Niche shift of tephritid species after the Oriental fruit fly (Bactrocera dorsalis) invasion in La Réunion

In a context of successive fruit fly invasions (Tephritidae), this study investigated how the top invader, Bactrocera dorsalis, displaced established fruit fly populations. We focused, particularly, on how this invasion impacted the host range and climatic niche of each resident species.

They can lead to a decrease in gene pools by causing the extinction of native species and alter habitat and ecosystem functions (Simberloff et al., 2013;Vilà et al., 2010). Biological invasions disrupt ecosystem services, such as provisioning services, which also has an important economic impact (Colautti et al., 2006;Olson, 2006;Pimentel et al., 2001;Pimentel, Zuniga, & Morrison, 2005).
Invasive species can interact with established species at different trophic levels. Authors frequently describe interspecific competition, which is widespread among insects and is one of the primary biotic factors that significantly influences their distribution, abundance and diversity in ecological communities (Denno et al., 1995;Reitz & Trumble, 2002). One of the potential outcomes of an interspecific competition event is the competitive displacement of one of the species. DeBach (1966) defined the competitive displacement principle as follows: 'different species having identical ecological niches cannot coexist for long in the same habitat'. The superior competitor can cause the local extinction of the weaker competitor, although this is rare, that is competitive exclusion. In general, the less competitive species uses 'refuge niches', and coexistence continues. Competitive displacement is generally observed between closely related species.
In most cases, it is triggered by the invasion of an exotic species, which displaces an indigenous species or an established exotic species (Reitz & Trumble, 2002). Different niche-based hypotheses have attempted to explain mechanisms of successful biological invasions and coexistence. For example, the use of an empty niche by an invader species may reduce competition with native species and allow the coexistence of species. On the contrary, if non-native species are superior competitors, they may compete for resources and cause a niche shift or the competitive exclusion of the native species, which is less common (Amarasekare, 2003;Blonder, 2018;Musseau et al., 2016;Peterson, Rice, & Sexton, 2013;Ricciardi, Hoopes, Marchetti, & Lockwood, 2013). Moreover, the outcome of interspecific competition can be modulated by abiotic conditions, such as temperature and humidity (Rwomushana, Ekesi, Ogol, & Gordon, 2009;Tilman, Mattson, & Langer, 1981). Thus, differential climatic tolerance among competitors can allow species coexistence across environmental gradients (Czárán, 1991). Duyck and Quilici (2006) define an invasive series as a succession of invasions by closely related taxa in the same territory. In this case, the new invader often replaces the existing species as the dominant species Vila & Weiner, 2004). Invasive series have been observed in the fruit fly community (Diptera: Tephritidae) in La Réunion (Indian Ocean), where nine fruit fly species of economic importance coexist. This community consists of generalist species: Bactrocera dorsalis, B. zonata, Ceratitis catoirii, C. capitata, and C. quilicii, whose larvae feed on the fruit of various plant families. Furthermore, there are more specialized species, such as Dacus demmerezi, Dacus ciliatus and Zeugodacus cucurbitae, whose larvae feed predominantly on the fruit of Cucurbitaceae; and Neoceratitis cyanescens, whose larvae feed on the fruit of the Solanaceae family.
Apart from the two endemic species, C. catoirii and D. demmerezi, fruit flies have successively invaded La Réunion. As far as the generalist species are concerned, C. capitata was introduced in 1939 and C. quilicii (formerly Ceratitis rosa) in 1955. As these species became widespread on the island, the endemic species, C. catoirii, became rarer (White, De Meyer, & Stonehouse, 2000). Bactrocera zonata invaded La Réunion in 2000. This species became competitively dominant over the other established species, thus, modifying the host range of the other three generalist species (Charlery de la Masselière et al., 2017;Joomaye, Price, & Stonehouse, 2000). A previous study, based on experimental tests of exploitative competition between larvae of B. zonata, C. quilicii, C. capitata and C. catoirii showed that the competitive hierarchy of fruit flies reflected their order of invasion (Duyck & Quilici, 2006;.

Bactrocera dorsalis is the most recent tephritid invader in La
Réunion and it was first detected in May 2017. This species is native to India, Southeast Asia and southern China. It is unusually polyphagous and is regarded as one of the top invaders in the world (Clarke et al., 2005). It has spread rapidly throughout Africa. It was first detected in Kenya in 2003 (Lux, Copeland, White, Manrakhan, & Billah, 2003), and has since invaded all countries in sub-Saharan Africa, the Indian Ocean Islands in the Malagasy subregion (De Villiers et al., 2015;Zeng et al., 2019). It was detected in Comoros in , Mayotte in 2007, Madagascar in 2010, Mauritius in 2015and La Réunion in 2017Mauremootoo, Pandoo, Bachraz, Buldowoo, & Cole, 2019). Despite the disastrous economic impact due to loss of fruit production and the associated export markets, the invasion of B. dorsalis provides a unique opportunity for observing and evaluating the role of niche differentiation in community assembly in real-time. The impact of B. dorsalis on Ceratitis species has been observed in other regions, where the dominance of B. dorsalis caused a niche displacement but never to the point of extinction because established insect populations were generally maintained in 'refuge niches' (Duyck, Sterlin, et al., 2004;Ekesi, Billah, Nderitu, Lux, & Rwomushana, 2009;Hassani et al., 2016;Mwatawala et al., 2009aMwatawala et al., , 2009b . So far, no studies have described the effect of the introduction of B. dorsalis on the population dynamics and host range of a resident B. zonata population. In La Réunion, the changes in the Tephritidae community caused by the invasion of B. dorsalis are hard to predict because they depend on the structure of the invaded community and the invader's competitiveness in specific environmental conditions. A comparative analysis of interspecific interactions before and after the invasion is necessary to determine how invasive species impact the ecological network. However, few studies include a detailed description of the community structure prior to invasion (Charlery de la Masselière et al., 2017). In La Réunion, this comparison is possible because long-term field databases were compiled from 2001 to 2009 (after B. zonata, but before the B. dorsalis invasion) and recent data were collected in 2018 (one year after the B. dorsalis invasion) and 2019. Drawing on the existing databases, we determined how the top invader, B. dorsalis, affected a resident fruit fly community.
We focused on the following points: i) the distribution and host range of this polyphagous species in La Réunion; ii) How the host range has evolved and iii) how the climatic niche of each species changed after this invasion.

| Study site
La Réunion is located in the Southern Indian Ocean (55°30′E; 21°10′S), approximately 700 km off the coast of Madagascar and covers an area of 2 512 km 2 . This volcanic island is mountainous, rising to an altitude of 3,100 m, with very rugged topography and a heterogeneous climate. It has a humid tropical climate with two main seasons: a dry season, from May to October, mainly cold and dry with trade winds; and a wet season, from November to April, which is hotter and wetter with light winds. There are two main climatic zones delimited by the central mountain range. The east is exposed to trade winds and has high precipitation (more than 2-3 m per year).
In contrast, in the west, the coast is characterized by less humid, even arid, climatic conditions (less than 1 m per year) (Grünberger, 1989).

| Sampling
We collected fruit samples to monitor Tephritidae infestation in La Réunion. This allowed us to establish a specific link between the host plant and fruit fly species, which is not possible when adult flies are caught with a trap. Agents from CIRAD (a French Agricultural Research Centre for International Development) conducted campaigns from 2001 to 2009. After the invasion of B. dorsalis, a further field campaign was conducted in 2018 and 2019. No field collection was conducted between 2009 and 2018. During this period, we consider that the fruit fly community was stable because no new species were introduced (fruit fly and parasitoid species) and the studied abiotic parameters did not change significantly (Appendix S1). Field collection covered the entire island and included cultivated, ornamental and wild plant species. Fruits were randomly collected regardless of the presence or absence of potential punctures. Whenever possible, 15 fruit samples were collected per plant species, site and date. We We collected fruit from 70 potential host plants in the first period and 112 potential host plants in the last period (Table 1). Forty-eight host plant species were the same for both periods.

| Laboratory rearing of fruit flies
At the end of each day of field sampling, fruits were taken to the laboratory and subjected to a standardized protocol (Boinahadji et al., 2019; FAO/IAEA, 2019; Leblanc, Vueti, Drew, & Allwood, 2012;N'Dépo, My, & Nl, 2019). Fruits were weighed and individually placed in plastic boxes, containing sand as a pupation substrate, and covered with fine-mesh cloth. Fruit samples were kept in a maturation room at 25°C ± 2°C and 70% ± 20% humidity until pupation.
These conditions were chosen because they are favourable to the proper development of all fruit fly species of economic importance in La Réunion (Duyck & Quilici, 2002;Duyck, Sterlin, et al., 2004).
Over a 3-week period, fruit samples were regularly inspected and the sand was sifted for Tephritidae pupae. Pupae were kept in a climatic room in plastic boxes until emergence. They were taxonomically identified to species level. We collected data on the number of emerging individuals of each fruit fly species for the different fruit (species and weight), site and date. We calculated (a) the infestation rate as the number of fruit fly individuals per kg of collected fruits and (b) the proportion of infested fruit as the number of fruits with at least one fruit fly emergence divided by the number of fruits collected. The proportion of co-infestation was defined as the number of individual fruits with two or more fruit fly species out of the total number of infested fruits.

| Statistical analysis
Statistical analyses were performed with R version 3.6.2 (R Core Team, 2019). Unless indicated otherwise, data are presented as means ± SE. Carpomya vesuviana was only observed once on Ziziphus mauritiana and was not included in the following analyses.

| Host range
Only the 48 species that were the same for both periods were kept for analyses (see Appendix S2 for the geographic distribution of samples). For each fruit fly species and each sampling period, the extent of the host niche was calculated as the number of host plant species and by estimating host richness with a Jackknife estimator. Host diversity was also measured with the Shannon index. To estimate changes in the host niche between the first and the second sampling period, we calculated two dissimilarity indexes: Index of Sorensen (Sørensen, 1948), which measures dissimilarity based on presence/ absence data (host diversity) and Bray-Curtis (Bray & Curtis, 1957) based on abundance data (host proportion).
We constructed two matrices of interaction between fruit flies and host plant species, one for the historical sampling period (2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)) and one for the recent sampling period (2018-2019). For each matrix, rows were normalized by dividing the infestation rate of one fruit fly species for a given host plant species by the global infestation rate of the fruit fly species. We used the 'ggbipart' package to create the bipartite network diagram and the 'FactoMineR' package for the principal component analysis (PCA) from the two interaction matrices.
For each resident species (except C. catoirii), we realized a generalized linear mixed model (GLMM) with a negative binomial to test

| Climatic range
Climatic data (maps with mean annual temperature and mean annual precipitation) were drawn from the AWARE Atlas (https://smart is.re/p/AWARE), which was developed by CIRAD in La Réunion. We used precipitation and temperature as environmental predictors.
These factors impact fly development (Eskafi & Fernandez, 1990;Mahmoud, 2016;Shoukry & Hafez, 1979;Teruya, 1990;Vargas, Walsh, Jang, Armstrong, & Kanehisa, 1996;Yang, Carey, & Dowell, 1994) and, therefore, influence the distribution and abundance of Tephritidae (De Villiers et al., 2015;Ni et al., 2012). Extrapolated temperature data were derived from 73 meteorological stations evenly distributed across La Réunion and collected between 1997 and 2017. Precipitation data were obtained from 143 stations and collected between 1986 and 2016 by Météo-France and CIRAD. Prior to the analysis of climatic niches, we analysed data from METEOR (https://smart is.re/METEOR) to verify the absence of climatic changes between the two studied periods.
METEOR provides information on daily temperature, precipitation  and solar radiation. We selected data for 30 sites in 10 different municipalities and for three different elevation ranges (0-300 m; 301-600 m and > 600 m), and compared mean values for the two studied periods (Appendix S1).
We only studied the climatic range for generalist fruit fly species because of the uneven distribution of the host fruit of specialist species (Cucurbitaceae and Solanaceae). To study the distribution of each fruit fly species, we reduced the data set and focused on host fruit with high infestation rates and broad distri- For each fruit fly species, a niche comparison between the two studied periods was performed using the 'ecospat' package (Cola et al., 2017). Niche functions in the 'ecospat' package provide tools to quantify and compare species niches with an ordination approach. Niche was described in relation to precipitation and temperature. The global overlap between niches was calculated using metrics of Schoener's D or Hellinger's I, ranging from 0 (no overlap) to 1 (complete overlap) (Broennimann et al., 2012). We performed tests of niche equivalency and similarity. The niche equivalency test assesses, through 1,000 random permutations of occurrences between ranges, whether the two niches are equivalent. The niche similarity test assesses, through 1,000 random shifts of the niches within available conditions in the study area, whether the species niches are more or less similar than expected by chance.

| Tephritidae community structure
Among the nine tephritid species analysed in this study, three were

| Host range
The principal component analysis (PCA, Figure 2) allowed us to de-

| Bactrocera dorsalis
Among the 112 potential host plant species sampled, 52 were infested by B. dorsalis (Table 2). This tephritid was found in fruit from many cultivated species of economic importance in La Réunion,

| Ceratitis quilicii
The species richness of the host range was similar for the two periods (Table 3)

| Ceratitis capitata
Host range species richness was similar for the two periods (Table 3)

| Other Tephritidae species
Diversity and species richness of the host range was similar for the two periods for D. ciliatus, D. demmerezi, Z. cucurbitae and N. cyanescens (Table 3, Figure 1a, 1b). Nevertheless, we observed some differences in host niche in terms of diversity and proportion of species (Bray-Curtis: 0.29-0.84 and Sorensen: 0.38-0.50), probably due to the lower host diversity.

| Bactrocera dorsalis
One year after it was first detected, B. dorsalis was found all over the island, at a range of from 0 to 1 600 m, the maximal altitude sampled (Figures 3a and 4a)    The niche equivalency test showed that the ecological niche was similar for the two studied periods (Niche overlap D = 0.61, I = 0.78, P D = 0.06, P I = 0.06, Figure 4b).

| Ceratitis quilicii
Ceratitis quilicii was present throughout the island and found in fruit The niche equivalency test showed that the ecological niche was significantly different between the two studied periods (Niche overlap D = 0.51, I = 0.67, p D = .35, p I < .001). For the second sample period, C. capitata was less present in sites with higher temperatures (lower altitude) than for the first period (Figure 4c).

| Ceratitis capitata
Ceratitis capitata was more frequent in the west of the island and found in fruit harvested between 0 and 850 m altitude (Figure 3f The niche equivalency tests showed that the ecological niche differed significantly between the two studied periods (Niche overlap D = 0.24, I = 0.034, p D = <.001, p I < .001, Figure 4). For the second sampling period, C. capitata was found in sites with lower precipitation and temperature than for the first period (Figure 4e).

| Ceratitis catoirii
Ceratitis catoirii was found in the north and south of the island in fruit harvested between 0 and 760 m altitude. There were not enough data to study the preference and niche modification of this rare endemic species (Figure 4d).

| D ISCUSS I ON
One and two years after the Bactrocera dorsalis invasion, respectively, we observed a shift in the host range and spatial distribution of the established species. In the case of specialist species, which shared few host plant species with B. dorsalis, no significant change in host range was observed. On the contrary, generalist species such as B. zonata, C. quilicii and C. capitata modified their host range (diversity and proportion) and we observed a shift in their climatic niches.

| Host range of B. dorsalis
Our as essential hosts for this invasive species in different sites (Goergen, Vayssières, Gnanvossou, & Tindo, 2011). Their nutritional value maximizes larval development and survival in generalist species (Hafsi et al., 2016).  (Duyck, Sterlin, et al., 2004;Ekesi, Mohamed, & De Meyer, 2016;Mwatawala et al., 2009aMwatawala et al., , 2009b (Duyck, Sterlin, et al., 2004). Bactrocera zonata has proven to be more sensitive than Ceratitis species to the invasion of B. dorsalis. One hypothesis could be that these closely related species suffered from greater competition because their niches were too similar. For a stable coexistence, species require different niches.

| Climatic niche shift
In our study, in addition to the host range shift, we observed a shift in the climatic niches after the B. dorsalis invasion. Ceratitis quilicii and C. capitata were less present at low altitude (higher temperature) and C. capitata was less present in the east of the island (higher humidity) following the B. dorsalis invasion. Similar niche partitioning associated with the B. dorsalis invasion was observed in Eastern Central Tanzania, where C. rosa became predominant at a higher elevation (Geurts, Mwatawala, & De Meyer, 2012) and in Hawaii, where C. capitata populations were only maintained in peach and other fruit at high elevations (Keiser et al., 1974), while B. dorsalis was dominant in lowlands.
These results seem to demonstrate that established species are found in areas where they perform better (i.e. climatic optimum). to the optimum temperature for larval development, which is between 25°C and 30°C .
Ceratitis capitata was less abundant in humid and warm areas of the island after the B. dorsalis invasions. Duyck and Quilici (2006) showed that this species is more adapted to a dry climate than other Ceratitis species. It can tolerate all temperatures between 15 and 30°C, although it develops more slowly than C. quilicii in lower temperatures. The infestation rate of C. quilicii decreased at low altitude after the invasion of B. dorsalis. This appears to be consistent with the fact that this species has a higher tolerance to low temperatures; its temperature threshold for larval development is 3.1°C (Duyck & Quilici, 2006). Thus, this observed shift could be due to niche-dependent competition, whereby each species becomes dominant in its optimum environment. Numerous models have shown that the environment has a considerable impact on the outcome of competition and tends to shift the balance in favour of one of the species (Snyder, 2008;Velázquez, Garrahan, & Eichhorn, 2014). Climatic niche displacement was probably one parameter that allowed the coexistence of B. dorsalis and the two Ceratitis species.
However, B. zonata is more sensitive to cold than the other two species, with a 12.6°C temperature threshold for larval development (Duyck & Quilici, 2006;Duyck, Sterlin, et al., 2004). In La Réunion, this species probably did not have the opportunity to escape from B. dorsalis at higher altitudes.  (Duyck et al., 2008;Duyck & Quilici, 2006).

| The competitive displacement
However, the invasion of B. dorsalis affected this balance. We have shown that the coexistence between B. dorsalis and C. quilicii was possible because the species have a different response to temperature; that is, C. quilicii demonstrates a niche shift to a higher altitude than B. dorsalis. The coexistence of C. capitata with B. dorsalis and the other resident species was possible because of its ability to develop at a lower temperature (high altitude) and to exploit fruit species that are not host to other fruit fly species.
Bactrocera zonata and B. dorsalis have similar ecological requirements for climatic and host range. Both species prefer high temperatures and have a similar range of host plants, such as mango, Indian almond or guava. In La Réunion, the coexistence of B. zonata with B. dorsalis seems to be compromised (DeBach, 1966;Hardin, 1960).
However, other studies showed that the coexistence of these two species is possible in other parts of the world, including non-native areas like Sudan (Agarwal et al., 1999;Mahmoud et al., 2020).
According to climatic models, B. zonata seems a little less sensitive to dry stress than B. dorsalis (De Villiers et al., 2015;Ni et al., 2012).
We suppose that differences in tolerance allow the coexistence of B. dorsalis and B. zonata within the limits of climatic suitability for

B. dorsalis.
Many parameters could influence the competitive outcomes.
For example, previous studies have shown that oviposition competition occurs between adult females. Liu et al. (2017) showed that B. dorsalis species has a clear advantage when competing with C. capitata for egg-laying. Aggressive behaviour has been observed in some fruit fly species, including B. dorsalis. Females are reported to defend their oviposition sites from other females (Benelli, 2014;Shelly, 1999) and may be the cause of agonistic interference competition. In addition, competition between females for egg-laying sites could be an issue if B. dorsalis has a greater capacity for locating or exploiting the resource or if B. dorsalis uses the resource at an earlier stage than other fruit fly species ). In the case of co-infestation of the same fruit, interactions between larvae could create interference or competition (Duyck et al., 2008;Rwomushana et al., 2009;Shen et al., 2014). In larval competition, the short duration of larval development of Bactrocera species appears to be an advantage (Duyck, David, & Quilici, 2007).
Other mechanisms could promote coexistence or competitive displacement, such as apparent competition. This occurs when a natural enemy increases in number or becomes more efficient at attacking a given species in the presence of a third species (Holt, 1977).
Some cases of niche shift, which were interpreted as competitive displacement, may actually involve apparent competition (David et al., 2017). In La Réunion, generalist species share the same parasitoid, Fopius arisanus. This parasitoid could have a different effect on fruit fly species that coexist in the same biotope (Rousse, Gourdon, & Quilici, 2006).

PE E R R E V I E W
The peer review history for this article is available at https://publo ns.com/publo n/10.1111/ddi.13172.