Abstract
Most animals depend on olfaction for survival and procreation. Odor-guided behavior is a quantitative trait, with phenotypic variation due to multiple segregating quantitative trait loci (QTL). Despite its profound biological importance, the genetic basis of naturally occurring variation in olfactory behavior remains unexplored. Here, we mapped a single Drosophila QTL affecting variation in avoidance response to benzaldehyde, using a population of recombinant inbred lines. Deficiency complementation mapping resolved this region into one female- and one male-specific QTL. Subsequent quantitative complementation tests to all available mutations of positional candidate genes showed that the female-specific QTL failed to complement a P-element insertional mutation, l(3)04276. The P-element insertion was in the intron of a novel gene, Vanaso, which contains a putative guanylate binding protein domain, is highly polymorphic, and is expressed in the third antennal segment, the major olfactory organ of Drosophila. No expression was detected in the fly brain, suggesting that Vanaso plays a role in peripheral chemosensory processes rather than in central integration of olfactory information. QTL mapping followed by quantitative complementation tests to deficiencies and mutations is an effective strategy for gene discovery that allows characterization of effects of recessive lethal genes on adult phenotypes and here enabled identification of a candidate gene that contributes to sex-specific quantitative variation in olfactory behavior.
ALL organisms interact with their environment and physiological responses to environmental stimuli are, ultimately, expressed as behaviors driven by the nervous system and enabled by the genome. Behavioral adaptations to environmental cues determine survival and reproductive success; indeed, behavior supports the stage on which the interplay between the environment and the genome guides evolution.
Among environmental cues on which organisms depend, none are more important than chemical signals. Odor-guided behavior is essential for most organisms for food localization, avoidance of environmental toxins or predators, oviposition site selection, kin recognition, species recognition, mate selection, and reproduction. While great advances have been made in understanding the neural mechanisms that mediate recognition of odorants and pheromones (reviewed by Mombaerts 1999; Firestein 2001), including the recent identification of multigene families of olfactory (Clyneet al. 1999; Gao and Chess 1999; Vosshallet al. 1999) and gustatory (Clyneet al. 2000; Dunipaceet al. 2001; Scottet al. 2001) receptors in Drosophila, the genetic architecture that enables the expression of odor-guided behavior at the organismal level has remained virtually unexplored (Anholt and Mackay 2001).
It is intuitively evident that the response to odorants involves a vast ensemble of genes and that variation in the expression of any of these genes can generate individual variation in olfactory behavior within a natural population. Thus, odor-guided behavior is a quantitative trait and understanding its complex genetic architecture requires identification of genes that contribute to naturally occurring variation. Previously, 43 X chromosomes and 35 third chromosomes were extracted from a natural population of Drosophila melanogaster and substituted into a common inbred genetic background (Mackayet al. 1996). Quantitation of avoidance responses to the repellent odorant benzaldehyde revealed significant genetic variation in olfactory behavior. Intriguingly, the genetic correlations between the sexes for olfactory avoidance behavior were extremely low, suggesting that different genes contribute to variation in odor-guided behavior in males and in females. Thus, the genetic architecture for olfactory behavior is sex specific, indicating that the olfactory subgenome in males and females may evolve along different evolutionary trajectories (Mackayet al. 1996). This observation was corroborated by the finding that a significant fraction of transposon-tagged smell-impaired mutants show sex-specific olfactory impairments (Anholtet al. 1996).
Despite its profound biological importance, no genes have yet been identified that contribute directly to phenotypic variation in odor-guided behavior. Here, we utilized a stepwise quantitative genetic approach that enabled us, first, to identify a quantitative trait locus (QTL) responsible for sexually dimorphic variation in chemosensory behavior; second, to narrow the QTL interval to regions containing only a small number of candidate genes; and, finally, to identify one novel candidate quantitative trait gene (QTG) responsible for the QTL effect. As would be predicted from a gene that contributes to phenotypic variation in olfactory behavior, this novel gene is, even for Drosophila, exceptionally polymorphic and is selectively expressed in the third segment of the antenna, the main olfactory organ of adult flies.
MATERIALS AND METHODS
Olfactory behavior: Olfactory avoidance responses were measured as described previously (Anholtet al. 1996). Single-sex groups of five individuals were placed for 2 hr in empty culture vials marked with two lines, 3 and 6 cm from the bottom. A cotton swab saturated with 1% benzaldehyde (Aldrich Chemical, Milwaukee) was inserted such that the tip was aligned with the 6-cm mark. Ten counts of the number of flies in the compartment delimited by the bottom and the 3-cm line were taken at 5-sec intervals, starting at the 15-sec time point. The avoidance score of each replicate was the average of these 10 counts.
QTL mapping: A mapping population of 98 recombinant inbred lines (RILs) was derived between two unrelated homozygous strains, Oregon and 2b, that were not selected for olfactory behavior (Nuzhdinet al. 1997). Cytological insertion sites of 81 polymorphic roo transposable elements were determined for each of the RILs, to generate a dense molecular marker map with an average spacing of 3.2 cM (Nuzhdinet al. 1997; Leips and Mackay 2000; Vieiraet al. 2000). The avoidance response to benzaldehyde was quantified for each RIL from five replicate measurements per sex. Mixed model analysis of variance (ANOVA) was used to partition variation in avoidance response according to the model Y =μ + L + S + L × S + E, where μ is the overall mean, L (random) and S (fixed) are the cross-classified effects of line and sex, respectively, and E is the error variance. ANOVAs, F-ratio tests of significance using type III mean squares, and estimates of variance components were computed using SAS software (SAS Institute, 1988).
QTL contributing to variation among the RIL for mean avoidance response were mapped using composite interval mapping, as implemented by the Ri2 design of QTL Cartographer software, Version 1.14 (Bastenet al. 2000). Six marker cofactors were chosen by forward selection-backward elimination stepwise regression. Likelihood ratio (LR) test statistics were evaluated every centimorgan. Marker cofactors within 10 cM of a test location were excluded from the model. The significance threshold was determined by randomly permuting the data 1000 times and calculating the maximum LR statistic across all intervals for each permutation. LR statistics that exceeded the 50th highest permuted value were deemed significant at α= 0.05 under the null hypothesis (Churchill and Doerge 1994; Doerge and Churchill 1996).
Quantitative complementation tests: All deficiency (Df) stocks and mutations (m) used were obtained from the Bloomington Drosophila Stock Center and were maintained against a marked balancer (Bal) chromosome. Each Df (or m) stock was crossed to Oregon (Ore) and 2b, and five replicate measurements of avoidance behavior per sex were made on each of the four progeny genotypes (e.g., Df/Ore, Df/2b, Bal/Ore, and Bal/2b). If the difference in avoidance score between Ore and 2b in the Df (m) background is significantly greater than the difference in avoidance score between Ore and 2b in the Bal background, there is quantitative failure of Ore and 2b QTL alleles to complement the Df (m) (Pasyukovaet al. 2000; Mackay 2001a,b). These data were analyzed by three-way fixed effects ANOVA, according to the model Y = L + G + S + L × G + L × S + G × S + L × G × S + E, where L denotes the effect of line (Oregon vs. 2b), G is the effect of genotype [Df (m) vs. Bal], S is the effect of sex, and E is the error variance. Reduced models were also computed separately for each sex using the model Y = L + G + L × G + E. Significant L × G or L × G × S interaction terms indicate quantitative failure to complement. All cases of quantitative failure to complement were confirmed by five additional replicate assays per sex per genotype. ANOVAs and F-ratio tests of significance were computed using SAS software (SAS Institute 1988).
DNA sequencing: DNA sequence data were obtained for 6.6 kb including the Vanaso gene from the Oregon and 2b alleles. PCR primers were designed to amplify partially overlapping 500-bp fragments of the Vanaso gene region. Several 50-μl reactions from each primer pair were pooled for each line to minimize the contribution of polymerase errors to sequence variation and purified using Qiaquick columns (QIAGEN, Valencia, CA). PCR products were sequenced directly from both strands with internal primers and ABI big dye terminator chemistry. Sequences were aligned using Vector (Burlingame, CA) NTI programs. Chromatograms were checked for polymorphic sites and manually edited where necessary.
In situ hybridization: In situ hybridization was performed on 10-μm formalin-fixed and paraffin-embedded sections of 50 randomly oriented heads with a digoxigenin-dUTP-labeled 500-bp probe generated by PCR amplification of genomic DNA using as primers 5′-CATAATACGCGGCCCCAGT and 5′-TTG GTGCCCAACGTACTACGT and cloned into the pGEM-T vector (Promega, Madison, WI). The slides were treated with acetic anhydride and triethanolamine-HCl to reduce background and incubated with 5 mg/ml proteinase K for 1 hr at 37°. The sections were hybridized for 16 hr at 55° in hybridization buffer (50% formamide, 5× SSC, 100 μg/ml tRNA, 50 μg/ml heparin, 0.1% Tween 20) with either a heat-denatured antisense or a sense probe. The slides were washed at 55° with hybridization buffer for 1 hr followed by 20-min sequential washes at 55° with 75% hybridization buffer/25% PBT (4.3 mm Na2HPO4, 1.4 mm KH2PO4, 137 mm NaCl, 2.7 mm KCl, and 0.1% Tween 20, pH 7.4), 50% hybridization buffer/50% PBT, and 25% hybridization buffer/75% PBT. Hybridization products were visualized with an alkaline phosphatase-conjugated anti-digoxigenin antibody (Roche Molecular Biochemicals, Indianapolis), using nitro blue tetrazolium chloride and 5-bromo-4-chloro-3-indolyl-phosphate as substrates.
Bioinformatics analysis: Analysis of domains in the predicted Vanaso protein was done using DART (Domain Architecture Retrieval Tool) software (http://www.ncbi.nlm.nih.gov), with the low complexity filter off and the P value set at <0.01. The human genome database was surveyed for homologs of Vanaso with the BLASTp program.
RESULTS AND DISCUSSION
A single QTL affects variation in olfactory behavior between two Drosophila strains: To identify QTL that contribute to variation in olfactory behavior we quantified avoidance responses to the repellent odorant, benzaldehyde (Anholtet al. 1996), in a population of 98 RILs derived from two homozygous strains of D. melanogaster, Oregon and 2b (Nuzhdinet al. 1997; Leips and Mackay 2000; Vieiraet al. 2000). There was significant genetic variation in olfactory behavior among the RILs and parental lines (F99, 99 = 2.67, P < 0.0001) and between males and females (F1,99 = 17.47, P < 0.0001), but the nonsignificant sex by line interaction term (F99,800 = 1.02, P = 0.42) suggests that underlying loci have the same effects in males and females.
—QTL for olfactory avoidance behavior. (a) Genome scan. Likelihood ratio (LR) test statistics are plotted against genomic location (in centimorgans) for the X, second, and third chromosomes (C1, C2, and C3). Marker positions are indicated by triangles. The horizontal dashed line denotes the experiment-wise 5% significance threshold, determined by permutation. (b) Deficiency complementation mapping. Cytological map positions of the deficiencies used are shown relative to Bridges’ drawing of chromosome 3L (Bridges 1935), where the length of each line corresponds to the size of the deleted segment. Blue and red lines denote significant failure of the deficiency to complement the olfactory phenotypes of the Oregon and 2b alleles in males and females, respectively; black lines indicate complementation.
Cytological insertion sites of polymorphic roo transposable element markers were determined for each line, providing a dense molecular map (Nuzhdinet al. 1997). A genome scan for QTL affecting variation in avoidance response was conducted using composite interval mapping (Zeng 1994) to evaluate marker genotype-phenotype associations (Figure 1a). A single QTL with a peak LR test statistic = 21.3 exceeded the significance threshold derived by permutation. The 95% confidence interval for the location of this QTL, detected in both males and females, was at cytological position 61A-66B on the left arm of the third chromosome. Two additional putative QTL on the left arm of the second chromosome between cytological positions 27C-34A (LR = 9.5 and 10.7) did not reach formal significance on the basis of the permutation test.
High-resolution mapping reveals two closely linked sex-specific QTL for olfactory behavior: We used deficiency complementation mapping to confirm the existence and refine the map positions of the QTL (Pasyukovaet al. 2000). Oregon and 2b males were crossed to females from each of 10 overlapping deficiency strains spanning cytological locations 27C—34A and to each of 13 deficiency stocks together uncovering 61A-66C (Table 1). Olfactory avoidance behavior was quantified for males and females for each of the four resultant progeny genotype classes. Failure of Oregon and 2b alleles to complement a deficiency was inferred if the difference in mean avoidance score between the two hemizygous genotypes was greater than the difference between the two heterozygous genotypes, as determined by a significant line (L, Oregon vs. 2b) by genotype (G, deficiency vs. balancer chromosome) interaction term from analysis of variance (Pasyukovaet al. 2000).
Failure to complement was not observed for any of the chromosome 2 deficiencies (Table 1). However, patterns of failure to complement and complementation for the chromosome 3 deficiencies indicated that the single QTL affecting olfactory behavior in both sexes detected by composite interval mapping was in fact attributable to two linked QTL, one with a male-specific and the other with a female-specific effect (Table 1; Figure 1b). The L × G interaction was significant for Df(3L)pb1-X1 (cytological breakpoints 65F5-66B10) averaged across both sexes and for males, but not for females. Similarly, the L × G interaction was significant for Df(3L)ZP1 (66A17-66C5) averaged across both sexes and for males, but not for females. None of the L × G interactions were significant for overlapping deficiencies Df(3L)RM5-2 (65E1-66B2), Df(3L)pb1-NR (66B1-66B2), and Df(3L)66C-G28 (66B8-66C9) placing the left border of the male-specific QTL at 66B2 and the right border at 66B8. The L × G interaction for Df(3L)Aprt-32 (62B1-62E3) was significant averaged over both sexes and for females, but not for males. None of the L × G interactions were significant for the overlapping Df(3L)R-G7 (62B8-62F5); therefore, the female-specific QTL is localized between 62B1-8.
Quantitative complementation test results
Vanaso is a candidate gene for the female-specific QTL: A total of 10 genes and predicted genes are in the region of the male-specific QTL for olfactory behavior (http://flybase.bio.indiana.edu), including nemo (a protein serine/threonine kinase) and Arp66B (a cytoskeletal structural protein); the remaining 8 are of unknown function. Mutant stocks were available for nemo and spindle-C (one of five Drosophila spindle genes required to generate polarity during axis formation). Both mutations complemented the olfactory phenotype of Oregon and 2b (Table 1).
A total of 39 genes and predicted genes are in the region corresponding to the female-specific QTL (http://flybase.bio.indiana.edu); mutant stocks were available for only 4: male-sterile(3)neo4, Cdc37 (a chaperone localized to the cytoplasm that is expressed in larval eye-antennal discs), lethal(3)04276, and Roughened (encoding a RAS small monomeric GTPase, expressed in the adult brain and larval eye-antennal discs). Quantitative complementation tests of mutations of these loci to Oregon and 2b alleles revealed significant failure to complement only for the homozygous lethal P-element insertional mutant, lethal(3)04276, but not for the other mutations, including Cdc37, located immediately downstream of lethal(3)04276 (Table 1).
The P-element insert in lethal(3)04276 was in the intron of a novel gene (CT41026) with a 3904-bp predicted message and a predicted protein of 1240 amino acids (Figure 2). This gene has been incorrectly annotated in the Berkeley database as an isoform of α-spectrin, encoded by a gene immediately upstream. We have named this gene Vanaso (Van), since segregation of putative Van alleles in Oregon (VanOre) and 2b (Van2b) causes additive genetic variance (VA; Falconer and Mackay 1996) in olfactory avoidance behavior (“naso” is a Spanish term for nose).
—Sex-specific failure of mutations (Mutant) and deficiencies (Df) in the 62-66 cytological interval to complement QTL for olfactory behavior. Bal denotes the balancer chromosome. Crosses to Oregon are indicated as circles and to 2b as triangles; male scores are in blue and female scores are in red. (a) Df (3L) pb1-X1. (b) Df (3L) ZP1. (c) Df (3L) Aprt-32. (d) l(3)04276 (Vanaso).
As for any genetic complementation test, failure of QTL alleles to complement a deficiency or a mutation of a candidate gene is indicative of a genetic interaction, but the test cannot discriminate whether the interaction is allelic or epistatic. In this case, however, several lines of evidence point to allelism as the most likely cause of the interaction. Interactions involving other QTL in Oregon and 2b cannot be invoked to explain the observed patterns of failure to complement, since there is no evidence for additional QTL affecting olfactory behavior outside the 61A-66B region segregating between these strains. Epistasis between the Oregon and 2b QTL alleles and genes outside the deleted regions of the deficiency chromosomes is very unlikely to give rise to internally consistent patterns of failure to complement using different deficiencies or to the observed replication of highly similar but sex-specific interaction effects in different genetic backgrounds. The test for quantitative failure to complement controls for interactions between Oregon and 2b QTL and QTL on the balancer chromosome or background genotype of the tester strains. Polymorphisms between Vanaso alleles from Oregon and 2b and expression of Vanaso in chemosensory organs provide corroborating evidence that variation in Vanaso contributes to the variation in olfactory behavior between these strains.
—Molecular variation in Vanaso. (A) Single nucleotide and insertion/deletion polymorphisms between Oregon and 2b Vanaso alleles. Position 1 of the alignment corresponds to position 260,423 (AE003472.1) of the 3L sequence of Berkeley Drosophila Genome Project genome sequence release 2.0 (http://www.fruitfly.org/). Differences in the alignment between the published sequence and Oregon and 2b alleles are due to insertion/deletion polymorphisms. The Berkeley sequence has small deletions at positions 254,363 (2 bp), 254,606 (33 bp), and 260,020 (4 bp) relative to the Oregon and 2b alleles. (B) The Vanaso protein (1240 amino acids) has a putative guanylate binding protein motif (underlined). We show the protein for 2b; nonsynonymous amino acid polymorphism sites in the Oregon protein are indicated in blue. The 2b Vanaso protein differs from the protein predicted from the Berkeley sequence by a leucine (2b)/serine (Berkeley) polymorphism at amino acid position 425. Asterisks indicate amino acid residues that are identical between the Drosophila protein and its human homolog.
Vanaso is highly polymorphic: Sequence analyses of 2b and Oregon alleles revealed a remarkable degree of polymorphism (Figure 3A). Twenty single nucleotide polymorphisms (SNPs) were in the 3.7-kb coding region of Vanaso, 10 of which were nonsynonymous amino acid polymorphisms, including substitution of a charged aspartic acid residue to a neutral alanine (nucleotide position 5331 in Figure 3A). The 5′ and 3′ flanking regions (∼2.4 kb) together contain 11 SNPs and, of particular interest, a 67-bp insertion/deletion (indel) near the 5′ promoter region. Additional SNPs and indels are evident between the Oregon and 2b sequences and the published sequence. The amino acid sequence of the 2b allele differs from the published sequence only at position 425 where the 2b allele contains a leucine instead of a serine residue (data not shown).
The Vanaso protein has a putative guanylate binding protein motif between residues 853 and 1075 (Figure 3B) and is homologous to predicted proteins, also of unknown function, on human chromosome 15 (gi16041130; Figure 3B) and of Macaca fascicularis (accession no. BAB69732). The C-terminal region of the human protein shows 23% amino acid sequence identity with residues 87-988 of Vanaso; the N-terminal region (not represented in Vanaso) contains an intermediate filament domain. None of the nonsynonymous SNPs between the 2b and Oregon alleles (Figure 3A) are at positions of interspecies sequence identity.
Vanaso is expressed in the third antennal segment: A priori, it is expected that genes involved in olfactory behavior would be expressed in tissues that mediate odor recognition or processing of chemosensory information. Indeed, in situ hybridization (Figure 4) reveals that Vanaso is expressed in the third antennal segment, the principal olfactory organ of Drosophila, but not in other chemosensory organs, such as the maxillary palps or the proboscis, or in the fly brain. This suggests that Vanaso plays a role in peripheral chemosensory processes rather than in propagation or central integration of olfactory information.
Vanaso and the genetic architecture of olfactory behavior: Our observations show that QTL mapping alone may not reveal the full complexity of variation for quantitative traits. High-resolution mapping can fractionate a single QTL into multiple QTL (Pasyukovaet al. 2000). When this occurs, the QTL effect estimated from the larger interval is the net effect resulting from additive and epistatic interactions of constituent genes and can be misleading. Here, a single QTL affecting olfactory behavior in both sexes could be attributed to one male- and one female-specific QTL with effects of similar magnitude. Sex-specific effects appear to be a general property of genes affecting variation for olfactory behavior (Anholtet al. 1996; Mackayet al. 1996), life span (Nuzhdinet al. 1997; Leips and Mackay 2000; Pasyukovaet al. 2000; Vieiraet al. 2000), and sensory bristle number (Mackay 2001a,b) in Drosophila and have been observed with spontaneous and P-element-induced mutations, QTL, and even molecular polymorphisms in candidate genes (Mackay 2001b). Such sex-specific effects are unlikely to be confined only to quantitative traits in Drosophila, but are likely to be a general feature of the genetic architecture of complex traits. Although the underlying mechanisms that give rise to these sex-specific effects remain elusive, the recognition of the prevalence of sexual dimorphism in quantitative phenotypes is important, since the genome by sex environment interaction may constitute one evolutionary mechanism leading to the maintenance of quantitative genetic variation (Mackayet al. 1996).
—Expression of Vanaso in the third antennal segment of Drosophila. (a) In situ hybridization with an antisense Vanaso riboprobe. The second and third segments of the antenna are designated II and III, respectively. In adult flies expression of Vanaso is observed only in the third antennal segment. (b) In situ hybridization with a sense Vanaso riboprobe to demonstrate specificity of the staining seen in a.
In contrast to the relative ease in mapping QTL affecting variation for quantitative traits, it has proven exceedingly difficult to ascribe this variation to underlying QTGs. Quantitative complementation is an accurate and rapid method for identifying candidate QTGs corresponding to QTL and will be particularly valuable for Drosophila and other model organisms; efforts to produce targeted disruptions of all genes as a publicly available resource are underway (Spradlinget al. 1999). It has been suggested (Nadeau and Frankel 2000) that phenotype-driven mutagenesis screens can supplant mapping segregating QTL to determine the genetic basis of complex traits. While there is no doubt that analysis of quantitative effects of mutations is required to identify loci and pathways important for the normal expression of complex traits, most applications of quantitative genetics require that we understand what subset of loci affect naturally occurring variation in the trait. Quantitative complementation tests bridge the divide between mutagenesis and QTL mapping and are particularly useful for determining subtle pleiotropic phenotypic effects of loci, like Vanaso, at which extant mutations are homozygous lethal.
Vanaso is a candidate gene for the female-specific QTL affecting olfactory behavior that segregates between Oregon and 2b. Formal proof that Vanaso indeed corresponds to the QTL will require functional complementation of the female-specific olfactory phenotype by multiple independent transformations of the VanOre and Van2b alleles in the genetic background of a Vanaso null allele, while maintaining a controlled genetic background. This represents at present a difficult challenge, especially when alleles with subtle effects on quantitative phenotypes are being analyzed, and will be greatly facilitated when development of homologous gene targeting methods in flies advances to the point at which entire alleles can be replaced at their endogenous site. Finally, demonstration that variation in Vanaso contributes to naturally occurring variation in olfactory behavior will require linkage disequilibrium mapping in a large sample of alleles from nature.
Acknowledgments
This work was funded by grants from the National Institutes of Health to R.R.H.A. (GM-59469), T.F.C.M. (GM-45146 and GM-45344), and S.M.R. (GM-20897) and from the W. M. Keck Foundation. J.J.F. is the recipient of a CONICET fellowship from Argentina. J.J.F. and K.O.R. are recipients of W. M. Keck postdoctoral fellowships. This is a publication of the W. M. Keck Center for Behavioral Biology.
Footnotes
-
Communicating editor: M. A. F. Noor
-
Sequence data from this article have been deposited with the EMBL/GenBank Data Libraries under accession nos. AF487673, AF487674, AF487675, AF487676, AF487677, and AF487678.
- Received May 28, 2002.
- Accepted August 26, 2002.
- Copyright © 2002 by the Genetics Society of America
