Genetical analysis of intrapopulational variation in olfactory response in Drosophila melanogaster.

Genetical study of olfactory response to ethyl alcohol (8 per cent) has been carried out on inbred lines extracted from a natural population using biometrical and chromosome analyses. A full diallel analysis was made in seven endogamic lines (3 high, 3 low and 1 intermediate in response values). Our results show a highly significant (P less than 0.001) additive component of variance and a less significant (P less than 0.05) directional dominance component suggesting overdominance and some endogamic depression. No significant maternal or reciprocal components of variation appear in our population. Two chromosome substitution assays among 3 inbred lines of extreme olfactory response add up information to the results of diallel analysis. The differences in response are due to chromosome III with an additive effect highly significant (P less than 0.001) in both studies. The other major chromosomes do not show any effect themselves but some interactions with chromosome III are significant. They can explain the directional dominance that we find in the diallel study. Analysed olfactory variation in response to ethanol can be considered as unspecific in relation to various chemical compounds.

Genetical study of olfactory response to ethyl alcohol (8 per cent) has been carried out on inbred lines extracted from a natural population using biometrical and chromosome analyses. A full diallel analysis was made in seven endogamic lines (3 high, 3 low and 1 intermediate in response values). Our results show a highly significant (P <0.001) additive component of variance and a less significant (P < 0.05) directional dominance component suggesting overdominance and some endogamic depression. No significant maternal or reciprocal components of variation appear in our population. Two chromosome substitution assays among 3 inbred lines of extreme olfactory response add up information to the results of diallel analysis. The differences in response are due to chromosome III with an additive effect highly significant (P <0.001) in both studies. The other major chromosomes do not show any effect themselves but some interactions with chromosome III are significant. They can explain the directional dominance that we find in the diallel study. Analysed olfactory variation in response to ethanol can be considered as unspecific in relation to various chemical compounds.

I NTR OD U CTI ON
Olfactory response forms a part of the behaviour patterns which are responsible for survival and reproduction of Drosophila. In spite of this, the genetics of olfaction has been little studied.
At present, studies on genes which control this response and relations among them are almost completely limited to behaviour genetics work using induced mutants (Kikuchi, 1973;Rodrigues and Siddiqi, 1978;Aceves-Piña and Quinn, 1979). There has been little work on the parts of genome responsible for differences in olfactory response and the relations among them in natural populations. Only selection (Becker, 1970;Hoffmann, 1983) and partial chromosome substitution (Fuyama, 1978) experiments have been carried out. Response obtained by selection indicated additive variance in natural populations for olfactory behaviour to different chemicals. Moreover, Becker (1970) suggested that alleles for insensitivity to repellents found by him could be dominants. On the other hand, Fuyama (1978) located variability in response to esters and ketones on the right * Present address: Max-Planck-Institut für biologische Kybernetik, Spemannstrasse 38, 7400 Tübingen 1, Germany (FRG).
arm of the chromosome II. Differences among isochromosomal lines for the chromosome II were due to polygenic factors with additive effect. The objective of the present study was to survey, through classic methods of quantitative genetics, the genetical effects associated with phenotypic differences on olfactory response to ethyl alcohol (8 per cent) in a natural population. We could then compare the kind of relation among genes and the importance of this character in fitness (Breese and Mather, 1960;Kearsey and Kojima, 1967), as has been made for other behavioural traits. Response to ethyl alcohol has been chosen for genetic analysis, as its variability in a natural population from Sandiche (Spain) shows up more clearly than reponses to other stimuli we have assayed in the same population.

Olfactometry
The olfactometer used was a Y-maze with two ways of choice ( fig. 1). A tube with a filter paper impregnated with either 05 ml odorant or 05 ml distilled water was placed at the end of each arm. Mazes were set horizontally to the floor to avoid geotropic effects and they were lighted homogeneously. Flies tested were always 60 3-4 days old females which had been starved for 24 hours before the test. The olfactory index was: JO =No. flies in S/(No. in S+No. in C), where S indicates stimulant and C indicates control. 10 values are between 0 and 1. Olfactory index measured in this way involves percentages over different totals. This fact makes the statistical analysis by normal parametric methods more difficult, since data error depends on the number of individuals considered (Snedecor and Cochran, 1984). The weighted and non parametric methods of statistical analysis that we used when we studied the JO values on a population have not been used here, because the analysis in this study is more complex. On the other hand, the individuals tested in each run are genetically homogeneous (endogamic lines or Fl between endogamic lines), and the different number of flies considered in various tests is not as important as when the line studied had genetically different individuals.

Flies and stimuli
In order to make a genetical study of a natural population from Sandiche (Asturias, Spain), which showed variation in olfactory response to ethyl alcohol (8 per cent) when we tested 51 isofemale lines, we have obtained inbred lines from it. These inbred lines were obtained from 11 isofemale lines whose olfactory response values represented the complete range of populational distribution. From each isofemale line we set up isofemale lines after 16 generations were considered genetically homogeneous. An olfactory study of them showed response values to ethyl alcohol (8 per cent) in the same range of population responses and JO means as 03320 and 03749 for inbred lines and population respectively.
Neither distribution of response values was significantly different when they were compared by a contingency x2 These facts indicate that the inbreeding process has not assembled gene combinations with phenotypical values more extreme than those of the population and populational distribution could be represented by inbred line distribution.
The response values observed are between O5326 and 02585, i.e. in the repellent response area. At this range of responses 10 high corresponds to indifferent response and 10 low to high repellency. This is to say, a low 10 represents more sensitivity to stimulus than a high 10 value. From the inbred lines we have chosen several lines for different experiments.  (b) Chromosomal substitution assay Three inbred lines, 45-1, 89-1 and 25-13 have been used for two chromosomal substitution experiments, one between 45-1 and 89-1 lines and the others between 45-1 and 25-13. Combinations of the three major chromosomes from both pairs of lines were synthesized following Kearsey and Kojima's (1967) crossing scheme. The inversion chromosomes used were Binscy, Cy 0 and TM3 Sb Ser. Each substitution is referred to by three numbers denoting the source (4 for 45-1 line, 8 for 89-1 and 2 for 25-13) of the first, second and third chromosome pairs. The letter X was used to refer to chromosomal heterozygotes. These two sets of chromosomal combinations were tested with ethyl alcohol (8 per cent). We made 10 replicate measures for each combination.
The significant effects of chromosomes and chromosomal interactions were studied later for additive and dominant contribution using the genetical model described by Mather and Jinks (1971) and explained for the case of three chromosomes by Albornoz eta!. (1987). Parameters which describe additive and dominant effects of the chromosomes as well as interchromosomal interactions are also the same as in Albornoz's work. D1, d2 and d3 represent the additive effects of chromosomes I, II and III; hi, h2 and h3 represent their corresponding dominance effects and the other 20 parameters describe interchromosomal interactions. Parameters i refer to d x d or d x d x d interactions, jto dx h (i.e., j1.2, interaction dl x h2), dxdxh (i.e., j12.3, interaction dlxd2xh3) or dx h x h (i.e.,j1.23 interaction dl x h2 x h3) interactions and 1 to h x h or h x h x h interactions. The parameter scores are estimated by solving equations obtained by equating observed cell means to their corresponding sum of components. Variances were homogeneous between cells within each substitution test. Errors of the genetical parameters are obtained from the variance-covariance matrix.
(c) Specific or unspecific character of olfactory response Lastly, specific or unspecific character of olfactory response differences was analysed using four synthetic lines A, B, C and D. They come from four inbred lines, the two highest (25-13 and 89-1) and the two lowest (16-7 and 45-1) response values.
Synthetic lines were obtained crossing endogamic lines in order to avoid endogamic depression effects. Lines A, B, C and D were the Fl generation of crosses between 16-7 x 45-i, 45-1 x 89-i, 45-i x 25-13 and 25-13 x 89-1, respectively. So we expected the maximum difference between lines A (Fl of the two inbred lines with lowest 10) and D (Fl of the two lines with highest JO).
Dose-response curves to ethyl alcohol were obtained in order to see whether the observed high 10 values, near 10 = 0.5, corresponded to no perception of stimulus or only to sensitivity differences. Moreover, we wanted to characterise the differences in response to this compound among synthetic lines to compare them with the other dose-response curves.
The shape of the dose-response curve was studied by a regression analysis for each line and stimulus. Variance of the data to different concentrations was divided in two components: (a) variance explained by adjustment to a regression line, with 1 degree of freedom and (b) variance of deviation to this regression line, with the other degrees of freedom (Sokal and Rohlf, 1979).

(a) Diallel analysis
Olfactory response values to ethyl alcohol (8 per cent) on the 49 lines obtained by crossing seven inbred lines in all the possible combinations are shown in table 1. 10 means of inbred lines are located on the diagonal of the table. The highest 10 values corresponded to inbred lines, 04908 and 04904 of the 25-13 and 110-6 lines respectively, but the lowest 10 were not on the diagonal. So, the lowest 10 value of inbred lines was 01420 of the line 16-7, but JO became 00386, 00549 and OO8O3 when this line was crossed with the 45-1 and 45-13 lines. Moreover, 10 values of the other crosses were often more extreme than 10 values of their parents. These facts indicate the existence of endogamic depression and overdominance.
Partition of variance according to Hayman's model (1954) (table 2) showed that almost all the   binations of the chromosomes II and III of the same original line were put together. The 112 interaction was significant (P <0.05) and negative, this is to say 10 values decrease when heterozygous combinations of the chromosomes I and II were together. The 1123 interaction was also significant (P <0.05) but positive in the direction of increasing 10 values when the three heterozygous were assembled. However this effect was not larger than the negative effect of the additive combination of double l interactions and the total effect when the three heterozygous combinations for the I, II and III chromosomes were together was negative in the direction of increasing sensitivity.
In the set 8-4 only the interaction i123 turned out significant (P <0.05) and positive among i and l interactions.
The results of these two chromosomal analyses agree with those of diallel analysis, the additive effect of the chromosome III, and directional dominance with the positive i interaction (endogamic depression) and the negative I interaction (heterosis). The other kind of interactions,], were not present in the diallel analysis because only homozygous combinations or heterozygous for the three chromosomes were obtained in the Fl generation.
(c) Specificity analysis The results of the four synthetic lines A, B, C and D to six different chemicals are presented in fig. 2. First we wanted to study the kind of differences found to ethyl alcohol (8 per cent) among lines.
The results obtained with different concentrations ( fig. 2) allowed us to reject the idea of anosmia in line D because in all the lines response depended on concentration. An analysis of variance showed highly significant differences due to concentration (P<0.001). Regression study (table 5)  Differences in response to ethyl alcohol between lines A and D can be summarised as follows: (a) line A was more sensitive than line D, because the same concentrations were more repellent for it than for line D and (b) the dose-response curve was linear in line A but it was not in line D, which showed an increase of 10 values at intermediate concentrations. Line B (45-1 x 89-1) and line C (45-1 x25-133) were intermediates.
We are going to examine these differences found mainly between line A and D for the response to the other chemicals studied. Responses to methyl alcohol and ethyl acetate agreed with those of ethyl alcohol in sensitivity as well as in the curve shape. On the regresion analysis (table  5) line A showed a good adjustment to regression line in response to methyl alcohol but it did not in line D. Line A in response to ethyl acetate showed the same behaviour, but line D gave a non significant regression line. This was due to the high value of deviation. Response to acetone was very similar to that of ethyl alcohol. Differences in response to every concentration were smaller than those of other chemicals and regression analysis did not give significant results, but the values went in the same direction as responses to ethyl alcohol, methyl alcohol and ethyl acetate.
Response to the two aldehydes tested agreed partially with the above results. The dose-response curve to acetaldehyde of line A showed a linear shape and line D a nonlinear shape, the same as the response to ethyl alcohol, but there were not the same differences in sensitivity. Response to benzaldehyde showed the same direction in response differences as that of ethyl alcohol but regression lines for A and D showed significant adjustment, however it was better for line A than for D. In response to both aldehydes we can find the characteristics of response to ethyl alcohol in the response of the lines B and C. All of this pointed to the existence of differences in olfactory response of unspecific components, but the response could be added to other specific components in a different direction as in the case of aldehydes.

DISCUSSION
Genetic control of the olfactory response variation in a natural population has been revealed to be quite simple. The biometrical analysis using diallel crossing showed additive control and some directional dominance or overdominance in the direction to decrease 10 values, as expected for a fitness  trait (Breese and Mather, 1960;Kearsey and Kojima, 1967). This means in the direction of increasing sensitivity in the range of the responses that were studied. The two chromosomal analyses agree with this first result and complement it. The major factor or factors controlling this olfactory response variation were located on chromosome III and their action was completely additive. This action was modulated by the genetic background as can be deduced from interchromosomal interactions. Some endogamic depression was found (i23 was significant in a substitution assay and i123 in the other) in the direction of increasing 10 values (to decrease sensitivity). Moreover, some significant heterosis has been found in one of the substitution sets (112 and 1123 were significant) and the global direction was to decrease JO values, this is to say to increase sensitivity.
This genetic control indicates that the selection responses are the same as those obtained by Becker (1970) and Hoffman (1983) on natural populations, and it coincides with the additive effect of genes for variability to esters and ketones found by Fuyama (1978).
In this population some directional dominance for increased sensitivity seems to be protected as it can be expected from the use of olfactory sense to find food, mates and oviposition sites (i.e., Par-Sons, 1973). On the other hand, complete chromosomal analysis shows not only the chromosome or chromosomes responsible for major differences but interchromosomal relations.
The study of specific or unspecific nature of variation among extreme response lines pointed to a main unspecific factor depending on stimulus concentration. This fact agrees with correlation in response to different chemicals found in a first study of the population using isofemale lines. This unspecific component of olfactory response could correspond to different points in the chemosensorial pathway. It has been shown that unspecific systems are responsible for stimulus response after stimulus capture by specific receptor molecules even from transduction processes of chemical stimulus to electrical stimulus (Norris, 1981). This work, besides Fuyama's isochromosomal work (1978), shows the existence of an important component of olfactory behaviour depending on autosomal chromosomes. Fuyama had already pointed out the existence of an unspecific factor of olfactory response on the chromosome III that modified the specific effect of chromosome II for esters and ketones, which he found in a natural population. This posibility has been confirmed by us. However, these two studies do not allow us to take general conclusions about the location of genes controlling specific and unspecific components of the complex olfactory behaviour.