated DNA polymorphism studies of trisomic fetuses or liveborns and their parents; subsequently, a considerable amount of information has accrued on the parent and meiotic stage of origin of human trisomy and, more recently, on the association of aberrant recombination and nondisjunction. Most human trisomies originate in maternal meiosis Molecular studies of the origin of trisomy are now available on over 1000 trisomic conceptuses (Table 1). The most extensively studied condition has been trisomy 21, with over 750 informative cases reported (36–38). Maternal errors predominate, accounting for over 90% of cases. It has not been possible to determine unequivocally the meiotic stage of origin due to the lack of a chromosome 21-specific centromeric polymorphism. Nevertheless, two groups (36,38) have used proximal 21q DNA markers to infer the meiotic stage of origin of trisomy 21 and have reported similar results. For maternally derived cases, over 75% are compatible with MI nondisjunction, while for paternal trisomy MII cases predominate (Table 1). A small proportion, approximately 5%, of cases are apparently due to mitotic nondisjunction (36). The other trisomies that have been studied display remarkable variability in origin (Table 1). For example, paternal MI nondisjunction is responsible for nearly 50% of cases of 47,XXY, while in trisomy 18 maternal MII errors predominate and for trisomy 16, all cases may derive from maternal MI nondisjunction. For the other trisomies, relatively few cases have yet been studied. Nevertheless, the available evidence suggests that maternal errors predominate for most trisomies, but that individual chromosomes may have specific liabilities to paternal and maternal nondisjunctional mechanisms. Aberrant recombination is a feature of human trisomy Several groups have used centromere mapping techniques (39) either (i) to compare the frequency and location of genetic exchanges occurring between nondisjoining chromosomes in trisomy-generating meioses with those occurring between normally disjoining chromosomes in normal meioses or (ii) to compare genetic linkage maps based on trisomy-generating meioses with those constructed from normal meioses (e.g. from conventional linkage studies of CEPH families). Using these approaches, data on recombination have been obtained for six nondisjunctional conditions: paternally derived cases of 47,XXY, maternally derived cases of sex chromosome trisomy and autosomal trisomies 16, 18 and 21, and maternal uniparental disomy 15 (Table 2). With the exception of trisomy 21, all mapping studies of individual trisomies have been based on a relatively small number of cases, with the reported effects of recombination appearing to vary among the different conditions. Thus, interpretations of the data must be made with caution. Nevertheless, all studies are in agreement in suggesting an association of abnormal recombination and human trisomy, and several general features are now beginning to emerge. A proportion of human trisomy is associated with failure of recombination. Similar to the observations of Merriam and Frost (17) for Drosophila, a significant excess of zero-exchange events has been reported for several of the human trisomies. This is especially true for paternally derived 47,XXYs, as indicated by studies of crossing-over in the XpYp pseudoautosomal region in 39 paternally derived 47,XXYs (40). Assuming that normally, a single chiasma within this region joins the MI XY bivalent, approximately one-half of the 47,XXYs would be expected to show evidence of recombination (in the other half, both recombinant or both non-recombinant meiotic products would be recovered, and the exchange would go undetected). Instead, Hassold et al. (40) found evidence for recombination in only six of the 39 cases, a significant reduction from expectation. Subsequently, they used the data to generate a genetic linkage map of the pseudoautosomal region and found it to be significantly shorter than the normal male map of the region (13 cM vs. 52 cM). 1501 Human Molecular Review Nucleic AcidsGenetics, Research,1996, 1994,Vol. Vol.5,22, No. 1 1501 Table 2. Summary of studies of aberrant recombination and the genesis of human trisomy Parental origin Trisomy (ref) of trisomy Stage of Observed effect Associated with parental age? Origin Paternal 47, XXY (40) MI No observable recombination in pseudoautosomal region in 33 of 39 cases of paternal 47,XXY; map length of 13 cM vs 52 cM in normal male meioses No Maternal 47, XXY, 47, XXX (19,43) MI Overall decrease in recombination, due to absence of recombination in some cases; however, other cases associated with increase in recombination in the pericentromeric region; overall map length = 98 cM vs 181 cM in normal female meioses Yes; cases with no detectable recombination have increased maternal age; cases with recombination near centromere not associated with increased age MII Overall decrease in recombination, but increase in pericentromeric region; map length = 140 cM vs 181 cM in normals No Uniparental disomy 15 (45,47) MI Reduced and possibly altered distribution of recombinational events; map length = 105 cM vs 174 cM in normals (47) Presumably, since UPD 15 is age-dependent Trisomy 16 (44) MI Reduced recombination restricted to proximal regions of chromosome; map length = 129 cM vs 172 cM in normals Presumably, since trisomy 16 is thought to be entirely age-related Trisomy 18 (46) MI Overall reduction in recombination; map length = 104 cM vs 163 cM in normals Probably, although only a small number (16) of MI cases have yet been studied MII Overall increase (although not significant) in recombination; map length = 191 cM vs 163 cM in normals Yes; MII cases elevated in maternal age by comparison with controls MI Reduced recombination, especially in proximal 21q; map length = 33 cM vs 72 cM for normals Yes; similar reductions in recombination seen in younger and older mothers MII Increased recombination, especially in proximal 21q; map length = 107 cM vs 72 cM for normals Yes; similar increases in recombination seen in younger and older women Trisomy 21 (38) Thus, the overwhelming majority of paternal 47,XXYs are apparently associated with failure of pairing and/or recombination. This observation is somewhat surprising, as evidence from the mouse suggests that abnormalities in XY pairing lead to germ cell death rather than to increased nondisjunction (e.g. 41). However, Ashley et al. (42) have recently suggested that, rarely, unrecombined sex chromosomes may complete meiosis to form functional (although frequently aneuploid) sperm. They studied male mice with a rearranged Y chromosome and a high frequency of sex chromosome univalency in which—using FISH—it was possible to distinguish between sperm or spermatids carrying recombinant and those carrying non-recombinant sex chromosomes. Most sperm were chromosomally normal and had normal levels of XY recombination, indicating that most cells with unrecombined sex chromosomes were eliminated at MI. However, a small proportion (4%) of gametes were aneuploid, either XY or ‘O’, suggesting that some of the cells with unrecombined sex chromosomes were able to proceed to MII and beyond. In humans, the available data suggest that a similar process may well be the most common cause of male sex chromosome nondisjunction. Failure of recombination also appears to be a feature of maternal nondisjunction, especially cases involving the X chromosome. In studies of sex chromosome trisomy of maternal MI origin (19,43), approximately 40% (32 of 82) of all examined cases showed no evidence of recombination between the nondisjoined Xs. As at least three to four exchanges are predicted for the X chromosome bivalent, this represents a significant departure from expectation. However, evidence for a similar effect involving autosomes is not nearly as compelling. In trisomies 16 and 21, an excess of achiasmate bivalents has been reported (44,38). However, in trisomy 16, exchanges were identified in approximately 80% (44 of 56) of cases, with the most significant difference from normal female meioses being an increased frequency of one-exchange events (44); and in trisomy 21 as well, studies of 173 cases of maternal MI origin indicated a significant increase in the frequency of one-exchange events (38). Similarly, there has been relatively little evidence for a significant increase in the frequency of zero-exchange events in either maternal uniparental disomy 15, based on studies of 27 cases (45) or maternal MI trisomy 18, based on studies of 16 cases (46). These observations are compatible with a role of reduced—but not absent—recombination in the genesis of most cases of autosomal trisomy. Thus, in humans the association between absence of recombination and nondisjunction may be largely restricted to the sex chromosomes. Reduced recombination is a feature of most, if not all, human trisomies of MI origin. Reductions in recombination have been observed in all trisomies of MI origin, with most trisomy-based genetic linkage maps being approximately one-half to two-thirds the length of the comparable maps from normal meioses (Table 2). However, there is mounting evidence that the reductions are not randomly located on the chromosome arms; instead, there 1502 Human Molecular Genetics, 1996, Vol. 5, Review Figure 3. The possible association of chiasma location and human nondisjunction. (A) Bivalents joined only by distally placed chiasmata may be susceptible to premature separation at MI. When followed by a normal MII, the resulting disomic gamete would contain nondisjoined chromosomes having genetically different centromeres; thus nondisjunction would be scored as occurring at MI. We suggest that any of several abnormalities—including impairment of meiotic motors or sister chromatid cohesion proteins—may increase the likelihood of nondisjunction of these bivalents. (B) Bivalents joined by proximally-placed chiasmata may be unable to separate at MI. At MII, the bivalent might undergo a reductional division, resulting in a disomic gamete in which the nondisjoined chromosomes have identical centromeres; thus, nondisjunction would be scored as arising at MII, even though the precipitating event was at MI. We suggest that some such nondisjunction is attributable to transposon mobilization (Fig. 1), and that this process may contribute to the maternal age effect on human trisomy. appear to be specific ‘cold spots’ in the trisomy-generating meioses. This is especially true for trisomies 16 and 21, the two autosomal conditions for which there are the most data (Fig. 2). In trisomy 16, there is little evidence for a reduction in recombination in either distal 16p or 16q. However, in the proximal regions of the chromosome Hassold et al. (44) reported a 20-fold decrease in recombination, with a map length of 77 cM in normals being reduced to only 4 cM in the trisomy 16 cases. Similarly, in trisomy 21 the reductions in recombination are restricted to proximal 21q; indeed in distal 21q the map is actually expanded by comparison with normals (38). Less information is available for other autosomal trisomies, but preliminary data from UPD 15 suggest that proximal exchanges may be diminished in this condition as well (45,47). Thus it may be that, at least for autosomal trisomies of maternal MI origin, the absence of a proximally located chiasma imparts an increased risk for nondisjunction. This mimics observations on the nod DTW mutation in Drosophila (see preceding discussion of model organisms), raising the possibility that—in humans as well as flies—there is a back-up system to normal chiasmate segregation. If components of this system (such as a human homolog of NOD) degrade, achiasmate bivalents or bivalents with distal-only exchanges may be preferentially affected; e.g. the centromeres may be less able to align when the exchange is physically distant, resulting in a bipolar attachment for one or both members of the bivalent. Alternatively, distal-only exchanges may increase the likelihood of premature separation of the bivalent—resulting in functional univalency—because of a decreased ability to maintain sister chromatid cohesion (Fig. 3). Impairment in the functioning of sister chromatid cohesion proteins (such as a human homolog of ORD) might well preferentially affect this type of bivalent. Regardless of the correctness of these or other models, the available evidence from humans suggests that it is altered placement of recombinational events, and not simply reduced recombination per se, which is the important determinant of maternal MI nondisjunction. ‘MII’ nondisjunction is associated with aberrant recombination. As genetic recombination occurs at MI, there was no obvious reason to suspect that nondisjunction at MII would be associated with aberrant recombination and this was consistent with early studies of recombination in maternal MII trisomy 21 (38). However, with the acquisition of new data, it is clear that this interpretation is incorrect. Studies of maternal MII trisomy 21 now indicate a significant effect of aberrant recombination although, in contrast to the situation with MI trisomy, the association is with increased recombination. Sherman (unpublished observations) has estimated the overall length of the MII map to be 107 cM, significantly elevated from the 72 cM for normal female meioses (Fig. 2). The distribution of exchanges is also different from normal, with an increased frequency of proximal exchanges in the MII population. There is also suggestive evidence that aberrant recombination is important in MII nondisjunction involving other chromosomes. In maternal MII sex chromosome trisomy, significantly increased pericentromeric recombination has been reported, although the overall length of the X chromosome map is shorter than the normal map (19). In maternal MII trisomy 18, there is an increase—although non-significant—in the overall length of the map (46). How can an event which occurs at MI affect segregation at MII? One explanation may be that the precipitating event occurs at MI, with aberrant segregation occurring at both divisions (Fig. 3). For 1503 Human Molecular Review Nucleic AcidsGenetics, Research,1996, 1994,Vol. Vol.5,22, No. 1 example, proximal chiasmata may predispose to ‘chromosome entanglement’ (48) at MI, with the bivalent being unable to separate and passing intact to the MII plate; at MII the bivalent divides reductionally, resulting in a disomic gamete. In this situation, the additional chromosomes would have identical centromeres, and thus be scored as an MII trisomy, even though the origin of the abnormality was at MI. One implication of this model is that ‘true’ MII errors may be relatively rare in humans. That is, in studies of trisomies of maternal origin, MII errors are typically reported to account for approximately 20–30% of all cases (Table 1). If these errors instead reflect abnormal processing of specific MI chiasmate configurations, it may be that virtually all human nondisjunction is attributable to errors at maternal MI, with errors at maternal MII and paternal MI and MII occurring at low, ‘background’ levels. This is intuitively an attractive idea, as the first meiotic division in females is unique in that it takes at least 10 and possibly as many as 40–50 years to complete. However, before seriously considering this posssibility, it will be important to verify the apparent association between aberrant recombination and MII nondisjunction. Maternal age-dependent trisomy is associated with aberrant recombination. An obvious implication of the above studies is an association between aberrant recombination and age-dependent nondisjunction. For example, in trisomy 16, thought to be completely maternal age-dependent (49), the genetic map is significantly shorter than the normal map (44); in trisomy 21, significant reductions in recombination have been identified in both older and younger women (38); and in maternal sex chromosome trisomy, the mean maternal age is significantly elevated for the zero-exchange cases (19). Thus, it is clear that aberrant recombination plays an important part in the genesis of the maternal age effect on trisomy. However, there is less certainty regarding the underlying basis of the association. An early, and still much debated, model was proposed by Henderson and Edwards over 25 years ago (50). Based on observations of declining chiasma frequencies and an increased frequency of univalents in aging mouse oocytes, they suggested that reductions in crossing-over with resultant univalent formation might explain age-dependent trisomy. However, as chiasma formation occurs prenatally in the female, they were forced to postulate the existence of a gradient, or ‘production line’, in the fetal ovary that causes the first formed oocytes to have a higher chiasma frequency than those formed later. Assuming that oocytes are ovulated in the same order that they enter meiosis, those that are ovulated last would be the most susceptible to nondisjunction due to a reduction in chiasmata. Recent studies of the mouse suggest that the last-formed oocytes are, indeed, the last to be ovulated (51), fulfilling one prediction of the ‘production line’ model. Studies of human trisomies fulfill a second prediction, namely, that age-dependent trisomy is associated with reductions in recombination. However, other predictions of the model are less easily reconciled with the mapping data from normal and trisomy-generating meioses. For example, conventional linkage studies of chromosome 21 (Audrey Lynn, personal communication) and chromosome 22 (52) provide no evidence that recombination declines with maternal age in normal meioses. Further, in studies of trisomies 16 and 21 of maternal MI origin, the magnitude of the reduction in recombination is similar in younger and older women (Hassold and Merrill, unpublished observations; Sherman, unpublished 1503 observations). Finally, cases of trisomy 21 of maternal MII origin are associated with increased recombination, despite the fact that this condition is clearly maternal age-dependent. Thus, the mapping data provide little evidence that the association of recombination and age-related trisomy is established prenatally. Instead, it seems more likely that the effect involves two ‘hits’: first, the establishment in fetal meiosis of a ‘susceptible’ meiotic configuration and second, age-related degradation of a meiotic process which increases the risk of nondisjunction of bivalents with susceptible configurations. For example, for the chromosome 21 bivalent the presence of a single, distally placed exchange may constitute a susceptible configuration (Fig. 3). In the ‘young’ ovary, the presence of a single chiasma—regardless of its location—may be sufficient to ensure normal segregation. However, with advancing age normal segregation may become increasingly dependent on the presence of a proximal, ‘anchoring’ chiasma. In its absence, the homologs may separate prematurely; for example, due to an age-related defect in a motor protein (such as NOD in Drosophila) which normally helps to hold the homologs in close register (2), or to age-related degregation of a protein involved in sister chromatid cohesion (such as ORD or MEI-S332 in Drosophila). As a consequence in the older ovary, bivalents with distal-only exchanges may function independently of one another at MI, increasing the likelihood that they will migrate to the same pole. However, what is the explanation for the increase in exchanges— particularly in the proximal regions of the chromosome— observed in some maternal age-dependent conditions (e.g. trisomy 21 of maternal MII origin)? One possibility is that such configurations also confer an increased susceptibility, but for a reason opposite that for distally placed exchanges; i.e. in these instances the configurations interfere with the ability of the bivalent to become disentangled at MI (Fig. 3). This might be a consequence of an age-related shortening in the meiotic cell cycle (53), which limits the amount of time to resolve chiasmata, or to age-related degradation of human homologs of proteins, such as NCD, which normally promote poleward movement of the chromosomes. As a result, the bivalent fails to separate and the homologs move together to the pole. At MII reduction occurs, leading to disomy and an apparent MII error. In the above two models of age-dependent trisomy it is assumed that aberrant recombination is correlated with, but is not the proximal cause of, age-related trisomy. Rather, it is assumed that with increasing maternal age, bivalents joined together by too few chiasmata (or by distally placed chiasmata) or by too many chiasmata (or proximally placed chiasmata), are more susceptible to nondisjunction. However, in trisomies associated with an excess number of exchanges, it may be that aberrant recombination is, indeed, the proximal cause of nondisjunction. As discussed earlier, Hawley et al. (18) have suggested that some nondisjunction in Drosophila results from the repair of double strand breaks generated by transposon mobilization and we suggest that some age-related human trisomy arises in a similar fashion. In this instance, the ‘recombinational’ events occur in the arrested (dictyotene) stage of oogenesis. They do not lead to functional chiasmata, but rather may cause the chromosomes to interlock and prevent their separation at MI (Fig. 1). If the bivalent passes intact to the MII plate, the nondisjunctional product may again appear to result from an MII event, as in Figure 3. One implication of this model is that there may be an environmental component to human 1504 Human Molecular Genetics, 1996, Vol. 5, Review age-related nondisjunction. That is, in addition to repair of tranposition events, other DNA damaging agents may lead to chromosome interlocking and eventual nondisjunction. With age, the likelihood of such an event presumably increases, generating an age effect on trisomy; and, if pericentromeric ‘exchanges’ promote interlocking, recombination in trisomy-generating meioses will appear to be increased in the proximal regions. However, such recombination should not display interference, providing a possible means for testing this model. SUMMARY Studies of Drosophila and humans indicate several similarities in nondisjunctional mechanisms in the two species. In both, a proportion of nondisjunction is associated with achiasmate bivalents and in both, exchanges which are either too distal or too proximal increase the likelihood of mal-segregation. Analyses of meiotic mutants in Drosophila have identified several likely candidates for these recombination-associated events: abnormalities in meiotic motors and sister chromatid cohesion proteins have been shown to increase the likelihood of nondisjunction in distal-only exchange bivalents, and transposon mobilization in the pericentromeric region may also be a source of nondisjunction in flies. The relevance of these mutational processes to human nondisjunction is not yet clear, and the technical and ethical difficulties associated with human meiotic analyses will complicate attempts to examine them. Nevertheless, recent advances in molecular cytogenetic technology now make it possible to study chromosome segregation in human gametes (35), and the utilization of knockout mice makes it possible to study the effect of specific mutations on meiotic chromosome behavior. By combining these with other emerging technologies, we may finally be able characterize the mechanisms underlying human nondisjunction; given the extraordinary incidence and clinical significance of trisomy to human well-being, it is important that such attempts begin. ACKNOWLEDGMENTS We gratefully acknowledge Drs Patricia Hunt and Helen Salz for their critical reading of the manuscript. Some of the research described in this manuscript was supported by NIH grants R01-HD21341 (to T.H.), P01-HD32111 (to T.H. and S.S.) and R01-GM51444 (to R.S.H.), and by American Cancer Society grant DB-39E (to R.S.H.). REFERENCES 1. Roeder, G.S. (1995) Proc. Natl Acad. Sci. USA 92: 10450–10456. 2. Hawley, R.S., McKim, K.S., and Arbel, T. (1993) Annu. Rev. Genet. 27: 281–317. 3. McKim, K.S., Jang, J.K., Therkauf, W.E., and Hawley, R.S. (1993) Nature 362: 364–366. 4. Jang, J.K., Messina, L., Erdman, M.B., Arbel, T., and Hawley, R.S. (1995) Science 268: 1917–1919. 5. McKim, K.S. and Hawley, R.S. (1995) Science 270: 1595–1601. 6. Nicklas, R.B., Ward, S.C., and Gorbsky, G.J. (1995) J. Cell Biol. 30: 929–939. 7. Nicklas, R.B. (1967) Chromosoma 21: 17–50. 8. Ostergren, B. (1951) Hereditas 37: 85–156. 9. Nicklas, R.B. (1974) Genetics 78: 205–213. 10. Hawley, R.S. and Theurkauf, W.E. (1993) Trends Genet. 9: 310–317. 11. Dernburg, A.F., Sedat, J.W., and Hawley, R.S. (1996) Cell (in press). 12. Lindsley, D.L. and Sandler, L. (1977) Phil. Trans. R. Soc. London. Series B: Biol. Sci. 277: 295–312. 13. Hawley, R.S. (1988) In: Genetic Recombination (eds R. Kucherlapati, G. Smith), pp. 497–525. American Society for Microbiology, Washington, D.C. 14. Ross, L.O., Maxfield, R., and Dawson, D. (1996) Proc. Natl Acad. Sci. USA 93: 4979–4983. 15. Carpenter, A.T.C. (1973) Genetics 73: 393–428. 16. Rasooly, R.S., New, C.M., Zhang, P., Hawley, R.S., and Baker, B.S. (1991) Genetics 129: 409–422. 17. Merriam, J.R. and Frost, J.N. (1964) Genetics 49: 109–122. 18. Hawley, R.S., Frazier, J.A., and Rasooly, R. (1994) Hum. Mol. Genet. 3: 1521–1528. 19. MacDonald, M., Hassold, T., Harvey, J., Wang, L.H., Morton, N.E., and Jacobs, P. (1994) Hum. Mol. Genet. 3: 1365–1371. 20. Petersen, M.B., Frantzen, M., Antonarakis, S.E., Warren, A.C., Van Broeckhoven, C., Chakravarti, A., Cox, T.K., Lund, C., Olsen, B., Poulsen, H., Sand, A., Tommerup, N., and Mikkelsen, M. (1992) Am. J. Hum. Genet. 51: 516–525. 21. Green, M.M. (1959) Proc. Natl Acad. Sci. USA 45: 16–18. 22. Theurkauf, W.E. and Hawley, R.S. (1992) J. Cell Biol. 116: 1167–1180. 23. Zhang, P. and Hawley, R.S. (1990) Genetics 25: 115–127. 24. Jang, J.K., Messina, L., Erdman, M.B., Arbel, T., and Hawley, R.S. (1995) Science 268: 1917–1919. 25. McDonald, H.B., Stewart, R.J., and Goldstein, L.S.B. (1990) Cell 63: 11591165. 26. Matthies, H.J.G., McDonald, H.B., Goldstein, L.S.B., and Therkauf, W.E. (1996) J. Cell Biol. (in press). 27. Goldstein, L.S.B. (1980) Chromosoma 78: 79–111. 28. Miyazaki, W.Y. and Orr-Weaver, T.L. (1992) Genetics 132: 1047–1061. 29. Bickel, S.E., Wyman, D.W., Miyazaki, W.Y., Moore, D.P., and Orr-Weaver, T.L. (1996) EMBO J. 15: 1451–1459. 30. Kerrebrock, A.W., Moore, D.P., Wu, J.S., and Orr-Weaver, T.L. (1995) Cell 83: 247–256. 31. Angell, R.R., Xian, J., Keith, J., Ledger, W., and Baird, D.T. (1994) Cytogenet. Cell Genet. 65: 194–202. 32. Maguire, M. (1978) Genome 29: 744–747. 33. Li, X. and Nicklas, R.B. (1995) Nature 373: 630–632. 34. McKim, K.S., Jang, J.K., Therkauf, W.E., and Hawley, R.S. (1993) Nature 362: 364–366. 35. Hunt, P.A., LeMaire, R., Embury, P., Sheean, L., and Mroz, K. (1995) Hum. Mol. Genet. 4: 2007–2012. 36. Antonarakis, S.E. (1993) Trends Genet. 9: 142–148. 37. Sherman, S.L., Takaesu, N., Freeman, S.B., Grantham, M., Phillips, C., Blackston, R.D., Jacobs, P.A., Cockwell, A.E., Freeman, V., Uchida, I., Mikkelsen, M., Kurnit, D.M., Buraczynska, M., Keats, B.J.B., and Hassold, T.J. (1991) Trisomy 21. Am. J. Hum. Genet. 49: 608–620. 38. Sherman, S.L., Petersen, M.B., Freeman, S.B., Hersey, J., Pettay, D., Taft, L., Frantzen, M., Mikkelsen, M., and Hassold, T.J. (1994) Hum. Mol. Genet. 3: 1529–1535. 39. Chakravarti, A. and Slaugenhaupt, S. (1987) Genomics 1: 35–42. 40. Hassold, T.J., Sherman, S.L., Pettay, D., Page, D.C., and Jacobs, P.A. (1991) Am. J. Hum. Genet. 49: 253–260. 41. Hale, D.W. (1994) Cytogenet. Cell. Genet. 65: 278–282. 42. Ashley, T., Ried, T., and Ward, C.C. (1994) Proc. Natl Acad. Sci. USA 91: 524–528. 43. Lorda-Sanchez, I., Binkert, F., Maechler, M., Robinson, W.P., and Schinzel, A.A. (1992) Hum. Genet. 89: 524–530. 44. Hassold, T., Merrill, M., Adkins, K., Freeman, S., and Sherman, S. (1995)Am. J. Hum. Genet. 57: 867–874. 45. Robinson, W.P., Bernasconi, F., Mutirangura, A., Ledbetter, D.H., Langlois, S., Malcolm, S., Morris, M.A., and Schinzel, A.A. (1993) Am. J. Hum. Genet. 53: 740–751. 46. Fisher, J., Harvey, J., Morton, N., and Jacobs, P. (1995) Am. J. Hum. Genet. 56: 669–675. 47. Mascari, M.J., Ladda, R.L., Woodage, T., Trent, R.J., Lai, L.W., Erickson, R.P., Cassidy, S.B., Petersen, M.B., Mikkelsen, M., Driscoll, D.J., Nicholls, R.D., and Rogan, P.K. (1993) Am. J. Hum. Genet. 53 (Suppl.): A260. 48. Bridges, C.B. (1916) Genetics 1: 1–52. 49. Morton, N.E., Jacobs, P.A., Hassold, T., and Wu, D. (1988) Ann. Hum. Genet. 52: 227–235. 50. Henderson, S.A. and Edwards, R.G. (1968) Nature 218: 22–28. 51. Polani, P.E. and Crolla, J.A. (1991) Hum. Genet. 88: 64–70. 52. Schlumpf, K.S., Kim, D., and Haines, J.L. (1994) Am. J. Hum. Genet. 55 (Suppl.): A202 53. Eichenlaub-Ritter, U. and Boll, I. (1989) Cytogenet. Cell Genet. 52: 170–176 54. Abruzzo, M. and Hassold, T. (1995) Environ. Mol. Mutat. Suppl 25: 38–47
© Copyright 2026 Paperzz