The standard slipped-strand mispairing (SSM) model for the formation of variable number tandem repeats (VNTRs) proposes that a few tandem repeats, produced by chance mutations, provide the “raw material” for VNTR expansion. However, this model is unlikely to explain the formation of VNTRs with long motifs (e.g., minisatellites), because the likelihood of a tandem repeat forming by chance decreases rapidly as the length of the repeat motif increases. Phylogenetic reconstruction of the birth of a mitochondrial (mt) DNA minisatellite in guppies suggests that VNTRs with long motifs can form as a consequence of SSM at noncontiguous repeats. VNTRs formed in this manner have motifs longer than the noncontiguous repeat originally formed by chance and are flanked by one unit of the original, noncontiguous repeat. SSM at noncontiguous repeats can therefore explain the birth of VNTRs with long motifs and the “imperfect” or “short direct” repeats frequently observed adjacent to both mtDNA and nuclear VNTRs.
VARIABLE number tandem repeats (VNTRs), especially microsatellites, have become the genetic marker of choice for forensics (Hagelberget al. 1991; Olaisenet al. 1997), genomic mapping (Dibet al. 1996; Dietrichet al. 1996), and quantifying intraspecific genetic variation (Bowcocket al. 1994; Schlöttereret al. 1997; Luntet al. 1998). Furthermore, many human genetic diseases are caused by microsatellite VNTR expansion (Mandel 1997) and an increase in VNTR variability is an indicator of mutations associated with several forms of cancer (Wadaet al. 1994). Despite the ubiquity and importance of VNTRs, little is known about the molecular mechanisms leading to their formation.
This lack of empirical data on repeat formation is surprising, because VNTR formation must be a frequent occurrence. This conclusion is based on the observation that, while all eukaryotic taxa surveyed possess VNTRs, studies where several species have been tested with the same VNTR primers reveal that a given repeat rarely occurs in more than a few closely related taxa. In some of these studies a repeat that is perfect and highly variable in one taxon is interrupted or very short and monomorphic in close relatives (Zardoyaet al. 1996; Brohede and Ellegren 1999). In many other studies, variable repeats in one taxon are unrecognizable in close relatives (Blanquer-Maumont and Crouau-Roy 1995; Angers and Bernatchez 1997; Tayloret al. 1999).
Slipped-strand mispairing (SSM; Levinson and Gutman 1987) is the most often proposed model for repeat formation, expansion, and contraction. A critical component of the SSM model for repeat formation is the occurrence of chance mutations that produce a few tandem repeats facilitating the first strand slippage event. Levinson and Gutman (1987) refer to these tandem repeats produced by chance mutations as the “raw material” for repeat expansion by SSM. A study of repeat evolution in primates demonstrated that chance mutations played a role in the formation of two VNTRs, one with a 2-bp motif and one with a 4-bp motif (Messieret al. 1996). By mapping sequence data onto a phylogeny, Messier et al. (1996) discovered that in the owl monkey an A-to-G mutation produced (GT)5, which expanded, presumably via SSM, to (GT)6. Messier et al. (1996) also discovered that a G-to-A mutation produced (ATGT)2 in Hominoidea, which expanded to (ATGT)4 in gorillas, bonobos, and chimpanzees and to (ATGT)5 in humans. Thus, the model proposed by Levinson and Gutman (1987) is supported by the microsatellite evolution data reported by Messier et al. (1996). However, it is unknown whether these two instances of VNTR birth are typical, or whether chance mutations are important in the formation of repeats with motifs longer than 4 bp.
While surveying guppy (Poecilia reticulata) populations for mtDNA control region sequence variation we uncovered a minisatellite with an 11-bp motif in individuals from a tributary of the Rio Grande in Trinidad. We investigated the evolution of this minisatellite by mapping mutations onto a population-level phylogeny. This led to a general hypothesis that SSM at noncontiguous repeats can lead to the birth of VNTRs with long motifs and characteristic flanking repeats.
MATERIALS AND METHODS
Sampling and molecular techniques: Our survey included 46 guppies from 33 sites in Trinidad, Venezuela, Guyana, and Surinam (Table 1). In the field, guppies were preserved in 95% ethanol. In the lab, DNA was isolated from the tail musculature using methods described by Fajen and Breden (1992). To produce sequencing templates, we amplified ~1000 bp of mitochondrial control region DNA using the primers L15926 (Kocheret al. 1989) and MRT2 (Ptacek and Breden 1998). PCR conditions included an initial denaturation at 94° for 1 min, followed by 35 cycles each consisting of denaturation at 94° for 1 min, annealing at 52° for 1 min 20 sec, and extension for 2 min at 72°. PCR products were purified on a 1% agarose gel. A small block of agarose containing the PCR product was cut out of the gel and frozen overnight. This gel block was then spun in a microcentrifuge at high speed for 7 min. Two microliters of the resulting liquid was used as a sequencing template. For 39 samples from 26 locations (Table 1), we sequenced both strands of the left domain or R1 portion (Fumagalliet al. 1996) of the control region using primers L15995 (Meyeret al. 1994) and MR1 (5′-TAT GGG TTT TGT CTA CCT TC-3′). For seven individuals from across the guppy geographic distribution (Table 1), we sequenced 911 bp of the mitochondrial control region using the following primers, which were spaced ~200 bp apart: L15926, L15995, MR1, 12 RS (5′-CAT TTG GTT CCT ATT TCA GG-3′), 13 (5′-CAT TTC ACA GTG CAT ACA CA-3′), 14 (5′-AGT ATC CCC CTC GGC TTT TG-3′), 15 (5′-AAT TTT GTT TAC ATA CTT TA-3′), and MRT2. These primers provided overlapping sequences for 80–90% of the control region. Sequences from repeat-possessing Rio Grande (RG) guppies were unreadable beyond approximately three repeat units, suggesting that these samples were heteroplasmic (i.e., possessed more than one mtDNA haplotype). To estimate the prevalence of the VNTR we end-labeled primer MR1 with γ33P and used MR1 and L15926 to amplify the R1 portion of the control region in 18 RG samples. These 18 samples included the 4 samples sequenced for either 911 or 150 bp (Table 1). End-labeled PCR products were electrophoresed on a 6% acrylamide gel and visualized by exposure to X-ray film.
Analyses: Sequence alignments were performed using CLUSTAL V (Higginset al. 1992). First, for all guppies (N = 46) the R1 portion (ca. 150 bp) of the control region was aligned. This alignment identified mutations occurring in the portion of the control region associated with repeat expansion. Second, complete control region sequences from seven P. reticulata samples and sequences from P. caucana (GenBank accession no. AF033057) and P. parae (accession no. AF03-3050) were aligned. Maximum parsimony analysis of this alignment produced a robust phylogeny of a subset of guppy populations upon which we mapped the changes in the R1 region. P. caucana occurs in the subgenus Poecilia and P. parae occurs in the subgenus Lebistes, along with the guppy (Bredenet al. 1999). In this second alignment P. reticulata and P. caucana sequences were truncated so that they could be aligned with the shorter (813 bp) P. parae sequences. In both alignments Rio Grande sequences were modified to include only one repeat motif.
A maximum parsimony analysis was performed on complete control region sequences (i.e., second alignment) using the heuristic search algorithm of PAUP version 3.1.1 (Swofford 1993). Deletions greater than one nucleotide long were treated as single characters in this analysis. PAUP settings included: addition sequence, stepwise (random seed number, 9); collapse zero-length branches; MULPARS option in effect. P. caucana was treated as the outgroup. Support for the tree topology was estimated using 500 bootstrap reiterations. Xiphophorus nigrensis sequence data (GenBank accession no. U06578) were added to the tree to polarize mutations.
Molecular results: Twenty individuals from 14 guppy populations possessed the 11-bp motif, CCAAAATCTGC, in the R1 portion of the control region (Table 2), but expansion of this motif was evident only in RG guppies. Deletions and duplications led to variation in the length of the R1 portion of the control region among the guppies surveyed (Table 2). Seventeen of the 18 RG samples surveyed using γ33P-labeled MR1 and L15926 were heteroplasmic, possessing four to eight different-sized mtDNA haplotypes each.
Phylogenetic analyses: A bootstrap 50% majority-rule consensus tree is shown in Figure 1. The seven P. reticulata samples form a monophyletic group. Within the P. reticulata clade there are two groups with bootstrap values ≥99%. One includes the Yarra, Arima, and Guanare river samples. The sample from the Essequibo River (BAR3) appears to be the sister taxon to this clade. The second well-supported clade includes samples from the Oropuche Drainage in Trinidad (RG1, OVR6, and QU48).
A model for the birth of a minisatellite: By mapping sequence data onto the phylogeny, we uncovered two substitutions that appear to have been important for the formation of the minisatellite in RG guppies. First, the repeat motif is the consequence of a G-to-A mutation changing CCGAAATCTGC to CCAAAATCTGC. There are two equally parsimonious reconstructions of the changes at this site. First, this mutation may have occurred in the common ancestor of guppies from the Rio Grande, Quare, and Oropuche rivers and again in guppies from the Essequibo River (Figure 1). Alternatively, this mutation occurred in the common ancestor of all guppies surveyed with a reversal in the common ancestor of the Yarra River + Arima River + Guanare River clade. Our conclusion that the G is the ancestral state for this nucleotide site is based upon the observations that CCGAAATCTGC occurs in guppies and P. caucana (Figure 1) and that CCGGACTCTGC occurs in X. nigrensis. The second substitution that appears to have been important for the formation of the RG minisatellite occurred in repeat flanking sequence. The guppies surveyed for mitochondrial sequence variation that have expanded repeats (RG1, 4, 6) possess a unique sequence, CCCAAATCT, adjacent to the repeat motif (Table 2; Figure 1). This repeat expansion-associated flanking sequence is apparently a consequence of an A-to-C mutation in the common ancestor of the RG guppies (changing CCCAAATAT to CCCAAATCT; Figure 1). We propose that the G-to-A and A-to-C mutations described above created an imperfect, noncontiguous, 9-bp repeat, (CCAAAATCT)GC(CCCAAATCT), in the inferred ancestor of the RG population (labeled “RG anc” in Figure 1) and that this noncontiguous repeat provided the raw material for repeat expansion due to SSM. A T-to-C mutation in one motif of this noncontiguous repeat appears to have prevented repeat expansion in the ancestor of RG5 (Figure 1).
We have not sequenced the entire control region for RG5, the “nonexpanded” individual from the Rio Grande. Phylogenetic analysis of 150 bp of all 46 guppies sequenced (Table 2) places RG5 in a monophyletic group that includes Oropuche River, Quare River, and other Rio Grande samples. Based upon the observation that RG5 shares the A-to-C mutation described above with other RG samples, we have drawn a tree showing RG monophyly in Figure 1. Our hypothesis for the formation of the noncontiguous repeat does not depend upon our assumption that the RG clade is monophyletic.
Our model for expansion at this locus is presented in Figure 2. This model is similar to the SSM model (Levinson and Gutman 1987) and the illegitimate elongation model (Burokeret al. 1990) but emphasizes misalignment at noncontiguous repeats. First (Figure 2A), replication of the mitochondrial heavy strand (H-strand) pauses after termination associated sequences, producing a stable triple-stranded structure referred to as the D-loop (Shadel and Clayton 1997). Second (Figure 2B), competition between the H-strand and the D-loop strand for light strand (L-strand) binding facilitates local melting and the D-loop strand becomes single stranded. Local melting is followed by “competitive misalignment,” i.e., reinvasion of the D-loop strand and misannealing between the D-loop strand and L-strand (Burokeret al. 1990) (Figure 2C). Misannealing of the D-loop strand may be enhanced by hairpins that reduce the “effective” length of the D-loop strand, preventing it from reannealing with its proper L-strand complement (Burokeret al. 1990). After competitive misalignment, the continuation of D-loop strand elongation leads to the duplication of the intervening base pairs and one motif of the noncontiguous repeat producing a heteroduplex (Figure 2D). D-loop strand elongation, i.e., formation of the nascent H-strand (H′ in Figure 2E), exposes the origin of replication for the L-strand (Shadel and Clayton 1997) leading to L-strand replication (Figure 2E). If the heteroduplex shown in Figure 2E is not repaired, then the next mtDNA replication produces three wild-type genomes (not shown) and a mitochondrial genome with a tandem repeat of an 11-bp motif flanked by one 9-bp motif from the original noncontiguous repeat (Figure 2F). We hypothesize that this is followed by expansion of the perfect tandem repeat via SSM. We refer to the remnant of the noncontiguous repeat as a 3′ “partial” repeat (Figure 2F) because it occurs upstream of the repeat on the L-strand (L″ in Figure 2F).
The birth of an mtDNA minisatellite in guppies: The discovery of a highly variable mitochondrial minisatellite in one Trinidadian guppy population provided us with an opportunity to study the processes responsible for the formation of VNTRs. By mapping sequence data onto a guppy phylogeny we discovered mutations that appear to have produced a noncontiguous, imperfect repeat in the ancestor of guppies from the RG population. We propose that SSM at this noncontiguous repeat produced a tandem repeat with an 11-bp motif that was flanked by a 9-bp partial repeat, (CCAAAATCTGC)2CCCAAATCT, and that subsequent SSM at the tandem repeat led to the mtDNA length variation currently observed in the RG guppy population. Thus, the first SSM mutation produces a perfect tandem repeat with a motif that is longer than the noncontiguous repeat motif formed by nucleotide substitutions. This model is similar to a model proposed by Torroni et al. (1994) to explain the formation of a rare 207-bp duplication in the mitochondrial genome of Caucasian humans.
A general model for minisatellite birth? A survey of published minisatellite sequence data suggests that SSM at noncontiguous repeats may be a general model for VNTR birth in both mitochondrial and nuclear DNA. As described above, a consequence of SSM at a noncontiguous repeat is the formation of a locus with long repeats flanked by one unit of the original noncontiguous repeat (i.e., a 3′ partial repeat). Fumagalli et al. (1996) noticed that in many taxa mtDNA VNTRs are flanked by “imperfect” or “degenerate” repeats. We compared the imperfect and “perfect” mtDNA repeats in shrews (Fumagalliet al. 1996), sturgeon (Brownet al. 1996), perch (Nesbøet al. 1998), the Taipei treefrog (Yanget al. 1994), and flounder (Leeet al. 1995) and discovered that in all cases imperfect repeats may be interpreted as 3′ partial repeats. That is, the first portion of the imperfect repeat matches the perfect repeats best (Figure 3). Similarly, partial repeats (called “short direct” repeats by Haber and Louis 1998) occur next to nuclear minisatellites in humans (Haber and Louis 1998; Murrayet al. 1999; Figure 4), birds (Gyllenstenet al. 1989), salmon (Goodier and Davidson 1998), and fungi (Giraudet al. 1998). These observations suggest that the mtDNA-specific components of our model (e.g., competitive misalignment) may not be critical and that the birth of VNTRs with long motifs in mitochondrial and nuclear DNA frequently involves SSM at noncontiguous repeats.
We thank Sampson Wu for technical assistance, Andrew T. Beckenbach and Michael J. Smith for valued discussions, and Ann E. Houde for sending additional Rio Grande guppies. This work was supported by a Simon Fraser University President's Ph.D. research stipend to J.S.T. and a Canadian National Sciences and Engineering Research Council grant to F.B.
Communicating editor: S. Yokoyama
- Received August 10, 1999.
- Accepted March 16, 2000.
- Copyright © 2000 by the Genetics Society of America