Exaptation and vulnerability to introduced mammal herbivores on Balearic endemic flora

Journal of Biogeography. 2023;00:1–12. | 1wileyonlinelibrary.com/journal/jbi Received: 25 May 2022 | Revised: 31 January 2023 | Accepted: 19 February 2023 DOI: 10.1111/jbi.14602 RESEARCH ARTICLE Exaptation and vulnerability to introduced mammal herbivores on Balearic endemic flora Miquel Capó1,2 | Rocío PérezBarrales3,4 | Joana Cursach1 | Jaume Garrido1 | Elena Baraza1 | Juan Rita1 This is an open access article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License, which permits use and distribution in any medium, provided the original work is properly cited, the use is non-commercial and no modifications or adaptations are made. © 2023 The Authors. Journal of Biogeography published by John Wiley & Sons Ltd. 1Research Group on Plant Biology under Mediterranean Conditions, Biology Department, University of Balearic Islands, Palma, Spain 2Departamento de Sistemas y Recursos Naturales, Universidad Politécnica de Madrid, Madrid, Spain 3School of Biological Sciences, King Henry Building, University of Portsmouth, Portsmouth, UK 4Botany Department, University of Granada, Granada, Spain Correspondence Miquel Capó, Departamento de Sistemas y Recursos Naturales, Universidad Politécnica de Madrid, Madrid, Spain. Email: miquel.c[email protected] Funding information Direcció General de Política Universitària i Recerca (Govern de les Illes Balears), Grant/Award Number: FPI/1925/2016; Ministerio de Economía, Industria y Competitividad, Gobierno de España, Grant/Award Number: CGL201570449R; European Union (NextGenerationEU) Handling Editor: Sandra Nogue Abstract Aim: Introduced mammal herbivores are predicted to negatively affect insular flora. However, disentangling which particular traits (1) developed from exaptations and (2) are functional to avoid herbivory remains mainly unknown. This study aims to assess if the flora of continental islands with historic native herbivores are exapted to the introduction of new mammal herbivores and to predict the potential vulnerability of endemic species from islands where mammal herbivores have not been introduced. Location: Balearic Islands. Taxon: 96 Balearic endemic plant species. Methods: We investigated whether the endemic flora on continental islands maintains functional traits that resist introduced mammal herbivores by analysing the chemical and morphological traits related to plant resistance of five individuals for each of 98 species. Also, we measured plantsize variables to assess plant escape strategies. Overall, we combined these traits with the accessibility to goats. Predictive models were generated for species that inhabit islands where goats have not been introduced to assess their potential vulnerability. Results: Endemic species may defend against new herbivores (e.g. goats) if they contain highly toxic compounds (alkaloids, glycosides, coumarins), spinescent and urticating structures, or specific plant architecture (low plant size, high specific leaf area). If such traits are absent, the species may become extinct— unless they inhabit areas inaccessible to goats. On continental islands, some endemic species are expected to resist the introduction of herbivores, while others may be significantly affected. Main Conclusions: From the ancient connection with the mainland, exaptations may allow the plants to resist the presence of introduced herbivores. However, nonexapted species could be threatened by the introduction of nonnative ungulates. KEYWORDS defence, escape, exaptation, insularity, plantherbivore interactions, resistance 2 | CAPÓ et al. 1 | INTRODUCTION Plantherbivore interactions represent an important ecological driver for the evolution of defence strategies to prevent, ameliorate and survive herbivory (Agrawal & Fishbein, 2006). Plant defence strategies against herbivory can be classified into three main groups: resistance, tolerance and escape (Mauricio, 2000; Strauss & Agrawal, 1999). Resistance strategies include the production of toxic compounds that make plants unpalatable, namely phenolic compounds (Baraza et al., 2004), condensed tannins (Cooper & OwenSmith, 1985), flavonoids (Pellissier, 2013), saponins (Ishaaya et al., 1969), triterpenoids (Zwenger & Basu, 2008), glycosides (Diner et al., 2009), alkaloids (Coley, 1987) and coumarins (Howery et al., 2016). Anatomical structures, including spines and high pilosity, provide another layer of defence against herbivory, primarily against mammals (Mauricio, 2000; Milton, 1991). Tolerance includes a wide variety of physiological mechanisms that act to rapidly compensate for the loss of biomass caused by predation (Agrawal, 1998; Rasmann et al., 2009). Finally, the escape strategy consists of avoidance of herbivory, for example, by reducing plant size (Alonso & Herrera, 1996; Gange & Brown, 1989) or inhabiting areas inaccessible to herbivores, such as cliffs or mountain walls (Pisanu et al., 2012). When herbivory disappears, either herbivores become extinct or plants colonize new environments free of herbivores, the selective pressure linked to resistance may also decrease or disappear, and allow plants to reallocate resources to other functions such as growth, competitiveness or reproductive output (HullSanders et al., 2007; Keane & Crawley, 2002; Pardo et al., 2004). Alternatively, resistance traits can remain in herbivorefree environments if the traits confer a complementary function or appear genetically fixed within the taxonomic group (Wink, 2008), as occurs for structural antiherbivore traits (Hanley et al., 2007). The hypothesis that resistance traits could disappear in herbivorefree environments has been tested in oceanic island ecosystems, where plant species might lack adaptive defensive traits due to the absence of mammalian herbivores, as exemplified by the flora on volcanic islands such as the Hawaii archipelago (Barton, 2016) or the Canary Islands (Cubas et al., 2019; Nogales et al., 2006). The introduction of exotic herbivores onto oceanic islands leads to severe perturbation to plant communities (Cubas et al., 2019; Nogales et al., 2006) and drives their extinction, and is currently one of the major threats to endemic species (Bowen & Van Vuren, 1997; Campbell et al., 2004; Médail, 2017). In contrast, less is known about the impact of herbivory on the flora of continental islands (Gizicki et al., 2018; Moreira et al., 2021). The plant communities on continental islands are mostly derived from the original flora present on the mainland to which the islands were previously connected. Hence, the flora of continental islands can possess defence traits that evolved in the context of the plantherbivore interactions present within the original continental flora. While the presence of herbivory defence traits may intuitively imply that plant communities should be resilient to herbivory, evidence of the negative impacts associated with the introduction of invasive mammals, such as goats, on continental islands worldwide is increasing (Carrion et al., 2011; Moreira et al., 2021), and eradication of introduced herbivores have been performed on several continental islands (Capizzi, 2020). More generally, biological invasions are one of the main causes of biodiversity loss worldwide (De Vos et al., 1956; McNeely et al., 2001; Médail, 2017), with exotic herbivores responsible for the extinction of endemic flora on island ecosystems (Bowen & Van Vuren, 1997; Campbell et al., 2004; Médail, 2017). The Balearic Islands Archipelago, located in the Western Mediterranean Basin, originated by fragmentation of the eastern coast of the Iberian Peninsula during the Oligocene, 25 Mya (Rosenbaum & Lister, 2002), and provides an example of continental islands inhabited by ancient mammal herbivores that evolved under insular conditions. The most recent connection between the islands and the Iberian Peninsula occurred during the Messinian Salt Crisis 5.9– 5.3 Mya when the last terrestrial vertebrate paleofauna arrived at Mallorca. During the first Quaternary glaciations, about 1– 2 Mya, the endemic bovid Myotragus present in Mallorca colonized the neighbouring island of Menorca (collectively known as the Gymnesic Islands). This bovid evolved under insular conditions, giving rise to the species Myotragus balearicus (Bover et al., 2008; Palombo et al., 2013; Winkler et al., 2013). However, there is no evidence of the presence of Myotragus in the archaeological records of Eivissa and Formentera (collectively known as the Pityusic Islands), although the fossil records show the presence of the Neogene bovid Ebusia moralesi until the Quaternary, when it disappeared and has not been detected during this period (MoyàSolà et al., 2022). The extinction of the native mammalian herbivore on the Gymnesic Islands coincided with the colonization by humans (Palombo et al., 2013) and the introduction of new fauna of mammalian herbivores, such as goats, rats and rabbits (4300– 4050 years BP; Muñoz et al., 2019; Seguí et al., 2005). The available data suggests that the introduced goats were managed and remained under human control through hunting or breeding purposes (Mayol et al., 2017). More recently, coinciding with landuse abandonment in the last five decades, the feral goat populations have exploded exponentially. Feral goat populations are mainly distributed across the mountains of Mallorca and represent a serious threat to the maintenance and conservation of native vegetation (Capó, Engelbrecht, et al., 2021; Mayol et al., 2017). Thus, the environmental authorities have carried out several programmes to fence areas to protect the flora threatened by goats (Moragues et al., 2015). The number of herds of feral goats on Menorca has also increased in the last decade, although their impact on the native vegetation is less harmful than on Mallorca (Menorca Insular Council, pers. com.). Fortunately, feral goats are not found freely in natural areas of Eivissa and Formentera and only managed individuals are found in farms or private proprieties (Capó et al., 2022). The paleontological history, introduction of feral goats and recent abandonment of the countryside of the Balearic Archipelago represent a unique opportunity to disentangle the interplay between the ecology and evolution of resistance traits to herbivory based on the knowledge already acquired about the endemic flora (Rita & 13652699, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jbi.14602 by Universidad De Granada, Wiley Online Library on [18/04/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License | 3 CAPÓ et al. Payeras, 2006; Roselló & Sáez, 2000, 2008, 2017; Sáez et al., 2011, 2017) and the exhaustive archaeological work to describe the native paleofauna of the entire archipelago (Bover et al., 2008, 2016, 2019; Palombo et al., 2013). Overall, the Balearic Archipelago represents an excellent study case because (i) it is a hotspot of insular endemic flora, (ii) the fossil record confirms the evolution of native mammalian herbivores after the formation of the archipelago, and (iii) the large population of feral goats is currently threatening the native flora within the natural landscapes on some of the islands. Considering this scenario, the objectives of the present study are to (i) evaluate if the flora of this archipelago, which has an evolutionary history with mammal herbivores, are exapted to the introduction of new herbivores— or whether it could be endangered— as reported for endemic species on other islands and (ii) use the data generated to establish a predictive model to evaluate the potential vulnerability of endemic species from islands without introduced ungulates. Our hypotheses are: (i) endemic species inhabiting areas accessible to feral goats will exhibit resistance traits against herbivory, (ii) resistance traits can be predicted based on the phylogenetic and ecological context, and (iii) islands not invaded by introduced ungulates might present flora without resistance traits against herbivory in areas potentially accessible to feral goats. Overall, by using the Balearic Archipelago as a representative case study, we hope to define a methodological and analytical approach to identify the levels to which the endemic flora of continental islands harbour mechanisms that may protect against new herbivores, and this information may help to inform the design of conservation and management strategies for natural resources. 2 | MATERIALS AND METHODS 2.1 | Study system The Balearic Archipelago (Spain) is located in the central part of the Western Mediterranean Basin (Figure 1). The Archipelago is formed of four main islands with varied topographical and geological characteristics, covering areas from 3640 km2 (Mallorca) to 83 km2 (Formentera) and altitude ranges from 0 to 1445 m a.s.l. (Mallorca). The climate is typically Mediterranean, characterized by a dry summer season and rainfall events in spring and autumn (Homar et al., 2009). Owing to the variability in the geography and Mediterranean climate, the Archipelago contains a wide diversity of environments, from salt marshes or temporary ponds to woodlands and mountain gullies, and is home to 1551 plant species, of which 20 are endemic to the Tyrrhenian Islands (including Corsica and Sardinia) and 140 are exclusive to the Balearic Archipelago (Rita & Payeras, 2006; Sáez et al., 2013). The present study assessed 96 Thyrrenian or Balearic endemic taxa present in the Balearic Archipelago, representing 76 genera and 29 families that occur in various habitats (Table S1). For each endemic taxa, samples from five individuals (n = 480 samples in total) were collected during the springs of 2017 and 2018 to generate phenotypic data on herbivory resistance and escape traits (see below). To assess the variation of herbivory impact on endemic species depending on their ecological distribution, the potential accessibility of herbivores was described based on the environment that the endemic flora occupies. Specifically, plant species that inhabit cliffs and rocky mountain environments (rupicolous communities) were considered inaccessible, whereas species from shrublands and caespitose communities were considered accessible to mammalian herbivores. 2.2 | Resistance and escape traits against herbivory in endemic flora Two types of putative defence traits were assessed: chemical compounds and anatomic structures. We quantified the concentrations of chemical compounds commonly known to provide chemical protection against herbivory; specifically, total phenolic compounds, condensed tannins, triterpenoids, saponins, flavonoids, alkaloids, glycosides and coumarins. The chemical analyses are described in Supporting Information S1. Briefly, mature leaves were collected (up to 2 g dry weight), except for taxa with small plant sizes or those under threat, for which up to 100 mg dry weight was sampled. The leaf materials of all sampled accessions were stored at 40°C for 1 month to inhibit enzyme activity and avoid the hightemperature volatilization of polyphenolic compounds (Baraza et al., 2009). Samples were ground to a particle size of less than 1 mm and conserved in airtight tubes for chemical analysis. Total phenolic compounds were analysed following the Folin– Ciocalteu method (Baraza et al., 2009) with some modifications, and the final readings were performed using a Multiscan Sky Microplate Spectrophotometer (ThermoFisher Scientific Inc.). Tannic acid was used as a standard reference for total phenolic compounds and all readings were relativized to tannic acid equivalents. Tannins were quantified using the FIGURE 1 Map of the Balearic Islands archipelago. The islands are Mallorca and Menorca (Gymnesians), Eivissa and Formentera (Pityuses). Projection used: ETRS89/UTM zone 31N. Raster altitudinal layer MDT200 has been acquired from the national institute of geographic information (CNIG) from Spanish Government by free licence obtained in www.centr odede scarg as.cnig.es/. 13652699, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jbi.14602 by Universidad De Granada, Wiley Online Library on [18/04/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 4 | CAPÓ et al. proanthocyanidin assay (TomásBarberán, 1995) using cyanidin as a standard for condensed tannins; all readings are presented as cyanidin equivalents (CE). Saponins, flavonoids and triterpenoids were quantified using protocols obtained from AbdulAzeez et al. (2018), Sharma and Janmeda (2017) and Liang et al. (2014), respectively. As sufficient plant material was not available to carry out exhaustive quantification for all species, because most endemic species are small shrubs or herbs, the presence of defence compounds that are usually phylogenetically fixed— such as glycosides, alkaloids and coumarins— were determined using the literature on the presence of these compounds in closely related taxa (Table S2). The anatomical structures assessed included spinescence and pilosity. Spinescence was classified as spines covering the entire individual (e.g. as for Astragalus balearicus, Figure S1A); less abundant, but evident spines (i.e. Rhamnus bourgeana, Figure S1B); or as hairs with urticating proprieties (i.e. Urtica bianorii, Figure S1C). Pilosity was defined as a high density of hairs covering the leaves (i.e. Helichrysum crassifolium, Figure S1D). The plantsize escape traits assessed included average plant size, total leaf area and specific leaf area (SLA). The average plant size of each species was obtained from the literature (Castroviejo, 1986), total leaf area was determined from images of fully expanded leaves of the sampled individuals using ImageJ software (Abràmoff et al., 2004) and SLA (cm2/mg) was calculated by dividing the total leaf area by the dry weight of the same sample. 2.3 | Phylogenetic relationships between resistanceescape traits and the effects of goat herbivory As some resistance and escape traits were described using data from closely related species and some plant families appear overrepresented in the endemic flora, we analysed the raw data and phylogenetically independent contrasted data using pairwise comparisons between traits, according to the methods described by Agrawal and Fishbein (2006). We used the sequences of internally transcribed spacers and rDNA 5.8S from the GenBank® library at the genus level (Supporting Information S2). Once the phylogenetic reconstruction was complete, phylogenetic independent contrasts (PICs) were calculated using the Felsenstein's method (Felsenstein, 1985), as implemented in the ape v5.3 package (Paradis & Schliep, 2019) of R software v3.6.1 (R Core Team, 2022). Pearson correlation indexes were subsequently obtained using the ‘cor’ and ‘cor.mtest’ functions of the ‘corrgram’ package (Wright, 2018) for the raw and PIC data. A generalized linear model (GLM) was generated to investigate the potential functions of putative resistance and escape traits assessed as strategies to protect against herbivory. The accessibility of herbivores (1/0) was included as a response variable and modelled with a binomial distribution against all other previously described chemical and morphological traits as predictor variables. The analyses were conducted using data derived from the Gymnesic Islands (Mallorca and Menorca), where feral goats are present. Then, model selection was implemented to determine which traits best predict the accessibility of goats, and the bestfitting model was selected by applying the Akaike information criterion (AIC) using the ‘dredge’ function of the ‘MuMIn’ R package (Barton, 2020). The significance of the differences between resistance and escape traits depending on accessibility to goats were assessed in the selected model using the ‘Anova’ function of the R software package ‘car’. Resistanceescape traits were included in the bestfitting model as predictor variables. Traits that correlated significantly with resistanceescape traits in the PIC analysis were included in the principal component analysis (PCA), in which the accessibility of the plants to goats was assessed as a qualitative variable. PCA was performed using the ‘PCA’ function of the ‘FactoMineR’ package in R (Le et al., 2008). To evaluate the overall resistanceescape complex, we calculated trait diversity by considering Shannon's diversity (H′) to assess the balance between the richness and abundance of each trait, as described in Morris et al. (2014) for chemical diversity. Calculations were performed using the ‘vegan’ package v2.56 in R (Oksanen et al., 2008). 2.4 | Estimation of the potential vulnerability of plant species endemic to islands unaffected by herbivory Based on the bestfitting model obtained in the previous section, we analysed the potential vulnerability of endemic species to Eivissa and Formentera, where feral goats are not present. We extracted the regression equation of the model constructed for endemic species to Mallorca and Menorca. The model was adjusted to a binomial response variable to distinguish the probability of coexisting with goats. This equation was then implemented for endemic species to Eivissa and Formentera to evaluate the vulnerability of each species to the putative introduction of feral goats. Then, the output was compared with the actual distribution of the species on the islands at present. 3 | RESULTS 3.1 | Resistance and escape traits against herbivory among endemic flora Eightyfour species (87.5%) were collected from the Gymnesians and 12 species (12.5%) from the Pityuses. Of the species sampled on the Gymnesic Islands— where feral goats are present— 62 occur in areas accessible to goats and 22 were only located in inaccessible habitats. In the whole archipelago, the distribution of endemic species in the three main habitats was not equal: 47% of species are based in shrublands, 30% in caespitose communities and 23% in rupicolous communities. Of 96 endemic species, 95— all except Primula acaulis subsp. balearica— were found to contain phenolic compounds at 13652699, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jbi.14602 by Universidad De Granada, Wiley Online Library on [18/04/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License | 5 CAPÓ et al. concentrations ranging from 0.01 to 2.25 ng/mg and 92 species contained condensed tannins at concentrations that varied from 0.01 to 3.87 ng/mg. Quantitative analysis showed that 28.1% (27/96) of the species sampled contained saponins, 52.1% (50/96) contained flavonoids, and 26% (25/96) contained triterpenoids. The literature indicated that 17.7% (17/96) of endemic plants have closely related species with alkaloids; 18.7% (18/96) have glycosides; and 6.2% (6/96) have coumarins. In terms of anatomical traits, 10.4% (10/96) of the endemic species are spinescent or stinging and 18.7% (18/96) exhibited pilosity. The foliar areas ranged from 0.00001 to 648.23 cm2, SLA varied from 0.00001 to 1543.80 cm2/g and plant size ranged from 3 to 150 cm. 3.2 | Phylogenetic relationships between effects of goat herbivory and resistanceescape traits In PIC analysis (Table S3), the presence of alkaloids was positively correlated with the presence of glycosides (r = 0.48, p < 0.001), and both the presence of alkaloids and glycosides correlated with the absence of coumarins (r = −0.46, p < 0.001 and r = −0.39, p < 0.001 respectively). Additionally, the presence of alkaloids and glycosides both correlated positively with the foliar area (r = 0.42, p < 0.001 for both). Plant size correlated significantly with SLA in both the raw (r = −40, p < 0.001) and corrected data (r = −0.49), and also with the leaf area after applying PIC (r = 0.61, p < 0.001). While significant correlations between flavonoids and pilosity and between SLA and plant size were observed in the raw data, these correlations disappeared after correcting the species using PIC. Overall, 25% of endemic species from the Gymnesic Islands exhibited resistance to herbivory through chemical compounds and only 13.10% were spinescent (Capó et al., 2023). After fitting a model that included all the measured putative resistance and escape traits (setting coexistence with goats as the response variable) and using model selection, the variables selected for the bestfitting model were glycosides, coumarins, spinescence and plant size (Table 1). Moreover, although there were no significant differences in either alkaloids or SLA between species that inhabit areas accessible or inaccessible to goats, these factors were considered in the subsequent analysis as they are correlated with glycosides and plant size, respectively, as reported in Table S3. Principal component analysis was used to infer how the combination of traits explains the development of resistanceescape strategies among endemic species. The principal components separated groups of species based on the presence of resistance traits— as well as plant size and SLA, which are both measures of escape strategies (Figure 2). PC1 separated species that present resistance traits from species with escape traits, whereas PC2 separated the plants that exhibit resistance traits into species with anatomical resistance traits (spines) and chemical resistance traits (alkaloids, glycosides and/or coumarins). Species that lacked resistance traits were accessible to herbivores if the plant size was small and negatively correlated with high SLA, indicating a plantsize escape strategy. In contrast, species that inhabit areas inaccessible to herbivores exhibited an absence of chemical and anatomic structures and had medium plant size and SLA (Table 1). 3.3 | Estimating the potential vulnerability of endemic plant species to islands without introduced herbivores Principal component analysis separated the sampled species into four major groups, two of which corresponded to shrublands (chemical and anatomical resistance), one to caespitose communities (plantsize escape) and one to rupicolous communities (ecological inaccessibility). Overall, the flora of islands without feral goats showed lower trait diversity than the species from islands where goats are present (Figure 3). However, endemic species that occurred in accessible habitats on islands with goats had a larger plant size and displayed higher trait diversity than the groups of endemic species to islands without feral goats. Interestingly, the endemic flora to islands without feral goats lacked spines and coumarins. After considering all studied parameters and selecting the bestfitting model (Table 2), we extracted the regression equation obtained using the estimates of the model adjusted to a binomial response variable. Thus, the regression equation was: where gly = glycosides, cou = coumarins, spi = spinescence and size = plant size. Assuming a future scenario of the introduction of goats, this regression equation was used to predict their coexistence with endemic species to islands where goats are currently absent (Table 3). Based Coexistence =inv. logit(2.908+26.987×gly+22.276×cou +27.517spi−0.085size), TABLE 1 Species included in each group and proportions and values of resistance traits obtained by principal component analysis (PCA). PCA group Species included (%) Alkaloids (%) Glycosides (%) Coumarins (%) Spines (%) Plant size (cm) Specific leaf area (cm2/g) Chemical resistance 25.00 57.14 100.00 100.00 0.00 67.10 141.65 Anatomical resistance 13.10 7.14 0.00 0.00 100.00 57.27 133.63 Plantsize escape 34.52 28.57 0.00 0.00 0.00 5.00 564.48 No resistanceescape 27.38 7.14 0.00 0.00 0.00 51.52 134.46 13652699, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jbi.14602 by Universidad De Granada, Wiley Online Library on [18/04/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 6 | CAPÓ et al. on the regression model and according to their presence of herbivory resistance traits, Santolina vedranensis and Teucrium cossonii subsp. punicum were predicted to be able to coexist with herbivores as these genera of plants contain chemical protective compounds, such as glycosides or coumarins (De Marino et al., 2012; Oganesyan et al., 1991; Silván et al., 1996). Species that were predicted to not be able to coexist with introduced goats include Genista dorycnifolia and Biscutella ebusitana, which lack any resistance traits against herbivory and exhibit a large plant size. Additionally, Thymus richardii ssp. ebusitanus, which has a small plant size and high SLA, was predicted to have a high probability of coexisting with herbivores through an escape strategy. However, overall, 72.7% of species studied on the Pityuses have a 25% or lower probability of coexisting with feral goats. However, despite having resistance traits to facing herbivory, whether anatomical or chemical, other ecological factors can influence the risk of being predated by ungulates. For instance, in herbivore highdensity scenarios or in scarceresources environment, goats can feed on defended plants, as observed in Euphorbia dendroides (Capó, Engelbrecht, et al., 2021) and can be a serious threat to endemic species, as observed in previous studies on islets of the Balearic Islands (Capó et al., 2022). 4 | DISCUSSION 4.1 | Resistanceescape traits and their relationships based on phylogeneticindependent contrasts The endemic flora of the Balearic Archipelago contains a remarkable proportion of species with resistance and escape traits, with varied structures and chemical compounds that protect against herbivory. The endemic flora of the Gymnesic Islands (Mallorca and Menorca) exhibit two main functional strategies against herbivory: resistance, through highly toxic compounds or anatomic structures, and escape, through small plant size. Species without these traits would probably have otherwise become extinct, or their distribution reduced FIGURE 2 Principal component analysis of the resistance and escape traits of species endemic to the Gymnesic Islands selected in the bestfitting model of putative resistance traits and other correlated parameters after PIC correction. PC1 was mainly constructed of glycosides (24%), specific leaf area (28%) and plant size (38%), while PC2 was mainly constructed of alkaloids (29%), glycosides (15%) and spines (42%). FIGURE 3 Proportions of chemical compounds and anatomical defences present in species endemic to islands with feral goats (upper) and islands without feral goats (lower) separated by community type: shrublands (left), rupicolous (middle) and caespitose (right). Binary traits were relativized to 100%, with each trait represented in a different colour. Numbers indicate Shannon's diversity index. 13652699, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jbi.14602 by Universidad De Granada, Wiley Online Library on [18/04/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License | 7 CAPÓ et al. to islands where herbivores are absent or restricted to habitats inaccessible to goats. The endemic species that inhabit areas inaccessible to goats exhibit some putative resistance traits, such as phenolic compounds, tannins, saponins, triterpenoids, flavonoids or pilosity, but lack other traits such as glycosides, coumarins and spines, and their plant size is similar to other species from Gymnesians unable to coexist with goats. These putative resistance traits may confer additional functions, such as protection against insect herbivory (Hamilton et al., 2001; Waterman et al., 1984) or adaptation to high exposure to light radiation on cliffs (PereiraDias & Santos, 2015). Interestingly, some examples of endemic rupicolous species have been detected in accessible communities in areas from which goats have been excluded for long periods (e.g. 10 years after exclusion fencing, on mountains isolated from goat populations)— as represented by some Fabaceae species (i.e. Hippocrepis balearica) or Brassicaceae species (Brassica balearica), both of which exhibit a low capacity for resistance based on the traits analysed in this study (pers. obs.; data not shown). Also, endemic species from other islands of the Mediterranean with rupicolous habitats have been severely damaged by introduced mammal herbivores when they are in areas accessible to goats, as is the case of Centaurea horrida in the island of Sardinia (Pisanu et al., 2012). As a general trend, few biogeographical studies about plantherbivore interactions in insular conditions consider PICs (Moreira et al., 2021) and this study proves the importance of incorporating phylogenetic relationships in order to provide clearer results. When PIC corrections were considered, a strong positive correlation was observed between the presence of alkaloids and glycosides, while a negative correlation was detected between coumarins and alkaloids, and coumarins and glycosides. Alkaloids and glycosides are recognized to be more efficient toxic compounds against ungulate herbivory than other compounds (Majak, 1991; Mattocks, 1968), with the exception of coumarins (Lake, 1999). The presence of alkaloids and glycosides correlated positively with the foliar area, indicating that plants protected by these compounds exhibit wider leaves. Indeed, glycosides and alkaloids efficiently protect plants against mammalian herbivores (Burney & Jacobs, 2013; McNaughton, 1983), although the production of these compounds imposes a high resource cost (Vrieling & van Wijk, 1994; Zangerl & Berenbaum, 1997). Furthermore, the concentrations of phenolic compounds and flavonoids correlated negatively with the foliar area. Plants with low foliar areas were mainly distributed in caespitose communities, which allowed these plants to escape herbivores. Phenolic and flavonoid compounds are known to protect against invertebrate herbivory, but low concentrations of these compounds provide minimal protection against mammal herbivory (Summers & Felton, 1994; Treutter, 2006). Plant size was a good predictor of protection against mammal herbivory in the bestfitting model. Low plant size values correlated with low foliar area and high SLA, consistent with the hypothesis that plant architecture enables species to escape herbivory by ungulates (Boege & Marquis, 2005; Brown & Lawton, 1991). With regards to other anatomical traits, spines did TABLE 2 Selection of the five bestfit models for the analysis. Variables included in each model are marked with a cross. The variables were phenolic compounds (TAE), condensed tannins (CE), saponins (Sap), flavonoids (Fla), triterpenoids (Tri), alkaloids (Alk), glycosides (Gly), coumarins (Cou), specific leaf area (SLA), spinescence (Spi), pilosity (Pil) and plant size (size). The Akaike information criterion (AIC) value and its delta (∆) against the first model are shown in the last two columns. Model TAE CE Sap. Fla. Tri. Alk Gly Cou SLA Spi Pil Size AIC ∆AIC M01 X X X X 10.8 0 M02 X X X X X 13.1 2.32 M03 X X X X X 13.1 2.32 M04 X X X X X 13.1 2.32 M05 X X X X X 13.1 2.32 Endemic species Distribution Threatened Coexistence B(0,1) Asperula paui Ei, Fo, Is No 0.23 Beta maritima ssp. marcosii Ei, Fo, Is Yes 0.20 Biscutella ebusitana Ei, Fo, Is No 0.10 Diplotaxis ibicensis Ei, Fo, Is No 0.02 Euphorbia margalidiana Is Yes 0.003 Galium friedrichii Ei, Fo, Is No 0.10 Genista dorycnifolia Ei Yes >0.001 Hippocrepis grosii Ei No 0.02 Santolina vedranensis Is Yes 1 Teucrium cossonii ssp. punicum Ei, Is Yes 1 Thymus richardii ssp. ebusitanus Ei Yes 0.58 TABLE 3 Threatened status and probability of coexisting with introduced herbivores based on a binomial response variable in the model for species endemic to islands without herbivores. Eivissa (Ei), Formentera (Fo) and the surrounding islets (Is). Threatened status is according to the IUCN criteria (Sáez et al., 2017). 13652699, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jbi.14602 by Universidad De Granada, Wiley Online Library on [18/04/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 8 | CAPÓ et al. not correlate significantly with chemical or anatomical traits in either the raw data or after PIC, which suggests the absence of a relationship between spinescence and chemical protection. Spines may confer other functions, for example, avoidance of water loss in arid environments (Bagella et al., 2019); therefore, the presence of this trait may have been driven as a response to other abiotic factors, such as climate. Contrary to oceanic islands, the endemic flora of continental islands is derived from the flora present when the connection with the mainland was severed. Hence, it is not surprising to detect that the variation in the phenotypic traits is influenced by phylogenetic relationships and ancestry. If the continental flora evolved in the context of herbivores, and there is a strong phylogenetic signal for highly toxic compounds, it could be expected that the endemic flora of continental islands may be exapted to the introduction of feral goats or other herbivores. Examples of this hypothesis are given by species of the genera Paeonia (Wu et al., 2010), Phlomis (Amor et al., 2009), Digitalis (Ganapaty et al., 2003) and Helleborus (Colombo et al., 1990), which all include species that contain toxic compounds that are phylogenetically fixed and evolved over a long time scale, when the Balearic Islands were still connected with Iberian Peninsula. 4.2 | Estimation of the potential vulnerability of endemic plant species to islands without feral goats Endemic species to the Balearic Islands, where feral goats are absent, displayed fewer resistance and escape traits than similar species on islands invaded by feral goats. Plants of a large size from accessible areas on islands without feral goats exhibited fewer resistance traits than communities on islands where feral goats are present. Moreover, when considering the resistance and escape traits that enable plants to coexist with introduced ungulates, we found that a large proportion of endemic species on the noninvaded islands have a low probability of coexisting with herbivores, in agreement with a previous study (Moreira et al., 2021). This suggests that the plant species on islands without goats would be extremely vulnerable to potential introduction of ungulate herbivores, even those of continental origin, the introduction of herbivores in these islands would probably result on a reduction of the distribution of the flora, leaving inaccessible habitats as the only potential habitats to avoid herbivory. Differences in resistanceescape traits of endemic plants between invaded and noninvaded islands indicate that spatial distribution of endemic flora in Gymnesic Islands might be conditioned by the goat's accessibility and plant capacity to resist or escape. 4.3 | Endemic flora from continental islands do not always lack defences It has been argued that insular endemic species do not have defences against mammal herbivores (Bowen & Van Vuren, 1997) and lack tolerance to herbivory (Barton, 2016). However, as presented in this study, this might not be always true, and some species can exapt to herbivory. This exaptation depends on the historical and ecological context in which the endemic species evolved (Capó, RoigOliver, et al., 2021) and probably on other ecological factors, such as herbivory pressure by invertebrates or other plant stresses (Moreira et al., 2021). Plant defences are usually assembled through multiple traits of different origins. This coadapted complex must function by itself, otherwise tradeoffs may eventually compromise the evolutionary maintenance and stability of those traits, and the species may reinvest the resources into other ecological functions (Agrawal & Fishbein, 2006; Dobzhansky, 1970). The phylogenetic context supports the hypothesis that the ancient Balearic flora displayed traits to cope with ancient native herbivores, and in turn, these traits function as an exaptation to herbivory in the presentday endemic flora. As observed in other ecological contexts (Kursar & Coley, 2003), the endemic flora to the two Balearic Islands affected by feral goats is structured by resistanceescape strategies. In this case, adaptation of the shrubland species on Majorcan mountains to ancient herbivores or exaptation in response to other evolutionary drivers (i.e. wind or high radiation exposure) that also function against herbivores, would determine the ecological distribution of the toxic and spinescent/stinging endemic species. The precise functions of individual chemical compounds in protection against ungulate herbivory are unclear, though synergistic effects may occur (Agrawal & Fishbein, 2006). Overall, a general pattern of exaptation was observed in the endemic flora of the Balearic Islands (Figure 4), namely resistance strategies based on (i) chemical resistance, that is, large species that produce highly toxic compounds and grow in habitats accessible to feral goats; (ii) anatomic resistance, that is, plants accessible to feral goats with leaves and flowers surrounded by spines or urticating hairs; (iii) plantsize escape strategies, that is, species accessible to feral goats that grow at ground level with small leaves and high SLA; and (iv) ecological inaccessibility, that is, species that lack any resistance traits to herbivory and have a large plant size, but live in areas inaccessible to herbivores. In areas where ungulate herbivores have not been introduced, species in the ecological inaccessibility category can be found in areas that are potentially accessible to goats. In fact, Moreira et al. (2021) found that island floras tend to present higher physical defences than continental relatives, which agrees with our findings that islands with high ungulate herbivory pressure present higher spinescent or stinging endemic species than those without ungulate herbivores. The insular effect on this plantherbivore interaction could also be linked to the variation found in that study, but further studies are needed to disentangle the real relationship between anatomical resistance. 5 | CONCLUSIONS The endemic flora to continental islands may be exapted to introduced herbivores, maintaining their traits by (i) pressure 13652699, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jbi.14602 by Universidad De Granada, Wiley Online Library on [18/04/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License | 9 CAPÓ et al. generated by other herbivores, (ii) a weak tradeoff of relocating these resources to face other ecological pressures, (iii) use for a secondary function or (iv) genetic fixation, regardless of ecological context. Exaptation to introduced herbivores is more likely to occur in lineages with a particular phylogenetic history where resistance traits are genetically fixed across evolution of the clade. However, other functional structures such as spinescence seem to be paraphyletic, and these traits may originate from both ancient herbivory pressure or other ecological features (i.e. high wind exposition or protection against radiation). As observed in this study, endemic plant species on islands noninvaded by feral goats are mainly located in accessible areas for ungulates and mostly not defended against herbivory than those occurring on invaded islands, which suggests that their flora is extremely vulnerable to future introductions of nonnative ungulates. The present study highlights that knowledge of the functional traits present in the endemic flora is essential to predict potential vulnerability and even estimate which taxa may be more likely to reduce their habitats and even become extinct in the event of the introduction of new herbivores. In turn, this information can be used as a tool to design optimal management strategies to protect endemic flora against introduced herbivores. ACKNO WLE DGE MENTS The authors are grateful to the Servei de Protecció d'Espècies (Govern de les Illes Balears) for their authorization to sample endemic and threatened species in their natural populations. The authors appreciate the valuable help of Joshua Borràs during the chemical analyses, Carles Cardona of the Centre Forestal de les Illes Balears (CEFOR) for his help sampling during the fieldwork and Beatrice Landoni for her help with statistical procedures. This work was funded by Project CGL201570449R (Ministerio de Economía y Competitividad, Gobierno de España). M.C. was funded by PhD fellowship FPI/1925/2016 (Direcció General de Política Universitària i Recerca, Govern de les Illes Balears and the European Social Fund) and Margarita Salas postdoctoral fellowship (Ministerio de Universidades, Gobierno de España) through the Recovery, Transformation and Resilience Plan funded by the European Union (NextGenerationEU). Andrea Devlin from Science Editing Experts proofread the manuscript. CONFLICT OF INTEREST STATEMENT The authors declare they do not have any potential conflicts of interest. DATA AVAILABILITY STATEMENT Data are available from the Dryad Digital Repository (Capó et al., 2023) https://doi.org/10.5061/dryad.gxd25 47r4. ORCID Miquel Capó https://orcid.org/0000-0002-4394-7080 REFERENCES AbdulAzeez, I., Sulaiman Ayodeji, A., & Danjuma, B. (2018). Phytochemical and antimicrobial screening of the leaves of Crotalaria lachnosema against Staphylococcus aureus, Salmonella typhi, Escherichia coli and Klebsiella pneumoniae. Archives of Organic and Inorganic Chemical Sciences, 3(5), 421– 425. https://doi.org/10.32474/ aoics.2018.03.000173 FIGURE 4 Endemic flora complexes based on chemical and anatomical protection against herbivores and their accessibility to introduced herbivores. SLA, specific leaf area. 13652699, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jbi.14602 by Universidad De Granada, Wiley Online Library on [18/04/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License