Anim. Behav., 1971, 19, 448-453

THE EFFECTS OF ARTIFICIAL SELECTION FOR SLOW MATING

DROSOPHILA SIMULANS

H. GENETIC ANALYSIS OF THE SLOW MATING LINE

IN

BY AUBREY MANNING & JERRY HIRSCH Department of Zoology, University of Edinburgh; and Departments of Psychology and Zoology,

University of Illinois, U.S.A.

Abstract. The inheritance of the slow-mating characteristics of one selected line (SA) of Drosophila simulans has been studied by crossing it with a normal, fast-mating line, Slow-mating and fast-mating are distinct phenotypes with very few intermediates and the former behaves as a recessive character, whilst F2 and back-crosses to SA segregate for fast- and slow-mating females. Repeated back-crossing =of fast-mating females to SA leads to a decreasing proportion of fast-mating progeny; this suggests that many loci are involved in the development of SA characteristics. However the SA line does not respond to counter-selection and yields uniformly slow-mating sub-lines when inbred; this suggests that SA is homozygous for all genes that affect mating speed. These two, at first sight incompatible modes of inheritance: are discussed and a compromise model is proposed.

In the previous paper (Manning 1968)the behavioural changes produced by selection in Drosophila simulans were described and they may be summarized as follows:

(1) One line (SA) was produced in which flies mated far mote slowly than unselected controls from the same base population.

(2) This effect was shown to be due exclusively to the SA females, which did not become receptive at about 2 days of age in the normal manner. They refused to accept males even though they were courted persistently.

(3) The endocrine changes responsible for the assumption of receptivity in normal mature females also occur in SA females. It was sug- gested that the genes affecting receptivity which have accumulated in this line do not affect the endocrine system but change thresholds in some neural target sites upon which the hormones normally act.

The present paper describes our attempts to to analyse the genetic changes which have been produced in the SA line. Although our con- elusions remain tentative, they are of interest in that they illustrate some of the problems inherent in behaviour genetic analysis.

One of the foremost of such problems is the choice of suitable 'units' of behaviour for genetic studies. This choice will be influenced by the direction from which the problem is approached. One may begin with two populations which are known to differ genetically, two inbred lines for example, in which case the inheritance of any easily-measured aspect of behaviour can be

studied. Alternatively, as in the SA line of Drosophila simulans, one begins with a be- havioural change and attention is focussed on this particular aspect of the animal's repertoire. In either case the behavioural differences be- tween two populations may be expressed in a number of ways. Thus if one is considering motor performance, three common situations can be distinguished, not completely distinct:

(I) The frequency with which a pattern is performed, e.g. the number of intromissions before a male mouse ejaculates; the frequency of the courtship element 'licking' during a courtship bout of standard length by a male Drosophila; the number of times a mouse rears up during exploration of a novel environment.

(2) The latency or duration of a pattern, e.g. the time from first intromission to ejaculation in the mouse; the duration of the lordosis posture in a femal e guinea-pig; the delay before a mouse emerges from an enclosed space,

(3) The qualitative form of a pattern, e.g. the relative timing of head and tail movements comprising the 'head-up-tail-up' display of different duck species; whether a species of Attieine moth fastens its cocoon by sealing up a single leaf of its food plant or by drawing several leaves together around it.

Examples of all these categories have been used in genetic studies.

The differences between populations most commonly concern their distribution along a continuous scale of variation and there may be considerable overlap between the distributions.

448

MANNING & HIRSCH: ARTIFICIAL SELECTION FOR SLOW MATING IN DROSOPHILA 449

Alternatively, as with some examples involving the performance frequency of a particular be- haviour pattern, the differences may be so grea t as to justify a 'present' or 'absent' classification with the population falling into two distinct behavioural classes. This was the case with the performance of cell-uncapping and cell-cleaning patterns by 'hygienic' and 'unhygienic' strains of honeybee, investigated by Rothenbuhler (1964). Here it was possible to make a simple behavioural dichotomy into strains which did or did not uncap cells containing dead larvae and which did or did not remove the corpses. This behavioural simplicity was matched by an underlying genetic simplicity, for the strains were segregating at two loci, each with a major effect on the expression of one of the two pat- terns. However, we cannot necessarily expect exact correspondence between genotype and phenotype. The vagaries of gene action and of behavioural measurement will often mean that phenotypically continuous variation has a relatively simple genetic basis. Conversely, a division into two phenotypic classes may be produced by continuous genetic variation in a population. The case to be described here may be an example of this latter type.

Methods Tests were made over an extended period and

involved flies separated by many generations. Throughout, selection has been maintained for slow mating flies, at least every second gener- ation, using the mass-mating method described in the previous paper. In practice the flies mate so slowly, ten to fifteen pairs out of fifty within 30 rain is maximal, that selection consists of eliminating the few pairs that do mate from the fifty pair sample and breeding from the rest.

Flies were reared on standard agar, corn- meal, molasses medium at 25 ~ 4- I~ and on a 12 hr light : 12 hr dark cycle. All tests were made within 3 hr of the flies' 'dawn', which is the period of maximum activity.

For most of the tests single-pair matings were timed in small Perspex cells, using flies 3 to 4 days from eclosion. A reliable measure of a female's receptivity requires that she receives more or less continuous courtship throughout the test and when necessary males were replaced if they failed to court. SA males were used where possible because they are remarkably persistent courters, but for a few of the F2 and G I B + to normal samples (see

Fig. 2) males from the normal, unselected stock had i to be substituted. The proportion of receptive females in such samples fell inside the range for these genotypes and thus no major error appears to have been introduced.

Results The Classification of Phenotypic Classes

As described previously (Manning 1968) the SA line differs from normal in that the majority of females do not become receptive and close observation during single-pair matings reveals that they extrude their ovipositor in the manner of fertilized females when males court them; they remain unmated after 30 min of continuous courtship. Since the majority of normal females when receptive, accept males well within this period, the criterion of '30 rain without accep- tance' seemed a suitable one to distinguish the receptive from the unreceptive phenotype.

4O.

~ ~0.

tt.

.e

IO,

o

o -$ s - t o t o - I s 1~-2o 2o-z$ 2~-3o >3o

Courtship time in minutes

Fig. 1. The frequency distribution of courtship times for Drosophila simulans females of all genotypes used in the present study (>30 category refers to females unmated after 30 min of courtship when observations were ended). N=1758.

Figure 1 illustrates the justification for this criterion. It is a frequency histogram of court- ship time to acceptance in 1758 Single pair matings of D. simulans made during the course of the tests to be described. More than 95 per cent of females accept males within 15 rain or reject them for more than 30 min producing a marked bimodal distribution.

Genetic Analysis Cross-breeding studies. Figure 2 summarizes

the results of a whole series of tests made over many generations of the SA line. Each point represents the percentage of receptive females in a sample of a particular cross all tested at the same time; samples varied in size from twenty to fifty. Crosses were made between SA and a

4 5 0 A N I M A L B E H A V I O U R , 19, 3

IOC

7c

~ 4c

3o

G~B+ G~B+ G~4 GIB+ F z fonormolNormol F= loSA roSA roSA

SA

Fig. 2. The percentages of receptive females in samples (N=20 or more) of different genotypes. The two parental classes are represented at the top and bottom of the same column; there is no overlap between them. In the F2 column, black circles represent samples derived from the Ft between SA males and normal females, open circles represent those derived from normal males and SA females.

normally receptive D. simulans stock. Most commonly an F1 was produced by mating SA males to normal females, but the reciprocal cross was made several times with no consistently different results; the data from both crosses are given in Fig. 2. There is often significant hetero- geneity between different samples of a given cross and for this reason no means are calculated. Certainly the genotype of a female is not the only factor that determines her receptivity as measured in this way. However, the overall effect of the genotype is clear enough.

There is no overlap in receptivity between samples of SA females and those from the normal stock, and the F 1 though more variable, closely resembles normal, with the majority of samples showing more than 90 per cent receptive females. The SA phenotype thus appears to be 'recessive'. F2 samples never reach the high level of receptivity shown by the F1 and in some of the earlier tests few F 2 samples differed significantly from a 3 �9 1 ratio of receptive to non-receptive females. This, coupled with the fact that the majority of samples of the first generation back cross to SA did not deviate significantly from a 1 : 1 ratio, at first suggested that a single recessive gene was responsible for the failure of SA females to become receptive.

However ratios of this simple Mendelian type can also result from polygenic systems, and the best check is to examine further the segregating classes among the progeny of the first back cross to the recessive parent. Here we have two phenotypes in roughly equal pro- portions which on the single recessive gene hypothesis represent 50 per cent individuals heterozygous for the gene and therefore recep- tive, and 50 per cent homozygotes which are unreceptive. I f this is the case then crossing the receptive females to SA males should again yield progeny which segregate in 1 " 1 ratio. This latter prediction was not fulfilled. This back crossing technique relies on the correct identification of genotypes and since a small proportion of SA females do mate within 30 min, one might find homozygotes that were nevertheless receptive turning up in the F 1 back cross. We cannot dismiss this possibility. How- ever it did appear that at least some of the SA females that mated were, in fact, raped for they struggled hard to dislodge the male for some time after he mounted. In the tests being described, receptive back cross females were selected with care and flies were used for breed- ing only if they accepted SA males rapidly and with no signs of struggling once the male had mounted. In spite of these precautions the pro- portions predicted by a single gene hypothesis were not found. Figure 2 shows samples from the progeny of receptive F1 back cross females and SA males.

One model to account for this result is of the type originally discussed by Sewall Wright (1934) in relation to the inheritance of the capacity to form a fourth digit on the hind limbs of guinea-pigs. Certain inbred strains have a high proportion of individuals with fourth digits, other strains resemble the wild guinea-pig and fourth digits are extremely rare. As a result of extensive hybridization ~tests Sewall Wright proposed a model of continuous variation on whose hypothetical scale there is a normal distribution of genetic and non-genetic factor combinations contributing in additive fashion to the capacity to form a fourth digit. The ge~aes operate along a 'developmental scale' with a threshold whose propert ies are such that a genetic value to one side the thresh- old value leads to development of three digits, but one to the other side leads to four. Thus two phenotypes result and hybrid populations whose distributions of genetic values lie athwart the threshold will show various proportions of

MANNING & HIRSCH: ARTIFICIAL SELECTION FOR SLOW MATING IN DROSOPHILA 451

three- and four-toed phenotypes. The guinea- pigs showed some variation in the formation of the fourth digit; in a few individuals it was small and weakly developed. Accordingly Sewall Wright proposed two thresholds, close together on the developmental scale and suggested that these few individuals whose genetic value fell between them showed weak development of the fourth digit. Clearly this is not a critical feature of the model because the precise nature of the developmental processes will determine how many phenotypes can be distinguished.

FI

Scale of 'genetic values'

J ~ GtB*to SA sA

. l i t

Developmental threshold

Receptive " * ' ~ ' - ~ Unreceptive

Fig. 3. Sewall Wright's developmental threshold model as applied to the present data. (Further explanation in tex0.

Figure 3 represents a similar model as applied to our data, assuming normal distribution of values (for genetic and non-genetic factor combinations) around a single developmental threshold determining whether females become receptive or not. As we discussed earlier, there is a clear dichotomy between receptive and un- receptive females and there seems no reason to invoke intermediate phenotypes and a second developmental threshold. The positioning of the distributions for each population in Fig. 3 are in accordance with the data of Fig. 2 but it is impossible to make a precise fit.

Further evidence supporting this hypothesis comes from our failure to find significant associ- ation between any one chromosome and the segregation of receptivity in the G 1 B + gener- ation. Such association would be predicted on a single gene hypothesis or on a modification of this which invokes a major locus with modifiers. Our evidence is that the genes responsible for

the abnormal behaviour of SA females are distributed throughout the three major chromo- somes.

Any major effect of the X chromosomes is discounted, by the similarity of the reciprocal F2 generations. Those samples derived from SA males and normal females show the same range of receptivity as do samples derived from SA females and normal males. The same phen- omenon was observed in the reciprocal F1 back crosses to SA. To test for any major effect of the second and third chromosomes, marker genes had to be employed. These have their disadvantages because some mutants tend to depress the receptivity of females and may thus confound the effect one is trying to measure. The range of mutant stocks in Drosophila simul- ans is only a fraction of those available for Drosophila melanogaster but a number were kindly supplied by Dr E. Novitski. From these we chose black (b) a recessive body-colour mutant to mark chromosome II and pink peach (pP) another recessive, to mark the third chromo- Some. The original mutant stocks were crossed into our normal Drosophila simulans stock for three generations and the mutants then re- extracted. After this mixing, mutant females showed normal patterns of receptivity. SA males were mated to homozygous mutant females and the F t males back crossed to SA females. Here we rely on the convenient fact that there is no crossing over in male Drosophila and thus a male F 1 hybrid between SA and the marker stock will produce only two types o f gamete, bearing SA chromosomes and marked chromosomes in equal proportions. When these combine with SA chromosome in a back cross generation there will be two types of female; (1) those with two SA chromosomes of the type being tested, and (2) those carrying one marked (but otherwise normal chromosome) and one SA chromosome.

I f the marked chromosome serves to mask the effects of a recessive gene with large behavioural effects on the equivalent SA chromosome, then the genotypes (1) and (2) above will differ markedly in receptivity, type (1) being unre- ceptive and type (2) receptive. If, on the other hand, genes affecting receptivity are distributed more evenly through the SA genome, then no clear correspondence between behavioural and chromosomal segregation will be found.

Females of genotypes (1) and (2) look alike and in order to identify them they had to be mated to homozygous mutant males after their

452 ANIMAL B E H A V I O U R , 19, 3

receptivity had been tested with SA males in the normal way. Females which were heterozygous for the marker gene produced 50 per cent mutant progeny from such a cross.

Table I. The Distribution o f Behavioural and Chromo- somal Segregant Classes in F1 Back Cross Populations to SA when Chromsomes 1I and HI were Marked

Receptive Un- Total receptive

Chromosome II (marked with black)

Heterozygous for SA 22 13 35 chromosome Hom0zygous for SA I3 19 32 chromosome

Chromosome lII (marked with pink peach)

Heterozygous for SA 26 27 53 chromosome Homozygous for SA t2 33 45 chromosome

There is an association between chromosome III and receptivitY; the excess of unreceptive females homo- zygous for the SA chromosome III has P<0.05, Z2 =4.24. However, with neither chromosome does the association approach 10O per cent as would be predicted on a single gene hypothesis.

Table I shows that whilst there was the typical behavioural segregation in the GI back cross populations, there was no clear association be- tween a female being unreceptive and carrying two SA second or third chromosomes. This evidence seems to exclude the occurrence of any gene with major effect and the data give some further direct support to the threshold hypo- thesis. Females homozygous for the SA chromo- some being tested would not all be expected to be unreceptive but since they will, on average, carry one fifth more of SA genes (discounting the fourth dot chromosome) they should provide a higher proportion of unreceptive females than those heterozygous for the SA chromosome and also carrying a normal marked chromo- some. This proves to be the case; female homo- zygous for the SA chromosome II are 50 per cent unreceptive, the heterozygotes are 37 per cent unreceptive, a difference in the right direction, but not significant on these figures. The equivalent classes involving chromosome III are 72 per cent unreceptive and 50 per cent unreceptive respectively and this difference is significant at the 5 per cent level, So far as this evidence goes, it suggests that chromosome III is the most important carrier of SA genes.

Counter selection and inbreeding. Whilst the evidence from the cross-breeding experiments described above all tends to support the thres- hold hypothesis, experiments involving counter selection and inbreeding are not consistent unless further assumptions are made.

Figure 3 makes the assumption that there is considerable genetic variability in all the cross- bred populations, including the SA line itself and for a small proportion of most SA samples receptivity seems to develop normally. Conse- quently it would be predicted that SA should respond to counter-selection, i.e. breeding each generation from the few females that do accept males ought to increase the proportion which becomes receptive. This is not the case; three separate attempts at counter selection have all failed. After fifteen or more generations of selection the proportion of receptive females remains as low as in the parent SA stock.

This suggests that, contrary to expectations, there is little genetic variability affecting re- ceptivity in SA. This conclusion is supported by two other lines of evidence.

(1) Lapsing selection for up to thirty gener- ations has little or no effect on the slow mating characteristics of SA.

(2) Four separate inbred lines taken from SA showed, after brother-sister mating for eighteen generations, uniformly slow mating, using the mass-mating technique. I f there were significant heterozygosity in SA, one would expect considerable differences between inbred lines according to which loci become fixed. Admittedly poor receptivity and slow mating are common manifestations of inbreeding de- pression in Drosophila. However parallel inbred lines taken from the unselected Drosophila simulans stock do show diversity. Only two lines survived eighteen generations of inbreeding and one of these showed unaffected fast mating, the other was slowed but still faster than SA. Slow mating is thus not an invariable resuR in inbreeding in Drosophila simulans.

Discussion

The evidence from the genetic analysis is not consistent. On the one hand we conclude from the cross-breeding tests and the failure to find a marked association between any one of the three major chromosomes and female behaviour, that genes affecting receptivity in the SA line are distributed throughout the genome. How- ever the evidence from counter-selection and

MANNING & HIRSCH: AKTIFICIAL SELECTION FOR SLOW MATING IN DROSOPHILA 453

from inbreeding suggests that there is con- siderable homozygosity at loci affecting recep- tivity in SA flies.

Whilst these two conclusions are not theor- etically incompatible (since the genes accumul- ated in the SA line may not act in a simple additive fashion), their association would not be predicted by the simplest form of the thresh- old model, outlined earlier, which postulates a scale of eontinuous variation in each popu- lation associated with many genes each of small effect. Perhaps the most likely explanation is a type of compromise. I f a relatively small number of loci is involved, say five to seven, this might be small enough for them all to have become fixed in the SA line, but too many for their effects to be detectable individually without unusually large samples, particularly since environmental and behavioural factors also affect the develop- ment and expression of receptivity in females. A more refined genetic analysis involving, for example, the successive introduction of un- selected chromosomes into the SA genome, would help to clear up this situation. However the marvellous variety of speeial stocks which would make this a straightforward, if laborious task in Drosophila melanogaster (Hirsch 1967) are not available for Drosophila simulans.

It seems probable that threshold systems of the type discussed here are not an uncommon way in which genetic variability becomes translated into behavioural tendencies. Fuller, Easler & Smith (1950) suggested that a system of this type underlies the tendency to develop audiogenic seizures in mice. They, and others, have dis- cussed the particular interest that attaches to populations whose distributions of genetic values appear to lie completely across a developmental threshold. Such populations should be extremely susceptible to environmental influences tending to push development one way or the other and ought to offer excellent material for the study of

gene-environment interaction in development. In our case the first generation back cross

to SA may be such a population. We have been studying the effects of various environmental factors, crowding, temperature, etc., on the proportion of receptive flies which develop. So far we have had little success in identifying environmental conditions which have a clear effect of this type. When one considers the num- ber of faetors that may be operating to influence this particular piece of behavioural develop- ment it is probably over-optimistic to expect that we shall identify those that are critical, until we know much more about the mechanisms at work.

Acknowledgments We wish to express our thanks to Miss M. C. Hill for her excellent technical assistance throughout this work. We are most grateful to Dr R. C. Roberts for criticism of the manuscript and to Dr E. Novitski for the supply of mutant stocks. We were generously supported by the Science Research Council.

REFERENCES Fuller, J. L., Easier, C. & Smith, M. E. (19505. Inherit-

ance of audiogenic seizure suseeptibihty in the mouse. Genetics, 35, 622-632.

Hirsch, J. (1967). Behavior-genetic analysis at the chromosome level of organization. In: Behavior- genetic Analysis (Ed. by J. Hirsch), pp. 258-269. New York: McGraw-HilL

Manning, A. (1968). The effects of artificial selection for slow mating in Drosophila simulans. I. The be- havioural changes. Anita. Behav., 16, 108-113.

Rothenbuhler, W. C. (1964). Behavior genetics of nest cleaning in honey bees. IV. Responses of F1 and backcross generations to disease-killed brood. Am. Zool., 4, II1-123.

Wright, S. (1934). The results of crosses between inbred strains of guinea pigs differing in number of digits. Genetics, 19, 537-551.

(Received 20 July 1970; revised 15 January 1971; MS. number: 993)