Haplotype and interspersion analysis of the FMR1 CGG repeat identifies two different mutational pathways for the origin of the fragile X syndrome
Haplotype and interspersion analysis of the FMR1 CGG repeat identifies two different mutational pathways for the origin of the fragile X syndromeEvan E. Eichler, James N. Macpherson1, Anna Murray1, Patricia A. Jacobs1, Aravinda Chakravarti2 and David L. Nelson*
Department of Molecular and Human Genetics, Human Genome Center, Baylor College of Medicine, Houston, TX 77030, USA, 1Wessex Regional Genetics Laboratory, Salisbury District Hospital, Salisbury, Wiltshire SP2 8BJ, UK and 2Department of Genetics, Case Western Reserve University, 10900 Euclid Avenue, Cleveland, OH 44106-495, USA
Received October 24, 1995;Revised and Accepted December 20, 1995
To understand the origins of the fragile X syndrome and factors predisposing alleles to instability and hyperexpansion, we have compared the haplotype (using markers FRAXAC1, FRAXAC2, and DXS548) and AGG interspersion patterns of the FMR1 CGG repeat for 214 normal and 16 premutation chromosomes. Association testing between interspersion pattern and haplotype reveals a highly significant (P <0.002) non-random distribution, indicating that all three markers are useful in phylogenetic reconstruction of mutational change. Parsimony analysis of the FMR1 CGG repeat substructure predicts that loss of AGG interruptions has occurred independently on many haplotypes associated with the fragile X syndrome, partially explaining the haplotype diversity of this disease. Among haplotypes found in linkage disequilibrium with the fragile X mutation, two different modes of mutation and predisposition to instability have been identified. One pathway has involved the frequent and recurrent loss of AGG interruptions from rare asymmetrical ancestral array structures. Intergenerational transmission studies suggest that these predisposed chromosomes progress relatively rapidly to the disease state. In contrast, the second mutational pathway involves a single haplotype which has maintained two AGG interruptions. Parsimony analysis of CGG repeat substructure within this haplotype suggests that larger alleles have been generated by gradual increments of CGG repeats distal to the most 3' interruption. Pedigree analysis of the intergenerational stability of alleles of this haplotype confirms a gradual progression toward instability thresholds. As a result, a large reservoir of chromosomes carrying large repeats on this haplotype exists. These chromosomes are predisposed to disease. The present data support a model in which there are at least two different mutational pathways predisposing alleles to instability and hyperexpansion associated with the fragile X syndrome.
Investigations into the origin of the fragile X syndrome have traditionally involved the use of three polymorphic loci (DXS548, FRAXAC1, FRAXAC2) found within 150 kb of the FMR1 CGG repeat (1 ,2 ) (Fig. 1 ). Haplotype analysis of normal and mutant chromosomes among diverse population groups has indicated that a substantial proportion of all fragile X chromosomes shows linkage disequilibrium with a small subset of DXS548-FRAXAC2-FRAXAC1 haplotypes (3 -15 ). In more genetically isolated populations (such as Finland), linkage disequilibrium has been shown to be even more pronounced with 75% of all fragile X chromosomes occurring on single rare DXS548-FRAXAC2 haplotype (9 ,12 ). Genealogical studies of fragile X kindreds previously considered to be unrelated further confirmed the existence of founder effects for the fragile X syndrome, by demonstrating common descent for the fragile X chromosome from shared ancestors in the 17th and 18th centuries (11 ,16 ,17 ). These findings suggested that FMR1 CGG repeat alleles predisposed to the development of the fragile X syndrome could be carried silently through human pedigrees from five to 100 generations (4 ,12 ,15 ) prior to hyperexpansion and hypermethylation associated with disease (18 -22 ). The conclusion that new mutations are rare was difficult to reconcile with an X-linked dominant disorder in which the affected males rarely reproduce (3 ,15 ,23 ).
Due to our observation that the longest tract of pure repeats was a more suitable indicator of FMR1 CGG repeat allelic stability, we speculated that the number of AGG interruptions contained within a repeat might determine the longevity of predisposed alleles within a population (30 ). To test this model, we have compared the AGG interspersion pattern and DXS548-FRAXAC1-FRAXAC2 haplotype of 200 normal chromosomes (10 ) and 30 premutation and high-end normal chromosomes in which intergenerational stability or instability had been ascertained (30 ,33 ). Association testing between the haplotype markers and interspersion patterns revealed a significant non-random distribution, which could be used in the phylogenetic reconstruction of mutational events of FMR1 CGG repeat alleles. Parsimony analysis predicts that the loss of AGG interruptions has occurred independently and frequently among many of the haplotypes associated with the fragile X chromosome. In particular, two of the most prominent fragile X haplotypes (6-4-5 and 6-4-4; Table 1 ) show a correlation between the recurrent loss of AGG interruptions and asymmetrical array substructures. In contrast, one haplotype (2-1-3) is predicted to have been refractory to the loss of AGG interruptions. This haplotype's unique constraint to maintain two AGG interruptions has allowed FMR1 CGG repeat alleles to increase in length likely by replication slippage-like events distal to the last AGG interruption, resulting in the accumulation of high-end normal alleles in the population. Our data provide a molecular basis for the existence of two distinct mutational pathways for the origin of the fragile X mutation and suggest haplotype-specific influences in both the maintenance and loss of AGG interruptions.
Table 2 summarizes the distribution of 52 FMR1 CGG repeat AGG interspersion patterns among 28 DXS548-FRAXAC1-FRAXAC2 haplotypes for 200 unrelated male subjects from the Salisbury district of Wessex, England (Materials and Methods). The nomenclature and position of haplotype markers and the FMR1 CGG repeat interspersion patterns is described in Figure 1 . In order to assess the randomness of distribution (Table 2 ), 5000 computer simulations were performed using the program ASSOC (15 ). The [chi]2 statistic was employed to determine the significance of association between rows (interspersion pattern) and columns (haplotype) in Table 2 . Association testing revealed a highly significant non-random distribution between haplotypes and interspersion patterns ([chi]2 = 1381; P <0.002; n = 5000). In addition, each individual cell entry (association between row and column) was tested for significance. Significant associations (P <0.002) between interspersion pattern and haplotype are indicated in Table 2 by an asterisk (*). Due to the complex polymorphic nature of FRAXAC2 and its reported potential for microsatellite hypermutability (34 ), various compressions of Table 2 were analysed to assess the utility of each marker independently against the interspersion configurations. All three markers demonstrated significant independent associations with FMR1 CGG repeat interspersion patterns: FRAXAC1, [chi]2 = 102.5, P <0.0001; FRAXAC2, [chi]2 = 409.7, P <0.006; DXS548, [chi]2 = 256.3, P <0.004). In a similar fashion a highly significant (P <0.0001) non-random association was determined between FRAXAC1 and FRAXAC2.
Table 2 . FMR1 CGGrepeat interspersion pattern distribution amongDXS548-FRAXAC1-FRAXAC2haplotypesFifty-two AGG interspersion configurations for the FMR1 CGG repeat are compared against 28 chromosomal haplotypes. Haplotypes are organized in columns and are grouped using the FRAXAC1 marker (since it shows the strongest association with interspersion pattern, P <0.00001). Interspersion patterns are arranged in rows based on increasing length of the 5' tract of CGG repeats proximal to the first interruption. Significant positive associations (P <0.02) between interspersion pattern and haplotype were determined from ASSOC computer simulations and are indicated by bold and asterisk in the table. A single significant deficiency was observed and is indicated by a minus sign (haplotype 7-3-4+; interspersion pattern 9+9+9).
The three most common interspersion patterns (10+9+9, 9+9+9 and 10+9) account for 51.5% (103/200) of all X chromosomes and are distributed non-randomly among the various haplotypes (Table 2 ). The most abundant interspersion configuration (10+9+9) is concentrated primarily on FRAXAC1 haplotype 3 with the most significant associations (P <0.002) occurring on DXS548-FRAXAC1-FRAXAC2 haplotype, 7-3-4+ (40/55 or 72.7% of all 10+9+9 chromosomes). Similarly, the 10+9 interspersion pattern shows a similar restriction to FRAXAC1 haplotype 3 with the most significant associations occurring on haplotypes 7-3-4+ and 6-3-4+ (Table 1 ). The 9+9+9 interspersion pattern demonstrates the greatest haplotype diversity being distributed among 9/28 (32.1%) of all observed haplotypes. Of these, the association of 9+9+9 with haplotypes 7-4-6+ and 1-1-3 are the most significant (P <0.002).
Based on association testing between interspersion patterns and haplotype as well as visual inspection of the data in Table 2 , it is apparent that haplotypes fall into one of two categories based on the position of the first AGG interruption. Both FRAXAC1 alleles 1 and 4 predominantly (51/53 or 96%) show interspersion patterns in which the first AGG interruption occurs at position 10 (9+n configurations). In contrast, haplotypes with FRAXAC1 marker 3 associate strongly with FMR1 CGG repeat interspersion patterns which possess the first AGG interruption at position 11 (10+n configurations) (Table 2 ). One notable exception is haplotype 7-3-4 which shows a substantial number of chromosomes with the 9+9+9 configuration and a significant association (P <0.002) with the 13+9 interspersion pattern.
Although the position of the first AGG interruption generally divides haplotypes into one of two groups (relative to haplotype FRAXAC1), other significant associations are observed when one considers the other microsatellite haplotype markers and the position of the second AGG interruption. A total of 14 significant associations were identified using the DXS548-FRAXAC1-FRAXAC2 haplotype and the FMR1 CGG repeat interspersion pattern. (Only half as many significant associations were identified, using exclusively the FRAXAC1 marker). Using all three markers, 10 out of 14 of significant associations occur on FMR1 CGG repeat alleles with two AGG interruptions, and four out of 14 occur among alleles with a single interruption. For example, haplotypes 6-4-4 and 6-4-5 are virtually unique in possessing asymmetrical array patterns of the types, 9+10+9, 9+11+9 and 9+12+9 (Table 2 ). Haplotype 7-4-5, a relatively rare haplotype in this population survey, demonstrates a significant (P <0.002) association with an interspersion pattern devoid of a second AGG interruption (9+21). Similarly, another rare haplotype, 2-4-6+, is associated with an equally rare interspersion pattern (9+15+9).
Due to the prevalence of 7-3-4+ haplotype in this survey (89/200 = 44.5% of all chromosomes), several significant interspersion configurations were identified, namely; 10+9, 10+9+9 and 10+9+10. In addition, a single significant deficiency was found with interspersion pattern 9+9+9, confirming once again a partition of haplotypes based on the position of the first AGG interruption. It is interesting that the majority of FMR1 CGG repeats completely devoid of interruptions occur also on haplotype 7-3-4+ and other closely allied haplotypes (7-3-4 and 6-3-4+).
It should be noted that many of the DXS548-FRAXAC1- FRAXAC2 haplotypes (17/28) do not demonstrate non-random associations with particular FMR1 CGG repeat interspersion patterns due to the under-representation of these haplotypes in this population survey. One haplotype, 2-1-3, is interesting in this regard. Although no significant association was found, five out of the six interspersion patterns occurring on this haplotype are unique, due to the presence of a long distal tract of CGG repeats (Table 2 ). Furthermore, the position of the second AGG is remarkably conserved among all FMR1 CGG repeat alleles on this chromosomal background, generating the general configuration pattern 9+9+n.
FMR1 CGG repeat alleles may be categorized into different classes based on the total number of AGG interruptions which they possess (29 -32 ). In this population survey of 200 random chromosomes, alleles were distributed as follows: nine (4.5%) with no interruption; 59 (29.5%) with a single interruption; 129 (64.5%) with two AGG interruptions; and three (1.5%) with more than two AGG interspersions. Table 3 compares the most frequent and most significant 2-AGG configurations with 1-AGG interspersion patterns (>25 total repeats) within each haplotype. In most cases, the total overall repeat length of the 1-AGG class does not differ dramatically from the total repeat length of alleles with two AGG interruptions. Interestingly, the position of the first AGG interruption is also generally conserved between 2-AGG and 1-AGG classes within the same haplotype (Table 2 ). With one exception (Table 3 ; haplotype 7-3-4+; allele 26+9), there appears to be a considerable bias among normal alleles to lose the most distal AGG interruption.
Table 3 . Intra-haplotype comparisons between 2-AGG and 1-AGG allele configurations (>25 total repeat length)FMR1 CGG repeat interspersion configurations of alleles with a single AGG interruption (1-AGG class) and alleles with two or more interruptions are compared. Only alleles whose total repeat length was greater than 25 triplet repeat units are considered. Total repeat length is indicated in brackets. In most cases, only the most significant (indicated by an asterisk) and/or the most frequent 2/3-AGG alleles are considered.
In an attempt to reconstruct the phylogenetic progression of normal alleles to the fully expanded state, we examined an additional 30 unrelated premutation and high-end repeat length `normal' alleles in which intergenerational stability/instability had been ascertained (30 ). The FMR1 CGG repeat substructure, the haplotype and the stability for each premutation and evolving premutation are summarized in Table 4 . The majority of haplotypes at high-risk for the fragile X syndrome (Table 1 ) possess FMR1 CGG repeat alleles which have a single AGG interruption or are completely devoid of AGGs (Table 4 ). Among those alleles in which a single interruption predominates (haplotypes 6-4-5, 6-4-4, 8-3-4+ and 6-4-6+), the position of the most 5' AGG has been conserved between normal and premutation alleles (Table 2 ). Similarly, unstable high-end normal alleles, which likely represent intermediate states between normal stable alleles and premutation alleles, show a similar maintenance of the position of the first AGG interruption (Table 4 ).
Table 4 . AGG interspersion patterns among premutation and `protopremutation' haplotypesThe DXS548-FRAXAC1-FRAXAC2 haplotype is compared with the FMR1 CGG repeat interspersion pattern of 15 premutation (documented progression to full mutation) and 15 protopremutation (evolving premutation) alleles, based on large total repeat length or observation of unstable intergenerational transmissions (30). Samples are derived from 30 unrelated pedigrees. The study of origin for each allele is indicated in the reference column (10,18,30,31,33). Samples are arranged according to haplotype and are grouped relative to the FRAXAC1 marker. The number of documented stable (s) and unstable (u) transmissions in each pedigree is summarized in the column labelled stability. Alleles which have been observed to progress to full mutation are denoted in this column as fm.
Although the reduction or complete absence of AGG interruptions is common to most of the fragile X haplotypes, a specific subset of DXS548-FRAXAC1-FRAXAC2 haplotypes (2-1-3, 7-1-3 and 1-1-3) have interspersion patterns in which the position of the second AGG interruption has been maintained (Table 4 ). Intergenerational transmission studies of these FMR1 CGG repeat alleles indicate that these 2-AGG configurations can be remarkably stable (Table 4 ). One of the largest stable premutation-size alleles identified to date (66 total repeats; interspersion pattern 9+9+46; sample 15,135; Table 4 ), belongs to this conspicuous fragile X haplotype. Although this allele has a pure repeat tract which is much greater than the postulated instability threshold (34-37 CGG repeats) (30 ), only stable transmissions have been observed to date within this pedigree over two generations (data not shown). Despite this unusual high degree of stability, alleles carrying 2-AGG interspersions have been observed to progress to full-mutation (samples 80-813 and 16,415; Table 4 ).
Parsimony analysis was performed using both heuristic, branch-and-bound and exhaustive tree searches on aligned normal and premutation CGG repeat sequences (Table 2 , Materials and Methods). Majority rule (>50%) and strict consensus trees were generated for haplotypes which had at least four different sequences and multiple, equally parsimonious topologies. Based on character-state reconstructions for trees of the shortest length, parsimony predicts that the loss of AGG interruptions has occurred frequently and independently within eight out of the nine fragile X haplotypes which could be examined (data not shown). The phylogenetic relationship of interspersion patterns for four haplotypes is depicted in Figure 2 .
Several groups have reported a significant founder effect phenomenon for the fragile X syndrome (3 -15 ) with the mutation being enriched from two- to five-fold on specific DXS548, FRAXAC1 and FRAXAC2 haplotypes. Dependent upon the genetic homogeneity of the population under study (9 ,11 ,12 ), 50% of all fragile X chromosomes on average, share a few common genetic backgrounds (9 ,11 ,12 ). The remaining fragile X chromosomes either exist at equilibrium with common haplotypes in the population (10 ) or occur on rare haplotypic variants (10 ,28 ). These observations have prompted speculation that recurrent mutational events may occasionally occur on genetic backgrounds not at risk for the development of the disease, resulting in the formation of de novo fragile X lineages (10 ). Furthermore, based on multistep progression models for the progression of predisposed alleles to the disease state (25 -27 ), it has been predicted that among haplotypes in which clear linkage disequilibrium was demonstrated, one might expect an enrichment of alleles with large repeats among normal chromosomes of that haplotype. Surprisingly, only one of the three major fragile X haplotypes (Table 1 ; 2-1-3) demonstrates such an enrichment (Table 2 ). This suggests haplotype-specific influences other than total repeat length are playing a role in predisposing chromosomes to the development of the fragile X syndrome (10 ,13 ,29 ).
In order to test the hypothesis that the loss of AGG interruptions might resolve these two founder effect discrepancies, we compared the AGG interspersion patterns of 200 normal and 30 premutation and high-end repeat alleles within various haplotypes (Table 2 ). Unlike previous investigations (32 ), we have found a significant (P <0.002) non-random distribution between haplotype and FMR1 CGG repeat interspersion patterns. Furthermore, the use of all three markers was found to be more informative than any single haplotype marker. Phylogenetic reconstruction of CGG repeat mutational events within haplotype lineages, using association testing and parsimony analysis, predicts that the loss of the most 3' AGG interruption has occurred frequently and independently on many normal and at-risk fragile X haplotypes (Fig. 2 b,c,d, as examples). Parsimony analysis indicates that alleles in the normal population which have been subjected to the loss of AGG interruption belong to the same clades as alleles which are predisposed to instability and hyperexpansion (Fig. 2 ). Our data suggest that the loss of AGG interruptions likely accounts for the occurrence of the fragile X mutation among haplotypes found in equilibrium within the normal population (Table 1 and Fig. 2 c). The frequent loss of AGG interruptions, thus, may account for the considerable haplotype diversity of the fragile X mutation.
Parsimony analysis and intra-haplotype comparisons of 1-AGG and 2-AGG class of alleles (Table 3 and Fig. 2 b,c,d) confirms that there has been a preferential bias to lose the most 3' AGG interruption (Eichler, in press). This appears to have occurred by a mechanism which tends to preserve the overall length of the repeat (Table 3 ). Previously it was suggested that the loss of AGG interruptions may be mediated by one of several different mutational processes such as unequal-sister chromatid exchange, AGG deletion or AGG to CGG transversion (29 -32 ). Since the loss of AGG interruptions by non-reciprocal recombination events would be expected to alter the total number of CGG repeats (35 ), it is unlikely that this is the predominant mechanism. Our data would best support the deletion or conversion of an AGG to a CGG, resulting, perhaps, from deficiencies in mis-match repair processes (36 ,37 ). It should be noted, however, that not all 1-AGG and 2-AGG intra-haplotype comparisons show a conservation of total repeat length in the human population (6-4-6+ and 8-3-4+; Table 3 ). Indeed, parsimony analysis reveals that the loss of AGG interruptions may occur by the occasional deletion of an entire AGG(CGG)9 array (Fig. 2 c). Such an event could likely be mediated by non-reciprocal recombination processes.
Although the loss of AGG interruptions occurs frequently from a phylogenetic perspective, such losses could not occur with equal frequency among all haplotypes without completely obscuring founder effect phenomena. Differential propensities for the loss of AGG interruptions, then, must exist in order to account for observed haplotype linkage disequilibrium with the mutation (3 -14 ). Two haplotypes in this study (6-4-4 and 6-4-5) appear particularly prone to the loss of AGG interruptions. Parsimony analysis indicates that a substantial proportion of unstable and premutation alleles have likely progressed from a single AGG interruption ancestral state (Fig. 2 b,c), with the position of the 5' AGG interruption being remarkably conserved (Table 2 ). Not surprisingly, DXS548, FRAXAC1 and FRAXAC2 haplotypes equivalent to 6-4-5 and 6-4-4 have been shown to be enriched for fragile X chromosomes in virtually all human populations studied to date. These haplotypes generally account for 15-40% of all fragile X chromosomes among Japanese, American, Australian, Finnish, English and French populations (3 ,4 ,6 ,7 ,10 ,12 ,13 ). The strong association with the loss of AGG interruptions within these two haplotypes may reflect either an `ancient' or recurrent event; two possibilities which parsimony analysis can not distinguish. The conspicuous absence of high-end normal FMR1 CGG repeat alleles among these haplotypes (10 ,13 ,24 ,29 ), however, strongly argues against the former hypothesis, suggesting that alleles progress relatively rapidly through instability and hyperexpansion thresholds without the accumulation of large intermediate (S) alleles (25 ). Another argument favoring the recurrent loss of the 3' AGG interruption is the finding of global linkage disequilibrium among these haplotypes with the fragile X mutation. Conservative estimates based on multistep progression models of total repeat length (25 ) calculate that the longevity of predisposed (S) alleles in the population is approximately 90 generations or 2000 years (15 ). The Japanese and caucasian populations likely diverged from a common Homo sapiens stock, 150 000 to 200 000 years ago (38 -41 ). The most prosaic explanation, then, for linkage disequilibrium for the same haplotypes within these two populations would be a recurrent mutational event (such as the loss of an AGG interruption), rather than a common ancestral mutation.
Due to the absence of normal alleles with large total repeat lengths (>40 repeats) among the 6-4-5 and 6-4-4 haplotypes, it was originally speculated that a recurrent mutational event might allow alleles on these haplotypes to `leap-frog' toward instability and hyperexpansion thresholds associated with disease (10 ). Our analysis indicates that the loss of AGG interruptions could clearly provide the molecular basis for this `leap-frog' mutational event. Occurring by a mechanism which tends to maintain the overall length of the repeat, the loss of AGG interruptions would generate alleles which `leap-frog' in cognito by 10 and 12 repeat units toward their instability (35 pure repeats) and hyperexpansion thresholds (70 pure repeats) in a single generation (30 ). The longevity of such alleles due to increased pure repeat length would be dramatically reduced, obviating the accumulation of high-end normal repeat length intermediate alleles. The propensity for these chromosomes to frequently incur such mutations would confirm previous suggestions of haplotype-specific influences for these fragile X genetic backgrounds (29 ).
Although it is difficult at this point to identify the nature of these haplotype-specific influences, our analysis suggests that symmetry of repeat configurations may play a role in predisposing alleles to the loss of AGG interruptions. In both 6-4-5 and 6-4-4 haplotypes, significant associations were observed for configurations in which symmetry has been disrupted for the middle tract of the FMR1 CGG repeat (9+10+9 and 9+12+9; Table 2 ). Common symmetrical arrays such as 9+9+9 are conspicuously absent among these haplotypes. Parsimony analysis and association testing indicate that 9+12+9 and 9+10+9 have likely been the original precursors to unstable and premutation alleles within haplotypes 6-4-5 and 6-4-4; respectively (Table 3 ; Fig. 2 b,c). Conversely among haplotypes,1-1-3 and 7-4-6+, in which there is a strong association for symmetrical arrays (9+9+9), fragile X chromosomes have rarely been observed or are found in equilibrium within the normal population (Table 1 ). Interestingly, one allele with a large tract of pure repeats (9+26; Table 2 ) has been observed in haplotype 7-4-6+. Although far too few alleles exist in this haplotype to reconstruct an unambiguous phylogeny by parsimony, due to the similar conservation of repeat size, the 9+26 may have originated from the asymmetrical allele, 9+9+5+9, by the consecutive loss of 2-AGG interruptions (9+9+5+9 to 9+15+9 to 9+26 see Table 2 ; 7-4-6+ and 6-4-6+ haplotypes). Recent repeat length (13 ,28 ) and interspersion analysis surveys of Asian populations show a preponderance of 9+9+5+9 (35 total repeat) interspersion configurations on DXS548-FRAXAC1 haplotype (7-4) and FRAXAC1-FRAXAC2 haplotype (4-6+) (Composite 7-4-6+) (13 ,28 ). Among the Japanese, at least, 15% of all fragile X cases occur on the 4-6+ haplotype, suggesting that once again a correlation exists between asymmetrical array patterns, loss of AGG interruptions and progression to disease. An alternative explanation for the correlation between asymmetric interspersion configurations and the loss of AGG interruptions may however be that asymmetry is simply a consequence rather than a cause of increased instability on these particular haplotypes. If this were true, other cis-acting factors closely linked to the CGG repeat would have to be invoked to explain the propensity for haplotype 6-4-5 and 6-4-4 to lose AGG interruptions.
In many respects, the fragile X founder haplotype 2-1-3 (Table 1 ) represents the antithesis of 6-4-5 and 6-4-4. Computer simulation testing for non-random distributions of alleles within this haplotype reveals no significant association between haplotype and interspersion patterns, despite the fact that the majority of AGG configurations are unique. This lack of association is likely a reflection of the extreme FMR1 CGG repeat sequence heterogeneity of alleles within this haplotype (Table 2 ). As has been suggested earlier for the FRAXAC1 haplotype 1 (29 ), the positions of both AGG interruptions appear to have been highly conserved with most changes in repeat length occurring distal to the second 3' AGG interruption. Character state reconstructions of the 2-1-3 phylogenetic tree confirms that the 9+9+n interspersion pattern represents the ancestral configuration of this haplotype (Fig. 2 a). Parsimony analysis further suggests that 2-1-3 is the only haplotype in which FMR1 CGG repeat alleles have progressed toward instability and disease with the 2-AGG configuration. Although the loss of one or both AGG interruptions has occurred in association with instability in this haplotype, it should be noted that unique 2-AGG containing premutations have been observed among haplotypes closely allied to 2-1-3 (Table 4 ; haplotype 7-1-3 and 6-1-3). In toto, these observations suggest that the 2-1-3 haplotype has been particularly refractory to the loss of the most 3' AGG interruption.
The constraint of the 2-1-3 haplotype to maintain two AGG interruptions defines a second mutational pathway for the origin of the fragile X syndrome (Fig. 3 ). Unlike other founder haplotypes, the 2-1-3 haplotype initially has not jumped in increments of 10 CGG repeat increments toward instability and hyperexpansion thresholds by the loss of an AGG interruption. Most changes in repeat length have occurred distal to the last AGG interruption, likely in small increments of one or two repeat units, by a mechanism similar to replication slippage (42 ). As a result, a continuum of repeat lengths has been generated (Table 2 ). Due to the persistence of 2-AGG interruptions, longer total repeat lengths are attained before uninterrupted pure CGG repeat instability and hyperexpansion thresholds can be reached (30 ). This pathway accommodates the postulated multistep progression of alleles toward disease (25 ,26 ) and explains why the 2-1-3 founder chromosome is the only haplotype which shows an enrichment for high-end normal FMR1 CGG repeat alleles (3 -5 ).
Phylogenetic reconstruction of fragile X lineages based on parsimony analysis, association testing and intrahaplotype comparisons of interspersion patterns among normal and premutation alleles, reveals that the origins of the fragile X syndrome are complex. Our analysis predicts that fragile X mutations have arisen by both founder effects and a recurrent mutational event involving the loss of the most 3' AGG interruption. Among haplotypes which demonstrate linkage disequilibrium with the fragile X mutation (Table 1 ), two distinct mutational pathways have been delineated (Figs 2 a,b,c, 3 ). The DXS548-FRAXAC1-FRAXAC2 haplotype 2-1-3 (Fig. 1 ) appears refractory to the loss of AGG interruptions, progressing relatively slowly toward instability and hyperexpansion thresholds associated with the disease. As a result, the 2-1-3 haplotype is enriched for high-end repeat length alleles, generating a large founder pool of chromosomes which are predisposed to the development of disease (Fig. 3 ). In contrast, haplotypes 6-4-5 and 6-4-4, based on parsimony analysis, appear particularly prone to the loss of the most 3' AGG interruption from asymmetrical 2-AGG interspersion configurations (Fig. 2 b,c). This occurs by a mechanism which tends to preserve the total overall length of the repeat (Fig. 2 and Table 3 ). Due to the observation of unstable transmission in `normal' human pedigrees on this genetic background and the dearth of large intermediate alleles in random population surveys of these haplotypes, we propose that chromosomes which have lost an AGG interruption at the FRAXA locus progress rapidly toward the disease state (Fig. 3 ). The frequent and recurrent loss of AGG interruptions, is thus likely responsible for the observation of linkage disequilibrium with the 6-4-5 and 6-4-4 founder haplotypes (Fig. 3 ). Parsimony analysis, in addition, predicts that the loss of AGG interruptions has occurred independently on most DXS548-FRAXAC1-FRAXAC2 haplotypes. The sporadic loss of AGG interruptions and the genesis of de novo fragile X lineages (Figs 2 c, 3 ) can account for the observation of disease among haplotypes found in linkage equilibrium with the mutation and partially explain the considerable haplotype diversity observed with the fragile X syndrome. Although our model of founder effects and recurrent loss of AGG interruptions can adequately explain the origin of 70% of all fragile X chromosomes, further haplotype analysis, interspersion determination, and repeat-length transmission studies will be necessary to confirm our observations and to explain the frequent occurrence of fragile X syndrome on haplotypes rare in the normal population.
DNA samples from the normal control population, consisting of 200 unrelated non-fragile X male subjects, were collected from blood donors living in Wessex, Southern England as previously described (10 ). The remaining 30 premutation and large repeat-length normal alleles were derived from unrelated fragile X pedigrees or from families in which the intergenerational stability of the FMR1 CGG repeat had been documented (10 ,18 ,30 ,31 ,33 ). Six of the 15 premutation alleles were ascertained from the same population as the control samples (Wessex, England). The remaining premutation and unstable `normal' FMR1 CGG repeat alleles were collected from North American populations of caucasian ethnic origin (Table 3 ).
PCR amplification of the FMR1 CGG repeat was performed using a previously developed protocol which replaces exo(-) Pfu polymerase (Stratagene) with Taq DNA polymerase (48 ). In order to reconstruct the AGG substructure of the FMR1 CGG repeat, PCR products were digested with Mnl I restriction enzyme (New England Biolabs), electrophoresed, transferred to nylon membrane (GeneScreen plus), and probed with a [gamma]32P end- labelled (CGG)10 oligonucleotide as previously described (30 ). Based on the size of the FMR1 PCR product, the pattern of Mnl I digestion and the sizes of Mnl I fragments which hybridized to the CGG repeat oligonucleotide, the AGG interspersion configuration for each FMR1 CGG repeat allele was deduced. FMR1 CGG interspersion nomenclature is summarized in Figure 1 . In our convention, a `+' sign designates the position of an AGG interruption and the number refers to the length of uninterrupted repeats.
Three polymorphic microsatellite markers, DXS548, FRAXAC1 and FRAXAC2, which span 150 kb of the FMR1 CGG repeat locus were used to reconstruct the chromosomal haplotype of each normal, premutation and fragile X allele (Fig. 1 and Table 1 ). The genotypes for DXS548, FRAXAC1 and FRAXAC2 were determined using previously described PCR conditions and primers (4 ,6 ,10 ). For each polymorphic marker, alleles were designated using numbers in descending order of repeat length (Fig. 1 ). Due to the complex nature of the FRAXAC2 marker, which consists of two dinucleotide polymorphic repeat tracts and a polymorphic poly T tract (34 ) care was taken to distinguish between FRAXAC2 alleles which differ by a single bp. These are indicated in this study by a `+' sign. For example, a FRAXAC2 genotype `4+' is intermediate in size between genotype 3 and 4 differing by a single bp (Fig. 1 ). Composite haplotypes using all three markers are considered in this analysis and are configured based on the order of each marker (Fig. 1 ).
A total of 28 composite haplotypes (DXS548-FRAXAC1-FRAXAC2) were compared against 52 different AGG interspersion patterns (Table 2 ). The randomness of distribution between rows (AGG configuration) and columns (haplotypes) was tested using the association program ASSOC (AC, unpublished). Briefly, 5000 tables (replicates) were generated by computer simulation using row and column totals and assuming a random distribution between haplotype and interspersion pattern for Table 2 . A [chi]2 value was then calculated comparing these randomly-generated tables to the observed table (Table 2 ). Significant departures of randomness (P <0.002) for each cell entry (intersection between haplotype and interspersion pattern) were determined by analysing the individual ranges of distribution for each cell entry. The most significant association between haplotype and interspersion pattern (P <0.02) are summarized in Table 1 . In order to test the usefulness of the composite haplotype against each haplotype marker independently, three compressions of Table 2 were also generated (one for each marker) and non-random distributions were tested as described above (see Results).
PAUP (phylogenetic analysis using parsimony) version 3.1.1 (Illinois Natural History Survey) was employed to derive phylogenetic relationships among various AGG interspersion FMR1 CGG repeat patterns within each haplotype. Only those haplotypes which exhibited four or more different interspersion patterns were analysed due to software requirements for at least four distinct taxa. Deduced FMR1 CGG repeat sequences were encoded (Let C = CGG and A = AGG) to eliminate the possibility of gap formation within a trinucleotide repeat unit. Alignment of encoded data was performed using the ClustalW software package (default setting: gap penalty = 15.0 and gap extension penalty = 1.0). Reduction of gap penalty assignments below default setting parameters resulted in the generation of sequence alignments which were inconsistent with characteristic intra- haplotype AGG interspersion patterns. Phylogenetic trees were generated from aligned sequences using heuristic and exhaustive tree searches (Fig. 2 ). In cases where multiple, equally parsimonious trees were generated, a consensus tree was constructed using majority-rule (>50%) methods (Fig. 2 d). Ancestral states were defined using the ANCSTATES command under the assumptions block of PAUP (49 ) and were chosen using the most significantly associated interspersion configuration (P <0.002) within each haplotype. When no significant association could be found (haplotype 2-1-3), the likely ancestral state was deduced using the character-state reconstruction option (Fig. 2 a). When ambiguous character states were obtained (i.e. either the presence or absence of a CGG trinucleotide were equally parsiminious), ancestral states were reconstructed excluding these character assignments. In such situations, the shortest, most parsimonious ancestral state was considered.
We would like to thank Drs P. Ward, B. Popowich, J. Holden, A. Reiss, S. Richards and S. Thibodeau for providing some of the DNA samples used in this study. This work was supported in part by NIH grants (HD29256 and GM52982) to DLN.
1 Richards, R.I., Holman, K., Kozman, H., Kremer, E., Lynch, M., Pritchard, M., Yu, S., Mulley, J. and Sutherland, G.R. (1991) Fragile X syndrome: genetic localisation by linkage mapping of two microsatellite repeats FRAXAC1 and FRAXAC2 which immediately flank the fragile site. J. Med. Genet.28, 818-23.MEDLINE Abstract
2 Riggins, G.J., Sherman, S.L., Oostra, B.A., Sutcliffe, J.S., Feitell, D., Nelson, D.L., van Oost, B.A., Smits, A.P.T., Ramos, F.J., Pfendner, E., Kuhl, D.P.A., Caskey, C.T. and Warren, S.T. (1992) Characterization of a highly polymorphic dinucleotide repeat 150 kb proximal to the fragile X site. Am. J. Med. Genet.43, 237-43.MEDLINE Abstract
3 Richards, R.I., Holman, K., Friend, K., Kremer, E., Hillen, D., Staples, A., Brown, W.T., Goonewardena, P., Tarleton, J., Schwartz, C. and Sutherland, G.R. (1992) Evidence of founder chromosomes in fragile X syndrome. Nature Genet.1, 257-60.MEDLINE Abstract
4 Oudet, C., Mornet, E., Serre, J.L., Thomas, F., Lentes-Zengerling, S., Kretz, C., Deluchat, C., Tejada, I., Boue, J., Boue, A. and Mandel, J.L. (1993) Linkage diequilibrium between the fragile X mutation and two closely linked CA repeats suggests that Fragile-X chromosomes are derived from a small number of founder chromosomes. Am. J. Hum. Genet.52, 297-304.MEDLINE Abstract
5 Buyle, S., Reyniers, E., Vits, L., De Boulle, K., Handig, I., Wuyts, F.L.E., Deelen, W., Halley, D.J.J., Oostra, B.A. and Willems, P.J. (1993) Founder effect in a Belgian-Dutch fragile X population. Hum. Genet.92, 269-72.MEDLINE Abstract
6 Jacobs, P.A., Bullman, H., Macpherson, J., Youings, S., Rooney, V., Watson, A. and Dennis, N.R. (1993) Population studies of the fragile (X): a molecular approach. J. Med. Genet.30, 454-59.MEDLINE Abstract
7 Hirst, M.C., Knight, S.J.L., Christodoulou, Z., Grewal, P.K., Fryns, J.P. and Davies, K.E. (1993) Origins of the fragile X syndrome mutation. J. Med. Genet.30, 647-50.MEDLINE Abstract
8 Arinami, T., Asano, M., Kobayashi, K., Yanagi, H. and Hamaguchi, H. (1993) Data on the CGG repeat at the fragile X site in the non-retarded Japanese population and family suggest the presence of a subgroup of normal alleles predisposing to mutate. Hum. Genet.92, 431-6.MEDLINE Abstract
9 Haataja, R., Vaisanen, M.L., Li, M., Ryynanen, M. and Leisti, J. (1994) The fragile X syndrome in Finland: demonstration of a founder effect by analysis of microsatellite haplotypes. Hum. Genet.94, 479-83.MEDLINE Abstract
10 Macpherson, J.N., Bullman, H., Youings, S.A. and Jacobs, P.A. (1994) Insert size and flanking haplotype in fragile X and normal populations: possible multiple orgins for the fragile X mutation. Hum. Mol. Genet.3, 399-405.MEDLINE Abstract
11 Malmgren, H., Gustavson, K.H., Oudet, C., Holmgren, G., Petterson, U. and Dahl, N. (1994) Strong founder effect for the fragile X syndrome in Sweden. Eur. J. Hum. Genet.2, 103-9.MEDLINE Abstract
12 Oudet, C., von Koskull, H., Nordstrom, A.M., Peippo, M. and Mandel, J.L. (1993) Striking founder effect for the fragile X syndrome in Finland. Eur. J. Hum. Genet.1, 181-189.MEDLINE Abstract
13 Richards, R.I., Kondo, I., Homan, K., Yamauchi, M., Seki, N., Kunikazu, K., Sutherland, G.R. and Hori, T. (1994) Haplotype analysis at the FRAXA locus in the Japanese population. Am. J. Med. Genet.51, 412-16.MEDLINE Abstract
14 Zhong, N., Ye, L., Dobkin, C. and Brown, W.T. (1994) Fragile X founder chromosome effects: Linkage disequilibrium or microsatellite heterogeneity? Am. J. Med. Genet.5, 405-11.
15 Chakravarti, A. (1992) Fragile X founder effect? Nature Genet.1, 237-8.MEDLINE Abstract
16 Smits, A.P.T., Dreesen, J.C.F.M., Smeets, D.F.C.M., de Die-Smulders, C., Spaans-van der Bijl, T., Govaerts, L.C.P., Warren, S.T., Oostra, B.A. and van Oost, B.A. (1992) The fragile X syndrome: no evidence for any recent mutations. J. Med. Genet.30, 94-6.
17 Drugge, U., Holmgren, G., Blomquist, H.K., Dahl, N., Gustavson, K.H. and Malmgren, H. (1992) A study of individuals possibly affected with the fragile X syndrome in a large Swedish family in the 18th and 20th centuries. Am. J. Med. Genet.43, 353-4.MEDLINE Abstract
18 Fu, Y.H., Kuhl, D.P.A., Pizzuti, A., Pieretti, M., Sutcliffe, J.S., Richards, S., Verkerk, A.J.M.H., Holden, J.J.A., Fenwick, R.G., Jr., Warren, S.T., Oostra, B.A., Nelson, D.L. and Caskey, C.T. (1991) Variation of the CGG repeat at the fragile X site results in genetic instability: resolution of the Sherman paradox. Cell67, 1047-58.MEDLINE Abstract
19 Oberle, I., Rousseau, F., Heitz, D., Kretz, C., Devys, D., Hanauer, A., Boue, J., Bertheas, M.F. and Mandel, J.L. (1991) Instability of a 550-base pair DNA segment and abnormal methylation in fragile X syndrome. Science252, 1097-102.MEDLINE Abstract
20 Yu, S., Pritchard, M., Kremer, E., Lynch, M., Nancarrow, J., Baker, E., Holman, K., Mulley, J.C., Warren, S.T., Schlessinger, D. et al. (1991) Fragile X genotype characterized by an unstable region of DNA. Science252, 1179-81.MEDLINE Abstract
21 Heitz, D., Rousseau, F., Devys, D., Saccone, S., Abderrahim, H., le Paslier, D., Cohen, D., Vincent, A., Toniolo, D., Della Valle, G., Johnson, S., Schlessinger, D., Oberlé, I. and Mandel, J.L. (1991) Isolation of sequences that span the fragile X and identification of a fragile X-related CpG island. Science251, 1236-9.MEDLINE Abstract
22 Kremer, E.J., Yu, S., Pritchard, M., Nagaraja, R., Heitz, D., Lynch, M., Baker, E., Hyland, V.J., Little, R.D., Wada, M. et al. (1991) Isolation of a human DNA sequence which spans the fragile X. Am. J. Hum. Genet.49, 656-61.MEDLINE Abstract
24 Snow, K., Doud, L.K., Hagerman, R., Pergolizzi, R.G., Erster, S.H. and Thibodeau, S.N. (1993) Analysis of a CGG sequence at the FMR-1 locus in the fragile X families and in the general population. Am. J. Hum. Genet.53, 1217-28.MEDLINE Abstract
25 Morton, N.E. and Macpherson, J.N. (1992) Population genetics of the fragile x syndrome: a multiallelic model for the FMR1 locus. Proc. Natl Acad. Sci. USA89, 4215-17.MEDLINE Abstract
26 Morris, A., Morton, N.E., Collins, A., Macpherson, J., Nelson, D.L. and Sherman, S. (1995) An n-allele model for progressive amplifcation in the FMR1 locus. Proc. Natl Acad. Sci. USA92, 4833-7.MEDLINE Abstract
27 Kolehmainen, K. (1994) Population genetics of fragile X: a multiple allele model with variable risk of CGG repeat expansion. Am. J. Med. Genet.51, 428-35.MEDLINE Abstract
28 Zhong, N., Liu, X., Gou, S., Houck, G.E., Li, S., Dobkin, C. and Brown, W.T. (1994) Distribution of FMR1 and associated microsatellite alleles in a normal Chinese population. Am. J. Med. Genet.51, 417-22.MEDLINE Abstract
29 Kunst, C.B. and Warren, S.T. (1994) Cryptic and polar variation of the fragile X repeat could result in predisposing normal alleles. Cell77, 853-61.MEDLINE Abstract
30 Eichler, E.E., Holden, J.J.A., Popovich, B.W., Reiss, A.L., Snow, K., Thibodeau, S.N., Richards, C.S., Ward, P.A. and Nelson, D.L. (1994) Length of uninterrupted CGG repeats determines stability in the FMR1 gene. Nature Genet.8, 88-94.MEDLINE Abstract
31 Snow, K., Tester, D.J., Kruckenberg, K.E., Schaid, D.J. and Thibodeau, S.N. (1994) Sequence analysis of the fragile X trinucleotide repeat: implications for the origin of the fragile X mutation. Hum. Mol. Genet.3, 1543-51.MEDLINE Abstract
32 Hirst, M.C., Grewal, P.K. and Davies, K.E. (1994) Precursor arrays for triplet repeat expansion at the fragile X locus. Hum. Mol. Genet.3, 1553-60.MEDLINE Abstract
33 Reiss, A.L., Kazazian, H.H., Jr., Krebs, C.M., McAughan, A., Boehm, C.D., Abrams, M.T. and Nelson, D.L. (1994) Frequency and stability of the fragile X premutation. Hum. Mol. Genet.3, 393-8.MEDLINE Abstract
34 Zhong, N., Dobkin, C. and Brown, W.T. (1993) A complex mutable polymorphism located within the fragile X gene. Nature Genet.5, 248-53.MEDLINE Abstract
35 Dover, G. (1982) Molecular drive: a cohesive mode of species evolution. Nature299, 111-17.MEDLINE Abstract
36 Fishel, R., Lescoe, M.K., Rao, M.R.S., Copeland, N.G., Jenkins, N.A., Garber, J., Kane, M. and Kolodner, R. (1993) The human mutator gene homolog MSH2 and its association with hereditary nonpolyposis colon cancer. Cell75, 1027-38.MEDLINE Abstract
37 Liu, B., Nicolaides, N.C., Markowitz, S., Willson, J.K.V., Parsons, R.E., Jen, J., Papadopolous, N., Peltomaki, P., de la Chapelle, A., Hamilton, S.R., Kinzler, K.W. and Vogelstein, B. (1995) Mismatch repair gene defects in sporadic colorectal cancers with microsatellite instability. Nature Genet.9, 48-55.MEDLINE Abstract
38 Dorit, R.L., Akashi, H. and Gilbert, W. (1995) Absence of polymorphism at the ZFY locus on the human Y chromosome. Science268, 1183-5.MEDLINE Abstract
39 Horai, S., Hayasaka, K., Kondo, R., Tsugane, K. and Takahata, N. (1995) Recent African origin of modern humans revealed by complete sequences of hominoid mitochondrial DNAs. Proc. Natl Acad. Sci. USA92, 532-6.MEDLINE Abstract
40 Paabo, S. (1995) The Y chromosome and the origin of all of us (men). Science268, 1141-2.MEDLINE Abstract
41 Bowcock, A.M., Ruiz-Linares, A., Minch, E., Kidd, J.R. and Cavalli-Sforza, L.L. (1994) High resolution of human evolutionary trees with polymorphic microsatellites. Nature368, 455-7.MEDLINE Abstract
42 Levinson, G. and Gutman, G.A. (1986) Slipped-strand mispairing: a major mechanism for DNA sequence evolution. Mol. Biol. Evol.4, 203-21.
43 Rousseau, F., Morgan, K., Rouillard, P. and Khandjian, E.W. (1995) Prevalence of carriers of premutation-size alleles of the FMR1 gene and implications for the populaton genetics of the fragile X syndrome. Am. J. Hum. Genet. in press.MEDLINE Abstract
44 Fry, M. and Loeb, L.A. (1994) The fragile X syndrome d(CGG)n nucleotide repeats form a stable tetrahelical structure. Proc. Natl Acad. Sci. USA91, 4950-4.MEDLINE Abstract
45 Gacy, A., Goellner, G., Juranic, N., Macura, S. and McMurray, C. (1995) Trinucleotide repeats that expand in human disease form hairpin structures in vitro. Cell81, 533-40.MEDLINE Abstract
46 Kuryavyi, V.V. and Jovin, T.M. (1995) Triad-DNA: a model for trinucleotide repeats. Nature Genet.9, 339-41.MEDLINE Abstract
47 Smith, S.S., Laayoun, A., Lingeman, R.G., Baker, D.J. and Riley, J. (1994) Hypermethylation of telomere-like foldbacks at codon 12 of the human c-Ha-ras gene and the trinucleotide repeat of the FMR1 gene of fragile X. J. Mol. Biol.243, 143-51.MEDLINE Abstract
48 Chong, S.S., Eichler, E.E., Hughes, M.R. and Nelson, D.L. (1994) Robust amplification of the fragile X syndrome CGG repeat using Pfu polymerase: ethidium bromide detection of normal and premutation alleles. Am. J. Med. Genet.51, 522-6.MEDLINE Abstract
49 Swofford, D.L. (1993) PAUP: Phylogenetic Analysis using Parsimony. Illinois Natural History Survey, Champaign, Illinois.
*To whom correspondence should be addressed
This page is maintained by OUP admin. Last updated Thu Oct 31 15:22:29 GMT 1996. Part of the OUP Journals World Wide Web service.Copyright Oxford University Press, 1996
