Disentangling the coexistence strategies of mud-daubing wasp species through trophic analysis in oases of Baja California peninsula

Species within the same trophic level show different strategies to avoid competition. Among these mechanisms, differences in body size, spatio-temporal segregation, and diet preference often leads to a niche partitioning. Nonetheless, little attention on coexisting predatory insects and their network interactions has been paid. In this study, we analyzed the strategies to avoid competition among three sympatric mud-daubing wasps of the genus Trypoxylon (Hymenoptera: Crabronidae) in oases and their surrounding xeric area from the Baja California peninsula, Mexico. We compared the prey richness, composition and proportion of spider guilds that were captured by the wasps. We tested whether the differences in wasp body size explained the niche breadth, niche overlap and the size of spider prey. We assessed the spider-wasp interactions through a network analysis. With the use of trap-nests, we collected 52 spider species captured by the wasps. Both the guild and species composition of preyed spiders was different between the three wasp species. Differential proportions in the capture of spider guilds and a little diet overlap were found among the wasp species. We found that the wasp body size was positively correlated with prey size, but it was not a proxy of niche breadth. Moreover, the largest wasp species was able to nest in both mesic and xeric habitats, while the two smaller species were restricted to the oases. This study reveals that the diversity of spiders in oases of Baja California peninsula is crucial to maintain highly specialized oasis-dependent wasp species. The niche partitioning between mud-daubing wasps can be shaped by their inherent body size limitations and hunting strategies through foraging specialization for specific spider guilds. Food selection and slight differences in body size reduce competition and allow the coexistence of sympatric wasps. Our study is the first approach exploring the interaction networks between mud-daubing wasps and their spider preys, highlighting new insights into the morphological and ecological factors that shape antagonistic interactions, and allow the coexistence of predators in deserts.

Introduction the wasp body size plays a role in the size of spider prey, with the prediction that wasp body size is positively correlated with prey size.

Ethics statement
Field work was conducted with permission of Secretaría del Medio Ambiente y Recursos Naturales, in agreement with the Subsecretaría de Gestión para la Protección Ambiental and the Dirección General de Vida Silvestre who issued the Scientific Collector Permit: SGPA/DGVS/ 09769/15. Our study does not involve any endangered or protected species.

Study area
Sampling was carried out from April to September during 2016 and 2017 on six localities along the Baja California peninsula (BCP), in northwest Mexico. The north portion of the BCP belongs to the Nearctic region, which is represented by xeric scrublands typical of the Sonoran desert. In the southernmost area, there are taxa with Neotropical affinity and vegetation such as tropical deciduous dry forest [37]. The sampled areas included oases located between 23˚N and 30˚N, with sizes ranging from 0.06 to 2.6 km 2 in size (Table 1).

Sampling protocol
We focused our analyses on three wasp species that occupied the nests (i.e. 133 in total): Trypoxylon bridwelli (31 nests), T. dubium (47) and T. tridentatum (55). We used artificial cavities to collect the wasp´s nests and identify their composition (i.e. spider prey). We set the traps in both the oases and the surrounding desert to evaluate the prey composition between habitats. The traps were set in gradual distances from the edge of the waterbody (inside the oasis) towards the desert area. That is, from around 5.0 m to 3,500 m away from the waterbody.
In total, we offered 2,430 cavities equally distributed in 162 trap nests set in the six localities. Trap nests consisted in three wooden blocks (10x17x2.5 cm each) set 1.5-2.0 m height, piled and gripped together. Each nest contained five rows of tunnels with different diameter (3.1, 6.3, 1.9, 9.5, 12.7 mm) x 150 mm long [25]. From April to September 2016 we sampled the southernmost oases: Santiago (SA), El Pilar (EP), and La Purísima (LP). From April to September 2017, the northernmost oases were sampled: El Sauzal (ES), San Borja (SB), and San Fernando (SF). In both years, we monthly replaced the occupied traps.
The morphological differences of adults, immature larvae and external appearance of pupae allowed the identification of wasp species and thus the association with their spider prey [26]. Inside the nests, the brood cells were carefully inspected, sorting and counting the spiders for taxonomical identification. Taxonomic identification was done in the laboratory of Arachnology and Entomology (CARCIB) of the Centro de Investigaciones Biológicas del Nororeste (CIBNOR). Since the great majority of spiders were juveniles, identification to species level was not always possible. However, the vast material of spiders from Baja California Sur deposited in the CARCIB, allowed the correct distinction of most morphospecies. Because in some cases we found the remaining parts of spider prey within the cells (i.e. already devoured by the wasp larva), the identification was possible to family level.

Diversity and composition of prey
Spider prey was categorized into the trophic guilds suggested by Cardoso et al. [28] and Uetz et al. [40]. These guilds were proposed as surrogates of spider families based on their foraging strategy, prey range, and vertical stratification. Based on these categories, we classified the spiders captured by the Trypoxylon species within the following guilds: ambush hunter, orb web, spatial web, stalker, ground hunter, and specialist. For each wasp species, we calculated alpha diversity of prey using Hill numbers and estimated the number of spider species through Chao-1 [41,42]. We compared the diversity of the overall spider prey captured by the wasps. The rarefaction curves for each wasp species were estimated. These analyses were computed on R v3.5.1 using the iNEXT package [43,44].
To test the differences in composition of the spider communities captured by the three wasp species, we used a Permutational Multivariate Analysis of Variance (PERMANOVA) [45]. A multivariate analysis of homogeneity of group dispersions (PERMDISP) was used to test heterogeneity of the prey community and as a measure of beta diversity [46]. We used a Non-metric Multidimensional Scaling (NMDS) measured by Bray-Curtis distances to visualize the dissimilarities in the composition of spider species captured by the wasp species [47]. These analyses were done with vegan package in R [44,48].
To evaluate if there is an effect of habitat and time in the amount of food the wasps can provide to the offspring, we used generalized linear mixed models (GLMM) to assess if the length of brood cells depends on the habitat (oasis or desert) or if it varied through the time (i.e. monthly variation). Models were fitted with lme4 package [49] in R [44].

Niche partitioning
In order to determine the relative level of dietary specialization and niche segregation of the species, the niche breadth and niche overlap were calculated for the three Trypoxylon wasps [50]. Hurlbert's formula [51] was applied to obtain the standardized niche breadth (Bj) [52]. We used the Pianka´s index to measure niche overlap. Niche breadth index ranges from 0 to 1, where values close to 1 indicate more specialization. Pianka´s index also ranges from 0 to 1, where values close to 1 indicate a higher diet overlap. We used the spaa package to compute these indexes [53]. We built a bipartite network (predator-prey) to calculate the complementary specialization index (H 2´) and Shannon diversity of interactions between both trophic groups [54]. H 2´i ndex ranges between 0 (no specialization) and 1 (complete specialization). Since this index can be sensitive to matrices constructed with few species, we compared our data against 1000 random null models that avoid biases regardless the matrix size. With the null model approach, we assessed that the spider-wasp interactions are not being reflections of the sampling properties, thus producing random links between predator and prey [55, 56]. The network analysis was done with the package bipartite in R [44,57,58].
To test whether the size of predators influenced the niche breadth and prey size selection, we measured morphological attributes of size on both wasps and spiders. We obtained the values of body length, facial and intertegular distance from the three Trypoxylon species. The spider size was represented with the values of cephalothorax´s length and width [59]. Only female wasps were considered since they are the responsible of hunting and transportation of spiders to the nests [24]. The correlation between the wasp size and cephalothorax´s length-width ratio of spiders was tested. We used a Linear Discriminant Analysis (LDA) to evaluate whether the three wasp species were well separated according to the morphological measurements [60]. LDA was calculated with the psych package in R [44,61]. Measurements were done under a stereo microscope Nikon SMZ25, with 11x magnification, 1.1x zoom and a Nikon SRH 1x objective.
The bipartite interactions between prey and predators showed a high degree of both overall specialization (H 2´= 0.89) and diversity of interactions (H´= 2.6). Our network significantly differed from expected null models (Mean ± S.D.: 0.09 ± 0.01, p<0.001) (Fig 5).
The spider prey diversity (H´) was significantly different between the three wasp species and the accumulation curves of prey diversity was asymptotic in T. dubium (Fig 7). Trypoxylon tridentatum showed the richest composition of prey and nocturnal species, whilst T. dubium showed the highest values of diversity, evenness and niche breadth. Moreover, T. bridwelli (the smallest species) had the highest dominant prey composition and the poorest species richness (Table 3). Diet overlap and similarity of prey species were very low among the three species.
The highest values of both indices were found between the two largest wasps, T. dubium and T. tridentatum (Table 4).
There was a significant difference in the wasp body size that allowed the separation of the three Trypoxylon species (F 2, 69 = 773.6, p<0.001) (Fig 8A). Likewise, a significant difference in the selection of prey size was found between wasp species (F 2, 69 = 123, p<0.001) (Fig 8B). The wasp body size was positively and highly correlated with the spider size (r = 0.86, p <0.001), having T. dubium the widest range of prey size on its diet (2.0 to 4.75 mm length of cephalothorax) (Fig 8C).

Discussion
According to our proposed hypotheses, each wasp species has a differential composition of spider prey. Therefore, there is a low diet overlap between wasps. Moreover, we confirmed that slight changes in the wasp body size (± 2 mm) play a role in the size of spider prey they hunt. As it was predicted, larger wasp species captured larger prey.
Usually, isolated habitats tend to harbor less species that display more specialized interaction networks [30]. As we found in this study, the wasp-spider interactions showed a highly specialized network, where the wasps have a strong prey preference. The use of resources in islands can be the result of selective pressures, since organisms undergo similar environmental pressures and dispersal limitations [30,31]. Such pressures can also be found in isolated patchy environments (e.g. oases, caves, or sky islands), where coexisting species have developed several mechanisms to avoid competition [3,32].
Although niche specialization can be high in insular habitats due to the restricted amount of resources [62], species can coexist through the niche segregation into multidimensional axes such as space, time, strata, or diet [37]. The three wasp species coexisting in the oases of Baja California Peninsula are displaying differential food specialization, which is probable the result of isolation processes that have led to evolutionary selective pressures.
Trypoxylon (Trypargilum) species within the Nitidum Species Group seem to be specialized for orb-weavers (Araneidae) and space web (Theriididae) species within the genera Eustala, Metepeira, Eriophora (Araneidae), and Theridion (Theriidiae) [26]. Similarly, we found that most of the diet of T. tridentatum and T. dubium (both within the Nitidum group) included araneid and theridid spider species. However, it is likely that both species richness and abundance of local prey at oases could have influenced the wasp´s prey choice. For example, the prey composition of T. tridentatum seems to be variable across localities: Theridion submissum (25% of total prey) in the southern Baja California peninsula [63,64] and Eustala rosae (21%) in Arizona [25]. In our work, we found that Metepeira arizonica (46%) and T. submissum (33%) where the main target. Our results reinforce the idea that local composition of spiders may affect the wasp´s prey selection. Therefore, it is likely that T. dubium and T. bridwelli show a similar response, especially among oases from the BCP, where the spider composition is highly variable [36,65].

Selection of spider prey
It has been proposed that three-dimensional (3-D) webs are more efficient than two-dimensional (2-D) web architectures to avoid predation from mud-daubing wasps [27]. In some cases, 2-D weavers such as Metepeira species can build 3-D barrier webs [66]. However, whilst 2-D species rest at the center of webs, 3D-weavers remain protected inside the silk network. This defensive mechanism may explain why Trypoxylon are more prone to catch orb-weaving species. Although, spiders use other strategies to avoid predators (e.g. cryptic of coloration, silk retreats near webs, dropping from webs), web architecture is an important strategy of weaver Niche partitioning of mud-daubing wasps in oases spiders [67]. Even though the three wasp species preyed on web-building species, the specialization of T. bridwelli upon jumping spiders is noteworthy (>90% of prey species).
It has been suggested that both visual and chemical cues are likely the main mechanisms to locate either web-building or non-web-building species [68][69][70][71][72]. In our study, the inclusion of ant-mimic spiders in the diet of T. bridwelli, convey to the idea that probably chemical signals are being followed to locate its prey. Moreover, since spiders occupy different microhabitats and strata [73], it is likely that wasps are displaying specific patrolling behavior to locate their prey. In this sense, stratified sampling of spiders may help to understand the patterns of waspś foraging. More research is needed to determine the intricate combinations of chemical, aggressive mimicry, shape, and color perception involved during the spider recognition [74].
The nutritional composition of prey also plays a role in the food preferences of predators [75]. Among wasps, nutrition of larvae has a strong effect on reproductive fitness, foraging, and brain gene expression of adults [76,77]. It makes conceivable that protein, lipid and carbohydrate supply is likewise critical in the fitness of adult Trypoxylon species. Especially because their offspring need to storage energy and protein before the diapause period [24]. Female Trypoxylon wasps usually prevent nutritional unbalance of larvae by providing more small body size juvenile spiders until the biomass reaches the required energy [78]. This is perhaps the explanation of the large number of juvenile spiders within the cells. However, the high amount of juvenile prey may be also an artifact of population dynamics of spiders or environmental effects on local the abundance [79].

Effect of body size on prey preference
Since larger Trypoxylon wasps hunted larger spider prey, the effect of body size fits well with the assumption of cascade model (a positive predator-prey correlation) [10,11]. However, this model is not always true for arthropods in particular assemblages [80], and the small amount of experiments is still a limitation to further assumptions. In our results, the body size was positively correlated with niche overlap but was not with the niche breadth. In this case, the overlap between wasps increased with the body size but the mid-sized species (T. dubium) showed  Table 3. Summary of diversity and functional parameters between the three wasp species and their prey. The habitat indicates where the nests of each wasp species were found. The wasp body length is the measurement from the frons to the last segment of metasoma (only females). All estimators and diversity indices were calculated for the total number of spider species each wasp captured. Niche partitioning of mud-daubing wasps in oases the widest niche breadth and range of prey size. Similarly, Polidori et al. [81] showed that the body size between crabronid wasps was positively correlated with the prey size, but with the niche breadth. These authors suggest that most solitary wasps experience a strong individual specialization, which broadly influences the prey size selection. In addition, according to the flight muscle ratio, thus the load-lifting capability, only a few wasp species are optimal foragers, which is attributed to several biotic factors [82].

T. bridwelli T. dubium T. tridentatum
On the other hand, the seasonal variation can be a factor shaping the range of prey size and niche breadth of wasps. In northern Brazil, Araújo & Gonzaga reported variation in the prey selection of Trypoxylon (Trypargilum) albonigrum, finding a broader range of prey size and niche breadth during the wet season [83]. However, the relationship between the wasp, prey size and niche breadth is still under debate, especially since the prey preference of several wasp species remain unknown. Although the identification of spider prey could be challenging considering the large amount of juveniles captured by the wasps, the appropriate identification of morphospecies did not alter the analyses. In this sense, DNA approximations would be an option to avoid misleads in the prey identification.

Habitat effect on niche partitioning
An alternative explanation for niche partitioning could be the effect of habitat use among wasp species [25]. In other Trypoxylon species, habitat is an important factor explaining the composition and abundance of spider prey [84,85]. In our study, the highest abundance and richness of spiders were found in oases. The largest wasp species (T. tridentatum) used both oases and desert habitats and in general hunted larger spider prey. Among insects, physiological limitations have implications on their distribution and habitat use [86]. In a broad sense, insects with larger body size are able to tolerate more environmental variability [87]. Moreover, flying insects with larger body mass tolerate higher temperatures in extreme habitats (e.g. deserts) due to higher surface area-volume ratios [88]. In dry environments, water loss rates are lower in large bee species thus being less vulnerable to desiccation [89]. Although, temperature could explain the microhabitat selection of some bees and wasps [90,91], the thermoregulatory physiology of wasps is still poorly understood [92]. Nonetheless, it seems that the body size of T. dubium and T. bridwelli may be a physiological boundary in both thermal tolerance and preylifting capacity.
Beyond microclimatic preferences, the coexistence of wasps could be mediated by temporal segregation. For example, in Northeast Brazil, Santos & Presley [93] observed that social vespid wasps had slight changes in their peak activity across the day, suggesting that environmental factors such as temperature and humidity are playing a role in the foraging activity, thus, reducing the interspecific competition. In other cases, the coexistence of social wasps is allowed by broad differences in their diet (i.e. specialist vs opportunistic species) [94].

Conclusions
With these results, we disentangled the strategies that three mud-daubing wasp species have to coexist in isolated and small mesic environments into a desert ecosystem. Differential use of resources, either food or habitat (oasis or desert) allows the coexistence of ecologically similar wasp species. In this regard, each wasp species is showing a high guild preference, presenting a specialized predator-prey network to hunt certain spider species. Moreover, the wasp body size is positively correlated with prey size, allowing the ecological differentiation for the three Trypoxylon species.
On the other hand, the oases seem to work as islands for the studied wasp species, either as the product of physiological boundaries or food preference. Therefore, the conservation of isolated habitats become crucial for many species that depend on specific and limited resources. Moreover, habitat characteristics can be crucial for cavity-nesting species [95], but the direct effects on these guilds are poorly known. Since anthropogenic pressure is the main cause of disturbance in oases of Baja California [29], highly specialized sympatric insects could be threatened if habitat loss and fragmentation modify the structure of these insular-like environments.