Exploration for olive fruit y parasitoids across Africa regional distributions and dominance of co-evolved parasitoids

The olive fruit y, Bactrocera oleae, has been a key pest of olives in invaded regions Europe and North America. We conducted the largest modern exploration for the y’s co-evolved parasitoids across Sub-Saharan Africa (Kenya, Namibia, and South Africa) and some of the y’s expanded regions (Canary Islands, China, India, Morocco, Pakistan, Réunion Island and Tunisia). From Sub-Saharan regions, four native braconids, Psytallia lounsburyi, P. humilis, Utetes africanus and Bracon celer were collected. Principal Component Analysis showed that the regional dominance of these parasitoid species was related to climate niches, with P. lounsburyi the dominant species in the more tropical areas of Kenya, P. humilis dominant in the hot semi-arid areas of Namibia and U. africanus prevalent in Mediterranean climates of South Africa. Psytallia concolor was found in the Canary Islands, Moroccan and Tunisian, and the Afrotropical braconid Diachasmimorpha longicaudata in Réunion Island. In South Africa, seasonal monitoring of B. oleae showed consistently low infestation in unripe or ripe fruits. Multivariate analyses suggest that fruit maturity, seasonal climates and interspecic interactions shape the local parasitoid diversity that effectively regulates y populations at low levels. The results are discussed with regard to ecological adaptations of co-evolved parasitoids, and how their adaptations impact biocontrol.


Introduction
Exotic insect pests often thrive in their invaded regions due to the absence of co-evolved natural enemies and lack of effective indigenous natural enemies 1,2 . Classical biological control (CBC) by the introduction of coevolved natural enemies from the exotic pest's native range is an attempt to restore the pest-natural enemy balance after an invasion event [3][4][5] . Economic returns on successful programs are overwhelmingly positive 6 , but CBC programs require proper steps to be successfully implemented and to reduce inconsequential natural enemy releases or negative nontarget impacts 7,8 . For herbivorous invaders, this requires a fundamental understanding of the natural enemy impact in its native range, biology and host speci city, as well as potential tri-trophic interactions that develop from a high degree of co-adaptation between plant-herbivore-carnivore and the impact of habitat and environment on the selected natural enemy 7 . One aspect is matching the climatic niches occupied by the natural enemies in the native range to the invaded range 9 . Climate matching has been particularly important in fruit y biological control programs 10,11 . Hymenopteran parasitoids from the braconid subfamily Opiinae have been used worldwide in CBC programs to control fruit-infesting Tephritidae [12][13][14][15][16] . The vast majority of utilized braconid parasitoids are koinobiont endoparasitoids that oviposit in the host egg or larval stage and emerge from host pupae 17 . Therefore, the adult female parasitoid must rst locate and attack the concealed immature stages of host y inside the fruit, then bypass the host immune response, and successfully develop. For these reasons, opiine parasitoids are generally highly co-evolved with their associated host species.
The olive fruit y, Bactrocera oleae (Rossi) (Diptera: Tephritidae) has been a key pest of cultivated olives throughout the Mediterranean Basin and North America, largely due to a lack of effective natural enemies in these invaded regions 18 . The y larvae feed exclusively in olives 19 , both cultivated olives, Olea europaea ssp. europaea (Wall ex G. Don), and wild olives, of which various subspecies, occur widely in parts of Africa, southern Europe and southwestern Asia 20,21 . The y's current range extends throughout the Mediterranean Basin, northern and Sub-Saharan Africa, southwestern Asia (parts of India, Pakistan and China), and North America (California and Mexico) 18,21 . Population structure and genetic analyses suggest that B. oleae is native to Sub-Saharan Africa and then likely moved into North Africa and later the Mediterranean Basin, then proceeded westward through Europe and eventually North America [22][23][24] . The close association of the y and olives suggests the existence of highly co-evolved parasitoids associated with B. oleae. In 25,26 . Similarly, the indigenous parasitoid attacking B. oleae in California, Pteromalus nr. sp. myopitae (Pteromalidae), is also a generalist 27 . These species are idiobiont ectoparasitoids, placing their eggs on the host surface and may not need to overcome internal host defenses, thus are polyphagous, attacking even unrelated insect hosts. While present, these generalist parasitoids do not provide effective B. oleae control.
The rst major attempt to introduce coevolved parasitoids to suppress B. oleae populations dates to the early 1900s with the exploration for natural enemies in Africa to be released in Italy by Filippo Silvestri 28 . The early explorations discovered and described several braconid species collected from B. oleae including Psytallia concolor (Szépligeti), P. lounsburyi (Silvestri), Utetes africanus (Szépligeti) and Bracon celer Szépligeti collected in South Africa, Kenya and Ethiopia reviewed in 18,29,30 . However, none of these parasitoids were successfully cultured by Silvestri and only small numbers of some of these parasitoids were released in Italy without subsequent establishment 31 . Only P. concolor, obtained from Tunisia, was repeatedly introduced since the early 1900s and extensively released in the Mediterranean Basin, but this species has established only in some southern regions and does not provide effective control 32,33 . Still, there has been continued interest in massrearing and releasing P. concolor and/or P. lounsburyi to improve sustainable y management in Europe [34][35][36] .
The invasion and widespread establishment of B. oleae in California and northwestern Mexico initiated renewed interest in the classical biological control of this pest 37 . Modern exploration for effective natural enemies was designed to include the y's likely native ranges in Sub-Saharan Africa (Kenya, South Africa and Namibia) and some expanded regions in Africa (Canary Islands, Morocco, Tunisia and Réunion Island). Here we present a comprehensive analysis of the regional distribution, diversity, and dominance of braconid B. oleae parasitoids in these seven African regions and examine how the regional dominance might be related to regional climatic variables. Furthermore, we analyzed the seasonal dynamics of B. oleae and its co-evolved parasitoids in South Africa, one of the native regions with a high diversity of host-speci c parasitoids, and examined how some biotic and abiotic factors might have shaped the local diversity of the parasitoid complex that effectively regulates B. oleae populations at low levels. This framework may provide new insights into the nature of climate niches of different parasitoid species and their associated tri-trophic interactions in guiding the design of ongoing classical biological control programs in California and the Mediterranean Basin.

Results
Parasitoid regional distribution and diversity. Surveys from 110 sites of wild olives, O. e. nr. ssp. cuspidata, in seven African regions yielded a total of 443,308 olive berries (Fig. 1), of which 72,453 y pupae, 27,848 adult B. oleae and 22,576 adult braconid parasitoids were obtained (Table 1). Two closely related African Bactrocera species, B. biguttula (Bezzi) (1.2%) and B. munroi White (2.6%) were also recovered, but in low numbers and with B. biguttula found only in South Africa and B. munroi only in Kenya (Table 1). Five Opiinae braconid wasps were recovered: P. concolor, P. lounsburyi, P. humilis (Szépligeti), U. africanus and Diachasmimorpha longicaudata (Ashmead); one Braconinae braconid wasp, B. celer, was also recovered. Psytallia concolor was the only species found in the Canary Islands, Tunisia and Morocco, whereas D. longicaudata was the only species recovered in the Réunion Island (Fig. 1). The other four species were found in Namibia and South Africa and three of them (except P. humilis) were found in Kenya, with P. lounsburyi, P. humilis and U. africanus being the predominant parasitoid species in Kenya, Namibia and South Africa, respectively (Fig. 1).  4). Despite the overlap of a few sites, the explored regions represented clearly different climate types. The climates in the Canary Islands, Morocco and Namibia were similar and are characterized by high temperatures and low precipitation. However, the climates in Kenya were related positively to the precipitation but negatively to the maximum temperature of the warmest month with South Africa falling between these two climate types. The Réunion climates were highly correlated to precipitation. The regional dominance of the parasitoid species was re ected in the PCA ordination (eigenvalues: component 1= 5.62, 51.1% of variance; component 2 = 2.17, 19.7% of the variance) (Fig. 5). Sites in Kenya were assigned on the left while sites in Namibia were assigned on the right, and those in South Africa were in the middle. There was a positive relationship between the annual mean temperature or maximum temperature of the warmest month and the relative abundance of P. humilis, however this relationship was negative for P. lounsburyi. The relative abundance of U. africanus was negatively correlated with the minimum temperature of the coldest month and strongly dependent on the precipitation during the wettest quarter.
Mean monthly host density (or fruit infestation rate) on the unripe and ripe fruit were 0.5-11.7% (mean = 4.5 ± 0.7 %) and 4.6-41.2%, (mean = 14.8 ± 2.0%), respectively (Fig. 6A). Host density increased with fruit maturity and was affected by the interaction between fruit maturity and seasonal temperature (Table 3). Most emerged ies were B. oleae. Only 0.87 ± 0.39 % and 1.90 ± 0.66 % (n =24) of the emerged ies from the unripe and ripe fruit were B. biguttula. Mean combined parasitism of B. oleae and B. biguttula were 28.6 ± 2.8% and 25.4 ± 2.6% on the unripe and ripe fruit, respectively. The parasitism was not affected by fruit maturity but rather negatively related to mean temperature (Table 3), decreasing only during mid-summer months (Fig. 6B). All four braconid parasitoids, P. lounsburyi, P. humilis, U. africanus and B. celer, were found in both unripe and ripe fruit. Diversity was generally higher in ripe than unripe fruit (Fig. 7A). U. africanus was the predominant parasitoid, followed by P. lounsburyi while both P. humilis (mean 1.8% and 2.5% on the unripe and ripe fruit, respectively) and B. celer (mean 0.1% and 4.9% on unripe and ripe fruit, respectively) were much less common in both ripe (Fig. 7B) and unripe (Fig. 7C) fruit. GLM analyses showed that diversity was not affected by mean temperature but was positively related to fruit maturity and host density. The relative abundance of U. africanus was affected negatively by fruit maturity and the presence of other parasitoid species but was positively related to host density ( Table 4). The relative abundance of P. lounsburyi was affected only by the presence of other parasitoids (Table 4).

Discussion
We conducted the largest modern exploration for olive fruit y parasitoids in Africa. Our surveys reveal remarkable differences in distribution, diversity and dominance of braconid parasitoid guilds from wild olives across the African continent. The sub-Saharan regions of Namibia, South Africa and Kenya maintained the highest diversity of braconid B. oleae parasitoid species, supporting the argument of a Sub-Saharan origin of B.
oleae [22][23][24] . We found only one native braconid parasitoid (P. concolor) in northern Africa, despite climates in the sampled regions being similar to that of Namibia. We recovered only the introduced parasitoid, D. longicaudata, in Réunion Island where the native Afrotropical species Diachasmimorpha fullawayi (Silvestri) was reported from other tephritid fruit ies 38 44 . In our collections, parasitoids were obtained from pupae collected after exiting fruit or by rearing adults from infested fruit. It is likely that parasitoids that locate and attack hosts in the soil after larvae drop from fruit, or following pupation, have been underrepresented 43 . Some pupal parasitoids such as Pachycrepoideus vindemiae Rondani (Hymenoptera: Pteromalidae) were known to attack B. oleae 28 . Other chalcidoid parasitoid species were reported previously from Africa attacking fruit ies, but they are considered to be generalists and would not be recommended for introduction for biological control e.g., 45,46 . The other two closely related y species recovered, B. biguttula in South Africa and B. munroi in Kenya, were also collected from wild olives. The collected parasitoids may also attack these two y hosts, but B. oleae is thought to be their major host species as the number of the other two y species were extremely low in South Africa and Kenya and not recovered in Namibia during our collections.
Four parasitoid species, P. lounsburyi, P. humilis, U. africanus and B. celer, were sympatric in the sub-Saharan regions surveyed. However, their dominance varied among regions with different climate types, as determined by PCA. In central Kenya where P. lounsburyi was the dominant species, the climate is characterized by mild tropical weather with relatively limited uctuations in temperature extremes but ample precipitation during the rain months. In contrast, in Namibia where P. humilis was the dominant species the climate is typically hot and dry during summer and cold and humid during the winter. Indeed, laboratory studies con rmed that P. humilis was more heat-tolerant, yet less cold tolerant, than P. lounsburyi 47,48 , which may impact their establishment in regions with either hotter summers or colder winters. Although little is known about U. africanus' temperature tolerance, the current surveys showed U. africanus was more abundant in the Mediterranean-like climates. Many other biotic and abiotic factors could also affect the distribution of these parasitoids. Rainfall patterns would strongly in uence the seasonal occurrence and abundance of fruit availability, and consequently the abundance of ies and their parasitoids. In drier habitats, the fruit is likely to be small and ripen slowly, offering little food for y larvae. Annual precipitation was consistently highest in Kenya, as were olive y populations and their parasitoids (Table 1). Interspeci c competition may occur and coexistence between these species is likely facilitated by niche segregation through differentiation in biological or ecological traits. As shown in South Africa, U. africanus was more dominant on small and unripe fruit whereas P. lounsburyi was more dominant on ripe fruit. Large ripe fruit may limit the access of U. africanus, which has the shortest ovipositor among all ve larval parasitoids 10 , but other parasitoids such as P. lounsburyi ll the niches. If interspeci c competition shapes the parasitoid guilds, it likely would show a similar dominance across different regions.
Thus, adaptation to abiotic conditions is likely a major force underpinning diversi cation and dominance of these species.
All ve braconid parasitoids have been imported and evaluated in classical biological control of B. oleae in California 49 . In addition, D. longicaudata was also found to readily attacks B. oleae 10,40,50 , but it is a generalist parasitoid of tephritids 38 . Among the co-evolved African braconid parasitoids, the relatively shorter length of U. africanus ovipositors match with lower pulp thickness of wild olives. This parasitoid is ineffective on cultivated olives that has higher pulp thickness through breeding programs. This thicker pulp allows B. oleae y larvae to move deeper into the olive pulp to escape attack from larval parasitoids that have short ovipositors 10,50 . Bracon celer is able to attack the Cape ivy y, Parafreutreta regalis Munro (Tephritidae: Tephritinae), which itself was introduced from South Africa into California for the control of the invasive Cape Ivy weeds 40,51 . Psytallia concolor is also a common parasitoid of the Mediterranean fruit y, Ceratitis capitata (Wiedemann) (Diptera: Tephritidae) in eastern Africa 12,38,52 . Although the current surveys did not nd it on B.
oleae in Kenya, it has been previously collected from coffee-infesting C. capitata in other parts of Kenya 52 . Genetic analysis showed clear separation of the North African populations from the Sub-Saharan populations and thereafter referred the Sub-Saharan P. concolor populations (often described as P. cf. concolor 52,53 as P. humilis 54,55 ). P. lounsburyi has been reared only from B. oleae 38,52,56,57 and is the most host-specialized parasitoid among all parasitoid candidates 58 . In the Mediterranean Basin, P. concolor is the only parasitoid that has been extensively studied e.g., 59 and widely released with partial establishment in the southern regions 60 .
In California, two P. humilis populations, originated from B. oleae on wild olives in Namibia and C. capitata on coffee in Kenya, were released without subsequent establishment 37,61-63 . However, two populations of P. lounsburyi, originating from Kenya and South Africa, were released and successfully stablished along the California coastal regions 37 . Low winter survival may contribute to the failure of establishment of P. concolor in northern Mediterranean Basin 60 and P. humilis in California 47,64 . For this reason, P. ponerophaga from Pakistan is being considered for release in California as it was found to have higher rates of low temperature survival than P. humilis 65 . Other factors such as availability of olive fruit for the host-speci c B. oleae and alternative hosts could also restrict the establishment of its specialized parasitoids in introduced regions. In South Africa, wild olives are su ciently available all year, and alternative hosts may also help parasitoid populations to survive periods when local B. oleae populations are sparse 29,52 .
The current study showed that fruit maturity, seasonal climates and interspeci c competition likely shape seasonal host-parasitoid dynamics in South Africa. Collectively, the co-adaptation of parasitoids and hosts has resulted in balanced population densities in its native range. Fruit infestation rate was generally less than 15%.
In contrast, untreated olives in California can reach 100% infestation 66 . Although olive fruit y larvae are not tolerated in fruit used for canning, 10-30% infested fruit can be tolerated in olives that are pressed for oil in Collected fruits were kept at room temperature (20-23 °C) in collaborating laboratories or hotel rooms near collection sites. The y larvae often pupate inside unripe fruit but will exit and pupate outside of ripe fruit (typically in the soil underneath the tree in situ). When available, the majority of collected fruit were ripe, this allowed an easier collection of the y puparia emerging from fruit, although both unripe and ripe fruit were collected. Larval ectoparasitoids, such as B. celer, emerge directly from fruit, which were held for up to one month for maximum emergence of ies or parasitoids. When possible, the emerged pupae were returned with the collector or sent by cooperators to the ARS European Biological Control Laboratory (EBCL), otherwise the material was held at collaborating laboratories for emergence of ies or parasitoids. All emerged insects were identi ed to species and gender. Stellenbosch and Wellington (Fig. 2). These sites were located within 200 km of each other and ranged in elevation from 77-823 m. Approximately 900 fruits of wild olive were collected at each site once every 2-4 weeks, depending on the availability of fruit. Collected fruit were processed at Stellenbosch University and sorted by size and condition. Fruit size, or pulp thickness, was assumed to affect some parasitoid species ability to nd and oviposit into y larvae feeding deeper inside the fruit because of their short ovipositors 10 .
Fully ripe (black) fruit is generally larger than unripe (green) fruit and y larvae will feed deeper inside the softer fruit. Therefore, green and black fruit were sorted and assessed separately. Subsamples of unripe and ripe fruit were measured to estimate the pulp thickness of each fruit by inserting an insect pin trough the pulp to the seed three times at randomly selected points on the fruit. The mean depth (pin length minus the exposed portion of the pin) of the three measurements was used to estimate fruit pulp thickness. Collected fruit and emerging puparia were kept at room temperature (20-23°C) until the emergence of wasps and ies.
Data analysis. The relative abundance of each parasitoid species (i.e. percentage of each parasitoid species emerged), total parasitism by all parasitoids and diversity were estimated for each sample in each site and region. Total parasitism was calculated by dividing the total number of emerged parasitoids by the sum of the number of emerged parasitoids and flies. The Shannon index (H) was used to estimate the diversity: where pi is the proportion of each parasitoid species. Sex ratio (% females) of each parasitoid species was pooled from different regions because initial analyses did not detect signi cant differences for any parasitoid among different regions. Mean parasitism and diversity among different regions and sex ratio among different parasitoid species were compared using one-way ANOVA. All data were rst inspected for normality and error variance for homoscedasticity and all percentage data were logit transformed as needed before analysis.
Principal Component Analysis (PCA) was conducted to compare climate among the sampled regions (Tunisia was excluded due to the small samples and its climatic similarity to Morocco) and to analyze potential relationships among regional dominance of parasitoid species and bioclimatic variables in the three Sub-Saharan countries. A set of eight bioclimatic variables were selected for the analyses: annual mean temperature (Ann tem), maximum temperature of the warmest month (Max tem), minimum temperature of the coldest month (Min tem), mean temperature of the warmest quarter (Warm tem), mean temperature of the coldest quarter (Cold tem), annual precipitation (Ann prec), precipitation of the wettest quarter (Wet prec) and precipitation of the driest quarter (Dry prec). These bioclimatic variables were extracted from the WorldClim Global Climate Database 1.3 (http://www. worldclim.org) using the R 3.1.3 release. These variables are considered biologically relevant and used commonly in species distribution studies. A biplot analysis was conducted to characterize the relationships.
For the analyses of seasonal host-parasitoid dynamics in South Africa, host y density was estimated as the number of y puparia per fruit. Because wild fruit is smaller than cultivated fruit, each wild fruit supports fewer ies, commonly one y larvae per fruit 43 ; therefore the y density per fruit approximately matches the percentage of infested fruit (i.e. fruit infestation rate). Data were pooled from different sites to estimate monthly mean host density or fruit infestation rate, total parasitism by all braconid parasitoids, parasitoid diversity, and the relative abundance of each braconid parasitoids on unripe and ripe fruit, respectively. Mixed models were used to analyze the effects of fruit maturity (unripe vs. ripe) and seasonal climate (both were xed effects) as well as year (random effect) on monthly host density and total parasitism. Monthly mean temperature was used to represent a seasonal climate variable as precipitation was considered similar within the surveyed areas and other temperature parameters (e.g. maximum or minimum temperature) are highly correlated with the mean temperature. The temperature data were obtained from Weather Information (https://us.worldweatheronline.com/) from the closest cities (Stellenbosch, Paarl, Citrusdal, Cape Town, Bonnievale or Wellington) of the sampled sites. Generalized linear models (GLM) were applied to analyze the effects of (1) fruit maturity, mean monthly temperature, host density and parasitism on diversity, and (2) fruit maturity, mean monthly temperature, host density, and incidence of other parasitoids on the relative abundance of two major parasitoids (U. africanus and P. lounsburyi). For GLM analyses, fruit maturity was coded categorically as 1 and 2 for unripe and ripe fruit, respectively, and parasitoid species incidence was coded as 1 (present) and 0 (absent). Percentage data were modelled with binomial distribution and a logit link function while the diversity data was modeled with Poisson distribution and a log link function. Statistical analyses were performed using JMP Pro ver13 (SAS 2013, Cary, NC).