Processes of de novo duplication of human α-globin genes

Abstract

Ectopic recombination between repeated but nonallelic DNA sequences plays a major role in genome evolution, creating gene families and generating copy number variation and pathological rearrangements in human chromosomes. Previous studies on the α2- and α1-globin genes have shown that de novo deletions common in α⁺-thalassemics can be directly accessed in human DNA and provide an informative system for studying deletion dynamics and processes. However, nothing is known about the reciprocal products of ectopic recombination, namely gene duplications. We now show that molecules carrying three α-globin genes can be detected in human DNA by using physical enrichment plus an inverse PCR strategy. These de novo duplications are common in blood and sperm and appear to arise by two distinct mechanisms: meiotic exchanges between homologous chromosomes that generate a minority of sperm duplications, plus mitotic ectopic exchanges that occur in the soma and germ line and can show erratic fluctuations in frequency most likely caused by mutational mosaicism. The dynamics and processes of duplication are very similar to those of deletion, particularly for meiotic exchanges. This result suggests rearrangement pathways dominated by fully reciprocal ectopic exchange, with nonreciprocal pathways such as intramolecular recombination and single-strand annealing playing at best only a minor role in the generation of deletions. Finally, the high level of instability at the α-globin locus contrasts with the rarity in most populations of chromosomes carrying duplications or deletions, pointing to strong selective constraints that maintain α-globin gene copy number in human populations.

Very little is known about the dynamics of DNA duplication and the reciprocity of unequal exchange in human DNA, despite the importance of these processes in the creation of new genes and the generation of copy number variation and pathological DNA changes (1, 2). Examples do exist of reciprocal duplications and deletions detected in patients and in human populations (3–5). However, the dynamics and reciprocity of the processes that generate these rearrangements cannot be investigated from population data, because factors such as selection can influence the population frequencies of rearrangements. Ectopic recombination between misaligned homologous DNA sequences on different DNA molecules, whether sister chromatids or homologous chromosomes, should generate both deletions and duplications as reciprocal exchange products. However, other processes can generate deletions, including single strand annealing during repair of a broken DNA molecule (6), as well as recombination within single molecules to produce deletions plus excised DNA circles (7) known to exist in mammalian cells (8). Furthermore, gene duplications can also in principle arise in a nonreciprocal fashion by gene conversion events between chromatids or homologous chromosomes that leave the donor chromosome unaltered (9).

The best studied example of ectopic recombination is provided by the disorders Charcot–Marie–Tooth 1A (CMT1A) and hereditary neuropathy with liability to pressure palsies (HNPP). These genomic disorders result from unequal crossover between distant DNA repeats leading to duplication or deletion, respectively, of the intervening 1.4-megabase interval (10, 11). Unequal crossover breakpoints map to the same narrow recombination hot spot within the repeats in patients with de novo CMT1A duplications and in individuals with HNPP deletions (12, 13), consistent with reciprocal exchange (14). However, there is evidence for sex-dependent pathways of ectopic exchange (15), and it remains unknown whether ectopic recombination in CMT1A/HNPP is an exclusively intermolecular process that generates duplications and deletions in a fully reciprocal fashion.

Ectopic exchanges can also occur between locally repeated DNA sequences separated by kilobases rather than megabases; such exchanges play a major role in the evolution of gene families. Classic examples are provided by the human globin gene clusters in which numerous copy number changes resulting in hemoglobinopathies have been described (16–18). De novo deletions between misaligned δ- and β-globin genes, analyzed by single DNA molecule methods, are extremely rare in sperm DNA, preventing further analysis of the reciprocity of ectopic recombination (19). In contrast, the α-globin gene cluster shows a far greater propensity for deletion, most likely promoted by the extended homology blocks that encompass the α2- and α1-globin genes on normal αα chromosomes. Chromosomes with a single α-globin gene (−α) are very common in some populations and are most likely maintained by malaria selection (20). De novo −α deletions can also be detected in human DNA by physical enrichment of deletion molecules (21). These deletions proved to be common in both sperm and blood DNA. Some sperm deletions arise by unequal exchange between homologous chromosomes, as shown by exchange of flanking SNP markers. These exchanges most likely arise by unequal crossover at meiosis and should generate reciprocal duplication products, namely ααα chromosomes carrying triplicated α-globin genes. In contrast, all blood deletions and the majority of sperm deletions arise by a pathway that does not involve interaction between homologous chromosomes. These intrachromosomal deletions could arise either by reciprocal sister chromatid exchange or by other nonreciprocal processes of intramolecular deletion (6, 22).

Chromosomes carrying three α-globin genes do indeed exist in human populations, although the dynamics of the duplication processes that generate them remain completely unknown. These ααα chromosomes are less prevalent than −α chromosomes in most populations (23–25), but this disparity does not necessarily indicate nonreciprocity in the duplication/deletion process because −α and ααα chromosomes are likely to be affected differently by selection. Direct analysis of duplication is only possible by accessing de novo rearrangements directly in human DNA. We now show that such duplicated DNA molecules can be recovered by physical enrichment and amplified by inverse PCR, allowing detailed analysis of the dynamics and processes of gene duplication.

Results and Discussion

Recovery of de Novo Triplicated α-Globin Gene Molecules.

Related DNA sequences in the α2- and α1-globin genes can be grouped into associated homology blocks (Fig. 1A). Various types of deletion can occur depending on the point of ectopic exchange. By far the most common in blood and sperm is the −α^3.7 deletion that arises by unequal exchange between homology blocks Z2 and Z1, losing 3.8-kb DNA (21). We therefore analyzed the reciprocal ααα^anti3.7 duplication mutants that have gained 3.8-kb DNA. By digesting human genomic DNA with SphI plus XbaI, it was possible to release the target from αα chromosomes on a 7.3-kb DNA fragment. Duplication will increase this fragment size to 11.1 kb, allowing separation of progenitor and duplication molecules by gel electrophoresis. Analysis of different electrophoretic size fractions for progenitor molecules and for a control genomic DNA fragment matched in size to duplication molecules showed substantial separation of progenitor and control DNA, with >95% depletion of progenitor molecules from fractions that could contain duplications (Fig. 2B).

Fig. 1.

Strategies for detecting de novo duplications in the α-globin gene region. (A) Organization of normal αα chromosomes, with Y and Z homology blocks shown in color, and divergent primers used to amplify duplications in man 1 indicated by red arrows. SNP sites used to characterize recombinants are shown as gray circles. Key restriction sites are also shown. −α^3.7 deletions could be generated by various pathways involving exchanges between Z homology blocks; only the intermolecular pathway generating the reciprocal ααα^anti3.7 duplication and the intramolecular pathway that produces a deletion plus an excised circle are shown. (B) Detecting duplications in man 1. Genomic DNA was digested with SphI plus XbaI and fractionated to deplete progenitor molecules. Duplication molecules were recovered by nested inverse PCR amplification, using divergent primers that, after duplication, become convergent. Typing SNPs in the exchange interval (white and black circles for haplotypes A and B, respectively) allows intrachromosomal recombinants (intra AA, intra BB) to be distinguished from interchromosomal exchanges (inter AB, inter BA). (C) Detecting duplications in man 2, with digestion of fractionated DNA with AflII before inverse PCR to eliminate progenitor artefacts. (D) Alternative approach for detecting de novo ectopic recombinants. Blood DNA was digested with AflII only and size fractionated to recover 3.8-kb DNA fragments derived from duplicated ααα^anti3.7 chromosomes and from any linearized extrachromosomal circular DNAs, before amplification by inverse PCR. Loss of a 5′ SNP site in man 1 after AflII digestion prevented the distinction of intrachromosomal vs. interchromosomal recombinants.

Fig. 2.

Detection of de novo ααα^anti3.7 duplication molecules. (A) Examples of ααα^anti3.7 duplication molecules recovered by inverse nested PCR amplification of sperm DNA fractions 6 and 7 from man 1. Each PCR contained DNA derived from 1.8 × 10⁴ and 1.1 × 10⁵ amplifiable haploid genomes, respectively. PCR products were analyzed by agarose gel electrophoresis. M, λDNA × HindIII. (B) Cumulative frequencies of duplication molecules (in total, 249 mutants recovered from 4.0 × 10⁶ amplifiable haploid genomes), across size fractions of SphI–XbaI-digested sperm DNA of man 1. DNA size ranges covered in each fraction are shown by gray bars. Fragment lengths of progenitor, control, and ααα^anti3.7 duplication molecules are indicated by dotted lines. The control molecule is an 11.0-kb SphI–XbaI genomic fragment from chromosome 11 matched in size to the ααα^anti3.7 duplication.

The α-globin gene region is very GC-rich and difficult to amplify by PCR, and the recovery of entire ααα^anti3.7 duplication mutants by long PCR was therefore not possible. Instead, we developed an inverse PCR strategy by using primers located between the α2- and α1-globin genes that point away from each other in αα chromosomes (Fig. 1A). After duplication, these primers point toward each other and allow amplification of a 3.3-kb interval that spans the site of ectopic exchange (Fig. 1B). Analysis of genomic DNA from an αα/αα homozygote and an ααα/αα heterozygote showed that this strategy could efficiently recover triplicated α-globin gene molecules in the presence of excess progenitor molecules (data not shown). Analysis of the fractionated DNA revealed de novo duplication mutants correctly distributed across the size fractions (Fig. 2). These mutants must therefore correspond to authentic ααα chromosomes.

Two men were analyzed for duplications. By careful choice of the location of the inverse PCR primers, it was possible to include SNP heterozygosities in the test interval in each man, allowing the haplotype of origin of each mutant to be determined. However, the shifted primer sites in man 2 resulted in PCR artefacts from αα chromosomes (see Materials and Methods); these artefacts were eliminated by digestion of fractionated DNA with AflII (Fig. 1C) to reveal authentic duplication molecules.

Duplication Rates.

These ααα duplications were analyzed in DNA from the same two αα/αα homozygotes analyzed for de novo α-globin gene deletions (21). As with deletions, duplication molecules were detected in both sperm and blood DNA, with mutants 10- to 20-fold more common in sperm than in blood (Table 1). Man 1 showed 2- to 5-fold more mutants than man 2 in both tissues analyzed (P < 0.001 for each tissue), indicating significant interindividual variation in duplication frequency.

Table 1.

Frequencies of ectopic ααα^anti3.7 duplications and −α^3.7 deletions in sperm and blood DNA in two men

Exchange type	Mutation frequencies per haploid genome, ×10−6
	Man 1		Man 2
	αααanti3.7	−α3.7	αααanti3.7	−α3.7
Sperm
Interchromosomal	9.3 (15%)	14.8 (22%)	7.1 (27%)	6.9 (51%)
Intrachromosomal	52.5 (85%)	52.0 (78%)	19.3 (73%)	6.7 (49%)
Total	61.8	66.8	26.4	13.6
Blood
Total	5.5	6.4	1.1	6.6
AflII selection	5.5	—	3.2	—

Percentages of interchromosomal vs. intrachromosomal exchanges in sperm are indicated in parentheses. AflII selection on blood DNA was capable of recovering duplications plus any potential extrachromosomal circles generated by intramolecular recombination (Fig. 1D). Dashes indicate not applicable.

The frequencies of ααα duplications and −α deletions were very similar, both in the sperm and in the blood of man 1 (P = 0.34 and 0.37, respectively). This is consistent with reciprocal ectopic exchange and thus a predominantly if not exclusively intermolecular recombination pathway. In contrast, man 2 showed disparities in duplication/deletion frequencies; the 2-fold excess of duplications over deletions seen in sperm and the 6-fold excess of deletions in blood are highly significant (P < 0.001 in each tissue) and could point to factors that perturb mutation frequency in the germ line and soma.

Blood mutant frequencies were further checked by using a different size fractionation strategy capable of detecting not only duplications but also any extrachromosomal circles generated by intramolecular recombination (Fig. 1 A and D). For man 1, this approach gave an overall mutant frequency very similar to the duplication frequency (P = 1). For man 2, the overall mutant frequency increased significantly (P < 0.001) but not to the level of the −α^3.7 deletion frequency (P = 0.005). It is not known whether this shift in frequency reflects the existence of genuine extrachromosomal circles. However, the structures of mutants recovered by this approach showed no significant differences from those isolated as genuine duplications (Fig. 3B).

Fig. 3.

Distribution of ectopic exchange points in ααα^anti3.7 duplications and −α^3.7 deletions across the Z2/Z1 homology blocks. (A) Structures of duplications and deletions. The location of PSVs and SNPs within the aligned Z2 and Z1 homology blocks are shown below as black bars. A SNP present only in man 1 is marked with an asterisk. (B) Location of exchange points. The homology block can be divided into various intervals determined by PSV distribution, with SNPs further dividing the region in a haplotype-specific fashion. The resulting intervals of sequence identity (horizontal lines) mark regions to which ectopic exchanges can be mapped. The numbers of sperm and blood exchanges mapping to each interval are shown in black and red, respectively, under the line (−, no mutants). Additional blood mutants recovered by AflII digestion (Fig. 1D) are indicated in purple. Duplication mutants were recovered from 4.0 × 10⁶, 8.3 × 10⁶, and 10.9 × 10⁶ amplifiable haploid genomes of DNA from sperm, blood, and blood (alternative approach that could also detect extrachromosomal circles), respectively, from man 1, and from 9.0 × 10⁶, 13.9 × 10⁶, and 7.7 × 10⁶ haploid genomes from man 2. For deletions, mutants were recovered from 6.5 × 10⁶ and 12.9 × 10⁶ haploid genomes from sperm and blood DNA, respectively, from man 1 and 6.7 × 10⁶ and 7.1 × 10⁶ molecules from man 2. Deletion data are taken from ref. 21. (C) Cumulative number of exchanges, per 10⁶ haploid genomes, across the homology block for each class of ααα^anti3.7 duplication and −α^3.7 deletion. Intrachromosomal ααα^anti3.7 duplications in both men and intrachromosomal −α^3.7 deletions in man 1 identified in sperm DNA were relatively abundant and are shown separately with different scaling.

Duplication Exchange Breakpoints.

Each duplication mutant was typed for paralogous sequence variants (PSVs) that distinguish the α2- and α1-globin homology blocks. As with deletions (21), almost all mutants (99.4%) were simple duplications with exchange points mapping to a single interval of sequence identity shared between homology blocks. Only four mutants were complex, with switching of PSVs near the site of ectopic exchange that presumably arose by patchy repair of heteroduplex DNA generated during recombination; a similar low frequency of complex events has been found with deletion mutants (21). Four exchanges were located in the very short Y2/Y1 blocks (Fig. 1A) and mapped to one or other of the longest regions of sequence identity (25, 39 bp) in these blocks (Fig. 4). Such Y2/Y1-driven duplications have yet to be reported in human populations. All remaining exchange points mapped within the long homology blocks Z2 and Z1 and generated ααα^anti3.7 duplications. As with −α deletions, unequal exchange points were fairly randomly distributed across the Z homology blocks (Fig. 3), suggesting that most regions are equally prone to exchange, irrespective of the presence of PSVs, and arguing against the presence of a local recombination hot spot of the type that drives unequal exchange in genomic disorders such as CMT1A (12, 13).

Fig. 4.

Ectopic duplications arising from misaligned Y homology blocks. Four duplicated mutant molecules generated by Y exchanges were identified in man 1. Y2 and Y1 sequences are shown aligned over three lines, with PSVs marked in green. Ectopic exchange intervals are indicated by arrowed lines, with the number of exchanges seen in sperm (black) and blood (red) indicated above.

Sperm Duplications Arising by Meiotic Recombination.

All ααα^anti3.7 mutants were also typed for informative SNPs (Fig. 1 B–D). Roughly 20% of sperm mutants showed exchange of these flanking markers (Table 1), establishing that they had arisen by ectopic recombination between homologous chromosomes. In contrast, only two blood mutants of 145 (1.4%) showed marker exchange. This germ-line specificity of interchromosomal exchange has also been seen for −α deletions (21) and is fully consistent with these exchanges arising primarily by unequal exchange at meiosis. Furthermore, interchromosomal deletion and duplication frequencies in sperm in man 2 were indistinguishable (Table 1), consistent with a fully reciprocal unequal exchange process. Although man 1 showed a 1.6-fold excess of deletions, this difference is of marginal significance (P = 0.026). As expected for products of meiotic recombination, these exchanges were also symmetric with respect to haplotype (Fig. 3B), with for example man 1 showing very similar numbers of AB- and BA-type sperm duplications (20 and 17, respectively). The frequency of these duplications and deletions would therefore simply reflect the frequency of unequal crossover between Z blocks on homologous chromosomes at meiosis (≈10⁻⁵ per sperm).

Mechanisms of Intrachromosomal ααα^anti3.7 Duplication.

Most duplications in blood and sperm are intrachromosomal without flanking marker exchange (Fig. 3B and Table 1) and must therefore have arisen by unequal sister chromatid exchange involving crossover or gene conversion. For blood, these events must result from mitotic recombination, and their prevalence in sperm suggests a premeiotic component to germ-line duplication occurring at some stage before spermatogenesis. Unlike meiotic events, these duplications can show substantial distortions in frequency between haplotypes. For example, sperm mutants in man 1 mapping to the longest interval of sequence identity (Fig. 3B) were more common on haplotype B than on haplotype A (108 vs. 55; P < 0.001).

The most reasonable explanation for these haplotype asymmetries is mutational mosaicism whereby an early mitotic recombination event can spread to multiple descendant cells. We have already noted clear instances of very unusual deletions that should be rare but that are detected repeatedly in deletion surveys, signaling mosaicism (21). We see a similar phenomenon for duplications, for instance the six haplotype B sperm duplications in man 1 that all mapped to an interval just 8 bp long at the beginning of the Z homology block (Fig. 3B). Given the expected rarity of exchange in such a small region, these six mutants most likely derived from a single ancestral exchange event. Mosaicism is therefore not unusual and could provide a reasonable explanation for other haplotype distortions in intrachromosomal rearrangement frequencies. Sperm mosaicism also suggests that a significant proportion of intrachromosomal sperm events must be premeiotic in origin rather than arising by meiotic sister chromatid exchanges.

Although mosaicism could play a major role in influencing the frequency of germ-line and somatic rearrangements, other factors could also contribute to these haplotype asymmetries. Perhaps SNPs that can act as PSVs and disrupt homology blocks in a haplotype-specific fashion could perturb the ectopic exchange process. However, there is no obvious effect of sequence interruptions on unequal exchange rate and distribution for interchromosomal duplications and deletions (Fig. 3B). There are instances of an apparent effect on intrachromosomal rearrangements (for example, the reduction of 108 haplotype B sperm duplications in the longest region of sequence identity in man 1 to 55 haplotype A duplications in the presence of two disrupting paralogous SNPs). These are, however, countered by instances of exchange frequencies showing elevation despite the presence of a disruption (for example the 5′ adjacent interval shows 4 haplotype A duplications and 27 haplotype B duplications, despite an interruption in the latter). Haplotype disparities could also be caused by epigenetic marks or distal regulators in cis that influence ectopic recombination frequencies in a haplotype-specific fashion. There is, however, no clear evidence for haplotypes that show a consistent enhancement of both duplication and deletion predicted from such regulators (Fig. 3B). The dominant factor creating haplotype asymmetries therefore appears to be mutational mosaicism.

Duplication Dynamics and the Population Incidence of ααα^anti3.7 Chromosomes.

As with −α deletions, the high frequency of de novo ααα^anti3.7 duplications in sperm (6.2 × 10⁻⁵ per sperm in man 1, 2.6 × 10⁻⁵ in man 2) predicts a correspondingly very high incidence of ααα chromosomes in human populations. We have previously shown that the low incidence of −α chromosomes in populations not affected by malaria can only be maintained by significant selection against −α/αα carriers and/or −α/−α homozygotes (21). Precisely the same arguments apply to ααα duplications, which also show a very low incidence in most populations (0.002–0.006) (24), which can only be maintained in the face of the strong duplication pressure by selection against ααα carriers and/or homozygotes, with strengths of selection similar to those reported for −α deletions (21). Given the normal hematological profiles of individuals with ααα chromosomes (23, 26), the nature of this selection, whether through subtle imbalances between α- and β-globin levels or through chromosomal processes such as segregation distortion, remains completely unclear.

Conclusions

Reciprocal ectopic recombination has long been postulated as a major driving force in human genome rearrangements and in the evolution of gene families (27–30). Although what appear to be reciprocal products of unequal exchange can be found in human populations, for example −α and ααα globin chromosomes, these do not exclude the existence of distinct deletion and duplication pathways. The same reservations apply to genomic disorders such as CMT1A/HNPP, where reciprocal exchange points in duplications and deletions point to a common initiating mechanism for unequal exchange but not necessarily to a fully reciprocal exchange process (12–14, 31). The present report shows that de novo gene duplications can be recovered from human DNA and used to analyze processes of duplication. The overall picture in the α-globin gene cluster is one of substantial reciprocity in the dynamics and processes of duplication and deletion, with instability occurring both in the germ line and in somatic DNA. Unequal exchange between homologous chromosomes at meiosis plays only a minor role in establishing overall germ-line instability levels. Although we cannot exclude a possible contribution from meiotic sister chromatid exchange, the dominant germ-line process appears to be mitotic recombination. This process generates rearrangements whose frequency can be significantly influenced by factors such as mutational mosaicism, which in turn can lead to erratic inflation of various classes of rearrangement, apparently affecting the overall incidence of duplications and deletions. This study also establishes that other pathways such as intramolecular recombination play at best only a minor role in the generation of deletions. Finally, the α-globin genes present a picture of a gene cluster being subjected to very strong forces of duplication/deletion that contrast strongly with the rarity of rearranged chromosomes in most populations, giving clues about the strength of selective forces that must have acted to stabilize gene copy number.

Materials and Methods

DNA Samples.

The two men studied (man 1, man 2) were the same as in ref. 21 and were 56 and 45 years old, respectively, at the time of sample collection. Blood and semen samples were collected with approval from the Leicestershire Health Authority Research Ethics Committee and with informed consent. Blood and sperm DNAs were extracted and manipulated under conditions designed to minimize the risk of contamination as described in ref. 32.

Physical Enrichment of Duplication Mutants.

Genomic DNA (89–200 μg) was digested with SphI (New England Biolabs, Beverly, MA), ethanol precipitated, redigested with XbaI (New England Biolabs), again ethanol precipitated, and dissolved in 400 μl of 5 mM Tris·HCl (pH 7.5). This double-digested DNA was mixed with 100 μl of loading dye [44 mM Tris·borate (pH 8.3), 1 mM EDTA, 30% vol/vol glycerol, and bromophenol blue], ethidium bromide was added to 400 μg/ml, and the sample was loaded into a 5 × 0.3-cm slot in a 40-cm-long, 1.4-cm-deep 0.8% SeaKem HGT agarose gel (Cambrex Bio Science, Rockland, ME). DNA was electrophoresed until a 6.6-kb λ DNA × HindIII marker had migrated 32 cm. Twelve size fractions were collected ranging from 6.8 to 22.7 kb to include all ααα^anti3.7 duplication molecules. DNA in each fraction was recovered by electroelution onto dialysis membrane, ethanol precipitated, and dissolved in 50 μl of 5 mM Tris·HCl (pH 7.5).

An alternative approach of size fractionation was performed with blood DNA (104 μg) after digestion with AflII (New England Biolabs). Digested DNA was loaded as above into a 0.8% agarose gel but electrophoresed until the 2.3-kb λ DNA × HindIII marker had migrated 16 cm. Eight size fractions were collected over the size range of 2.2–5.8 kb, to include the 3.8-kb molecules derived from any ααα^anti3.7 duplications and from any linearized extrachromosomal circles.

DNA Recovery and Size Validation.

Gel electrophoretic comparison of DNA pooled from all fractions with known amounts of digested genomic DNA showed a total DNA recovery of 40–60%. DNA recovery and size distribution in each fraction was estimated by PCR amplification of a 5.5-kb interval from a control 11.0-kb SphI–XbaI double-digest DNA fragment from chromosome 11, using PCR primers target4F (5′-CCA GAC TCT TGA GGT GGA GG-3′) and target4R (5′-CTG CCC CTC AGT GAT CTT CC-3′). Fractions from the alternative approach were similarly analyzed but by using PCR primers R85.7F (5′-CCA TTA CTG GCT TGC AGA AA-3′) and R87.5R (5′-AAG AGA CTG GGA GTG GCT GG-3′) to amplify a 1.8-kb interval from a 3.8-kb AflII DNA fragment from the MHC region. These control molecules identified fractions that should contain mutants and also allowed overall DNA recovery to be estimated, again at ≈40%. The frequency of contaminating progenitor molecules in each fraction was determined by PCR amplification of a 4.9-kb segment lying within the SphI–XbaI interval of the α-globin gene region, using PCR primers A18.1F (5′-GAA GGG TTT GGA ACT CAG CC-3′) and A22.9R (5′-CAA TAG CTG GAA CCG GCT GG-3′).

Recovery of Duplication Mutants.

Duplication molecules in size-fractionated genomic DNA were recovered by inverse PCR performed in a PTC 240 Tetrad 2 thermal cycler (MJ Research, Cambridge, MA) in 0.2-ml PCR tubes or 96-well plates. Inverse PCRs used 0.2 μM PCR primers, 0.03 units/μl Taq polymerase (ABGene, Surrey, U.K.), and 0.003 units/μl cloned Pfu polymerase (Stratagene, La Jolla, CA) in 0.5 M betaine, 41 mM Tris·HCl (pH 8.8), 12.5 mM Tris base, 9.9 mM ammonium sulfate, 4.1 mM MgCl₂, 6.0 mM 2-mercaptoethanol, 4.0 μM EDTA, 0.9 mM dNTPs, and 102 μg/ml BSA. Primary inverse PCR primers were A21.8R2 (5′-TGC TCT GGG CTG TGA GGC G-3′) plus A22.3F (5′-TCT GCC CAA GGC AGC TTA CC-3′) for man 1, and A22.6R2 (5′-GGG TGG ACT TCT GGG GAA G-3′) plus A22.8F (5′-AAG TCT CAC CTC CTC CAG G-3′) for man 2. Secondary nested PCR primers were A21.8R3 (5′-TGT GAG GCG CAG GAA GAG C-3′) plus A22.4F (5′-GCT TGC TCC TGG ACA CCC AG-3′) for man 1, and A22.6R (5′-AGG TGC AGG AGT GCC AGT G-3′) plus A22.8F2 (5′-CCT CCA GGA AGC CCT CAG-3′) for man 2. Cycling for primary inverse PCR was at 96°C for 1 min, followed by 26 cycles of 96°C for 20 s, 68°C (man 1) or 56°C (man 2) for 30 s, and 68°C for 4 min (man 1) or 5 min (man 2). Primary PCR products were diluted 200-fold with water plus 1 μg/ml salmon sperm carrier DNA, and 10 μl of secondary nested PCRs were seeded with 0.7 μl of this diluted DNA. Cycling for secondary PCR was at 96°C for 1 min, followed by 34 cycles of 96°C for 20 s, 68°C (man 1) or 58°C (man 2) for 30 s, and 68°C for 4 min (man 1) or 5 min (man 2).

The DNA input for each PCR was adjusted to contain at most 1.3 amplifiable duplication molecules as established from pilot experiments performed on each fraction to obtain an initial estimate of duplication frequency. Secondary PCR products were analyzed by agarose gel electrophoresis and visualized by staining with ethidium bromide. A total of 565 positive reactions of 5,325 nested PCRs of blood and sperm DNA from both men were collected for further analysis.

Poisson analysis of limiting dilutions of genomic DNA (33) from an αα/ααα heterozygote showed that the nested inverse PCR strategy used for man 1 could amplify a single duplicated DNA molecule from 11.3 pg of DNA, indicating a 53% efficiency of PCR in recovering duplication molecules. The efficiency of the protocol used for man 2 was similar, at 66%.

Elimination of Inverse PCR Artefacts.

Pilot PCR assays showed PCR artefacts in DNA fractions from man 2 increasing with the degree of progenitor contamination. It seemed that the 3′ end of an extending DNA strand in homology block Z1 could loop back into block Z2 during early stages of PCR to create an amplicon from progenitor molecules identical in size to that expected from a ααα chromosome. PCR analysis of genomic DNA from a αα/αα homozygote showed that these artefacts could be eliminated by digestion with AflII, which presumably separates the Z1 and Z2 blocks in any progenitor molecules and prevents this intramolecular looping back. Each DNA fraction was therefore overdigested with AflII (4–6 units of enzyme per microliter of fractionated DNA) before inverse PCR. Control overdigestion of genomic DNA from an αα/ααα carrier showed no significant effect on the efficiency of recovery of duplication molecules.

Characterization of Duplication Mutants.

Ectopic exchange points were mapped by typing each mutant by dot-blot hybridization with oligonucleotides specific to PSVs and to SNPs (33). Haplotypes A and B in man 1 and man 2 differed at 8 and 11 SNP sites, respectively, 3 of which mapped into the amplified interval. A total of 30 mutants with exchange points that could not be located unambiguously by hybridization were sequenced by using BigDye Terminator, version 3.1 (Applied Biosystems, Foster City, CA). An additional 53 mutants (≈8.4% of the total) were also sequenced. None showed any rearrangements in addition to exchange of PSVs, and only four single base misincorporations were seen over 158 kb of DNA sequenced. Seven intrachromosomal mutants with abnormal lengths were also detected and sequenced; these contained 0.3- to 2.2-kb deletions extending over the site of exchange. The authenticity of these mutants was uncertain given the broad DNA size coverage in most fractions and were discarded from further analysis despite some being in correct size fractions.

A total of 68 positive reactions containing mixed mutants as shown by mixed PSV and/or SNP sites were separated by sequence-specific PCR directed to a mixed PSV or SNP site before characterization. All mixtures were successfully resolved into their two to three constituent molecules. A full inventory of all mutant types over all PCRs was used to Poisson correct for instances of a PCR containing more than one molecule of a given type of recombinant (details are in ref. 33). These corrections were modest, with only a 1.3-fold increase for the most abundant type of mutant.

Abbreviations

CMT1A

Charcot–Marie–Tooth type 1A

HNPP

hereditary neuropathy with liability to pressure palsies

PSV

paralogous sequence variant.