Quantitative Trait Loci for Cuticular Hydrocarbons Associated With Sexual Isolation Between Drosophila simulans and D. sechellia

Abstract

The identification of genes with large effects on sexual isolation and speciation is an important link between classic evolutionary genetics and molecular biology. Few genes that affect sexual isolation and speciation have been identified, perhaps because many traits influencing sexual isolation are complex behaviors. Cuticular hydrocarbons (CHs) of species of the Drosophila melanogaster group play a large role in sexual isolation by functioning as contact pheromones influencing mate recognition. Some of the genes that play key roles in determining species-specific CHs have been identified. We have performed separate quantitative trait locus (QTL) analyses of 7-tricosene (7-T) and 7,11-heptacosadiene (7,11-HD), the two major female CHs differing between D. simulans and D. sechellia. We find that ∼40% of the phenotypic variance in each CH is associated with two to four chromosomal regions. A region on the right arm of chromosome 3 contains QTL that affect both traits, but other QTL are in distinct chromosomal regions. Epistatic interactions were detected between two pairs of QTL for 7,11-HD such that if either were homozygous for the D. simulans allele, the fly was similar to D. simulans in phenotype, with a low level of 7,11-HD. We discuss the location of these regions with regard to candidate genes for CH production, including those for desaturases.

STUDIES of the genetics of speciation have, until very recently, been hindered by a paucity of candidate genes for detailed analysis (Swanson and Vacquier 2002; Noor 2003). Reproductive isolation, responsible for speciation, is likely to involve complex, coevolved, polygenic traits (leading to a “type I” genetic architecture in the terminology of Templeton 1981, i.e., numerous genes of small effect). Most empirical studies of reproductive isolation, especially of sexual isolation, have confirmed polygenic effects (Hollocher et al. 1997; Ritchie and Phillips 1998; Ting et al. 2001); however, a few large effect genes have been identified for both postmating (Ting et al. 1998; Barbash et al. 2003; Presgraves et al. 2003) and premating isolation (Wheeler et al. 1991; Greenberg et al. 2003). In Drosophila, most of these large effect genes have been identified by mutagenesis of D. melanogaster or by association analysis with geographic variation. Quantitative trait loci (QTL) studies, in contrast, can identify the minimum number and location of genes influencing complex traits (Mackay 2001) and therefore identify the traits that are likely to be influenced by genes of large effect underlying naturally occurring variation. QTL studies may also indicate the potential candidate genes underlying genetic differentiation.

Cuticular hydrocarbons (CHs) affect mating behavior in many Drosophila species and are traits with the potential for a type II (large gene effect) genetic architecture (Templeton 1981). CHs are long-chain hydrocarbons located on the cuticle surface that prevent desiccation and can function as contact pheromones (Blomquist et al. 1987). During courtship, male Drosophila perform several different behaviors, including orienting toward a female, vibrating a wing (producing courtship “song”), and tapping her abdomen with his foreleg. In tapping the female, the male detects her CHs. Drosophila species are either monomorphic, with both males and females having the same CH composition (e.g., D. simulans and D. mauritiana, Jallon and David 1987), or dimorphic with differing profiles (e.g., D. melanogaster and D. sechellia, Cobb et al. 1989). CH profiles affect interspecific mating asymmetrically: males of monomorphic species will not court females of dimorphic species, whereas males of dimorphic species will court females of monomorphic species (Cobb and Jallon 1990).

The major CH affecting mate recognition in D. simulans is 7-tricosene (7-T, Péchiné et al. 1985), which is produced by both males and females. Male D. sechellia have several tricosenes, including the same isomer, whereas females have very little 7-T and high levels of 7,11-heptacosadiene (7,11-HD, Cobb et al. 1989). In crowding experiments, in which one female of a species was housed with females of the other species, the CH of the abundant females was transferred to the single female. In subsequent mating tests, the behavior of the males toward the females was changed such that D. simulans females housed with D. sechellia were courted less frequently than control D. simulans females by D. simulans males. In contrast, D. sechellia females housed with D. simulans were courted more vigorously than D. sechellia controls (Coyne et al. 1994).

Through study of the biochemistry of CH production in Drosophila, several genes that affect CH production have been identified. In insects, especially Dipterans, long-chain hydrocarbons with odd numbers of carbons are derived from even-numbered fatty acids by reduction and decarboxylation (Blomquist et al. 1987), a pattern confirmed for the biosynthesis of the major CHs of D. melanogaster and D. simulans (Jallon 1984; Chan Yong and Jallon 1986; Pennanec'h et al. 1997). Early pathway steps may be common to both sexes and both species. The main substrates are myristate and palmitate, which are produced by a fatty acid synthetase (de Renobales and Blomquist 1984). From labeling experiments, two types of desaturases are implicated and molecular studies have discovered two closely linked genes that code Δ9 desaturases, but use different substrates: palmitate for desat1 (Wicker-Thomas et al. 1997; Dallerac et al. 2000) and myristate for desat2 (Dallerac et al. 2000). The desat1 gene is involved in the first desaturation of these saturated fatty acids leading to hydrocarbons bearing a ω7 double bond common to males and females of both species (Labeur et al. 2002). The desat1 gene affects the level of 7,11-HD and the desat2 gene affects that of 5,9-HD in D. melanogaster females (Wicker-Thomas et al. 1997; Coyne et al. 1999; Dallerac et al. 2000; Takahashi et al. 2001; Fang et al. 2002; Greenberg et al. 2003; Marcillac et al. 2005). The desat2 gene, functional in African D. melanogaster populations, is not functional in Cosmopolitan strains because of a small deletion in the promoter region and results in females that are rich in 5,9-HD (Takahashi et al. 2001; Wicker-Thomas and Jallon 2001). The desat1 gene differs between D. melanogaster types in only three amino acid replacements and its function is not greatly altered, leading to the major production of 7-monenes in both sexes and to 7,11-dienes in non-African females (Dallerac et al. 2000). Environmental selection upon the CH components may have driven the change at desat2 (Greenberg et al. 2003), but whether this has also influenced assortative mating within D. melanogaster (including the premating isolation seen between Zimbabwe and Cosmopolitan strains) is uncertain (Coyne et al. 1999; Greenberg et al. 2003; Ritchie and Noor 2004).

The diene second double bond of D. melanogaster females requires a second desaturation step, which is probably catalyzed by a different desaturase (Wicker-Thomas and Jallon 2001). Flies carrying various deficiencies in the third chromosome region 67E–69B have variation in unsaturated hydrocarbon levels in both sexes, although the difference is much larger in females than in males (Wicker-Thomas and Jallon 2000). One gene in this region, Enhancer of zeste [E(z)], decreases the level of female 7,11-HD and has a correlated increase in 7-heptacosene, depending on the strength of the allele (Wicker-Thomas and Jallon 2000). The same type of phenotype has been observed in mutations in two temperature-sensitive endocrine genes, ecdysoneless (ecd, Wicker and Jallon 1995) and Dopa decarboxylase (Ddc, Marican et al. 2004) when females were bred at restrictive temperature.

Other genes have also been implicated in CH variation and possibly mate choice. Two second chromosome genes, seven pentacosene (sept) and small monoene quantities (smoq), affect the balance between 7-T and 7-pentacosene (7-P) in D. melanogaster males (Ferveur and Jallon 1996). The nerd locus affects the level of 7-T production in D. melanogaster (Ferveur and Jallon 1993b) and has been mapped to the third chromosome. Although sept and smoq affect the relative production of 7-T and 7-P in D. melanogaster, in D. simulans, a single locus on the second chromosome, Ngbo, has a dose-dependent effect on production of 7-P. Ngbo has been proposed to be a structural gene that changes enzyme specificity (Ferveur 1991). Also in D. simulans, the sex-linked kété locus affects overall levels of 7-T, 7-P, and all linear hydrocarbons (Ferveur and Jallon 1993a).

Genes involved in somatic sexual differentiation may also affect CH composition. The gene doublesex (dsx) affects the production of female-specific dienes in D. melanogaster (Jallon et al. 1988). The female-specific transcript DSX F acts as a dominant regulator of female dienes (Waterbury et al. 1999). The intersex (ix) locus affects the ratio of production of the female-specific 7,11-HD and 7,11-nonacosadiene (7,11-ND) to the production of 7-T and 5-T, which are more prominent in males (Waterbury et al. 1999). Differential expression of transformer (tra) affects the production of sex-specific pheromones (Savarit et al. 1999) as does transformer-2 (tra2, Jallon et al. 1986) and Sex-lethal (Sxl, Tompkins and McRobert 1989; Tompkins and McRobert 1995). The homeotic genes Antennapedia (Antp) and Ultrabithorax (Ubx) also influence the relative levels of monoenes vs. dienes (Wicker-Thomas and Jallon 2001).

With the completion of the D. melanogaster genome sequence, additional genes are identifiable through domain similarity with genes of known function. Two such genes are Fad2 and infertile crescent (ifc), both of which have stearoyl CoA-desaturase activity (FlyBase Consortium 2003). Additional unknown genes with this function include CG8630, CG9743, and CG15531. Other activities include a SUR2-type hydroxylase/desaturase catalytic domain (CG1998, CG11162, and CG30502), acyl-CoA Δ11-desaturase activity (CG9747), and a fatty acid elongase (CG2781).

In this study, we use quantitative trait loci (QTL) mapping to refine the number and locations of loci affecting 7-T and 7,11-HD, the most important CHs differing between D. simulans and D. sechellia, and compare the position of QTL with that of the candidate genes. Previously, the location of genes involved in the difference in CH composition between D. simulans and D. sechellia has been studied in crossing experiments with limited markers. Using a D. sechellia strain with only one marker per chromosome, Coyne et al. (1994) studied the ratio of 7-T to 7,11-HD in the female progeny of hybrid females backcrossed to D. sechellia. The third chromosome had the largest effect on the trait. The X chromosome had an effect only on the amount of 7-P, a polymorphic CH found only on D. sechellia, which also induced male courtship by D. sechellia males (Cobb and Ferveur 1996; Coyne et al. 1994).

MATERIALS AND METHODS

Strains and crosses:

One strain each of the two species was used in this study. A D. sechellia strain was kindly provided by Jean R. David. The strain was inbred for 18 generations of brother-sister mating to produce the strain DavidA4 used in the subsequent crosses. The D. simulans line (f²;nt, pm;st,e), kindly provided by Jerry Coyne, had five morphological markers, one per chromosome arm (Table 1).These are the same strains used in a QTL analysis of courtship song differences between these species (Gleason and Ritchie 2004).

TABLE 1.

Markers used and their resulting map order

Ordera	Geneb	Abbreviation	Locationc	Size relationd	Sourcee
X chromosome
1	scute	sc	1A8	sec > sim	1
2	Phosphogluconate dehydrogenase	Pgd	2D4	RsaI cuts sim	Intron
3	period	per	3B1–2	sim > sec	Repeat
4	defective chorion	dec1	7C1	DdeI cuts sim	X96929, X96931
5	DS01391 (microsatellite)	DS01391	9A1–2	sec > sim	2
6	cacophony	cac	10F8–11A1	BstUI cuts sec	Intron
7	no on or off transient A	nonA	14B18–C1	sec ≫ sim	Intron
8	forked	f2	15F4–7	Morphological
9	Zwischenferment (G6PD)	Zw	18D13	sec > sim	Intron
Second chromosome
1	expanded	ex	21C4–6	sec > sim	3
2	net	nt	21B1	Morphological
3	odd skipped	odd	24A1	sim > sec	4
4	Mst26Aa	Mst	26A1	sec > sim	X70899, X72630; intron
5	big brain	bib	30F5	sec > sim	Repeat
6	spalt	sal	32F1–2	sim > sec	Intron
7	Supressor of Hairless	Su(H)	35B8	sim > sec	4
8	caudal	cad	38E9–10	sec > sim	Repeat
9	Phosphoglucose isomerase	Pgi	44F6	TaqI cuts sim	Intron
10	G protein oα47A	Dgα	47A7–9	sim > sec	1
11	slit	sli	52C9–D1	sec > sim	2
12	grainy head	grh	54E10–F1	sim > sec	4
13	plum	pm	59E2–3	Morphological
14	twist	twi	59C2	TaqI cuts sim	Repeat
Third chromosome
1	veinlet	ve	62A2	sim > sec	1
2	Cdc37	Cdc37	62B4	Alw26I cuts sim	6
3	temperature induced paralytic E	tipE	64A10	sim > sec	Intron
4	Laminin B2	LanB2	67C2	MfeI cuts sec	7
5	Superoxide dismutase	Sod	68A7	HaeIII cuts sim	Intron
6	Esterase 6	Est6	69A1	BsrGI cuts sec	Intron
7	Accessory gland peptide 70A	Acp70A	70A4	sim > sec	X99414, X99417; intron
8	scarlet	st	73A3	Morphological
9	transformer	tra	73A10	sec > sim	Intron
10	Catalase	Cat	75E1	sim > sec	1
11	Glucose dehydrogenase	Gld	84D3	HaeIII cuts sim	Noncoding
12	ebony	e	93C7–D1	Morphological
13	nanos	nos	91F7	sim > sec	3
14	glass	gl	91A3	RsaI cuts sec	6
15	prospero	pros	86E2–4	CfoI cuts sim	1
16	Metallothionein A	Mtn	85E9	DraI cuts sec	Intron
17	slowpoke	slo	96A14–17	sec > sim	Intron
18	Myosin alkali light chain 1	Mlc1	98A14–15	Cuts sec	L49010, L49009; intron
19	similar	sima	99D3–7	Hsp92II cuts sec	8
20	janus	jan	99D3	DdeI cuts sec	5
Fourth chromosome
1	eyeless	ey	102C2	sim ≫ sec	Intron

^aThe order is that on the mapped chromosome (Fig. 1).

^bCandidate genes are identified by underlining.

^cCytological locations were obtained from Flybase (http://flybase.bio.indiana.edu/) and are for D. melanogaster.

^dSize of PCR products of D. simulans (sim) relative to D. sechellia (sec) or restriction enzyme used to digest PCR products. Five markers are morphological, not molecular, and are designated as such.

^eSequences of primers can be found in these references: (1) Schug et al. (1997), (2) http://i122server.vu-wien.ac.at/, (3) Goldstein and Clark (1995), (4) Michalakis and Veuille (1996), (5) Liu et al. (1996), (6) Schug et al. (1998), (7) Colson and Goldstein (1999), (8) Colson et al. (1999). Of these, numbers in italics designate primers for microsatellite sequences for which we did not find PCR product length differences. Instead a restriction enzyme was used to resolve the two species. The designations “intron,” “repeat,” and “noncoding” indicate markers that are new to this study and are PCR products crossing an intron, incorporating a repeat, or in noncoding sequence, respectively. GenBank accession numbers are given for sequences available from both species. Primer sequences and PCR conditions are available by request from the authors.

All fly culturing was at 25° using standard techniques and a 12 hr light:12 hr dark cycle. Female D. simulans were crossed to male D. sechellia and the female progeny were backcrossed to D. simulans males. For each cross, one female was paired with a single male in a vial (95 × 16.5 mm) for 7 days. A total of 44 crosses and 78 backcrosses were used to generate 487 females that were subsequently analyzed for cuticular hydrocarbons. Using the five morphological markers, we attempted to sample equally among the 32 backcross phenotypes to obtain representatives of all recombinants.

Cuticular hydrocarbon extraction and analysis:

Flies were collected within 8 hr of eclosion and phenotyped for the five morphological markers. Females were housed individually in small vials (95 × 16.5 mm). Five days post-eclosion, the female fly was placed for 5 min in 70 μl hexane containing 800 ng hexacosane as a control for recovery. The fly was then removed and DNA was extracted using the single fly extraction method and a total of 100 μl squishing buffer (Gloor and Engels 1992). The hexane containing the cuticular hydrocarbons was dried by evaporation and stored at −20° until gas chromatography.

Before gas chromatography, the sample was resuspended in 40 μl hexane. Five microliters was injected into a Perkin-Elmer (Norwalk, CT) gas chromatograph autosystem with split injection, a flame ionization detector, and a 25QC2/BP1 0.1 column (Scientific Glass Engineering, 25 mm × 0.22 mm × 0.1 μm) programmed to run from 180° to 270° with a 3°/min gradient.

The quantities of two cuticular hydrocarbons, 7-T and 7,11 HD, were calculated from the resulting chromatographs as the area under each peak. These were standardized by dividing by the quantity of hexacosane recovered, to account for differences in extraction and gas chromatography analysis. Eight data outliers were removed and the resulting values were natural log transformed to remove a right skew. Remaining residuals were normally distributed. All subsequent analyses were done with the transformed variables. Final sample sizes were 484 females for 7-T and 482 females for 7,11-HD.

Marker scoring:

Thirty-nine molecular markers were scored for each individual (N = 487, Table 1). These markers were all polymerase chain reaction (PCR) amplified and showed different size fragments for D. sechellia and D. simulans on 2% agarose, 4% Metaphor agarose (Cambex), or 8% acrylamide gels. Size differences were caused by natural variation in sequence length (indels or microsatellites) or differences in restriction enzyme sites (Table 1). Hybrids were easily distinguished from homozygotes.

Genetic mapping and QTL analysis:

With the morphological markers (see Strains and crosses), a total of 44 markers were scored. These markers were mapped using MAPMAKER (Lander et al. 1987). The map obtained was subsequently used in QTL analyses using QTL Cartographer v1.16c (Basten et al. 1997) to map QTL. Epistatic interactions between QTL were detected by a generalized linear model in SPSS 13.0 (SPSS) with Bonferroni correction for multiple tests, using allelic data at the nearest marker to the QTL.

Calculation of effects in terms of parental difference:

All analyzed data were natural log tranformed. The effect, as calculated, is the difference between the mean of the logs for the genotype (MLG) and the mean of the logs for the traits of all backcross individuals (MLT). This difference is equivalent to the log of ratios of the geometric mean for the genotypes (GMG) and the grand geometric mean (GMT); that is, MLG − MLT = log(GMG/GMT). Therefore, if exp(effect)−1 is multiplied by GMG, the result is GMT. Subtracting GMT from GMG yields the effect in the original units. The percentage of the parental difference was then calculated by dividing this effect by the difference in the mean value between the parental strains.

RESULTS

Marker mapping:

The markers were chosen to have an average spacing of approximately 7 cM on the basis of the D. melanogaster map. The average spacing realized between markers was 23.2 cM, because of segregation distortion and experimenter selection for recombinants. Most markers mapped in the same linear order as in D. melanogaster with the exception of two marker pairs on each end of the second chromosome (ex and nt; pm and twi). On the right arm of the third chromosome there is an inversion in D. simulans and D. sechellia relative to D. melanogaster and five markers (Mtn, pros, gl, nos, e) show this inversion with respect to D. melanogaster.

Trait values:

As expected, D. simulans females had more 7-T than D. sechellia and less 7,11-HD (Table 2). Values for hybrid and backcross females were intermediate, but hybrid females were more like D. sechellia for 7-T and more like D. simulans for 7,11-HD [similar to other crosses between D. mauritiana and D. sechellia (Coyne and Charlesworth 1997)], but different from crosses between D. simulans and D. melanogaster (Coyne 1996). Backcross females were more like D. simulans than D. sechellia for both CHs. CH components are potentially interrelated in that they are products of a common production pathway. The correlation between 7-T and 7,11-HD across recombinant females here was only −0.2, indicating the potential for different genes to independently influence the production of the compounds. We therefore analyzed QTL for the two traits separately (previous studies have analyzed the ratio of the two compounds).

TABLE 2.

Mean ± standard deviation (in nanograms) and sample sizes (n) for both cuticular hydrocarbons

Type	7-Tricosene	n	7,11-Heptacosadiene	n
D. simulans females	2445.6 ± 645.5	15	4.0 ± 15.2	15
D. sechellia females	8.8 ± 17.6	10	540.0 ± 296.0	10
Hybrid females	576.0 ± 204.8	20	84.0 ± 55.2	19
Backcross females	1847.2 ± 824.8	484	33.6 ± 64.8	482

QTL analysis:

The marker on the fourth chromosome (ey) was not significantly associated with either trait and thus results for this chromosome, which composes about 1% of the genome, are not shown. Composite interval mapping (CIM) was performed for the rest of the genome. CIM (Jansen and Stam 1994; Zeng 1994) combines interval mapping (Lander et al. 1987) with multiple regression. Each interval flanked by adjacent markers is tested for the presence of a QTL affecting the trait while statistically accounting for the effects of additional segregating QTL outside the interval. Significance levels of P = 0.05 for both CH were calculated from 1000 permutations of the trait data among marker classes (Churchill and Doerge 1994) and corresponds to a likelihood ratio of 14.33 for 7-T and 13.95 for 7,11-HD.

Parameters potentially affecting the detection of QTL using CIM are the size of the window around the tested interval, within which linked markers are excluded from multiple regression, and the number of background markers used. We tested a range of window sizes from 2.5 to 20 cM and this parameter did not influence the result. Figure 1 depicts the results using a backcross design, the Kosambi map function, all background markers, a walking speed of 2 cM, and a window size of 5 cM. Forward/backward stepwise regression resulted in nine significant markers for 7-T and eight significant markers for 7,11-HD that could be used in CIM. Results varied slightly with the addition of markers. For 7-T, QTL 6 narrowed with the addition of the fifth marker, Mtn, which bounds one edge of the QTL. Otherwise, all QTL were stable after the addition of the second marker for this trait. For 7,11-HD, adding the third marker (which is on the right boundary of QTL2) reduced QTL1 from approximately equal to QTL2 to a value of ∼26. Adding the sixth marker (which is on the right boundary of QTL 3) reduced it further to the presented value. For QTL3, adding the second marker (Mtn, in the middle of the QTL) narrowed this QTL to where it became stable. For these reasons, using all the markers for the final results gives a conservative estimate of the height of the QTL.

Figure 1.

QTL mapping of cuticular hydrocarbons. The amount of both 7-tricosene (7-T, shaded) and 7,11-heptacosadiene (7,11-HD, solid) in D. simulans-D. sechellia backcross hybrids was mapped using composite interval mapping. The positions of the markers, listed in Table 1, are denoted by circles along the x-axis. Triangles at the top of the graph indicate the positions of each QTL and are numbered for reference to the text. The significance level at P = 0.05 was determined by 1000 permutations of the data set and is shown for 7-T at a likelihood ratio of 14.34 (the level for 7,11-HD is 13.95). The locations of candidate genes are indicated by vertical lines along the x-axis, with the exception of the one candidate gene, tra, that was also a marker and is represented by a shaded circle. Abbreviations for candidate genes are given in Table 6.

Seven QTL were found for 7-T and three for 7,11-HD (Figure 1, Table 3). For 7-T, QTL 1 and 2 are located on the X chromosome (Figure 1), whereas the other five are on the third chromosome. Six of the QTL are in pairs with one contributing more to the phenotypic variance (V_p) than the other (QTL1 and 2, QTL6 and 7) or both having a small contribution (QTL3 and 4), and QTL5 is independent. Therefore, at least four distinct genomic regions are associated with the trait, with the greatest effects being on the right-hand sides of the first and third chromosomes. For 7-T, the proportion of V_p explained by each QTL varies from 2.29 to 17.68% (Table 3). Together, the QTL explain 44.10% of V_p. The effects for the QTL are all in the same direction, consistent with the trait being expressed in one species and not in the other. The effects, expressed as a percentage of the difference in means between the two parental species, range from 53 to 68% (Table 3).

TABLE 3.

QTL locations and effects

QTL	Chromosome	Position (cM)	Effect	%Vpa	%PDb
7-T
1	X	167.04	0.2606	17.68	53.23
2	X	175.76	0.1806	4.57	60.80
3	3	113.07	0.0964	2.30	68.13
4	3	115.93	0.1005	2.49	67.79
5	3	251.96	0.1021	2.29	67.66
6	3	335.68	0.2270	11.8	56.49
7	3	344.23	0.1567	2.97	62.94
7,11-HD
1	3	88.52	−0.0382	2.77	6.49
2	3	96.57	−0.0650	17.00	6.65
3	3	342.23	−0.0623	18.20	6.64

^aPercentage of phenotypic variance explained.

^bPercentage of parental difference.

For 7,11-HD, all of the QTL are on the third chromosome (Figure 1). Two QTL are paired, with one contributing more than the other (QTL1 and 2), and the other is independent, so that CIM identifies two regions of large effect. The QTL vary from 2.77 to 18.20% of V_p and together account for a total of 37.97% of V_p (Table 3). Again, all of the effects are in the same direction, which is opposite of that for the 7-T QTL. The effects are approximately the same when expressed as a proportion of the parental difference, ranging from 6.4 to 6.7%. This is a small proportion, heavily influenced by our crossing design (backcrosses to D. simulans). Hence, there was relatively little V_p, centered around the D. simulans parental value, so only a fraction of the parental difference could be explained here.

Epistatic interactions were tested using a generalized linear model in SPSS. No interactions between pairs of QTL for 7-T were significant after Bonferroni correction. For 7,11-HD, there were significant interactions between QTL3 and the other two QTL (Table 4). The interaction effect is such that both QTL regions need to be in a hybrid state to produce a significant amount of 7,11-HD (Table 5).

TABLE 4.

Significance of epistatic interaction of 7,11-HD

Source	Sum of squares	d.f.	Mean square	F	Significance
Corrected model	0.938	3	0.303	93.957	<0.001
Intercept	0.790	1	0.790	237.383	<0.001
QTL1	0.368	1	0.368	110.431	<0.001
QTL3	0.451	1	0.451	135.483	<0.001
QTL1XQTL3	0.209	1	0.209	62.842	<0.001
Error	1.571	472	0.003
Total	3.158	476
Corrected total	2.509	475
Corrected model	1.278	3	0.426	163.603	<0.001
Intercept	0.795	1	0.795	305.499	<0.001
QTL2	0.589	1	0.589	226.099	<0.001
QTL3	0.468	1	0.468	179.673	<0.001
QTL2XQTL3	0.359	1	0.359	137.836	<0.001
Error	1.221	469	0.003
Total	3.140	473
Corrected total	2.498	472

TABLE 5.

Trait values for epistatic interactions of 7,11-HD^a

	QTL1 X QTL3		QTL2 X QTL3
	QTL 1 homozygote	QTL 1 heterozygote	QTL 2 homozygote	QTL 2 heterozygote
QTL 3 homozygote	0.0031 ± 0.0011, 129	0.0169 ± 0.0035, 125	0.0018 ± 0.0010, 121	0.0174 ± 0.0033, 134
QTL 3 heterozygote	0.0229 ± 0.0056, 119	0.1208 ± 0.0093, 102	0.0097 ± 0.0041, 117	0.1360 ± 0.0090, 101

^aTrait values are means of the genotype ± standard error for natural log transformed data. The mean value of the trait was 0.0369 ± 0.0033. Sample sizes for each genotype are given.

Twenty-seven candidate genes for cuticular hydrocarbon biosynthesis were identified from the literature and from the available sequence data. The approximate locations of the candidate genes were estimated by their genetic location in D. melanogaster. Candidate genes do not fall directly with the QTL (Table 6, Figure 1), although several are on the edges of QTL including CG2781, dsx, desat1, and desat2 for 7-T. E(z) and Fad2 are on the edge of a 7,11-HD QTL.

TABLE 6.

Cuticular hydrocarbon biosynthesis candidate genes coincidence with cuticular hydrocarbon QTL

Gene	Abbr.a	Sourceb	Locationc	Cyt.d	7-Te	7,11-HDe
X chromosome
kété	kété	D. simulans	Left of dec1	No	No
Sex lethal	Sxl	D. melanogaster	Left of dec1	6F3-5	No	No
CG1998	Flybase	Right of cac	11F4	No	No
CG11162	Flybase	Between cac and nonA	12B2	No	No
Second chromosome
seven pentacosene	sept	D. melanogaster	To the left of ex	21C1	No	No
infertile crescent	ifc	D. melanogaster	To the right of Mst	26B2	No	No
Dopa decarboxylase	Ddc	D. melanogaster	To the left of cad	37C1	No	No
CG30502	Flybase	To the left of Pgi	43C1	No	No
intersex	ix	D. melanogaster	To the right of Dgα	47F5	No	No
Ngbo	Ngbo	D. simulans	Between Dgα and sli	No	No
small monene quantities	smoq	D. melanogaster	To the left of sli	49D3	No	No
transformer-2	tra-2	D. melanogaster	Between Dgα and sli	51B6	No	No
Third chromosome
ecdysoneless	ecd	D. melanogaster	To the right of Cdc37	62D2	No	No
Enhancer of zeste	E(z)	D. melanogaster	To the right of LanB2	67E5	No	On edge
Fad2	Fad2	D. melanogaster	Left of Sod	68A1	No	On edge
transformer	tra	D. melanogaster	Marker	73A1	No	No
Antennapediea	Antp	D. melanogaster	Left of Gld	84B1–2	No	No
doublesex	dsx	D. melanogaster	Right of Gld	84E5–6	On edge	No
CG2781	Flybase	Right of Gld	84E6–7	On edge	No
Ultrabithorax	Ubx	D. melanogaster	Right of gl	89D6–9	No	No
CG8630	Flybase	Left of pros	87E3–4	No	No
desat1	desat1	D. melanogaster	Left of pros	87B10–11	On edge	No
desat2	desat2	D. melanogaster	Left of pros	87B10–11	On edge	No
CG9743	Flybase	Right of jan	99E2	No	No
CG9747	Flybase	Right of jan	99E2	No	No
CG15531		Flybase	Right of jan	99E2	No	No

^aAbbreviations are used to identify position in Figure 1.

^bSpecies in which the gene has been demonstrated to affect cuticular hydrocarbon profiles or information from gene sequence in Flybase implies protein structure that might potentially affect CH synthesis.

^cChromosomal location on Figure 1.

^dCytological location for gene in D. melanogaster (from Flybase). Not all have been mapped and some are known only in D. simulans.

^eAssociation with QTL for each trait.

DISCUSSION

QTL studies of interspecific differences in adaptive quantitative traits have found QTL with both large (Laurie et al. 1991; Macdonald and Goldstein 1999) and minor effects (e.g., Fishman et al. 2002). Many interspecific trait differences have been shown to be polygenic (e.g., Kim and Rieseberg 1999; Zeng et al. 2000). The time since species divergence has been hypothesized to be positively correlated with the number of QTL found (Kim and Rieseberg 1999), although another possibility is that some traits are more prone to major gene effects than others. If a threshold of >25% of the phenotypic variance explained is used to designate a major QTL (Bradshaw et al. 1995, 1998), then neither trait studied here is influenced by major QTL. The majority of the phenotypic variation is not explained, indicating that, despite the large sample size, the study lacked sufficient resolution to detect additional small-effect QTL that also influence this trait difference. However, a high proportion of the parental difference is explained for 7-T. Given the values for the hybrids (Table 2), there are clearly dominance effects of alleles for both traits. For 7-T, D. sechellia alleles appear to be dominant, whereas for 7,11-HD, D. simulans alleles appear to be dominant. Using a backcross design, we cannot distinguish dominance from other genetic effects; because of the sterility of F₁ males, backcross analysis is the only crossing scheme possible with these species. Furthermore, we cannot detect all of the possible effects of D. sechellia alleles because our backcross was to D. simulans.

The effects of the QTL are consistent with the difference between the species: all effects are positive for 7-T and negative for 7,11-HD. This pattern is consistent with directional selection operating on these two species. The question remains open whether this might be from sexual selection, by male preferences for the female trait, or from natural selection, such as that found for different dienes in D. melanogaster females (Greenberg et al. 2003). The epistatic effect means that both members of interacting loci need to have an allele from D. sechellia to produce increased levels of 7,11-HD. This implies that each locus codes for an essential step in the production of 7,11-HD, which is absent in D. simulans. The combination of D. sechellia alleles at both loci produces a substantial change away from the D. simulans-like phenotype.

The large contribution of QTL on the third chromosome to both traits is probably what was detected by Coyne et al. (1994) in their study with one marker per chromosome. In that study, the ratio of 7-T to 7,11-HD was analyzed rather than each compound separately; thus the effect of the X chromosome, contributing only to 7-T production, was not detected. In a study using introgression lines (Civetta and Cantor 2003), a QTL for 7-T quantity in males of these species was found at approximately the same postion as our 7-T QTL 3 and 4. In contrast to our results, they did not find any QTL for this trait in females, probably because the study had low resolution for D. sechellia (Civetta and Cantor 2003).

A QTL study for a courtship song difference between D. simulans and D. sechellia (Gleason and Ritchie 2004) found that most of the QTL do not coincide with previously identified candidate genes. The results here are similar. In Figure 1, the 7-T QTL 6 appears to be adjacent to two fatty acid desaturases, desat1 and desat2. Because alleles of desat2 affect the ratio of 7,11-HD, in D. melanogaster females (Greenberg et al. 2003), this gene was thought to be a strong candidate gene for the species difference. Using a 2 LOD (9.22 likelihood ratio) confidence interval for the width of 7,11-HD QTL 3, these genes are not even adjacent to that QTL.

The desaturase family of enzymes do not colocalize. The complete D. melanogaster genome has revealed five more fatty acid desaturase genes, in addition to desat1 and desat2, on chromosome 3 and one on chromosome 2 (Jallon and Wicker-Thomas 2003). Markedly decreased levels of dienes were found in D. melanogaster females carrying deletions in the region 67E–69B [corresponding here to the region surrounding E(z) and Fad2, Figure 1]. Molecular work in progress in that region has characterized Fad2 as an additional desaturase gene, which, expressed only in females of the dimorphic species, acts on unsaturated fatty acids and leads to dienes (T. Chertemps and C. Wicker-Thomas, unpublished results).

Another group of enzymes (elongases) is involved in the elongation of the hydrocarbon chains and one of these (CG2781) is found on the third chromosome and on the edge of 7-T QTL 5. The same is true for doublesex. Although these two genes are unlikely to contribute to this QTL on the basis of these data, to distinguish the contribution of linked candidate genes to wide QTL peaks is notoriously difficult, and further resolution will require high-level recombination mapping and assessing allelic variation at the candidate loci themselves.

While there is partial overlap in some of the QTL identified for the two CH components studied here, at least three that affect 7-T are not implicated in the production of 7,11-HD. Thus, some steps in the biosynthesis of these compounds are affected by different genes. 7-T and 7,11-HD differ in chain length (23 vs. 27 carbons, respectively) and saturation (one vs. two double bonds, respectively). The genetic effects could therefore be in chain elongation, desaturation, or in overall levels of production. Only by further refined mapping to the genes underlying these traits will we be able to determine the contribution of each gene to hydrocarbon production. Despite the small effects relative to the phenotypic variance, the likelihood ratios of the QTL presented here are very large and indicate that it should be possible to map these traits more finely and identify the genes contributing to this important trait difference.