Taxonomy, Habitat Preference, and Niche Overlap of Two Arrow-Poison Flea Beetle Species of the Genus Polyclada in Sub-Saharan Africa (Coleoptera, Chrysomelidae)

Simple Summary The taxonomy of many African Coleoptera species is remains poorly known, and the knowledge of their ecological requirements is worse still. Starting with original data, we describe morphological differences and ecological data for two flea beetle species, Polyclada bohemani and P. pectinicornis, which traditionally have been used by the Bushmen people in sub-Saharan Africa to poison their arrows. Moreover, we evidence differences in the formations of vegetation used by these two species, which are known to occur mainly in savannah and open forest habitats. Also, we identify differently suitable areas in terms of climatic preferences, in addition to a common territory in East Africa. We also supply, for the first time, the description of the shape of the aedeagus and the spermatheca of both species, supplying important new diagnostic characters for their identification. Abstract Coupling the geographic distribution and the ecological requirements of species often supports taxonomy and biogeography. In this contribution, we update the distribution of two flea beetle species of ethno-entomological interest, Polyclada bohemani and P. pectinicornis, by analyzing original data. In addition, we supply their main morphological diagnostic characters, describing their aedeagal and spermathecal shapes for the first time. We also assess their niche differences in terms of climatic and vegetation needs, by means of ecological niche modelling and remote sensing techniques. Several new localities were identified to improve knowledge of the geographical distribution of both species. Moreover, we located a wide climatic suitability overlap in East Africa for these two flea beetle species, while in other areas they show a clear separation. Our analysis also reports that P. bohemani is associated with areas of denser tree cover than P. pectinicornis. Finally, the lectotypes of Diamphidia bohemani Baly, 1861, Clytra pectinicornis Olivier, 1791, and Diamphidia compacta Fairmaire, 1887 are here designated and the new synonymy Clytra pectinicornis Olivier = Diamphidia compacta Fairmaire syn.nov. is proposed.


Introduction
The Coleoptera, with about 380,000 described species, are the world's most diverse animal order. Their truly amazing morphological and functional solutions translate into critical ecological functions, ranging from plant pollination to nutrient recycling in various environmental matrices (e.g., soil, freshwaters), to secondary consumer food supply and pest biocontrol [1][2][3]. Despite their primary significance in many environments, the ecological and biogeographical variables that affect their distribution in many areas, especially in the southern hemisphere, have yet to be understood.
The family Chrysomelidae, known as leaf beetles, is among the most frequently represented of the Coleoptera families in the terrestrial habitats of all continents, except the the recently re-evaluated Blepharidina Bechyné and Calotheca Heyden [11,15,[18][19][20][21][22]. Polyclada species are associated with the Anacardiaceae (Sclerocarya birrea) and Burseraceae (Commiphora spp.) plant families, which are found in a variety of woodland and savannah ecosystems [22,23].
To explore and define environment-occurrence trends in the Afrotropical region for two species of this flea beetle genus, Polyclada bohemani (Baly) and P. pectinicornis (Olivier), we used state-of-the-art ecological niche models (ENMs), together with habitat and niche overlap analyses. Both species are well-known for their ethno-entomological importance, as the pupae and body fluids are employed as arrow poison by the San people (Bushmen) of southern Africa. The active molecule is the diamphotoxin, present also in the African flea beetle Diamphidia Gerstaecker, a toxalbumin that works similarly to various snake venoms, causing widespread paralysis, hemolysis, and death (cf. Chaboo et al., 2007).
In the present contribution, in addition to updated distributions, we also provide the main diagnostic characters for the identification of these two African species, reporting for the first time the description of the median lobe of the aedeagus and of the spermatheca. Moreover, we characterize their ecological requirements and interpret their pattern of occurrence at a continental scale.

Study Area, Species Database, and Vegetation Formations
The study area comprised sub-Saharan Africa. We performed analyses on a dataset including occurrences for two endemic flea beetle species (Chrysomelidae, Galerucinae, Alticini) of the genus Polyclada, P. pectinicornis (Olivier), with 80 occurrence localities, and P. bohemani (Baly), with 104. We retrieved data records of the two target species mainly from the examined material, consisting of 550 dried pinned specimens preserved in the institutions listed in the "Abbreviations Used" section below. The specimens were examined and dissected using a Leica M205C stereomicroscope. Photographs of the habitus and the spermatheca were taken using a Leica DMC5400 camera and compiled using Zerene Stacker software v. 1.04. Scanning electron micrographs were taken using a Hitachi TM-1000. Terminology follows D' Alessandro et al. (2016) [24] for the median lobe of the aedeagus and spermatheca. We recorded the geographic coordinates for the localities in decimal degree system using the WGS84 datum, and we added information included in square brackets to the label data, using the Google Earth website for coordinates and geographic information. Abbreviations for the depositories follow the list on the website The Insect and Spider Collections of the World [25]. Chorotypes follow Biondi and D'Alessandro (2006) [26]. We used the raster map of terrestrial ecosystems of Africa to gather spatial information (resolution: 90 m cell) about vegetation formations, classified in a hierarchical fashion (i.e., class, subclass, formation, division, macro-group) [27].
Moreover, we assessed tree cover within the habitats in which the two species occur. To assess this environmental predictor, we extracted the values of tree cover (%) from the dataset of Hansen et al. (2013Hansen et al. ( ) (updated to 2021 at a spatial resolution of 30 m. Tree canopy cover was defined as canopy closure for all vegetation taller than 5 m. Based on Landsat images, this data was acquired by the Google Earth Engine cloud platform using a Python-based code, which allowed us to extract the percentage value of tree cover for each occurrence of the two target species.

Model Building
To estimate the current suitable areas for Polyclada bohemani and P. pectinicornis, we built ecological niche models. For this purpose, we downloaded 19 temperature-and precipitation-related "bioclimatic" raster variables from the Worldclim 2.1 online repository [28] at cell resolution of 2.5 arcminutes (~5 km at the equator). To avoid possible correlation among predictors, which might lower the model's performance, we assessed both the variance inflation factor (VIF, threshold set = 10 following Guisan et al. (2017)) [29], and Pearson's r (|r| < 0.9, following Dormann (2007) [30] and Elith et al. (2006)) [31], using the 'vifstep' and 'vifcor' functions of the 'usdm' R package [32], then applying a subset of selected predictors to calibrate the models.
The building of the ENMs for both Polyclada species was completed through the "biomod2" package [33] in R environment [34]. Ensemble models (EMs, resulting from the combination of individual ENMs) were built for the current climatic conditions using the "BIOMOD_EnsembleModeling" function. We generated 10 sets of 1000 pseudo-absences using the Surface Range Envelope method [35,36], with quantile = 0.05. We parametrized models built for P. bohemani and P. pectinicornis as follows: generalized linear models (GLM): type = "quadratic", interaction level = 3; multiple adaptive regression splines (MARS): type = "quadratic", interaction level = 3; generalized boosting model, also known as boosted regression trees (BRT): number of trees = 5000, interaction depth = 3, cross-validation folds = 10; we performed five evaluation runs.

Model Evaluation and Ensemble Forecast
To assess the discrimination performance of the single models, we took advantage of two different evaluation metrics, the area under the curve (AUC) of the ROC curve [37] and the True Skill Statistics (TSS) [38]. We used 80% of the initial dataset to build the models, and the remaining 20% for their validation. Considering the five evaluations runs, 10 pseudo-absences sets, and three modeling algorithms chosen, 150 single models were finally generated for each species. We then built the EMs by selecting only the ENMs exceeding the thresholds TSS > 0.7 and AUC > 0.7; the "weighted mean of probabilities" (wmean), which averages the single models by weighting their AUC or TSS scores, was used for this purpose [33].

Environmental Niche Overlap
The extent of possible environmental niche overlap between Polyclada bohemani and P. pectinicornis was investigated through the PCA-Env approach [39], taking advantage of the 'ecospat' R package version 3.1 [40]. A principal component analysis (PCA) was calibrated by sampling the values of the considered covariates from occurrence localities of the target species, as well as from several background points across the study area. Subsequently, a two-dimensional gridded environmental space was defined based on the two principal components (hereafter, PrinComp) contributing the most to explained variance (i.e., showing the highest eigenvalues). Finally, 'niche occupancy' of each target species within this PCA-derived 2-D space was computed as a kernel-smoothed density of occurrence, correcting the observed density of occurrence by the density of environmental conditions possibly exploitable by the species (i.e., those within reachable areas). To calibrate the PCA-Env, we coupled a subset of uncorrelated bioclimatic variables (found by means of the VIF analysis described above), with cell-by-cell percentage cover of three  (2.A.1), and warm desert and semi-desert scrub and grassland (3.A.2) [27]. The chosen formations were those contributing most to the first two PrinComps of a preliminary PCA-Env, calibrated using only percentage cover of the seventeen African vegetation formations as covariates. Niche overlap between P. bohemani and P. pectinicornis was computed using Schoener's D metric and a modified version of Hellinger's distance metric, indicated as I [41]. The obtained niche overlap scores were tested for significant deviation from null expectation, considering possible differences between species in terms of exploitable environments, through the niche similarity test implemented in 'ecospat'. In this test, the observed overlap score between two species, for exaqmple j and k, is compared to a distribution of simulated overlap scores; such a distribution is derived by comparing, n times, the simulated niche occupancies of j and k, where simulated occupancies are defined by randomly sampling environmental conditions from a background area around the occurrence localities of each species. In this manner, environmental niches of the two species can be considered more or less similar than expected, based on differences between species in exploitable environments, if the observed overlap of the simulated scores is higher than the 95th percentile (i.e., niche conservatism) or lower than the 5th percentile (i.e., niche divergence), respectively [39]. Based on the limited dispersal capacity of P. bohemani and P. pectinicornis, we defined the respective background areas by drawing a circular buffer of 20 km around their occurrences. The distribution of simulated overlap scores was obtained by repeating the occupancy simulation process 250 times.

Morphological Remarks
Polyclada bohemani and P. pectinicornis are very similar in shape, proportions, and colour (Figures 1a and 2a) and have therefore very often been confused with each other. This is mainly due to their wide intraspecific variability, especially in P. pectinicornis, but also to the fact that adequate studies on the morphology of their aedeagus and spermatheca had not been carried out. The two species, however, exhibit useful diagnostic characters relating to their external habitus and show some constant chromatic differences. Although major morphological differences relate to the median lobe of the aedeagus (Figures 1 and 2), the two species can be reliably identified by focusing on the position of certain elytral patches: the hind sutural patches are generally aligned with the lateral patches in P. pectinicornis (Figure 2a), while in P. bohemani they are not aligned with the lateral patches and are placed closer to the elytral apex ( Figure 1a). This latter species tends to be stable in colour over its distribution range, with black patches, antennae, and legs, except for hind femora which are mostly brownish. P. pectinicornis is instead more variable, with specimens from different localities having antennomeres 1-3 and scutella that are partially yellowish, along with lighter femora; specimens from East Africa often display brownish, and not black, elytral patches.
With reference to the median lobe of the aedeagus, in Polyclada bohemani (Figure 1b) in the ventral view can be observed a sinuate shape that narrows in the distal fourth, an apical part that is medially very prominent, subromboidal, with a rounded median tooth; the ventral sulcus is visible in the apical third; in the lateral view can be seen a straight median lobe with the distal fourth clearly bent ventrally; dorsal ligula are elongated, moderately wide in basal 2/3 s and abruptly narrowed in the distal third; the apical part is bifid, dorsally curved and visible in the lateral view. In contrast, for Polyclada pectinicornis (Figure 2b) the median lobe of the aedeagus is fusiform in the ventral view, truncating apically, with a rounded median tooth; the distal surface is slightly depressed ventro-laterally, and medially prominent; in the lateral view, the median lobe is straight with the distal fourth slightly bent ventrally; dorsal ligula are elongate, wide but abruptly narrowed subapically; the apical part is formed by a wide medial lobe and two thinner lateral lobes, all dorsally bent and well visible in the lateral view. However, we observed some variability in the shape of the median lobe of the aedeagus, in particular in P. bohemani, where the specimens from more southern locations, for example KwaZulu-Natal, generally showed a more slender shape of the distal part. However, at present we prefer not to establish continuity solutions within this variability by the description of new taxa, pending future information deriving from phylogeographic assessment of these populations.  With reference to the median lobe of the aedeagus, in Polyclada bohemani (Figure 1b) in the ventral view can be observed a sinuate shape that narrows in the distal fourth, an apical part that is medially very prominent, subromboidal, with a rounded median tooth; the ventral sulcus is visible in the apical third; in the lateral view can be seen a straight median lobe with the distal fourth clearly bent ventrally; dorsal ligula are elongated, moderately wide in basal 2/3 s and abruptly narrowed in the distal third; the apical part  With reference to the median lobe of the aedeagus, in Polyclada bohemani (Figure 1b) in the ventral view can be observed a sinuate shape that narrows in the distal fourth, an apical part that is medially very prominent, subromboidal, with a rounded median tooth; the ventral sulcus is visible in the apical third; in the lateral view can be seen a straight median lobe with the distal fourth clearly bent ventrally; dorsal ligula are elongated, The spermatheca also displays some differences between the two species, although these differences are not always constant. In Polyclada bohemani (Figure 1c), the elongate basal part is generally slightly narrower towards the ductus, which is apically inserted, short and thickset; the distal part is thin, generally hook-shaped and apically rounded, with no distinct collum and no appendix. In P. pectinicornis (Figure 2c), the spermatheca is very similar but generally shows a subelliptical-elongate basal part and thin distal part, which is moderately curved and generally apically pointed.
Both species are also quite variable in size, although biometric ratios do not vary much within each species: Polyclada bohemani.

Ecological Niche Modelling and Habitat Preferences
After the VIF and Pearson's correlation analyses, we selected a set of nine uncorrelated bioclimatic variables (bio2: mean diurnal range, bio3: isothermality, bio8: mean temperature of the wettest quarter, bio9: mean temperature of the driest quarter, bio13: precipitation of the wettest month, bio14: precipitation of the driest month, bio15: precipitation seasonality, bio18: precipitation of the warmest quarter, and bio19: precipitation of the coldest quarter), which were then used to calibrate the models.
The ensemble models for both the target species resulted in high performance scores (AUC = 0.946 and TSS = 0.766 for P. bohemani, AUC = 0.954 and TSS = 0.810 for P. pectinicornis), indicating a wide and continuous suitable area for P. bohemani (Figure 3a) in the eastern part of sub-Saharan Africa, and a more irregular and narrower suitable area for P. pectinicornis (Figure 3b). Concurrently, a sub-Saharan suitability band, spanning from Senegal to Sudan, was found to be suitable for P. pectinicornis only, in addition to a suitable patch in Chad, territories where P. bohemani scored low suitability values (Figure 3).
When coupling this data with vegetation formations, a common Central African "core" is evident, along with a parallel division of sub-Saharan versus Southern African preference ( Figure 4).
Indeed, this translates into a higher preference of P. bohemani for denser savanna or forest environments, such as tropical lowland grassland savanna and shrubland (2.A.1) and tropical seasonally dry forest (1.A.1), while P. pectinicornis is more associated with dryer environments, such as warm desert and semi-desert scrub and grassland (3.A.2) and salt marsh (2.B.7) formations ( Figure 5).
This trend is also confirmed by the tree cover results, where P. bohemani showed a higher affinity to greater tree cover values (Figure 6a), while P. pectinicornis was found to prefer more open habitats (Figure 6b). When coupling this data with vegetation formations, a common Central African "core" is evident, along with a parallel division of sub-Saharan versus Southern African preference ( Figure 4). Indeed, this translates into a higher preference of P. bohemani for denser savanna or forest environments, such as tropical lowland grassland savanna and shrubland (2.A.1) and tropical seasonally dry forest (1.A.1), while P. pectinicornis is more associated with  When coupling this data with vegetation formations, a common Central African "core" is evident, along with a parallel division of sub-Saharan versus Southern African preference ( Figure 4). Indeed, this translates into a higher preference of P. bohemani for denser savanna or forest environments, such as tropical lowland grassland savanna and shrubland (2.A.1) and tropical seasonally dry forest (1.A.1), while P. pectinicornis is more associated with dryer environments, such as warm desert and semi-desert scrub and grassland (3.A.2) and salt marsh (2.B.7) formations ( Figure 5). This trend is also confirmed by the tree cover results, where P. bohemani showed a higher affinity to greater tree cover values (Figure 6a), while P. pectinicornis was found to prefer more open habitats (Figure 6b).

Environmental Niche Overlap
Within the PCA-Env calibrated using the same set of variables as the ENMs, together with percentage cover of the vegetation formations 1.A.2, 2.A.1 and 3.A.2, the first two principal components (PrinComp1 and PrinComp2) explained 56.8% of overall environmental variability across sub-Saharan Africa (Figure 7a). Percent cover of warm desert and semi-desert scrub and grassland (3.A.2) was strongly associated with the positive semi-axes of both the PrinComps, together with mean temperature of the wettest quarter (bio8) and precipitation seasonality (bio15). Meanwhile, percentage cover of tropical lowland humid forest (1.A.2) contributed to the negative semi-axis of PrinComp1, along with isothermality (bio3) and precipitation of the driest month (bio14). In contrast, percentage cover of tropical lowland grassland, savanna and shrubland (2.A.1) was mainly associated with the negative semiaxis of PrinComp2. In the resulting 2-D environmental space (Figure 7b), Polyclada bohemani and P. pectinicornis showed kernelsmoothed densities of occurrence which almost overlapped, particularly in their core

Environmental Niche Overlap
Within the PCA-Env calibrated using the same set of variables as the ENMs, together with percentage cover of the vegetation formations 1.A.2, 2.A.1 and 3.A.2, the first two principal components (PrinComp1 and PrinComp2) explained 56.8% of overall environmental variability across sub-Saharan Africa (Figure 7a). Percent cover of warm desert and semi-desert scrub and grassland (3.A.2) was strongly associated with the positive semi-axes of both the PrinComps, together with mean temperature of the wettest quarter (bio8) and precipitation seasonality (bio15). Meanwhile, percentage cover of tropical lowland humid forest (1.A.2) contributed to the negative semi-axis of PrinComp1, along with isothermality (bio3) and precipitation of the driest month (bio14). In contrast, percentage cover of tropical lowland grassland, savanna and shrubland (2.A.1) was mainly associated with the negative semiaxis of PrinComp2. In the resulting 2-D environmental space (Figure 7b), Polyclada bohemani and P. pectinicornis showed kernel-smoothed densities of occurrence which almost overlapped, particularly in their core zones (i.e., those with highest 'niche occupancy' values). However, P. pectinicornis niche space extended more than that of P. bohemani towards the positive half of PrinComp1 and PrinComp2 (Figure 7b), thus suggesting that the former species responds positively to higher temperatures during the wettest period and to more extensive cover by xeric vegetation formations. Nonetheless, niche overlap between the two species was quite high when measured with Schoener's D (D = 0.65), and even more so when using the modified Hellinger's distance metric (I = 0.78). Based on the performed niche similarity tests, the alternative hypothesis of niche divergence between the two species could be confidently rejected (p = 0.988), while that of niche conservatism received quite strong support (p = 0.02).
higher temperatures during the wettest period and to more extensive cover by xeric vegetation formations. Nonetheless, niche overlap between the two species was quite high when measured with Schoener's D (D = 0.65), and even more so when using the modified Hellinger's distance metric (I = 0.78). Based on the performed niche similarity tests, the alternative hypothesis of niche divergence between the two species could be confidently rejected (p = 0.988), while that of niche conservatism received quite strong support (p = 0.02).

Discussion
Polyclada bohemani and P. pectinicornis are well characterized from a morphological point of view, in particular as regards the shape of the aedeagus. However, this shows some small differences at the local level that should be better evaluated in the future with the support of molecular assessment. The wide chromatic variability of these two flea beetle species, especially in P. pectinicornis, has often been a cause of confusion in their identification, and until now has prevented correct definition of their distribution in the Afrotropical region. P. bohemani and P. pectinicornis are widely distributed in sub-Saharan Africa, with a large overlapping area in East Africa particularly in Kenya and Tanzania. However, while P. bohemani was found to have areas of high suitability in the more southeastern regions, including Mozambique, Zimbabwe, and the eastern part of the Republic of South Africa, the areas with higher values of suitability for P. pectinicornis are located in the more northern regions, in particular in the belt close to the Sahara. This implies that both species, while showing preferences for savannah and open forest habitats, maintain certain differences in their choice of habitat. P. pectinicornis, in fact, is more frequently present than P. bohemani in sub-desert environments, while the latter can tolerate environments with dense tree cover. That is also confirmed by the environmental niche overlap analysis, which suggests that P. pectinicornis responds more positively than P. bohemani to higher temperatures during the wettest periods, as well as to more extensive xeric vegetation.

Discussion
Polyclada bohemani and P. pectinicornis are well characterized from a morphological point of view, in particular as regards the shape of the aedeagus. However, this shows some small differences at the local level that should be better evaluated in the future with the support of molecular assessment. The wide chromatic variability of these two flea beetle species, especially in P. pectinicornis, has often been a cause of confusion in their identification, and until now has prevented correct definition of their distribution in the Afrotropical region. P. bohemani and P. pectinicornis are widely distributed in sub-Saharan Africa, with a large overlapping area in East Africa particularly in Kenya and Tanzania. However, while P. bohemani was found to have areas of high suitability in the more southeastern regions, including Mozambique, Zimbabwe, and the eastern part of the Republic of South Africa, the areas with higher values of suitability for P. pectinicornis are located in the more northern regions, in particular in the belt close to the Sahara. This implies that both species, while showing preferences for savannah and open forest habitats, maintain certain differences in their choice of habitat. P. pectinicornis, in fact, is more frequently present than P. bohemani in sub-desert environments, while the latter can tolerate environments with dense tree cover. That is also confirmed by the environmental niche overlap analysis, which suggests that P. pectinicornis responds more positively than P. bohemani to higher temperatures during the wettest periods, as well as to more extensive xeric vegetation.