Environmental parameters of species distributions
Overall, 23 sites in Taiwan, 89 sites in Java, and 89 sites in Sumatra were surveyed in this study (Fig. 1; see Table S1 in Additional file 1 for the detailed information on each sampling site). Only the black morph of D. elegans was found in Taiwan, ranging from 9 to 1366 meters above sea level, masl hereafter, with the most southern site of presence at 23°17' 73.13"N (areas of Taiwan further south were not surveyed). Although the higher altitude sites along Provincial Highway 7 (up to 1400 masl) and along Provincial Highway 18 (up to 2240 masl) were also investigated, no host plants and D. elegans were found over 600 masl in northern Taiwan or over 1366 masl in mid-Taiwan. Furthermore, no brown morph individuals of D. elegans were collected in the surveyed sites in Taiwan. Only one individual with an intermediate color similar to the body color of the F1 hybrids of brown and black morph flies was collected in Lantan during the course of several visits. The average temperature at Taiwan sites where D. elegans was collected was 25.8 ± 2.54 degrees Celsius. We further used the data acquired from weather stations nearby the surveyed sites to analyze the environmental temperature where we were able to find D. elegans. The daily average, maximum, and minimum temperatures of surveyed dates were 21.2 ± 4.14, 27.2 ± 4.07, and 17.2 ± 5.08 degree Celsius, respectively. The average, maximum, and minimum temperatures of 30 days before surveyed dates were 21.1 ± 3.79, 30.5 ± 3.84, and 13.9 ± 4.79 degrees Celsius, respectively. The data from weather stations suggest D. elegans may experience a temperature shift of over 10 degrees Celsius during the day in its natural habitat. During our survey, we also found that the flies tended to stay on the flowers in the shade during the summer, suggestive of a behavioral adaptation to avoid overheating.
D. elegans brown morph was found from 65 to 1658 masl in Java and from 7 to 1537 masl in Sumatra. No black morph individuals of D. elegans were found in Java or Sumatra during our survey. D. gunungcola was found from 939 to 1720 masl in Java and from 1015 to 1494 masl in Sumatra. D. elegans and D. gunungcola were found to co-localize at17 sites, ranging from 939 to 1658 masl. The altitudes at which D. gunungcola was found were significantly higher on average than those where D. elegans was found (Mann-Whitney test; P < 0.0001 and P = 0.0015 in Java and Sumatra, respectively). The temperature of sites at which D. gunungcola was found was significantly lower on average than those where D. elegans was found (22.9 ± 2.8 and 25.5 ± 3.4 degrees Celsius, respectively; P = 0.0021 by Student’s t-test). D. gunungcola was not found in sites where the temperature exceeded 28 degrees Celsius, but D. elegans with apparently normal activity was found in temperatures as high as 35 degrees Celsius. Given that several D. elegans laboratory strains do not survive well in temperatures higher than 25 degrees Celsius, it was surprising to find D. elegans active in many sites with temperatures above 30 degree Celsius. Different D. elegans populations may have evolved to adapt to high temperatures as the host plant flowers are available.
Among the sampling sites shown in Fig. 1, D. gunungcola was not found in the mountainous areas in northern hemisphere above 1°41'24.30"S, which was the northern limit of our sampling. This is despite some sampling sites in northern Sumatra being above 1200 masl and possibly suitable for D. gunungcola, e.g. Karo Regency (2° 54' 44.40" N, 1505 masl; indicated by a brown arrow in Fig. 2) and Central Aceh Regency (4° 38' 22.71" N, 1256 masl). The sites at which D. gunungcola was found tended to be at higher altitude when they are closer to the equator as shown in Fig. 2 (Pearson correlation, r2 = 0.805, P = 0.001), indicating that the distribution of D. gunungcola may be confined by the factors associated with both altitude and latitude, such as temperature, precipitation, or host plant distribution. Based on the trend that D. gunungcola appears at higher elevation toward to the equator, suitable habitat for D. gunungcola near the equator would need to be at least 1500 masl. There were no such sites at equatorial Sumatra. The limit of the D. gunungcola range in the Southern hemisphere suggests that this species might have originated in the Southern hemisphere and has remined restricted there. On the other hand, the overall D. elegans distribution does not appear to be restricted in the same way. D. elegans was found in a wider range of altitude once such landscape is present, i.e. from 3° to 2.0° S. The correlation between altitude and the distance from the equator among the sites with D. elegans was relatively weak (Fig. S1; r2 = 0.074, P = 0.0109).
The associations between latitudes and altitudes in the distribution of these two species were not detected in Java, which covers a narrower latitudinal range but wider longitudinal range. In Java, the local climates may be mainly influenced by altitude rather than latitude. Despite the limit of available surveyed sites at higher altitudes in some areas, D. elegans in Java tended to not co-localized with D. gunungcola in the higher latitude sites away from the equator (Fig. S2.). This indicates that the distribution of the D. elegans brown morph at higher altitudes may be limited as the latitude increases. The apparent niche partitioning of these two species at higher latitudes has implications for the origins of the body color morphs and of D. gunungcola. Moreover, the distribution of the host plants is undoubtedly affected by geography and plays a defining role in restricting the ranges of these two species.
Host plants and host species range
A total of 11 species from five plant families were recorded as frequently used as mating sites of D. elegans during our survey (Fig. 3 and Table S2). Five among them are newly recorded host plants, which brings to 12, the total number of species that have been observed as host plants of D. elegans to date (2–13). D. elegans was also occasionally found on the flower of Allamanda cathartica from Apocynaceae (J. True, personal communication) and Thunbergia erecta from Acanthaceae (S.-D. Yeh personal observation) but there is no evidence for these plants being mating or breeding substrates. On the other hand, four species belonging to two plant families, Ipomoea cairica, Ipomoea indica, Brugmansia x candida, and Brugmansia suaveolens, were found as mating and breeding sites of D. gunungcola in our observations and the previous studies (6, 7). These flower species are also used by D. elegans in Java and Sumatra, suggesting that D. gunungcola uses more a limited number of flower species and may frequently compete with D. elegans where the species are sympatric. Co-existence of two species on the same flower and interspecific courtship were observed on several locations during our survey (Fig. 3H) and have also been reported previously (14). However, interspecific copulation was not observed during our survey even though the two species will interbreed in the laboratory (15, 16).
The majority of the host plant species for D. elegans belongs to Ipomoea genus, with I. cairica and I. indica as the most frequently encountered hosts during our survey. D. elegans was found on the flowers of Ipomoea carnea frequently in Java and Sumatra and occasionally in Taiwan, as also recorded in New Guinea previously (4). A high density of D. elegans exhibiting sexual behavior was also found on Ipomoea batatas, also known as sweet potato, in Java and Taiwan as well as on Ipomoea aquatica, also known as water spinach, in Java during our survey (Fig. 3E and 3F), which brings to six, the total number of Ipomoea species have been observed as D. elegans hosts (Noted that the sixth species, I. alba, was not encountered during our survey but recorded in the previous study). Intriguingly, none of the six Ipomoea species is native in Taiwan (17, 18). These species are generally thought to be non-native in Java and Sumatra, either (19, 20). For instance, I. batatas was introduced to this region from South America and has become an important crop in many areas. The population density of D. elegans fluctuates coincidentally with the cultivation schedule of sweet potatoes as observed during our two visits in the Bandung region, suggesting that local D. elegans populations may constantly experience expansion and contraction cycles.
D. elegans and D. gunungcola are also commonly found associated with two Brugmansia species, Brugmansia suaveolens and Brugmansia x candida in Java and Sumatra. Although ornamental horticulture varieties of Brugmansia versicolor have been encountered frequently during our survey in Java and Sumatra, we did not find these two Drosophila species on those flowers with one exception of a plant that was immediately adjacent to B. x candida. Native to tropical regions of South America, the above species naturalized in Southeast Asia after being introduced as ornamental plants (17, 21). In Taiwan, a high density of D. elegans on B. suaveolens flowers in Taiwan has been observed during the winter when Ipomoea flowers were not available (S.-D. Yeh personal observation in 2005). Whether the flower is a breeding site or merely refuge for D. elegans in Taiwan is not clear, since the sexual behavior of black morph D. elegans on B. suaveolens has not been observed. No B. x candida plants were encountered during our survey in Taiwan. On the other hand, a large quantity of D. gunungcola adults can be found on single B. x candida flowers and the larvae also have been found on the flowers during our survey. Vigorous courtship behavior of D. elegans and D. gunungcola on B. suaveolens was also observed during our survey. These observations suggest that the D. elegans brown morph and D. gunungcola are frequently attracted by these two species of Brugmansia flowers.
D. elegans has not been found on the flowers of native Ipomoea species, such as Ipomoea obscura, Ipomoea pes-caprae, and Ipomoea imperati, in Taiwan. The D. elegans host plants native to Taiwan are Hibiscus mutabilis, Hibiscus taiwanensis, and Alpinia zerumbet (Fig. 3C and 3D)(22, 23). During the flowering season, we found a high density of flies on Hibiscus and Alpinia flowers. However, both Drosophila species in Java and Sumatra were not found on ornamental Hibiscus flowers, such as H. mutabilis and H. indicus, which are occasionally grown in residential gardens. We only encountered vegetative stage Alpinia spp. with no flowers during our survey in Java. Thus, it is not clear whether Hibiscus spp. or Alpinia spp. are host of these two Drosophila species in Indonesia. Moreover, we recently discovered that in Taiwan, D. elegans also appears on the ornamental Rhododendron pulchrum, which is non-native to Taiwan. Whether D. elegans breeds on the native Rhododendron species in Taiwan is still in question. Further investigations and experiments are required to more systematically test breeding suitability and host preferences of native and non-native species for Taiwan D. elegans populations. Overall, we found that D. elegans and D. gunungcola frequently shared the flowers of I. cairica and I. indica, whereas Brugmansia is predominantly used by D. gunungcola.
Body size variation
Variation in the body size is often observed in natural populations of Drosophila and many taxa and is thought to be involved in local adaptation to abiotic factors, such as temperature and ultra-violet radiation (24–26). Lighter body color and smaller body size tend to be found in the Drosophila wild populations inhabiting warmer areas (27, 28). In order to investigate variation of body size in nature populations of D. elegans and D. gunungcola, we measured the four proxies of body size (see Materials and Methods) in wild caught males from 92 sites with at least three males per site. As expected, there was substantial variation in body size among populations (Fig. 4A and Fig. S3). Overall, the body size of D. elegans brown morph was significantly smaller than D. elegans black morph and D. gunungcola (Fig. 4B; one-way ANOVA, P < 0.0001 for all four proxies). When the flies within islands are compared, the body size of D. elegans varied among sites. But such variation was not detected in D. gunungcola, likely owing to the limited number of sampled sites for this species (Table S3).
The variation of body size observed in wild-caught flies may be attributed to environmental or genetic variation. We tested for the presence of genetic variation for body size by comparing isofemale lines established from our collected samples. Although a high percentage of rearing attempts of collected flies failed, we were able to establish 23 isofemale lines from the survey collections (15 and 8 of D. elegans brown morph and D. gunungcola, respectively). The body size of these isofemale lines shows substantial variation (Fig. S4, one-way ANOVA, P < 0.0001 for both wing length and wing width in D. elegans; P < 0.0001 and P = 0.0003 for wing length and wing width in D. gunungcola, respectively). Even the body size of isofemale lines collected from the same site exhibited significant differences (Table S4 for ANOVA/Student’s t-test results), suggesting that the variation in body size in wild-caught males is partly due to the standing genetic variation.
Body size was frequently found to be positively corelated with latitude in several Drosophila species (29). We tested the influence of geographical origin on male body size by performing a simple linear regression. The sampling sites in Taiwan occupied a relatively limited latitudinal and altitudinal range. Male body size showed weak but significant negative correlation with latitude (Fig. 5A). The statistical significance is mainly due to the interaction between latitude and altitude as a multiple regression analysis with these two parameters indicated (P = 0.0002 and 0.0008 in latitude by altitude interaction for wing length and wing width, respective; P > 0.05 for both individual factors; See Table S5 for more details). Such an unexpectedly inverse relationship between body size and latitude may be due to various factors related to our sampling time. Temperature fluctuates widely from month to month in Taiwan (Fig. 5C), and the flies were collected in different months during our survey. Body size may be also influenced by the host plants in which the flies grew up and the availability of the host plant species also varies throughout a year. Thus, more intensive sampling in Taiwan is needed to understand this contradictory finding.
In Java and Sumatra, temperatures at most sites are more stable than in Taiwan, and the sampling period was limited (Fig. 5C). As expected, male body size of D. elegans showed no correlation with latitude but a positive correlation with altitude (Fig. 5B). On the other hand, male body size of D. gunungcola showed latitudinal and altitudinal clines despite a limited number of sampling sites. Peculiarly, male wing length and wing width in D. gunungcola were negatively correlated with altitude. We further performed pairwise comparisons of the flies from three sites, JV22 (1135 masl), JV23 (1299 masl), and JV27 (1438 masl) at similar latitudes since 10 or more flies from each sampling site were measured. Indeed, both wing length and wing width of D. gunungcola males from JV22 at 1135 masl were significantly greater than for males from the higher altitude (in JV22-JV23 and JV22-JV27, P = 0.048 and 0.004 for wing length, P = 0.072 and 0.023 for wing width, respectively; Student’s t-test, unequal sample variances). The difference in wing size is unlikely due to the host plants because flies were all collected from flowers of Brugmansia x candida. Another possible explanation is that the interspecific competition or reinforcement might drive differential phenotype, meaning larger body size in D. gunungcola, at where two species colocalize. But our dataset did not support this hypothesis, either. Only D. gunungcola was found in JV22 but D. gunungcola co-existed with D. elegans in JV27. More collections are needed to further investigate the relationship between body size variation and interspecific interactions.
Genetic correlation of body color and body size
Since similar cline patterns of body size and body color have been documented in various Drosophila species, we examined possible genetic correlations between these two traits in recombinant inbred lines generated from the hybrids of D. elegans HK, a brown-morph strain, and D. elegans TPCC, a black-morph strain. Contradictory to our findings from the wild-caught males, the body size of HK is significantly larger than that of TPCC (Table. S6). However, we noticed that the body size difference between these two strains decreased after both strains survived an environmental chamber malfunction in 2018, which involved a period of heat stress and high morality. Thus, the variation in body size among lab strains may be mostly due to laboratory evolution and may not reflect populations of origin.
A comparison of 35 light and 46 dark RILs revealed a significant association between body color and body size despite overlapping variation of body size between RILs with different body color (Fig. 6 and Table S7; P < 0.0001 in the nested ANOVAs in all the four body size proxies). Given that the flies were subjected to random mating before the RILs were established, this association between body size and body color was not expected and may reflect pleiotropy in the genetic underpinnings of the two traits. The possible genetic correlation of body color and body size may influence local adaptation to various environments in D. elegans.