Single Nucleotide Polymorphism Markers for Genetic Mapping in Drosophila melanogaster

Survey of Interstrain Polymorphism

We first evaluated several Drosophila strains for levels of sequence polymorphism. Many strains are in use in theDrosophila research community, and there are no widely used standard mapping strains. We therefore selected a few strains with the aim of identifying strains highly polymorphic relative to each other. The polymorphism rate in Drosophila varies among specific strain pairs, with the greatest variation observed between East African and other populations (Begun and Aquadro 1993). Therefore, strain selection might significantly affect the rate of SNP discovery. We also wanted to use a series of strains containing single mapped P-element insertions for fine-scale mapping (Cutforth and Rubin 1994; Spradling et al. 1995). Thus, another important consideration in our strain selection was the level of sequence polymorphism relative to these P-element strains.

Using genomic DNA from six wild-type strains (Barcelona, Capetown, Hikone, Pyrenees, w;iso2;iso3, and a P-element containing strain) as templates, we compared sequences from 24 third chromosome sequence tagged sites (STSs) from the P1-based physical map of the genome (Kimmerly et al. 1996). The results shown in Table1 indicate that the rate of sequence variation between any two strains ranges from 2.1 per kilobase (1 polymorphism/476 bp) to 5.2 per kilobase (1 polymorphism/192 bp, Table1A), in agreement with previous studies. Although the rate of sequence polymorphism varies depending on the strains compared, much of the observed polymorphism occurs in local clusters, a single STS containing as many as five SNPs. For purposes of genetic mapping, one needs only a single polymorphism per point in the genome. Thus, for the purposes of this study, a better measure of the interstrain polymorphism rate is the percentage of STSs that contain at least one SNP (Table 1B). Although the absolute polymorphism rate observed varies nearly 2.5-fold, from 2.1 differences/kb to 5.2 differences/kb, the percentage of polymorphic STSs varies less than twofold, from 33% to 63%. In addition, the polymorphism rates relative to the P-element containing strain vary less, ranging from 33% to 50%. Because we observed a largely random distribution of sequence variation and similar levels of relative polymorphism among pairwise sets of strains, we chose to focus our genome-wide discovery efforts on three commonly used strains- Canton S, Oregon R, and w; iso2; iso3. We also selected a series of 17 strains that contain single P-element insertions at evenly spaced intervals across the genome.

Table 1.

Analysis of Sequence Variation among Six Wild-Type

[i] Twenty-four STSs from the third chromosome were sequenced in the following strains: w;iso2;iso3 (ISO), Barcelona (BAR), Capetown (CAP), Hikone (HIK), Pyrenees (PYR), and a P-element strain (BEP). The total amount of sequence examined was 5.16 kb. Data are represented as (A) total number of polymorphisms per kilobase, and (B) percentage of STSs containing polymorphisms.

Open in new tabLink to table

Genome-Wide SNP Discovery

We set out to identify a set of SNP markers spanning the genome, with a goal of identifying at least one SNP every 500 kb. We used mapped STSs as a source of genomic sequences (Kimmerly et al. 1996). At the time this project was initiated, these STSs were the only mapped sequence elements with a genome-wide distribution of the required density. The cytological map positions of STSs were inferred from the P1 clones from which they were derived. We chose STSs evenly distributed across each cytological division for screening. We took two approaches to polymorphism identification. First, a set of 1016 STSs was amplified from isogenized versions of Canton S and Oregon R and then sequenced to identify candidate polymorphisms. Second, a partially overlapping set of ∼1050 STSs was screened for polymorphisms between w;iso2;iso3 and the 17 P-element strains by using denaturing high performance liquid chromatography (DHPLC; Underhill et al. 1997). DHPLC is a simple and robust screening tool and is also useful for genotyping SNPs in recombination mapping experiments (see Fig. 2 below;Spiegelman et al. 2000). Polymorphisms identified via DHPLC were verified and characterized by DNA sequencing.

Figure 2.

Recombination mapping of a recessive lethal mutation by using SNP markers. Chromosomes (bars) and molecular markers (vertical hatch marks) are shown. The mapping process occurs in two stages. (A) The mutation (asterisk) induced in the w;iso2;iso3 background (black bar) that has been mapped previously to a chromosome and balanced is mapped relative to a polymorphic mapping strain (open bar). Single flies heterozygous for the mutation-carrying chromosome and the mapping chromosome are crossed to flies homozygous for the parental w;iso2;iso3 strain to generate 96 recombinant flies. The four recombinant classes are represented. Each recombinant fly strain is assayed for a low-density set of markers that span the chromosome. These markers can be of any type, including SNPs or P-element insertions. We have typically tested six markers spaced at ∼10-Mb intervals on these 96 recombinants, for a total of 576 assays. Each recombinant also is assayed for presence or absence of the mutation by outcrossing. From the outcross data and initial marker data on this set of 96 recombinants, the mutation can be assigned to an interval of 10–20 Mb. (B) A higher density set of SNPs then is assayed on recombinants from A that break in the appropriate interval. An example of mapping a mutation in the hedgehog gene is shown. SNP markers are indicated by the STSs from which they are derived. Two P elements (Q 1059 and Q 1058) that were used to localize these mutations also are indicated. The chromosomal compositions of three recombinant (#6, #25, and #42) and two control (ISO and HET) flies are represented. Asterisks indicate chromosomes that carry the mutation as defined by outcrossing. The SNP markers shown were scored using DHPLC. The recombinants shown delimit the position of these mutations to between Dm1601 and Dm1655, a region of ∼984 kb. Subsequent complementation testing showed that these mutations are alleles of thehedgehog gene, which lies between Dm1601 and Dm1655. (C) DHPLC scoring of SNPs. PCR products were amplified from recombinants and analyzed under partially denaturing conditions. Data are shown for Dm1655, a C/T dimorphism. Run time in minutes is shown on the X axis and ultraviolet absorbance on the Y axis. Dm1655 was analyzed from the following strains: w;iso2;iso3 (ISO), a w;iso2;iso3/Q1059 heterozygote (HET), and hedgehogrecombinants 6 and 25 (#6 and #25). In this example, the sample from recombinant 25 shows a heteroduplex pattern and therefore is scored as a heterozygote. (DHPLC) denaturing high performance liquid chromatography.

Open in new tabDownload PowerPointLink to figure

Sequence data from a total of ∼1500 STSs from these two discovery approaches were assembled into a single data set. We identified polymorphisms both by visual inspection and automatically as high-quality discrepancies by using Phred quality scores in the Phrap assembly viewer Consed (Gordon et al. 1998). Table 2displays the sequences of 109 STSs for which sequence was derived from all strains used in the genome-wide survey. In addition, we aligned each consensus STS sequence to the Release 1 version of theDrosophila genome sequence to identify the allele in they;cn bw sp strain (Adams et al. 2000). These data define 279 polymorphisms, of which 225 are single nucleotide substitutions, 17 are small insertion/deletions, 8 are dinucleotide substitutions, and 29 are more complex substitutions. Among the single nucleotide substitutions, transitions outnumber transversions by 55% to 45%, and ∼22% of these substitutions occur at CpG dinucleotides. These data are similar to those observed in human and mouse genomes and may be evidence that similar mutagenic mechanisms are at work in Drosophila(Moriyama and Powell 1996), despite the fact that there is no deficit of CpG dinucleotides and little evidence for cytosine methylation that explains the bias seen in vertebrate genomes (Hacia et al. 1999). Note that the STSs analyzed here represent randomly sampled genomic sequence, so this analysis does not take into account effects of coding versus noncoding sequence. Approximately 40% of these SNPs create restriction site polymorphisms and thus are accessible to scoring by a simple restriction digestion (Table 2). In addition to the sequences provided in Table 2, the sequences for 365 STSs that define an additional ∼750 polymorphisms in pairwise subsets of the strains are available at http://www.fruitfly.org.

The chromosome-wide distribution of the polymorphisms identified in this study is represented in Figure 1. The distribution is relatively even across the genome, with an average density of one SNP marker per 225 kb on the autosomes and one per megabase on the X chromosome. The lower density of SNPs discovered on the X chromosome is due primarily to a lower density of available STSs. The region flanking the centromere of chromosome 3, including cytological divisions 77 through 83, contains far fewer polymorphisms per kilobase than the rest of the genome. All available STSs in this region were screened in the w;iso2;iso3/P-element strain comparison, and additional comparison of nearly 100 kb of sequence revealed a lack of polymorphism between these strains (data not shown). This observation is consistent with previous studies indicating that levels of sequence polymorphism correlate with recombination rates inDrosophila (Begun and Aquadro 1992; Charlesworth 1996). However, we do not observe a similar reduction of variation near the centromere of chromosome 2.

Figure 1.

Distribution in the Drosophila genome of SNPs identified in this study. The euchromatic portion of the genome is represented by horizontal bars, with the extent of each cytological division representing the genomic extent as estimated by Sorsa (1988). The positions of SNP markers identified in this study are represented by vertical hatch marks: SNPs for which sequence has been determined in all strains are represented by red hatch marks; those that were identified in pairwise subsets of strains are indicated by blue or green. The distribution is relatively even throughout the genome, with the exception of cytological divisions 78–83 near the centromere of chromosome 3. Strain designations (e.g., Q1040) and a small downward arrow indicate the positions of P elements useful for fine-scale mapping.

Open in new tabDownload PowerPointLink to figure

Mapping with SNPs

Having identified a sufficiently dense set of SNPs, we designed a simple strategy to use these markers for mapping single gene mutations (Fig. 2). This strategy requires two to four generations and informative assays for 10–20 SNPs. In the first step, the mutation of interest is mapped relative to six markers that span the chromosome at 8–10 Mb intervals (Fig. 2A), by using 96 unselected, chromosome-wide recombinants. These markers can be SNPs, transposon insertions, or visible markers. Once the interval in which the mutation lies is defined, SNPs are tested on the appropriate recombinants to further refine the position of the gene (Fig. 2B,C). Using this strategy, we mapped two independent recessive lethal mutations isolated in a screen for suppressors of a human p21 overexpression phenotype. These mutations were placed between the STSs Dm1601 and Dm1655 on chromosome 3, an interval of <1 Mb (Fig. 2C) and subsequently were shown to be alleles of the hedgehoggene (Tabata and Kornberg 1994).

To achieve the high-resolution mapping suitable for positional cloning, we have used the P-element strains listed in Table 2 in a strategy to select many white (−) recombinants in a small region (Cutforth and Rubin 1994; Spradling et al. 1995; see Discussion).

Drosophila Strains

Wild-type Drosophila strains were obtained from the Bloomington Stock Center. Canton S and Oregon R were obtained from the laboratory of G.M. Rubin; single P-element insertion strains were obtained from C. Goodman and G.M. Rubin and were derived from an enhancer trap screen performed in their laboratories. Canton S, Oregon R, and w;iso2;iso3 were made isogenic for the indicated chromosomes (X, 2, or 3) according to standard techniques, with the exception of Canton S and Oregon R isogenic chromosome 2 stocks, which were obtained from T. Laverty and G.M. Rubin. All P-element strains and newly isogenized wild-type strains are available from the Bloomington Stock Center.

Molecular Biology

Drosophila genomic DNA was isolated either from adult populations or single recombinant flies according to standard techniques. STSs were amplified from genomic DNA preparations using either standard or touchdown PCR (Don et al. 1991) to facilitate amplification of most STSs under a single set of conditions.

Identification of Sequence Variants

The STSs used in this study were developed by the BerkeleyDrosophila Genome Project (BDGP; Kimmerly et al. 1996), and were selected based on map position inferred from the P1 clones and contigs with which they are associated. STS sequences, primer sequences, and PCR conditions are available on the BDGP Web site (http://www.fruitfly.org). STS sequences were compared among different strains by using the two approaches discussed below.

Recombination Mapping

Alleles of the hedgehog gene were isolated in a screen for ethyl methanesulfonate-induced mutants that suppress a p21 overexpression phenotype. These mutations were induced in a w;iso2;iso3 strain, and balanced mutant stocks were established over TM3 or TM6 by standard methods. The mutations were localized initially by crossing to a mapping chromosome that contained six molecular polymorphisms at evenly spaced intervals. By analyzing 96 random recombinants for both (1) the allele at each marker, and (2) the presence or absence of the mutant gene, the position of the mutation along the chromosome can be established to within a 10–20-Mb interval of the chromosome between the molecular markers. SNP markers in the appropriate interval were amplified from each recombinant and analyzed using DHPLC to generate the haplotypes shown in Figure 2.