Male sterility at extreme temperatures: a significant but neglected phenomenon for understanding Drosophila climatic adaptations


Dr. J. R. David, CNRS, Laboratoire Populations, Génétique et Evolution, 91198 Gif-sur-Yvette Cedex, France.
Tel.: +33-1-69 82 37 15; fax: +33-1-69 07 04 21;


The thermal range for viability is quite variable among Drosophila species and it has long been known that these variations are correlated with geographic distribution: temperate species are on average more cold tolerant but more heat sensitive than tropical species. At both ends of their viability range, sterile males have been observed in all species investigated so far. This symmetrical phenomenon restricts the temperature limits within which permanent cultures can be kept in the laboratory. Thermal heat sterility thresholds are very variable across species from 23 °C in heat sensitive species up to 31 °C in heat tolerant species. In Drosophila melanogaster, genetic variations are observed among geographic populations. Tropical populations are more tolerant to heat induced sterility and recover more rapidly than temperate ones. A genetic analysis revealed that about 50% of the difference observed between natural populations was due to the Y chromosome. Natural populations have not reached a selection limit, however: thermal tolerance was still increased by keeping strains at a high temperature, close to the sterility threshold. On the low temperature side, a symmetrical reverse phenomenon seems to exist: temperate populations are more tolerant to cold than tropical ones. Compared to Mammals, drosophilids exhibit two major differences: first, male sterility occurs not only at high temperature, but also at a low temperature; second, sterility thresholds are not evolutionarily constrained, but highly variable. Altogether, significant and sometimes major genetic variations have been observed between species, between geographic races of the same species, and even between strains kept in the laboratory under different thermal regimes. In each case, it is easily argued that the observed variations correspond to adaptations to climatic conditions, and that male sterility is a significant component of fitness and a target of natural selection.

Introduction and historical overview

For most animal species and especially ectotherms, temperature is presumably the main environmental factor responsible for the distribution and abundance of individuals (Andrewartha & Birch, 1954; Precht et al., 1955; Cossins & Bowler, 1987; Hoffmann & Parsons, 1991,1997; Leather et al., 1993; Hoffmann et al., 2003). However, climatic adaptations are difficult to unravel for at least three main reasons. First, in a given place, thermal conditions are extremely variable according to microhabitat, daytime or season, often resulting in temporary mild or acute stresses. Second, temperature always interacts with other climatic variables, such as rainfall and humidity. Third, thermal variations are able to act directly on any kind of cell or function in an organism.

The interaction between temperature and spermatogenesis remains a poorly investigated field. In Mammals, it has long been known that normal spermatogenesis cannot occur at body temperature, around 37 °C, and that a lower temperature, achieved by the migration of the testes in scrotum, is necessary (Precht et al., 1955). There is no adaptive interpretation for this phenomenon. Indeed, an external location of the testes makes them prone to a diversity of wounds or accidents, and it is easy to argue that internal protection should be favoured by natural selection. In agreement with this idea, we know that, in several mammalian species, testes remain in the body cavity and migrate in the scrotum, i.e. to a lower temperature, only during the breeding season (Precht et al., 1955). So, the heat sensitivity of spermatogenesis in Mammals is generally considered as a strong functional constraint of unknown physiological basis. This is not an absolute constraint in sperm production, however, as demonstrated by the fact that, in Birds, normal spermatogenesis occurs in the body cavity, at a temperature of around 40 °C.

In Drosophila melanogaster, it was known by geneticists that permanent cultures could not be kept at 30 °C. Indeed, a nonpermissive temperature of 29 °C is generally used for the search of thermosensitive mutants (Suzuki, 1975). Long-term adaptation of laboratory cultures to heat conditions has been achieved using a constant, high temperature of 28 °C (e.g. Cavicchi et al., 1985). That male sterility was the main reason preventing a permanent culture at 30 °C was demonstrated by David et al. (1971). It was also shown that sterility was a transient phenomenon, and that spermatogenesis could be rescued after a return at a lower temperature. Quite surprisingly, a similar phenomenon was described 2 years later (Cohet, 1973) on the low temperature side that is the occurrence of sterile males after a development at 13 °C. The analogy between Drosophila and Mammals was merely interpreted as an indication that, in both cases, spermatogenesis was more thermosensitive than oogenesis.

In 1974 we discovered that a D. melanogaster strain collected by L. Tsacas in a hot and arid sub-Saharan locality (Fort Lamy, presently N'Djamena, Tchad republic) was fertile at 30 °C. This observation, suggesting an adaptation to a hot climate, was not published, although the strain was provided to several laboratories, and especially in the Soviet Union (now Russia). Russian investigators selected this strain for increased heat tolerance (Kuznetsova, 1994), and this strain was recently studied for the induction of heat-shock proteins (Zatsepina et al., 2001).

A comparative analysis of phenotypic plasticity of morphometrical traits in the two cosmopolitan species, D. melanogaster and Drosophila simulans (Morin et al., 1999) revealed serendipitously that the later could not be grown permanently at 28 °C, due to male sterility. Sterility thresholds, both at low and high temperature, were thoroughly investigated in these two species by Chakir et al. (2002). It was shown that in D. simulans spermatogenesis was more tolerant to cold and more sensitive to heat than in its sibling. The adaptive significance of this difference; however, was not straightforward, since it contradicted some ecological and ecophysiological observations (see David et al., 2004, for detailed comparisons and overall discussion). Recent new investigations on Drosophila buzzatii (Vollmer et al., 2004), D. melanogaster (Rohmer et al., 2004) and Zaprionus indianus (Araripe et al., 2004) have shown a large amount of genetic variability, both between and within species. These variations are correlated with local climatic conditions and contribute to an explanation of geographic distributions. In other words, male sterility thermal thresholds now appear as a significant component of fitness, are a target of natural selection and exhibit adaptive responses within and between species.

In this paper, we present a review of extant knowledge, add some new experimental data, and suggest future fields of investigation.

Functional aspects

Viability at different temperatures

For any species, it is possible to define a viability thermal range, that is, the interval of constant temperatures within which a complete development, from egg to adult, is possible. The response curve usually has a subrectangular shape, with a plateau in the middle and sharp decreases at low and high extreme temperatures. For example, in a D. melanogaster temperate strain (Fig. 1) viability is about 80% between 14 and 28 °C; it decreases very rapidly below 14 °C, becoming nil at 10 °C; a similar decrease occurs above 28 °C, with a null viability around 32 °C. At lethal temperatures (10 or 32 °C) development starts normally, mature larvae are observed and mortality occurs mainly in the late pupal stage. Interestingly, males exhibit a higher mortality than females at extreme temperatures (Pétavy et al., 2001). In natural conditions, temperature exhibits a daily periodic cycle, the amplitude of which may exceed 30 °C (Gibbs et al., 2003). The effects of a 12 h daily lethal stress, either cold (8 °C) or heat (33 °C) may be overcome by a daily return to an intermediate temperature. This recovery process permits an overall viability of more than 50% (Pétavy et al., 2004).

Figure 1.

Comparison of reaction norms for viability and male fertility according to growth temperature in a temperate strain of D. melanogaster. Notice the subrectangular shape for both traits. The thermal range for a permanent culture is narrower than the viability range, due to male sterility. Viability data for Pétavy et al. (2001), fertility data for Chakir et al. (2002).

Thermal ranges are variable across species and populations. Adult females emerging at extreme temperatures are generally able to lay eggs, and produce progeny when mated to fertile males. However, the progeny number is strongly affected by growth temperature (David et al., 1983), but this plasticity deserves more extensive investigations.

Male fertility: absolute versus median threshold

After development at 12 °C or 30 °C, a good viability is observed in D. melanogaster, but all males are sterile. Such an absolute thermal threshold is easily defined by keeping a large population at its developmental temperature for several weeks: no progeny will be observed. In an outbred strain, almost all males are fertile between 14 and 28 °C. Between 12 and 14 °C or 28 and 30 °C, only a part of them will be fertile. There are two general ways to estimate the proportion of fertile males. Each male may be mated with a normal female, and the cross examined for progeny production. This method is time and culture-vials consuming, however, and we favour an easier one, based on the observation of motile sperm in the seminal vesicles after dissection (Chakir et al., 2002). This method has often been used in interspecific hybrid studies to monitor the restoration of male fertility when F1 fertile females are backcrossed to parental species (David et al., 1976). It is however known that, in such cases, not all the males classified as fertile are able to produce progeny (Joly et al., 1997; Campbell & Noor, 2001). In our experiments, in which sterility was induced by a heat treatment, we always found a strong though not absolute correlation between progeny production and presence of motile sperm (Chakir et al., 2002; Araripe et al., 2004; Rohmer et al., 2004). In other words, male dissection data are a convenient predictor of male fertility, and a sample of 50 males is generally sufficient for estimating a percentage. In such investigations, different growth temperatures may be used, and the variation of fertility at extreme temperatures occurs very fast (Fig. 1). Temperature intervals of 0.5 °C provide good results, but a practical difficulty is to obtain a set of very precisely regulated incubators.

Experimental data on D. melanogaster and D. simulans at low temperatures are illustrated in Fig. 2. The experimental curves have a sigmoid shape, for which we recommend a logistic adjustment (Chakir et al., 2002; Rohmer et al., 2004). From such curves, we can define the absolute sterility threshold (11 °C in D. simulans, 12 °C in D. melanogaster), but also the temperature, which produces 50% sterility. This temperature corresponds to the inflection point of the curve; it can be estimated with its standard deviation and is called the median threshold. In D.simulans and D. melanogaster, they are 12.21 ± 0.06 and 13.24 ± 0.05 °C, respectively. It is also possible to calculate the slope at the inflection point, which shows the steepness of the logistic curve. Estimating a median threshold from a logistic adjustment appears necessary when very precise genetic comparisons are needed (see below).

Figure 2.

Variation of percentage of fertile males at low temperature in D. melanogaster and D.simulans (from Chakir et al., 2002). Curves are adjusted to a logistic model and allow the calculation of a median thermal threshold, which is the position of the inflection point, and corresponds to 50% fertility.

Cytological abnormalities in sterile males

A diversity of abnormalities occurring during spermatogenesis at high temperature has been described by Rohmer et al. (2004) in D. melanogaster. These include a slight increase in spermatid death rate within cysts of 64 cells, abnormalities in the position of sperm head within a cyst and in the state of chromatin condensation, but mainly an impairment of cyst elongation. All sterile males had cysts that were significantly shorter than normal, and the frequency distributions are illustrated in Fig. 3. For a given strain, there was practically no overlap between the distribution of sterile and normal cysts. As seen in Fig. 3, cyst length is quite variable even in normal males, with coefficients of variation of 4.4 and 7.9 in strains from Prunay and Delhi, respectively.

Figure 3.

Distribution of cyst length in normal (grown at 21 °C) and heat sterile males of D.melanogaster, grown either at 30 °C (Prunay, temperate strain) or at 31 °C (Delhi, tropical strain).

We may assume that, in sterile males, the abnormal elongation prevents a complete maturation of sperm and its migration from the basis of the testis toward the seminal vesicle. This hypothesis is supported by results obtained on other temperature sensitive mutants (Frankel, 1973) or triplosterile phenotypes (Timakov & Zhang, 2000). The precise mechanism of this process remains to be disclosed however.

Cyst and sperm length is highly variable among species (Joly et al., 1989; Joly & Bressac, 1994; Joly et al., 2004). We also know, mainly from wing investigations, that cell size is a plastic trait, which varies according to temperature (Robertson, 1959). So we asked the question: does the difference shown in Fig. 3 really reflect an abnormal, threshold phenomenon or only an extreme case of phenotypic plasticity?

It would be difficult to analyse this problem in D. melanogaster because cyst length distribution is much broader than that in its sibling species (Joly et al., 2004). We therefore chose D. simulans, in which cyst length is shorter and easier to measure, so that small differences can be measured precisely. Males of a French strain were grown between 12 and 30 °C and the results are shown Fig. 4. There was a highly significant temperature effect (anova, not shown), and the results could be adjusted to a quadratic polynomial (David et al., 1997). Maximum cyst length is observed at 18.7 ± 0.5 °C and, at that temperature, average length is 1.26 ± 0.04 mm. There is, however, no obvious discontinuity at the absolute heat sterility threshold, which, in this species, occurs at 28 °C. This observation suggests that the spectacular difference observed between fertile and sterile males in D.melanogaster (Fig. 3) might be more the consequence of a normal progressive plasticity than of the sudden disruption of a specific physiological process. Further studies are needed on this problem.

Figure 4.

Phenotypic plasticity of cyst length in D. simulans as a function of growth temperature. Data adjusted to a quadratic polynomial. Temperature of maximum length: 18.7 ± 0.5 °C; maximum length: 1.26 ± 0.04 mm.

In Drosophila males sterilized by a development at a low temperature, cytological abnormalities are also likely to occur during spermatogenesis, but their nature remains to be investigated.

Fertility recovery at a permissive temperature

The capacity of sterile males to recover fertility after a transfer at an intermediate temperature (David et al., 1971) was considered as evidence that male germ cells were not destroyed. This recovery capacity from a sterility induced either by cold or heat stress has been confirmed in all subsequent investigations (Chakir et al., 2002; Araripe et al., 2004; Rohmer et al., 2004; Vollmer et al., 2004).

Two experimental methods may be used for analyzing recovery time. The most informative is to isolate each male with several females and transfer the flies to a fresh culture vial daily. The appearance of larvae gives the recovery time, and the progeny number may also be analysed. Additionally it is also possible to check the appearance of sperm in the seminal vesicles as a function of the time spent at the permissive temperature (Rohmer et al., 2004).

Such investigations have led to two general conclusions. First, the average recovery time is proportional to the magnitude of the stress. For example, in a temperate French strain of D. melanogaster, the average recovery times were 3.5, 6, 7.5 and 10 days for developmental temperatures of 29.5, 30, 30.5 and 31 °C (Rohmer et al., 2004). Low temperatures give a symmetrical pattern for recovery (Chakir et al., 2002). Daily thermal stresses with a 12 h periodicity (8–25 °C or 18–33 °C) produced more than 50% viable adult flies (Pétavy et al., 2004). All the males, however, were sterile and, when transferred at 21 °C, the recovery times were 7 and 9 days for the heat and cold stress, respectively. These results again show that viability is less sensitive to temperature stresses than spermatogenesis. Moreover, it might provide a means for comparing the magnitude of stresses due either to constant growth temperatures or to daily periodic variations. A second conclusion is that the progeny number decreases with the magnitude of the stress (Chakir et al., 2002; Araripe et al., 2004). In other words, sperm number is very variable among individuals, but fertility remains less than normal in sterile males transferred to a permissive temperature: functional recovery is far from complete.

The physiological mechanisms responsible for the restoration of spermatogenesis at a permissive temperature are not known. It was recently shown, however, that a heat shock chaperone (Hsp 70) might have a protective role against heat sterility (Sarup et al., 2004).

Interspecific comparisons

The thermal range for viability is quite variable among Drosophila species, and differences generally match climatic adaptations. For example, the cold-adapted Drosophilasubobscura can be grown between 6 and 26 °C (Moreteau et al., 1997). The cosmopolitan tropical Drosophila ananassae may develop between 16 and 32 °C (Morin et al., 1997). We are not aware of any Drosophila species, which might be grown at a constant temperature below 6 °C or above 32 °C. Some species (e.g. D. melanogaster) have a broad thermal range, surpassing 20 °C, and may be described as eurytherm. Drosophila ananassae and some other purely tropical species (Cohet et al., 1980), have a narrower range, sometimes less than 15 °C, and may be considered stenotherm. Little doubt remains that most purely tropical species are cold-sensitive (Gibert et al., 2001) and do not grow at low temperatures.

In all published investigations so far, sterile males have been observed at low and high temperature, before reaching the lethality threshold. These results are summarized in Table 1, with addition of some unpublished observations. All species investigated produce sterile males below the heat lethality threshold, and the absolute sterility thresholds range between 23 and 31 °C. Results are less well documented on the low temperature side, but the occurrence of sterile males has also been generally observed. It is interesting to notice that, in three species, the absolute heat lethality threshold is 25 °C, while a few adults are obtained at 24 °C. One of the three (Hirtodrosophila confusa) is a fungus-breeding temperate species and its extreme heat sensitivity might be considered as normal. The other two, however, are tropical African species, but they both inhabit mountains, and thus are never subjected to high temperatures. For example, Zaprionus n.sp. B is endemic in the equatorial island of Sao Tome, but at an average altitude of 1600 m (D. Lachaise, personal communication). At sea level, average year temperature is around 26 °C, while in the mountains the average may be estimated to be below 20 °C.

Table 1.  Absolute thresholds for viability and male sterility in various Drosophilid species.
SpeciesLow temperatureHigh temperatureReference
  1. All values are in °C. n.a.: nonavailable. D.: Drosophila; H.: Hirtodrosophila; Z.: Zaprionus.

D. melanogaster10123032Pétavy et al. (2001) and Chakir et al. (2002)
D. simulans10112831Pétavy et al. (2001) and Chakir et al. (2002)
D. ananassae15163032Morin et al. (1997)
D. subobscura6n.a.2527Moreteau et al. (1997)
D. lamottei<11<112325Unpublished
D. buzzatii11n.a.3132Vollmer et al. (2004), and unpublished
H. confusa<11<112425Unpublished
Z. indianus12153032Araripe et al. (2004)
Z. n. sp. B<11112325Unpublished

Intraspecific genetic variations


Viability as a function of growth temperature was compared in Afrotropical and French strains of D. melanogaster and D. simulans (Cohet et al., 1980). It was found, in both species, that the tropical strains survived better at high temperature while the reverse was true at low temperature. This observation, which suggests adaptive changes in viability thresholds, should be confirmed by investigating more numerous populations.

High temperature sterility

The capacity of the N'Djamena strain of D. melanogaster to breed at a high temperature of 30 °C did not receive much attention until recently (see Introduction). A recent comparison of numerous world populations (Rohmer et al., 2004) revealed that most tropical strains could produce progeny at 30 °C, while all temperate strains were fully sterile. Several strains were investigated in more detail, determining their median threshold, and the results are illustrated in Fig. 5 for a French and an Indian population as well as for a laboratory selected Indian strain. For the French Draveil, we get a median threshold of 29.2 °C, while the value is significantly higher in the Indian strain from Delhi (30.5 °C). These two strains were kept in culture at a middle temperature, around 19–20 °C. Another Indian strain, Panipat, from vicinity of Delhi, was bred permanently at a constant temperature of 30 °C and later at 30.5 °C. Fertility data for this strain after 250 generations of heat adaptation are shown in Fig. 5. The median threshold is significantly higher (31.4 °C) than in the unselected Indian population, revealing the efficiency of laboratory heat adaptation.

Figure 5.

Variation of male fertility at high temperature in D. melanogaster. Draveil: temperate strain (median threshold 29.22 ± 0.03 °C); Delhi, tropical strain (median threshold 30.54 ± 0.01 °C); Panipat (near Delhi): Indian strain kept in the laboratory at high constant temperature for 250 generations (median threshold 31.41 ± 0.05 °C).

A precise genetic analysis was undertaken by crossing the unselected Delhi strain with another French strain (Prunay) from the vicinity of Paris (Rohmer et al., 2004). It was found that males from reciprocal F1 were significantly different and more alike to their father, suggesting a Y chromosome inheritance. This hypothesis was confirmed by successive backcrosses, introducing the Delhi Y into the Prunay background, and vice versa (see Rohmer et al., 2004, for more details). After 6 successive backcrosses, it was found that about half the difference between the parental strains was due to a specific effect of the Y chromosome. Since, within a population, the Y chromosome is generally considered as almost monomorphic (Clark, 1990; Zurovcova & Eanes, 1999), it is likely that the difference shown in Fig. 5 between Delhi and Panipat strains corresponds to a change in their genetic background, that is, the X chromosome and autosomes.

A genetic difference between two populations may also be shown by monitoring the recovery time after a return to a permissive temperature. In accordance with the median threshold difference, we found (Rohmer et al., 2004) that the Indian Delhi strain recovered more rapidly than the French strain at growth temperatures of 30, 30.5 and 31 °C (Fig. 6).

Figure 6.

Increase of recovery time for progeny production in males grown at various constant temperatures and transferred after emergence at 21 °C (from Rohmer et al., 2004). Recovery is faster in the heat-adapted Delhi strain. Arrows point to values which are significantly higher in the temperate Bordeaux strain.

A similar analysis was performed on D. buzzatii (Vollmer et al., 2004). Different geographic populations were studied, and according to the strain investigated, the absolute sterility threshold ranged between 30.5 and 31 °C. After a transfer to 25 °C, males grown at 31 °C recovered fertility within an average period of 8.26 days, while the normal age at maturity is 2.6 days. Significant differences among strains were also observed and, in this respect, populations originating from low temperature environments recovered faster than populations from warmer climatic regimes.

Low temperature sterility

Genetic variations for the low temperature sterility thresholds are, up to now, much less well documented and only unpublished observations were available. We have, however, long been aware that temperate populations could be grown permanently at 13.5 °C, while this was not possible for tropical populations. These could be kept permanently at 14.5 °C, however.

The median sterility threshold at low temperature has been measured precisely in a French temperate (Draveil) and an Indian tropical (Delhi) strain (Fig. 7). The results confirmed the expectation, with a median threshold at 13.3 °C in Draveil and 14.5 °C in Delhi. Interestingly, the value found for Draveil is close to that already found for a Moroccan strain (Fig. 2). North Africa belongs to the Palaearctic region and, in this respect, D. melanogaster results suggest that Morocco harbours temperate populations, in spite of the fact that the average year temperature is much higher in Marrakech (19.9 °C) than in the vicinity of Paris (13 °C). The well-known genetic differences (David & Capy, 1988) which distinguish Afrotropical (ancestral) populations and temperate (old derived) populations correspond to the occurrence of a geographic barrier, the Sahara desert.

Figure 7.

Variation of percentage of fertile males at low temperature in a temperate strain (Draveil) and a tropical strain (Delhi). Median thresholds values are: 13.30 ± 0.08 °C for Draveil and 14.50 ± 0.05 °C in Delhi.

In D. buzzatii (Vollmer et al., 2004) it was found that, at 25 °C, males reached maturity at an average age of 62 h. Males reared at 12 °C and transferred to 25 °C after emergence produced offspring after a delay of 119 h. This difference is likely due to a low temperature induced sterility.


The viability thermal range is variable among Drosophila species and also among geographic populations of the same species. Similar variations appear to be a general phenomenon in ectotherms (Andrewartha & Birch, 1954; Precht et al., 1955; Cossins & Bowler, 1987; Leather et al., 1993) and little doubt remains that they reflect climatic adaptations and play a significant role in explaining geographic distributions. What seems quite new, however, is that, a few degrees above or below from the absolute, low or high, lethality thresholds, we notice the occurrence of sterile males, so that the fertility range compatible with the persistence of a population is narrower than the viability range. Both kinds of thresholds are very variable across species. For example, heat sterility is observed at 31 °C in D. melanogaster and D. buzzatii, while in Drosophila lamottei, the threshold is around 23 °C (see Table 1). This broad variability prevents us from arguing that spermatogenesis exhibits a strong, developmental constraint with respect to temperature, as is generally argued for Mammals.

The parallel variation of viability and sterility thresholds, observed in Drosophila, may be interpreted in two ways: either both traits are independent targets of climatic selection, or they have the same genetic basis due to pleiotropic effects of the same set of genes. This is a general problem, for which it will be very difficult to find a solution. In this respect, geographic races of D. melanogaster provide important and original information. For this species, which proliferates under tropical and temperate climates, we know the direction of evolution (David & Capy, 1988; Lachaise et al., 2004). The species is native to tropical Africa, which harbours the ancestral populations. Ancestral characteristics are found in tropical populations and correspond to heat tolerant, but relatively cold sensitive phenotypes and genotypes. These characteristics are observed for viability (Cohet et al., 1980), chill coma tolerance (Gibert et al., 2001; Ayrinhac et al., 2004) and male sterility thresholds (Rohmer et al., 2004). The diversity of these phenotypes suggests that they are determined by different kinds of genes, but obtaining certainty would require a thorough investigation of genetic correlations, for example with the isofemale line technique (David et al., 2005). During their northern expansion toward colder, temperate regions, natural populations have been progressively selected for an increased cold tolerance, which has resulted in a better viability at low temperature, a better tolerance to chill coma and a lower male sterility threshold. It was not expected that their heat tolerance would be rapidly modified, in spite of the fact that these populations were no more, or only rarely selected by heat stress. Empirical observations have shown, however, that they lost a part of their heat tolerance, suggesting a kind of trade-off between cold and heat adaptation. The interpretation of this observation, which should be confirmed by investigating more populations, is not straightforward, and two kinds of explanation may be considered. First, heat tolerant phenotypes might be costly and would tend to disappear when the selection pressure is reduced. Second, there might be a negative pleiotropy between cold and heat tolerance, heat sensitive alleles providing a better cold tolerance, and vice versa. These alternative hypotheses, which are central in evolutionary theory, might be solved by the development of genetic analyses, for which D. melanogaster appears an ideal model.

The ultimate question is: shall we be able to identify the genes which function normally at a given temperature but fail at another, more extreme one? Such a task is obviously very difficult; complex phenotypes are generally assumed to be related to complex networks of interacting genes. Any general function (e.g. growth) will be affected by temperature, and a phenotypic change may be the consequence of a modification of numerous chemical reactions mediated by numerous genes. Polygenic phenotypes related to fitness are expected to change progressively with any small temperature variation. We know various phenotypes of this kind, for example duration of development (Pétavy et al., 2001) or various morphometrical traits (David et al., 2004). However, viability and male sterility exhibit very abrupt thresholds (Fig. 1) with subrectangular reaction norms. This is akin to the behaviour of thermosensitive mutants (Suzuki, 1975) and might suggest that only a few, relevant genes are involved. Indeed, our data on heat sterility in D. melanogaster have revealed a major role of the Y chromosome (Rohmer et al., 2004) and presumably of a single gene. Interestingly, heat sensitive mutants, affecting male sterility, were already described on the Y chromosome more than 30 years ago (Ayles et al., 1973; Frankel, 1973; Suchowersky et al., 1974). In a mutagenesis screen, Suchowersky et al. (1974) identified 32 Y-linked mutants which were heat-sensitive (sterile at 28 °C and fertile at 22 °C) and 19 Y-linked mutants which were cold-sensitive (sterile at 17 °C and fertile at 22 °C). These results suggest that the Y chromosome might be involved in the cold vs. heat-tolerance trade-off, which seems to exist in natural population. Other regulatory genes, such as heat shock protein genes, might also be involved. For example, it has been shown (Sarup et al., 2004) that an induction of Hsp 70 in D. buzzatii could reduce the duration of male sterility. It is our hope that investigating thermal thresholds for male sterility, which are obviously adaptive traits, will help to unravel the genetic bases of their variability among natural populations.

Finally, male sterility at extreme temperatures is a problem that is not well documented in insects. Nevertheless, some observations suggest that what has been found in drosophilids might also exist in other orders (Maisonhaute et al., 1999; Rinehart et al., 2000). In case of a general, positive answer, the analysis of male sterility thresholds might become a new paradigm for investigating thermal adaptation in ectotherms.