Contribution of population-level phenotypic plasticity to the invasiveness of Zaprionus indianus (Diptera: Drosophilidae)

Zaprionus indianus is a species of fl y native to the Afrotropical biogeographic region, which around twenty years ago invaded the American continent. Several studies have shown that local adaptation and phenotypic plasticity of an invasive species in its native range could favour the colonization of new environments. Zaprionus indianus is a holometabolo us generalist polyphagous species that breeds and feeds on the fruits of several different species, which constitute different environments. In this context, we performed a comparative analysis of the phenotypic plasticity of morphological and life history traits in response to seven different breeding environments (i.e. different breeding fruits). The comparison was of native (Africa) vs. invaded ran ge (South America) wild-derived populatio ns . The population-level phenotypic plasticity values related to heterogeneity in different breeding environments for most traits analysed were higher for one of the native range population. This differentiation was also recorded for the ranking across breeding environments of developmental time and wing length mean phenotypic values. In addition, mean phenotypic values pooled across fruit treatments were larger for individuals from the invaded range, which suggests local adaptation. Results defi ne a scenario in which, although not for all the populations analysed, phenotypic plasticity contributes to the invasiveness and local adaptation in native range population of Z. indianus.

characteristic of single genotypes. In this sensu lato conceptualization, any phenotypic change in a biological entity induced by the environment is legitimately considered as phenotypic plasticity and thus it includes the plastic responses of populations and species in their particular ecological contexts (see for example Pigliucci, 2001;Valladares et al., 2006;Gianoli & Vallarades, 2012;Forsman, 2015). For example, plastic responses of trait-mediated interactions among plants may allow them to adjust to the composition of their communities, promoting coexistence and community diversity (Callaway et al., 2003). It has also been shown that different levels of phenotypic plasticity at the population level in Mediterranean oaks favour their survival in fragmented habitats (Balaguer et al., 2001;Gratani et al., 2003). Therefore, the sensu lato consideration of phenotypic plasticity allows to evaluate this mechanism's relevance in ecological and phylogenetic contexts (Miner et al., 2005;Richards et al., 2006). Also, since this framework is suitable for comparisons of the magnitude and composition of phenotypic plasticity among populations or species it is possible to determine its role in adaptation or invasiveness by means of comparative studies.

INTRODUCTION
There is evidence that biological invasions involving phylogenetically distant taxa are rapidly increasing (Ricciardi & Atkinson, 2004;van Kleunen et al., 2010;Blackburn et al., 2011;Pimentel, 2011;Seebens et al., 2017). The proliferation of alien invasive species provides a unique opportunity to study ecological and evolutionary causes and consequences of a biological invasion. In this sense, numerous studies dealing with the role of phenotypic plasticity in biological invasions have shown that this mechanism could favour the colonization of new environments (Richards et al., 2006;Hulme, 2008;Zenni et al., 2014). Moreover, intraspecifi c comparative studies have shown that populations in the invaded range of an invasive species are more plastic than are native range populations, and this could facilitate the invasion process (Kaufman & Smouse, 2001;Sexton et al., 2002;Parker et al., 2003). The most generally accepted concept of phenotypic plasticity is that a single genotype can produce alternative phenotypes under different environmental conditions (Schlichting & Pigliucci, 1998). Nevertheless, a broader understanding of phenotypic plasticity does not limit it to being only a of Z. indianus focusing on nutrient plasticity. Specifi cally, we compared sensu lato phenotypic plasticity and adaptive responses of morphological and life history traits of fl ies reared in 7 different fruit diets. Our main hypothesis is that Z. indianus wild-derived populations from the invaded range will differ in the magnitude and composition of morphological and life history traits phenotypic plasticity when reared on different fruit. The prediction is that larger values of plasticity and different plastic response profi l es recorded in populations in invaded ranges indicate a signifi cant contribution of population-level phenotypic plasticity in determining the invasiveness of Z. indianus. Also, we tested the adaptive hypothesis, which states that under the adverse environmental conditions in the habitats in invaded ranges, organisms that mature early, i.e. have shorter developmental times, have a negative cost in terms of fi tness associated with a reduction in body size, which has an adverse effect on fertility (Roff, 1992;Stearns, 1992). Our related prediction is that natural selection favoured longer developmental times and large morphological traits in populations from the invaded range in South America.

Collection sites and establishment of laboratory cultures
Zaprionus indianus fl ies were collected from two populations in its invaded range in South America and two populations in its native range in Africa ( Fig. 1) 50001-1031.07). The four cultures were set up by massive breeding using the offspring of several Z. indianus single gravid females collected in the wild. Thus, these cultures represent wild-derived populations of Z. india nus and are equivalent samples of the natural genetic variation in each population. All cultures were maintained by full-sib mat ing for more than 20 generations before the experiments in the fi rst half of 2012. The cultures were kept in 300-ml bottles, 4 bottles per population and fed a standard fl y laboratory medium of cornmeal-sugar-agar and never exposed to a growth medium containing fruit (see below). Density was controlled by maintaining cultures stocks with ~50 adults per bottle as recommended for Z. indianus laboratory breeding to avoid negative effects of high population density on developing larvae (David et al., 2006b). All lines were kept at all times under controlled conditions of 25 ± 1°C, 60-70% of humidity and 12L : 12D photoperiod.

Experimental design
Zaprionus indianus were reared on one of seven different media that included different semi-natural fruit. Approximately 100 pairs of mature fl ies from each of the four cultures were each placed in separate oviposition chambers for 8 h where the females laid eggs in a 10 cm Petri dish containing 10 ml of 2.5% agar. Then, the eggs were left to hatch and 16 fi rst-instar larvae were transferred to individual vials containing 5 ml of one of the Gupta, 1970 is a species of fl y native to the Afrotropical biogeographic region (Chassagnard & Kraaijeveld, 1991;Yassin et al., 2008a, b), which about 40 years ago begun to extend its geographical distribution from its native range in Africa to other areas in the world (Commar et al., 2012). In South America it was found for the fi rst time in São Paulo city area near the Atlantic coast of Brazil in 1999 (Vilela, 1999). Since then, Z. indianus has been also detected in North and Central America (van der Linde et al., 2006;Castrezana, 2007Castrezana, , 2011Renkema et al., 2013;Joshi et al., 2014;Markow et al., 2014;Van Timmeren & Isaacs, 2014;Lasa & Tadeo, 2015;Holle et al., 2019). In South America it has been found in Ecuador (Acurio & Rafael, 2009), in many states of Brazil, both north and south from the initial point of detection (Castro & Valente, 2001;De Toni et al., 2001;Vilela et al., 2001;Santos et al., 2003;Tidon et al., 2003;Kato et al., 2004;Leao & Tidon, 2004;Chaves & Tidon, 2008;Furtado et al., 2009;Oliveira et al., 2009;Fernandes Rodrigues & Araújo, 2011;Pasini & Link, 2011;Ribeiro Barbosa et al., 2012;Poppe et al., 2014;Ferreira Mendes et al., 2017;Vasconcelos et al., 2017), and further south in Paraguay (Benítez Díaz, 2015), Uruguay (Goñi et al., 2001 and Argentina (Soto et al., 2006;Lavagnino et al., 2008). The most robust hypotheses about the introduction and subsequent spread of Z. indianus on the American continent points to human activity, more precisely fruit trade (Tidon et al., 2003;Galego & Carareto, 2007). Zaprionus indianus is classifi ed as a category E invasive species according to Blackburn et al. (2011), since it is fully invasive, with individuals dispersing, surviving and reproducing at multiple sites in many habitats.

Zaprionus indianus
An important characteristic of Z. indianus is that, both in its native and invaded ranges, it can use a wide variety of decaying fruit as breeding and feeding resources (Lachaise & Tsacas, 1983;Goñi et al., 2002;van der Linde et al., 2006;Schmitz et al., 2007;Lavagnino et al., 2008) what makes it a generalist polyphagous species (Aluja & Mangan, 2007). The different breeding resources represent different environmental patches where individuals spend their embryonic and larval stages. Due to its particular ecological characteristics, Z. indianus provides a unique opportunity to investigate the role of phenotypic plasticity in its invasion of the American continent. Studies on the phenotypic plasticity of Z. indianus have mainly focused on plastic responses of individual genotypes to changes in rearing temperature (Karan et al., 1999;Loh & Bitner Mathé, 2005;Loh et al., 2008;Bitner-Mathé & David, 2015). These studies have detected differences in phenotypic plasticity due to thermal variation and only focus on either invaded or native ranges. Testing hypotheses on what determines the invasiveness of a given species requires comparison of populations of the species in different stages of the invasion process (van Kleunen et al., 2010), for example those in the native range with those that invaded other areas. In this sense, we have performed an intraspecifi c comparison between native range (Africa) and invaded range (South America) wild-derived populations semi-natural fruit media. These media consisted of a mixture of fruit pulp and 5.10 -3 g/ml agar. The fruit pulp consisted of fruit liquefi ed in 1/5 H 2 O d . The fruit used were: and Citrus sinensis (Osbeck) ('orange'). Each of the different fruit media provided a different breeding environment. These fruit are present on the both continents from where Z. indianus fl ies were obtained for experiments (Morton, 2013). Five replicates were set up per culture per semi-natural fruit medium. Flies were reared in controlled conditions at a temperature of 25°C ± 1, humidity of 60-70% and a photoperiod (12L : 12D). Flies that emerged from each vial were collected every 12 h and sorted by sex.

Quantifi cation of phenotypic traits
Life history traits: developmental time (DT) was estimated as the time elapsed in hours from t 0 until t e , where t 0 is the time point exactly half way between the time the adults were put in the oviposition chambers and the fi rst-instar larvae were transferred to the vials and t e is the time point exactly half way between when the adult fl ies emerged and were collected and the last time the vial was checked . Viability (V) is the percentage of the total number of fi rst-instar larvae transferred to vials that completed their development to the adult stage. Differentiation based on sex cannot be measured for this trait, because it is not possible to determine the sex of the larvae when they are transferred to the vials.
Morphological traits: two fl ies of each sex were randomly taken from each replicate for measurement. Head, wings and thorax of each individual were removed and placed on a slide in their relative positions, except for the thorax, which was placed on its side. Images of all body parts were taken under a binocular microscope (10 ×) using a digital camera connected to a computer. Morphological traits of these digital images were measured using TpsDig software (Rohlf, 2001). Traits measured were: wing length (WL), wing width (WW), thorax length (TL), inter-ocular distance (ID) and head width (HW). Morphological traits in the images were quantifi ed in terms of pixels, which were then converted to millimetres (mm). All measurements are shown in mm × 100. These measurements are commonly used for quantifying the morpho-logical traits of drosophilids (Norry et al., 1994;David et al., 2006a, b;Carreira et al., 2009Carreira et al., , 2016Lavagnino et al., 2019).

Statistical analysis
Analytical and descriptive analyses of phenotypic variation and plasticity of morphological and life history traits of Z. indianus were carried out using R softwar e (R Core Team, 2016). First, the analytical analyses using generalized linear mixed models (GLMM) were done using the lme4 package (Bates et al., 2015). Models were constructed using the phenotypic values of each trait as variables and Origin [native range (Africa) vs. invaded range (South America)], Sex and Fruit (all 7 semi-natural media) as fi xed effects and Population (Origin) (the four populations analysed) as random effects nested in Origin. All variables except viability were modelled with a normal distribution. Viability was modelled with a binomial error distribution and a logit link function (Zuur et al., 2009) using the lme4 package (Bates et al., 2015). Over dispersion was corrected for by including a random variable at the level of observations (Harrison, 2014). Wald chi-squared tests were used to test signifi cance of fi xed effects using the car package (Fox & Weisberg, 2019). Likelihood ratio tests were used to test the signifi cance of random factors, for each factor the full model (including fi xed and random factors) was compared with the reduced model (without the random factor). Multiple testing was corrected using FDR correction (Benjamini & Hochberg, 1995). A signifi cant effect of Origin indicates that mean phenotypic values for individuals derived from the invaded range in South America differed from those from its native range in Africa without differentiating for the others factors. A significant Population (Origin) effect means that the mean phenotypic values of the trait analysed across breeding fruits and sexes differed between populations. If the Fruit effect is signifi cant, it means that population mean phenotypic values for this trait varies signifi cantly depending on which host fruit the fl ies were bred and our biological interpretation is that population-level phenotypic plasticity for different breeding resources exist. This is based on all cultures being equivalent samples of the natural genetic variation in each population and that the other environmental factors were controlled for. Since the plastic response is for a sample of similar genotypes within each population (each culture), we refer to it as population-level phenotypic plasticity. This conceptualization and estimate of plasticity is used in other studies (Pigliucci, 2001;Einhorn, 2005;Valladares et al., 2006;Gianoli & Vallarades, 2012;Forsman, 2015). In the cases where the interaction Population (Origin) × Fruit was signifi cant it means that population-level phenotypic plasticity varies between populations. If Origin × Fruit is also signifi cant, it indicates that phenotypic plasticity changes tend to be more similar for populations from the same origin than for those from the other continent. A signifi cant Sex term is interpreted as the existence of sexual dimorphism, and signifi cant interactions of Sex with the other effects represent variations in sexual dimorphism in relation to the origin of the fl ies, the population and fruit.
Secondly, descriptive analyses were carried out to compare the magnitude of phenotypic plasticity between populations. We used two quantitative estimators of population-level phenotypic plasticity for each trait analysed and fruit: Coeffi cient of Variation among the environments based on mea ns (CV m ) and Phenotypic Plasticity Index based on the maximum and minimum medians (PI md ). CV m differences between populations were defi ned by a descriptive criterion, which indicates there is an inter-population difference if a CV m value of population x falls outside the CV m 95% confi dence interval of population y, and the reciprocal is also true; i.e., CV m value of population y falls outside the confi dence interval of population x. CV m 95% confi dence intervals were estimated for each trait for each of the four populations studied. Intervals were estimated by means of a quantile function in the stasts package. PI md = (maximum median -minimum median) / maximum median); where maximum and minimum refers to the median phenotypic value for a population reared on a particular fruit, that is the largest or smallest for all the media used (Valladares et al., 2006). Finally, to compare if the composition of phenotypic plasticity varied among populations, rankings of mean phenotypic values of viability, developmental time and wing length in different breeding treatments were constructed and compared among populations. Wilcoxon ranked sum nonparametric tests for independent samples were performed for all pairs of populations.

Mean phenotypic values for life history and morphological traits from native and invaded range populations
Wild-derived Z. indianus fl ies from native range populations in Africa developed signifi cantly faster than fl ies derived from the invaded range in South America, with the mean developmental times of African and south American fl ies being 312 and 330.76 h, respectively (  Table S1). In contrast, fl ies from both origins had similar values for viability (Table 1, non-signifi cant Origin effect). Overall for all the different kinds of fruit used, 77% of the African larvae completed development to the adult stage and 78% of the South American larvae. Finally, mean values for all morphological traits were signifi cantly larger for individuals derived from the invaded range than for those from the native range, with the exception of thorax length that had a p-value of 0.0511 for the effect of Origin (Table 1, S1).
In fact, morphological traits of South American fl ies were between 6.5% and 6.8% larger than those of African fl ies.

Population-level phenotypic plasticity in life history and morphological traits of Z. indianus
Results show that the mean values of each trait for the populations varied signifi cantly depending on which fruit the fl ies were reared on (Table 1, signifi cant Fruit effect). However, the signifi cant Population (Origin) × Fruit interaction revealed signifi cant differences in the phenotypic plasticity between populations (Table 1). When considering fl ies derived from different origins without distinguishing between populations, differences in phenotypic plasticity in response to breeding fruit are not maintained for most traits (Table 1, non-signifi cant Origin × Fruit effect), with the exception of viability (Table 1, signifi cant Origin × Fruit effect). This means that the plastic responses for developmental time and morphological traits of native range populations did not differ in the same way from those of invaded range populations. While for viability, plastic responses between breeding fruits vary in a similar way for African populations and differently from American populations (see Table S1). Thus, the pattern of population-level phenotypic plasticity variation between populations from both ranges is quite complex and will be addressed in the following section.
Sexual dimorphism was recorded for all traits when sexes could be measured separately (Table 1, signifi cant Sex effect). In general, females developed faster (mean DT of females was 311.88 h and of males 330.87 h) and were smaller independently of their origins or the fruit they were reared on. Developmental time of females was 6.1% faster than that of males and males were 4.7% to 4.8% larger than females for all morphological traits. This dimorphism was independent of origin and the fruit fl ies were reared on. With the exception of developmental time, that signifi cant Origin × Sex interaction, showing faster development for fl ies populations form Africa (Table 1).

Comparison of population-level phenotypic plasticity of native and invaded range populations
First, we compared the magnitude of phenotypic plasticity among populations using the coeffi cient of variation (CV m ) and the Phenotypic Plasticity Index (PI md ) as estimators (Table S2). Yuto population in the invaded range had a larger population-level phenotypic plasticity, estimated using CV m , than any other population included in this study for all traits other than viability (Fig. 2). For PI md , although it could not be used for comparison, the pattern of population-level phenotypic plasticity was similar since the Yuto population had the largest PI md values for all traits. The Yokadouma population in the native range had the second largest PI md values for all traits other than viability (Fig. 2). In terms of median phenotypic values associated with rearing on different fruit, these two populations both had long developmental times and were larger when reared on 'kaki' (Table S1). Then, we compared the rankings of mean phenotypic values for viability, developmental time and wing length when reared on the different fruit. The Yuto population from the invaded range differed signifi cantly in developmental time and wing length from, Yokadouma and Lujeri, the two populations from the native range of Z. indianus (Fig. 3).

DISCUSSION
Studies dealing with phenotypic plasticity in Z. indianus have focused on plastic responses caused by temperature changes (Karan et al., 1999;Loh & Bitner Mathé, 2005;Loh et al., 2008;Bitner-Mathé & David, 2015). This is based on the reasonable premise that temperature is one of the most important environmental determinants of development and adult lifestyle of a holometabolous insect like Z. indianus. However, as this species is polyphagous and uses several different kinds of fruit for breeding and feeding, it is likely that these resources are also important ecological characteristics. In this context, our results indicate that there is a difference in the magnitude and composition of population-level phenotypic plasticity associated with feeding on different types of fruit between native and invaded ranges populations of this fl y. A lthough the phenomenon of different phenotypic plasticity between native and invaded populations was found, this may not be a general phenomenon as only the phenotypic plastic- ity of the Yuto population was greater than in one of the populations from native range. These differences were not found for the other invaded range population analysed, the Montecarlo population. It was also the Yuto population that had a different composition of phenotypic plasticity from both native range populations for developmental time and wing length. Differences in the composition of phenotypic plasticity were recorded in changes in the ranking of mean phenotypic values per type of fruit between populations. The lack of generality in the patterns identifi ed could be the consequence of differences in the genetic bases of phenotypic plasticity between populations, which probably resulted from drastic demographic events during the invasion of South America. In particular, reductions in the effective population size at the time of population foundation, or population bottlenecks in subsequent generations, may have affected the expression of phenotypic plasticity. This could be the case for the particular demographic history of the native range population Montecarlo, which differed in its plasticity pattern.
The occurrence of phenotypic plasticity in invaded ranges is generally interpreted as positive for a successful invasion because it could be benefi cial for coping with new and heterogeneous environments in invaded ranges (Kaufman & Smouse, 2001;Sexton et al., 2002;Parker et al., 2003;Fordyce, 2006;Richards et al., 2006;Chun et al., 2007;Matesanz et al., 2010;Davidson et al., 2011;Zenni et al., 2014). Also, following this scenario, greater values of phenotypic plasticity are expected in the invaded range, as was recorded in the present study for the Yuto population. However, a change in the composition of phenotypic plasticity could also indicate a contribution of this mechanism to invasiveness, as was found in the changes in the ranking of phenotypic values of traits between invaded range Yuto population and African populations. All in all, these differences in phenotypic variation related to heterogeneity in breeding substrates between native and invaded range populations could be a relevant factor enabling invasive Z. indianus to cope with new and heterogeneous environments (Sexton et al., 2002;Parker et al., 2003;Fordyce, 2006;Richards et al., 2006;Chun et al., 2007;Matesanz et al., 2010;Davidson et al., 2011;Zenni et al., 2014).
Our results revealed that all traits were sexually dimorphic when the sexes could be measured separately, with females developing faster and being smaller in terms of all the morphological traits measured. This pattern was unexpected because in most drosophilid species studied the females are larger and develop slower than males. Although previous studies on Z. indianus show the same expected direction for sexual dimorphism, they also indicate that the dimorphism is less marked than in other drosophilids (see Karan et al., 1999;Loh & Bitner Mathé, 2005;Bitner-Mathé & David, 2015). It is also reported that sexual dimorphism tends to disappear in laboratory cultures of Z. indianus (Loh & Bitner Mathé, 2005;Loh et al., 2008). Thus, it is possible that the sexual dimorphism recorded in our study was a consequence of the semi-natural fruit medium used in laboratory breeding of the fl ies. This hypothesis was not tested in the present study and should be further analysed. The comparisons we made also enabled the evaluation of whether natural selection had a role in determining the invasiveness of Z. indianus. Mean phenotypic values pooled across the fruit treatments for most traits analysed were larger for individuals derived from the invaded range on the American continent than those derived from the native range in Africa. Exceptions to this pattern were thorax length and viability, although thorax length was marginally signifi cant. Even if trait values are disaggregated between breeding fruit, larger phenotypic means were recorded for population in the invaded range for most morphological traits and types of fruit. For developmental time, larger values for invaded range populations were also found as a general trend. In the light of life history theory, these results could be interpreted as adaptive (Roff, 1992;Stearns, 1992). Under adverse environmental conditions, such as poor nutrients, different predators, competitors and/or extreme temperatures, there is a cost in terms of fi tness for organisms that mature earlier, i.e. have a shorter developmental time, which is associated with a reduction in body size and fertility (Roff, 1992;Stearns, 1992). In invaded ecosystems that are different from those in the native range, it is likely that either the physicochemical or biological conditions will be unfavourable for Z. indianus. So, in terms of life history theory natural selection is likely to favour a longer developmental time in populations in the invaded than in native ranges. The same can be proposed for morphological traits. In concordance with our results, previous surveys also report higher values for the morphological traits of the South American populations than the African populations (David et al., 2006a, b). Nevertheless, it should be noted that differences in the phenotypic values of morphological and life history traits among populations in different environments may only indicate the action of natural selection that results in local adaptation. But as Reznick & Travis (1996) point out, that although phenotypic differences between populations in different environments might indicate that adaptation is occurring in nature it must be confi rmed by another kind of evidence. At this point, it is worth mentioning that since the populations analysed were kept in the laboratory for several generations before the experiments we cannot rule out potential effects of laboratory adaptation and genetic drift. However, given that these populations were maintained at large numbers and reared on a standard laboratory medium it is unlikely that laboratory selection and/or genetic drift or bottlenecks affected the patterns recorded. In fact, a study by Maclean et al. (2018) report that laboratory maintenance does not affect comparisons of the patterns in the traits of fl ies similar to those used in this study.
Several studies that compare trait values of invasive species in their invaded and native ranges report large differences (Sakai et al., 2001;Tsutsui & Suarez, 2003;van Kleunen et al., 2010). However, few studies deal with whether the underlying mechanisms are phenotypic plasticity or adaptive evolution in the invaded range or a combination of both. In this sense, our results defi ne a scenario in which population-level phenotypic plasticity associated with heterogeneity in breeding substrates contributes to the invasiveness of Z. indianus as well as local adaptation of populations in their native range. All things considered, the results hint at the coexistence of adaptation and phenotypic plasticity being relevant for Z. indianus invasiveness. Therefore, when it comes to understanding and explaining the invasion of Z. indianus of the southernmost latitudes of the American continent, it is not possible to propose that one mechanism is of greater importance than another. In fact, it seems necessary to consider the possibility that both are acting simultaneously.
AUTHORS' CONTRIBUTIONS. NJL, MI and JJF planed, designed and did the experimental work, NJL, NF and VEO analysed the data, NJL, MI, NF, VEO and JJF wrote the manuscript.
ACKNOWLEDGEMENTS. This work was supported by grants from Agencia Nacional de Promoción Científi ca y Tecnológica ( Argentina) (PICT 2012-0640) and Universidad de Buenos Aires (Argentina) (UBACyT 20020100100482). NJL and JJF are members of Carrera del Investigador Científi co of CONICET (Argentina). MI, NF, and VEO are recipients of a scholarship from Carrera del Investigador Científi co of CONICET (Argentina). Table S1. Mean phenotypic values and standard errors (S.E.) for each trait analysed for each population reared on a particular type of fruit. DT -developmental time, V -viability, WL -wing length, WW -wing width, TL -thorax length, ID -inter-ocular distance, HW -head width. Values are based on 10 replicates for all traits, except for V for which 5 replicates were used.