The importance of threatened host plants for arthropod diversity: the fauna associated with dendroid Euphorbia plants endemic to the Canary and Madeira archipelagos

The arthropod fauna associated with seven endemic dendroid Euphorbia was sampled and studied in the Canary and Madeira archipelagos. The stem-diameter of the plants was considered, along with their genetic affinity, habitat and number of localities and islands where present. The arthropod assemblages and richness found on each Euphorbia species were statistically analysed, to determine which variables influenced the survey results. A total of 179 arthropod species were found, identified, and classified into characteristic or accompanying fauna, according to their relationship with the plants and their types of diet. Faunal assemblages and arthropod species richness differed among the Euphorbia species, each thus showing a strong and almost unique host relationship. Species richness increased with the architectural complexity of the host-plant species and number of localities and islands where present. The similarity of faunal assemblages was mainly related to stem diameter. Among diet types, a great number of exclusive and/or endemic taxa associated with this plant genus were xylophages. Our survey revealed that rich arthropod communities are associated with the genus Euphorbia, especially on its rare endangered species, and highlights the need to protect their host plants. This in turn will contribute to the conservation of their arthropod communities and their ecosystem functions.


Introduction
Macaronesia is a biogeographical province located in the eastern Atlantic Ocean between the Iberian Peninsula and West Africa. It includes five archipelagos (Azores, Madeira, Selvagens, Canaries and Cabo Verde), as well as small coastal areas of southern Morocco and the southwest Iberian Peninsula (Fernández-Palacios 2010). All the archipelagos have a volcanic origin, but due to their latitudes there are obviously significant climatic differences between them (Triantis et al. 2010). Other factors such as isolation, wind orientation, geological age and consequently surface area, relief and altitude, have led to a great variety of microclimates and potential niches. These have favoured speciation into a characteristic flora and fauna on each, with a high number of local endemics. In fact, the Madeira and Canary archipelagos are included in the Mediterranean Basin hotspot (Myers et al. 2000). They contribute more than 3500 plant species of its 20,000 total, of which more than 20% are exclusive (123 species to Madeira and 662 to the Canary Islands) (Borges et al. 2008;Gobierno de Canarias 2018).
The genus Euphorbia L. (Euphorbiaceae-spurge family) is one of the most diverse groups of flowering plants on earth and includes at least 2100 species. It presents a high variety of morphotypes, with herbaceous, dendroid, arboreal and cactiform species (Menier 1974;Bramwell and Bramwell 2001;Glimn-Lacy and Kaufman 2006). In the Macaronesian archipelagos there are 14 endemic species of Euphorbia, grouped into two subgenera: Esula and Electronic supplementary material The online version of this article (https ://doi.org/10.1007/s1084 1-020-00261 -z) contains supplementary material, which is available to authorized users.
Some dendroid spurges, like E. lamarckii in the Canary Islands and E. piscatoria in Madeira, are characteristic of the xerophilous and thermosclerophyllous vegetation (Sun et al. 2016) that constitutes a typical feature of the local arid and sub-arid landscapes (Molero and Rovira 1998). However, other endemic species are much less abundant, and even have restricted or threatened populations. Euphorbia bourgeauana is catalogued as 'Vulnerable' at Canarian regional level and 'In danger of extinction' in the Spanish Endangered Species Catalogue, and E. mellifera as 'In danger of extinction' at both levels but without threat status in Madeira. In addition, E. bourgeauana is classified as 'Vulnerable', and E. mellifera and E. piscatoria as 'Least concern' in the IUCN Red List of Threatened Species (Marrero 2011;Rivers et al. 2017;Menezes de Sequeira et al. 2017).
These Euphorbia plants are home to species-rich animal communities, consisting mainly of arthropods. The host specificity of arthropods on plants and their resulting assemblages depend on many factors, such as plant defence compounds, microhabitat structure, suitability for oviposition, shrub size and architectural complexity, and likelihood of mating encounters (Baker et al. 2012;Campos et al. 2006).
Such assemblages tend to be associated with particular parts of the plant (Bernays and Chapman 1994), such as roots, stems, leaves or bark.
Several authors such as Wollaston (1857Wollaston ( , 1864, Martínez de la Escalera (1923), Menier (1974), Jordal and Hewitt (2004), ), Jordal (2006, Stüben (2011) and Stüben and Behne (2015) have previously studied the entomofauna associated with Euphorbia in Macaronesia. However, much of this information is related to Coleoptera and more specifically to the true weevils, family Curculionidae. The present study was focused on determining the arthropod fauna found on different populations of endemic dendroid Euphorbia species in the Canary and Madeira archipelagos, with special interest for the conservation of entire arthropod assemblages. Specifically, the following questions were addressed: (1) Are different arthropod communities associated with each host-plant species? (2) Do arthropod assemblages depend on specific characteristics of each plant species? (3) Which harbour higher diversity, the rare or the abundant plants? Our initial hypotheses to be tested were that: (1) arthropod assemblages are exclusive to each Euphorbia host species, (2) specific characteristics of each plant species influence the arthropod assemblages, and (3) abundant plant species with greater stem diameter and architectural complexity should harbour higher arthropod diversity.

Sampling
Arthropods found on endemic dendroid Euphorbia species (henceforth Euphorbia) were sampled from 2008 to 2014 on three of the western Canary Islands (Tenerife, La Gomera and La Palma) and the Madeira archipelago (Madeira, Porto Santo and Deserta Grande) (see Fig. 1 and Table S1 in Supplementary Material for the list of species, To collect the highest number of arthropod species, different sampling methods were used: (1) direct search on the shrubs, mainly to more effectively detect scarce species in the woody parts (Leather 2005); (2) collecting shoots, stems, and trunks to detect those feeding or living inside plant tissues; (3) adapting the well-established and effective 'beating' technique used for collecting freeliving insects from accessible vegetation (Matthews and Reid 2002). This latter method was however carried out by gently agitating branches onto an umbrella to avoid damaging the plants, especially threatened species. This care also prevented captured specimens being spoiled with the abundant latex exuded by any wounded stems and leaves. The three procedures were attempted at each locality sampled. However, given the differences in abundance and conservation of the Euphorbia species, it was impossible to apply the same protocol to each. For instance, the Canarian E. bravoana and E. mellifera are extremely rare, with a single site for the former and a few for the latter but often with less than five plants. The specimens collected were preserved either in ethanol or as dry collections stored in the zoological collection of the Department of Animal Biology, Edaphology and Geology at the University of La Laguna (DZUL).

Species identification
The collected arthropods were identified using available identification keys and/or original descriptions of species. In addition, some individuals were directly compared to preserved DZUL collection specimens in order to check the identifications. We also searched the bibliography and revised the unpublished data on Euphorbia in the arthropod collection of the Department of Animal Biology, Edaphology and Geology (University of La Laguna). The arthropods were classified according to feeding behaviour (see Fig. 3) and as 'characteristic' or 'accompanying' fauna, depending on whether they have a direct trophic relationship with the host plant, or not. This was based on reviewing the information published on such relationships, consulting expert entomologists on certain taxonomic groups when necessary (see Supplementary Tables S2 and S3, and Acknowledgements).

Data analysis
A similarity matrix using the Sørensen index was obtained (hereafter, faunal assemblage) from the presence-absence data for the various Euphorbia. A matrix was also built from differences in the number of species (richness), comparing Euphorbia species in pairs. A correlation matrix was elaborated between these two dependent variables (faunal assemblage and richness) and five independent variables, in order to explore the relationship between the two kinds of variables. The following independent variables were analysed for each species: (1) stem diameter (weighted from 1 to 4 in increasing order, see Table 1), (2) number of islands where they were present, (3) number of sampled localities, (4) genetic affinity by the number of nodes separating two Euphorbia species, fide Riina et al. (2013) and Barres et al. (2017), and (5) type of habitat (identical, similar or completely different habitats weighted from 1 to 3, respectively). To analyse the effects on the independent variables, a GLM was applied to the dataset, following a Gaussian distribution for differences in faunal assemblages and a Poisson distribution for richness.
Finally, to visualise the relationships among the faunal assemblages found on Euphorbia species, Nonmetric Multidimensional Scaling (NMDS) based on dissimilarities calculated with the Bray-Curtis index was carried out. The scores for each plant species were correlated with the independent variables to infer the factors explaining the ordination. All statistical analyses were conducted in R version 3.3.1 (R Core Team 2017), taking into account only the characteristic species.

Arthropod biodiversity
A total of 179 species belonging to 81 families and 15 arthropod orders were obtained (see Supplementary Tables  S2 and S3 for the list of characteristic and accompanying species, respectively). The orders with the highest number of families and species were Coleoptera, Araneae, Diptera and Hemiptera, respectively (Table 2). More than 50% of the species found were endemic to their archipelago, 50% of these to the Canaries and 30% to Madeira.
According to diet, the most abundant arthropod groups were predators (52) and xylophages (43), followed by four others with similar diversity: saprophages, foliophages,  fungivores and sap-suckers (Fig. 3). Within characteristic species, the assemblages mainly consisted of xylophages (39) followed by fungivores (11), predators (7), saproxylophages (6), foliophages (6) and parasitoids (6). The diversity pattern of the arthropod species number per Euphorbia was similar for characteristic and accompanying fauna. However, the percentage of endemics was slightly higher than in the total pool (59.5% vs. 52.5%, respectively). The faunal composition varied between diet types, since some trophic groups had a higher percentage of characteristic species, such as xylophages (90.7%) or fungivores (78.6%). In other groups there were fewer, such as sap-suckers (30.7%) and predators (13.5%). Focusing on xylophages only, the greatest diversity was on E. mellifera, E. bourgeauana and E. lamarckii, with 13 species each. However, endemicity among xylophages was higher than in characteristic species, reaching 74.4%.

Analyses of Euphorbia species relationships
A negative correlation was found between faunal assemblage and all the independent variables. This means that the similarity among arthropod assemblages between Euphorbia species decreases when the differences between any of these variables increase. Except for number of localities, all tested variables explained the variation in faunal assemblage. A different pattern was found with the number of species, which showed a positive correlation between richness and all the independent variables, except for habitat (Table 3). Examining the relationships between the seven Euphorbia species, E. bravoana, E. berthelotii and E. atropurpurea are closely related, and also cluster with the slightly more distant E. lamarckii and E. bourgeauana. Arthropod assemblages were more similar within these latter two groups. In contrast, E. mellifera and E. piscatoria are the farthest away, being the most dissimilar from the rest (Fig. 4). This pattern was supported statistically as the data fit was appropriate (stress = 0.0325). A strong relationship was found between the first axis and the number of islands (0.8633), and between the second axis and the habitat category (0.8987).

Discussion
A study of the whole arthropod assemblage associated with a group of plants can reveal hidden richness. We found two types of arthropods: (1) those specifically linked to these plants that use them either as a trophic resource or as habitat for their life-cycle, and (2) others not strictly associated that may occasionally use them to locate potential prey or just to rest. Characteristic Euphorbia fauna need fresh, dry or decomposed plant material, either to satisfy their nutritional requirements or to establish nests and lay their eggs (Jordal 2006). The xylophagous species that particularly stand out are Cryptorhynchinae (Acalles and related genera), Cossoninae weevils (Mesites, Rhopalomesites, Pselactus, Caulotrupis and Pseudoploeophagus) and Aphanarthrum and Liparthrum bark beetles (Curculionidae, Scolytinae), since their whole biological cycle occurs in the plant . Other noteworthy xylophages are the larvae of longhorn beetles Deroplia albida and Lepromoris gibba (Coleoptera, Cerambycidae) (Sama 1996;Krátký and Aguiar 2019) and the false blister beetles Alloxantha seidlitzi and Nacerdochroa concolor (Coleoptera, Oedemeridae) (Kubisz and Borowski 1999), which are common in dead spurge stems. Among fungivores, the genus Tarphius (Coleoptera, Zopheridae) is noteworthy. It has five endemic flightless species with larvae developing in decomposed wood (Emerson and Oromí 2005). The family Latridiidae is also well represented by four species belonging to genera Corticarina, Corticaria and Metophthalmus. Regarding sap-sucking species, those of the genus Acrosternum (Hemiptera, Pentatomidae) that inhabit the Canaries are highly specialised and live throughout their life-cycle on Euphorbia species. The  (Belokobylskij and Zaldívar-Riverón 2014). They thus have great importance in the dynamics of these arthropod assemblages, since most beetles associated with Euphorbia are xylobionts. Some insects are strictly foliophagous with spurge leaves as their main trophic resources, such as the the Canary endemic genus Acrostira (Orthoptera, Pamphagidae) (López et al. 2007;Hernández-Teixidor et al. 2014), Aphthona leaf beetles (Coleoptera, Chrysomelidae) (Fornasari 1996) or the larvae of the hawkmoth Hyles tithymali (Lepidoptera, Sphingidae) (Gil-T 2010). Other species are root consumers during their larval stage but the adults feed on leaves, like Laparocerus distortus (Coleoptera, Curculionidae) (Machado et al. 2008). We thus report 61 newly recorded species that should be considered as characteristic Euphorbia fauna. Pooled with the 49 already described in the literature, these raise to 110 the number in Madeira and the Canaries. This is relatively high when compared to other studies. For instance, Mifsud et al. (2012) reported 33 characteristic arthropods associated with the genus Ficus in the Maltese islands, while Mecke et al. (2001) found 45 species related with Araucaria angustifolia in Brazil. Despite a search, we found no similar studies about characteristic arthropods in the Macaronesian archipelagos. In the Azores, long-standing monitoring efforts have explored their arthropod biodiversity. However, these did not distinguish between characteristic and accompanying fauna. Nunes et al. (2015) reported 161 spp. associated with Juniperus brevifolia Ribeiro et al. (2005). reported 65 herbivore insects for J. brevifolia and 53 for Erica azorica, among others, from a total of seven islands, and Rego et al. (2019) 35 for J. brevifolia and Laurus azorica, 34 for E. azorica, 30 for Ilex perado azorica and 26 for Vaccinium cylindraceum on Terceira. Traditionally, studies focusing on arthropod fauna related to a single host-plant species, genus or family (e.g. Nunes et al. 2015;Sasa and Samways 2015) combine both characteristic and accompanying fauna in their analyses. However, accompanying fauna may include species actually unassociated with the host plant but give rise to a stochastic record that leads to misinterpretated statistical results. Moreover, since such species have no trophic relationship with the host plant, they can move to other plant species within the same ecosystem. In formulating conservation plans, we therefore recommend discerning between characteristic and accompanying fauna. This would more closely estimate the real number of species that can be strongly affected by a decline or loss of their host plant.
The differences in distribution area, number of specimens per locality, stage of maturity and conservation status of these Euphorbia made it difficult to obtain a homogeneous sample. However, the data here presented are the most reliable information regarding their real arthropod assemblage. The less similar the host-plant characteristics, the more different was the richness pattern. In agreement with Campos et al. (2006), we found that spurges with greater stem diameter and architectural complexity, such as E. mellifera, E. bourgeauana and E. lamarckii, harboured a greater number of species. Conversely, species with smaller stem diameter  BERT E. berthelotii, BRAV E. bravoana, BOUR E. bourgeauana, LAMA E. lamarckii, MELL E. mellifera, PISC E. piscatoria (E. bravoana and E. atropurpurea) displayed less diversity. As expected, arthropod richness increased with the number of islands and localities in which the plant species was present, due to local faunal singularities. The genetic differences between spurge species also explained the different arthropod species found, since the subsections and species complexes to which they belong are also based on ecological and morphological characters [see Riina et al. (2013) and Barres et al. (2017)]. However, the high number of species detected in Euphorbia species from different habitats (E. mellifera, E. bourgeauana and E. lamarckii) made the habitat a non-determinant factor in the richness values. Greater faunal dissimilarity, with more different host-plants, is most prominent in species with more distant stem thickness measurements, since xylophages are very specific to their host plant. Exclusivity of arthropods to certain islands influenced the similarity of the assemblages from one locality to another. Species with highly restricted distributions have fewer possibilities to find another host plant. Not surprisingly, the faunal assemblage tended to be more closely comparable between similar than different habitats, as found by Gessé et al. (2014). All these results show that Euphorbia species with restricted distributions deserve conservation programmes, owing to their unique associated faunal assemblage with such a high proportion of endemic and characteristic arthropod species.
According to our data, E. mellifera and E. bourgeauana are hosts to the highest numbers of species overall and of endemics, together with E. lamarckii. This suggests they act as biodiversity reservoirs, since their arthropod assemblages include species unique to these endangered plants and their archipelagos. In fact, both species are the feeding source of Rhopalomesites proximus (Wollaston 1861), a threatened beetle catalogued as 'In danger of extinction'. Euphorbia bourgeauana is also a trophic resource for the endemic grasshopper Acrostira bellamyi (Uvarov 1922) (Hernández-Teixidor et al. 2014), catalogued as 'Critically endangered' by the IUCN. It is notable that endemic xylophages like Pseudophloeophagus tenax (Wollaston 1854) and those of the genus Aphanarthrum, as well as Leipaspis lauricola gomerensis Plata and Prendes, 1981-a predator of xylophagous larvae-were all detected only on E. mellifera. In addition, other species that spend their larval stage inside wood were only found in E. bourgeauana, such as Alloxantha seidlitzi Svihla, 1988 andSolva palmensis Báez, 1988. The parasitoid of this type of larvae, Spathius canariensis Hedqvist, 1976, was also present on both plants. Variation in arthropod assemblage composition is strongly influenced by host-plant identity or composition, particularly for phytophagous insects (Frenzel and Brandl 2001;Ødegaard et al. 2005). We found that xylophages are highly specialised to the species on which they develop. This pattern is quite common in xylobionts and, as they cannot move to other plant species, the conservation of their host is vital for them (Krivosheina 2016). This conclusion of our assessment is another reason to protect Euphorbia species.
Population losses will directly affect not just host survival but also each of their arthropod assemblages as a whole. Due to their abundance, arthropods play a major role in ecosystem dynamics and are good indicators for assigning conservation priorities based on monitoring programmes (Kremen et al. 1993). Our data on arthropod fauna indeed reveal an often hidden richness that reinforces the need to conserve these endemic spurges in Madeira and the Canaries, especially those officially under protection. Increased knowledge of associated fauna can thus serve to strategically redirect conservation efforts to protect both plants and animals and their respective ecosystem functions. Many species listed as threatened do not yet have recovery or management plans and actions are insufficient, due to lack of funding. In fact, none of these Macaronesian Euphorbia species have an approved plan. Budgets are limited, so it is essential to direct existing resources to key species such as these, given their impressive faunal assemblages. Through protecting these spurge populations and increasing their health status and number, the faunal populations linked to them will almost certainly be positively affected.
Ethical approval The authors declare that they have complied with ethical standards.