The CMT1A-REPs are two large directly repeating DNA sequences located on chromosome 17p11.2-p12 flanking the region duplicated in patients with Charcot-Marie-Tooth disease type 1A (CMT1A) and deleted in patients with hereditary neuropathy with liability to pressure palsies (HNPP). We have sequenced two cosmids, c74F4 and c15H12, which contain the entire proximal and distal CMT1A-REPs and determined that these repeats are ~99% identical across a 24 011 bp region. In addition, both contain an exon of the human heme A:farnesyltransferase gene (COX10). Hybridization studies revealed that COX10 spans the distal CMT1A-REP, while the proximal CMT1A-REP contains an isolated COX10 `pseudo-exon'. There is also a COX10 hybridization signal on chromosome 10 which appears to represent a processed pseudogene. We propose that the distal CMT1A-REP represents the progenitor copy of COX10 exon VI which was duplicated with surrounding intronic sequences during mammalian genome evolution and that the HNPP deletion results in a COX10 null allele.
The CMT1A-REP DNA sequence represents a large (>20 kb) highly similar repeat which flanks the 1.5 Mb region duplicated in Charcot-Marie-Tooth disease type 1A (CMT1A) and deleted in hereditary neuropathy with liability to pressure palsies (HNPP) (1 -3 ). The proximal and distal CMT1A-REP repeats on chromosome 17p are directly involved in the unequal crossing over event resulting in the 1.5 Mb tandem CMT1A duplication and 1.5 Mb HNPP deletion (2 ,4 ) associated with these two genomic diseases. We consider CMT1A and HNPP genomic diseases, more than single gene disorders, in the sense that they are usually caused by DNA rearrangements resulting in an alteration of the genome with a concomitant gene dosage effect, rather than mutation or disruption of a protein coding sequence.
Approximately 70% of inherited and 90% of sporadic cases of CMT type 1 are associated with the 1.5 Mb CMT1A duplication (5 -7 ), while >84% of unrelated and 86% of sporadic HNPP patients have the 1.5 Mb HNPP deletion (7 ). DNA probes from within the CMT1A-REPs have been used extensively to diagnose CMT1A or HNPP through the detection of rearrangement-specific junction fragments (8 ,9 ). Recently a hotspot for homologous recombination was detected within the CMT1A-REPs (10 ) and several groups have now confirmed that strand exchange events between these large misaligned directly repeating units consistently occur within a small region of the repeat in multiple unrelated CMT1A and HNPP patients of various ethnic backgrounds (11 -14 ). The current generally accepted model is that the CMT1A duplication and HNPP deletion are the reciprocal products of the same unequal crossing over event between misaligned CMT1A-REPs (4 ,10 ). The clinical consequences of these two molecular rearrangements and the fact that a single dosage-sensitive gene (PMP22) located between the CMT1A-REPs is responsible for both CMT1A via trisomic overexpression and HNPP via monosomic underexpression have been reviewed elsewhere (15 -18 ).
The frequency of recombination events between regions of the human genome can be directly correlated with the degree of homology between these regions (19 ). Such homologous recombination events occur in three stages: (i) misalignment of repetitive sequences; (ii) initiation of strand exchange; (iii) resolution of the Holliday structure. Although much has been published about the degree of homology between portions of the CMT1A-REPs (2 ,10 ,12 ,20 ), the exact size and overall extent of sequence identity is unknown. Recently a mariner DNA transposon has been hypothesized to play a role in the initiation of double-strand breaks near the hotspot (10 ), but how this remarkably precise recombination event occurs consistently within a 1.7 kb hotspot (10 ,14 ) remains to be elucidated.
In this study we have determined the extent and degree of homologous sequence between the proximal and distal CMT1A-REP elements and identified sequence features within and around the repeats which may influence homologous recombination at the hotspot. As a result of this analysis, an internal exon of the human COX10 gene was identified in the distal CMT1A-REP and a `pseudo-exon' in the proximal CMT1A-REP. We demonstrate that the COX10 gene spans the distal CMT1A-REP, providing definitive evidence that the distal repeat is the progenitor copy and that the molecular rearrangement resulting in the CMT1A duplication and HNPP deletion disrupts one copy of the COX10 gene on the recombinant chromosomes.
We previously identified cosmids spanning the proximal (c74F4) and distal (c15H12) CMT1A-REPs which were used to estimate the size of the repeat at 25 +- 5 kb by EcoRI restriction mapping and cross-hybridization of specific restriction fragments (2 ,10 ,20 ). To determine the exact sizes of the proximal and distal CMT1A-REPs, a shotgun sequencing strategy was used to elucidate the complete nucleotide sequence of these two cosmids with double-strand coverage (21 ,22 ). Figure 1 is a graphical representation of the results of a RepeatMasker (23 ) analysis used to identify SINE, Alu, MIR, MER, LINE, Transposon and AT-rich regions of the proximal and distal CMT1A-REPs. Note that this region is rich in both Alu elements and transposons. In addition to the mariner DNA transposon previously identified near the homologous recombination hotspot within the CMT1A-REPs (10 ), there are also remnants of a retrotransposon (THE1) and another class of DNA transposon (TIGGER1) on the telomeric side of the distal CMT1A-REP (Fig. 1 ).
A COX10 cDNA HindIII fragment from pG19/HT1 (27 ) containing the complete open reading frame (ORF) and the majority of the 3'-untranslated region (UTR) was hybridized to both PstI and BamHI-digested human monochromosomal hybrid panels (Oncor), identifying several restriction fragments on chromosome 17 and at least one hybridization signal on chromosome 10 (data not shown). Somatic cell hybrid mapping using a panel consisting of EcoRI-digested DNA from various deletions of human chromosome 17p in rodent backgrounds (2 ) with this COX10 cDNA probe further localized the gene to the CMT1A region (data not shown). Southern blot analysis was performed on EcoRI-digested YAC and mega-YAC DNA from previously published contigs (2 ,10 ,28 ) which span the proximal CMT1A-REP (yc49H7, yc915C12 and yc818H9), 1.5 Mb CMT1A duplication locus (yc804F11) and distal CMT1A-REP (yc225A3 and yc961F10). An ~1100 bp fragment was detected in all YACs and mega-YACs which contain the proximal and distal CMT1A-REPs (Fig. 3 A, lanes 4-6, 8 and 9) as well as human genomic DNA (Fig. 3 A, lane 10) and the chromosome 17 monochromosomal hybrid (Fig. 3 A, lane 14). The size of this fragment is consistent with the predicted size of 1031 bp for EcoRI fragments containing the internal COX10 exon identified in c74F4 and c15H12. Eleven additional fragments were detected in yc961F10, which spans the distal CMT1A-REP (Fig. 3 A, lane 9). Six bands detected in yc961F10 could be directly correlated with hybridization signals observed in human genomic DNA (Fig. 3 A, lane 9 versus lane 10) and originate from chromosome 17, as indicated by their presence in the chromosome 17 monochromosomal hybrid (Fig. 3 A, lane 14). In addition, other hybridization signals were detected in the YACs that did not match any bands in human genomic DNA (Fig. 3 A, lanes 4-5 or 6-9 versus lane 10). These signals appear to be a cross-hybridization of the human COX10 cDNA probe with the yeast genome, since they are present: (i) in total yeast genomic DNA (Fig. 3 A, lane 1); (ii) in a preparation containing a YAC from human chromosome 2 (Fig. 3 A, lane 2); (iii) in cosmid clone c8145 (29 ), which spans the Saccharomyces cerevisiae (yeast) cox10 gene (30 ; Fig. 3 A, lane 3); (iv) when individual COX10 exon PCR products are used as probe (data not shown). Therefore, if one adjusts for these background hybridization signals, only the 1031 bp fragment was detected in clones spanning the proximal CMT1A-REP region (Fig. 3 A, lanes 4-6). We have no compelling explanation, however, as to why the background signals appear somewhat different in multiple independent DNA preparations of YAC yc49H7 and yc915C12.
An additional hybridization signal was detected in human genomic DNA but not in YACs spanning the distal CMT1A-REP on chromosome 17 or the monochromosomal 17 hybrid MH22-6 (~3500 bp band, Fig. 3 A, lane 10 versus lanes 4-9 and 14). Since an additional hybridization signal was detected on chromosome 10 when the COX10 cDNA probe was used to screen the hybrid panel containing all human chromosomes (data not shown), DNA from a human chromosome 10-retaining hybrid in a hamster background was digested with EcoRI and included on the blot in Figure 3 A (lane 15), along with a hamster DNA control (lane 17). An ~3500 bp band was detected in human genomic DNA (Fig. 3 A, lanes 10-13) and the chromosome 10-retaining hybrid (Fig. 3 A, lane 15), but not detected in the hamster DNA control (Fig. 3 A, lane 17) nor in the chromosome 17 monochromosomal hybrid (Fig. 3 A, lane 14). This ~3500 bp hybridization signal was also detected in the chromosome 10 monochromosomal hybrid and human genomic DNA by PCR products from exons I, II and III, in addition to the complete cDNA, as a probe (data not shown). These data suggest that the hybridization signal originating from chromosome 10 may be a processed COX10 pseudogene which is identified by probes from exons I-III of the COX10 gene.
Since mapping (this study) and intron/exon sequencing studies (31 ) indicated that COX10 coding exons were contained within both CMT1A-REPs (exon VI) and internal to the duplication/deletion region (exon VII), we predicted that the homologous recombination event between misaligned CMT1A-REPs resulting in the CMT1A duplication and HNPP deletion would disrupt the COX10 gene (Fig. 4 ). Comparison of EcoRI-digested genomic DNA probed with the COX10 cDNA by Southern analysis of unaffected controls (2*), homozygous CMT1A duplication (4*), heterozygous CMT1A duplication (3*) and HNPP deletion (1*) individuals clearly indicates dosage differences in the ~900 and ~1800 bp fragments containing COX10 exon VII, which map between the CMT1A-REPs (Fig. 3 A, lanes 10-13). The 1031 bp cross-hybridizing fragment containing exon VI located within CMT1A-REP is predicted to have 4* copies in unaffected controls, 6* in homozygous CMT1A duplication and 3* in HNPP deletion individuals. These dosage differences support the contention that the COX10 gene is transcribed across the distal CMT1A-REP and that the exons of this gene are duplicated or deleted by the same unequal cross-over event which results in the CMT1A duplication and HNPP deletion.
The CMT1A-REPs are 98.7% identical over a 24 011 bp region and contain an internal exon of the COX10 gene. By sequencing two cosmids containing the proximal and distal CMT1A-REP repeats we were able to definitively show that the CMT1A duplication and HNPP deletion are generated by a strand exchange event between homologous misaligned, directly repeating 24 011 bp units (CMT1A-REPs). The regions surrounding both CMT1A-REPs appear to be an active area of the genome, as demonstrated by the insertion of three transposons and multiple MIR, MER, Alu and LINE elements (Fig. 1 ). Boundaries of the CMT1A-REPs could clearly be established by direct sequence comparisons of cosmids c74F4 (proximal) and c15H12 (distal) (Fig. 2 ). Interestingly, a coding internal exon of the human COX10 gene was identified in both CMT1A-REPs (Fig. 1 ). We had previously shown that at least one exon of the COX10 gene maps within the region between the CMT1A-REPs (32 ) but were surprised to find by direct sequence analysis an exon within both CMT1A-REPs. Further hybridization studies using COX10 cDNA probes on digested YAC DNA from the CMT1A locus indicated that this gene begins telomeric to the distal CMT1A-REP and spans the repeat, ending just centromeric to the distal CMT1A-REP (Fig. 3 B).
The clustering of COX10 exons around the distal CMT1A-REP clearly indicates that this repeat is the primordial copy of the CMT1A-REP and that the exon located in the proximal CMT1A-REP is a `pseudo-exon' created by the initial genomic duplication event. A similar conclusion regarding the evolutionary origin for the duplicated CMT1A-REP copy was proposed based on other analyses (12 ), but our identification of the COX10 gene spanning the distal CMT1A-REP and only a `pseudo-exon' in the ~2 Mb region surrounding the proximal CMT1A-REP provides substantive evidence that the distal copy of CMT1A-REP is the progenitor copy. These data are consistent with the original observation that the proximal CMT1A-REP is embedded in a complex mosaic repeat on chromosome 17p (2 ) and appears to have landed in this location as a direct result of the initial duplication of the distal CMT1A-REP. In addition, hybridization of the COX10 cDNA probe to EcoRI-digested DNA from several primate species (data not shown) indicates that exon VI of the COX10 gene, which is embedded in the CMT1A-REP, is present in only two copies in gorilla but four copies in chimpanzee and humans. This observation is consistent with the conclusions of Kiyosawa et al. that the CMT1A-REP was duplicated during primate genome evolution perhaps some time between the speciation from gorilla (Gorilla gorilla) to chimpanzee (Pan troglodytes) (12 ).
The yeast cytochrome oxidase (COX) complex is composed of three mitochondrial encoded subunits but is dependent on multiple nuclear encoded genes for proper assembly and function (33 ). Two such nuclear encoded genes are cox10 and cox11, responsible for the biosynthesis of the heme A prosthetic group of the cytochrome oxidase complex (34 ). A human homolog of the COX10 gene was cloned by rescuing the yeast cox10 respiration-deficient phenotype with a clone from a heart cDNA library (27 ). Based on the position of COX10 exon-containing fragments both outside and within the region which is duplicated in CMT1A and deleted in HNPP patients one may anticipate another consequence of the HNPP deletion through disruption of one copy of the COX10 gene. Although the phenotype associated with the HNPP deletion is a rather mild transient neuropathy (3 ,15 ,17 ), conceptual translation of the shuffled exons of the COX10 gene on the deleted chromosome indicates that this protein would be missing 130 amino acids from the C-terminus, including several catalytic domains conserved in protoheme farnesyltransferases from Escherichia coli and yeast (27 ,35 ), suggesting that this would result in a null allele of COX10 (Fig. 4 ). If an individual with the HNPP deletion also has a deleterious mutation in the COX10 allele on the non-deleted chromosome, then a severe COX deficiency phenotype should result.
COX deficiency has been associated with a variety of recessively inherited syndromes (36 ,37 ). To date, underlying molecular defects have only been found in a mitochondrial encoded gene (38 ), although there is indirect evidence that mutations in a nuclear encoded gene can result in some cases of Leigh syndrome (39 ). Mutations of a nuclear succinate dehydrogenase gene can also result in mitochondrial respiratory chain deficiency presenting as Leigh syndrome (40 ). It is therefore simplistic to propose that all individuals with COX10 deficiency will present clinically with Leigh syndrome, but it is plausible that a subset of patients with a clinical diagnosis of Leigh syndrome may have an underlying defect in the COX10 gene. Patients with the HNPP deletion in conjunction with a mutation in the COX10 allele on the homologous chromosome may, therefore, present with a more complicated clinical phenotype involving a mitochondrial myopathy in addition to a neuropathic process. Alternatively, what may appear to be a consequence of heteroplasmy (the involvement of isolated tissues due to unequal segregation of affected mitochondria) in these individuals with the HNPP deletion could actually be the result of somatic mutations in their functional COX10 allele in the early stages of development and a mosaic distribution of affected cells. If any of these potential possibilities exist, it may partially explain the differences in observed incidence between HNPP and the more common CMT1A, despite the rearrangements representing products of the same mutational event. Although the majority of individuals with the HNPP deletion probably remain undiagnosed due to ascertainment bias secondary to the mild phenotype (15 ), a compound heterozygote containing a null COX10 allele due to HNPP deletion may present with a more severe phenotype that masks the usually mild HNPP neuropathy or may even result in embryonic or perinatal lethality, depending on whether COX10 functions solely in COX assembly in humans.
The recombinant CMT1A-REP in CMT1A duplication patients also contains portions of the COX10 gene, as indicated by dosage differences in Figure 3 A and illustrated in Figure 4 . This rearranged COX10 gene lacks a promoter region and contains mostly 3'-UTR and thus would not be expected to produce a COX10 protein product. However, two normal copies of this gene would still be present in CMT1A duplication patients (Fig. 4 ).
In summary, we have shown that the CMT1A-REPs are ~99% identical sequences 24 011 bp in length. The CMT1A-REPs contain a COX10 coding exon and homologous reciprocal recombination results in the rearrangement of one copy of the essential COX10 gene. Extensive documentation exists in the literature that the CMT1A and HNPP demyelinating phenotypes result from dosage effects of the peripheral myelin protein 22 gene (PMP22) (reviewed in 15 -18 ). The data presented in this paper are not inconsistent with the dosage model, however, our present findings suggest a further complexity to the consequences of DNA rearrangements in the CMT1A/HNPP region and this may have ramifications for phenotypic expression in individuals with the HNPP deletion.
Cosmids c74F4 (proximal CMT1A-REP) and c15H12 (distal CMT1A-REP) were isolated from an arrayed flow-sorted chromosome 17-specific library as described (2 ,10 ,41 ). Shotgun sequencing of both cosmids was performed by standard methods (42 ,43 ) using automated fluorescent DNA sequencing chemistry. A gap in the 1.9 kb EcoRI end fragment from c74F4 could not be closed by this method. This fragment was subcloned into the low copy number plasmid vector pACYC184 and propagated in the recA- E.coli K-12 strain HMS174 (recA1 hsdR rifr) to obtain plasmid DNA template for gap closure. Sequencing reads were assembled using XGAP and the consensus sequences deposited in GenBank (21 ,22 ).
Consensus sequences were converted to Pearson/Fasta format and repetitive elements were identified and masked from the sequence using RepeatMasker (23 ). Masked and unmasked sequences were split into overlapping 4000 bp segments. These segments were transmitted for GRAIL, BLASTn and BEAUTY analysis via the BCM search launcher (24 ), available through IMGEN in the Molecular and Human Genetics Department (http://gc.bcm.tmc.edu:8088/search-launcher/launcher.html). The region of sequence alignment between CMT1A-REPs was identified using the programs GAP and BESTFIT (gap weight = 0.0, length weight = 1.0) from the Genetics Computer Group software package (44 ).
Cosmids, YACs and mega-YACs and monochromosomal hybrids used in this study have been previously published or are available in the public domain. Cosmid c8145 (29 ), which spans the yeast cox10 gene was obtained from the American Type Culture Collection (Baltimore, MD). CMT1A and HNPP patient samples have been described elsewhere (1 ,10 ). Cosmid, YAC, human and primate DNA samples were digested with restriction enzymes according to the manufacturers' instructions (New England Biolabs and Boehringer Mannheim). The Southern blot protocol used has been described (10 ). The hybridization illustrated herein was performed using a gel-purified 2.8 kb HindIII fragment from plasmid pG19/HT123 containing the complete ORF and the majority of the 3'-UTR of the COX10 cDNA. All COX10 cDNA cross-hybridizing fragments identified in genomic DNA Southern blots were confirmed using a probe consisting of a PCR fragment synthesized with primers to the ORF of COX10 (5'-ATGGCCGCATCTCCGCACAC-3' and 5'-TCAGCTGGGAGGGGGCCCTGC-3') and plasmid pG19/HT1 as template. To analyze the signals from chromosome 10 and to obtain the 5' -> 3' order of the COX10 exon-containing EcoRI restriction fragments illustrated in Figures 3 B and 4 these filters were stripped by boiling in 2* SSC + 0.1% SDS for 30 min and then rehybridized independently with a PCR product from each of exons I-VII of the COX10 gene amplified from YAC yc961F10 and cosmid c20G2 (data not shown).
We thank Dr Alexander Tzagoloff for providing COX10 cDNA clone pG19/HT1, Ying Shen and Patti Engler for their expert technical assistance, K.-S.Chen for providing purified mega-YAC clones and Bill Craigen and Lisa Shaffer for critical reviews. T.M. is supported by a Muscular Dystrophy Association (MDA) post-doctoral fellowship. This work was supported in part by National Institute of Neuromuscular Disorders and Strokes (NIH) (RO1NS2742) and MDA grants to J.R L. as well as a National Institute for Human Genome Research (NIHGR) grant to R.A.G.
*To whom correspondence should be addressed. Tel: +1 713 798 6530; Fax: +1 713 798 5073; Email: jlupski@bcm.tmc.edu
Human Molecular Genetics
Pages
Introduction
Results
Sequencing, analysis and size determination of the proximal and distal CMT1A-REPs
Physical mapping of COX10 exon-containing EcoRI restriction fragments to chromosome 17
Identification of COX10 hybridization signals on chromosome 10
COX10 gene disruption in CMT1A duplication and HNPP deletion patients
Discussion
Materials And Methods
Cosmid sequencing and analysis
Sequence analysis
COX10 cDNA hybridization
Acknowledgements
References
REFERENCES
This page is maintained by OUP admin. Last updated Wed Aug 13 15:52:16 BST 1997. Part of the OUP Journals World Wide Web service. Copyright Oxford University Press, 1996