An integrative approach to discern the seed dispersal role of frugivorous guilds in a Mediterranean semiarid priority habitat

Study area

The study was carried out in the SCI—Sierra de la Fausilla (Southeast of the Iberian Peninsula, Fig. 1), a coastal mountainous system of cliff morphology in the south-eastern region of the province of Murcia. This protected area located in the Mediterranean biogeographic region, belongs to the biogeographical province referred to as Murcia Almeriense, in which the most arid landscape in the Iberian Peninsula can be found (García et al., 2007). Covering 865.26 ha, the Sierra de la Fausilla was declared a SCI in the Natura 2000 Network (ES6200025: April, 1999) and a Special Protection Area (SPA ES000199: March 152 2000). Four types of priority habitat of community interest can be found in this protected area: 5220* (Mediterranean arborescent shrubland with Z. lotus), 1510* (Mediterranean salt steppes), 6110* (Rupicolous calcareous or basophilic grasslands of the Alysso-Sedion albi) and 6120* (Xeric sand calcareous grasslands) (Comunidad Autónoma de la Región de Murcia, 2005). In this study, we focused specifically on the semiarid shrubland habitat 5220*, which refers to pre-desert Mediterranean shrubland that grows from sea level to 300 m (Tirado, 2009) (Fig. S1), with limestone soils in depressions, riverbeds and zones of subsurface currents. The plant communities in this habitat are characteristically spiny species with tiny leaves that grow in the dry season, shaping islands of biodiversity that offer suitable microenvironments for the germination of other plant species (i.e., a nursery, Mendoza-Fernández et al., 2015), and shelter and food for animals. A large number of these Ibero-African plant species (i.e., species distributed in the North of Africa and in the South of the Iberian Peninsula) are fleshy-fruited shrubs of tropical or subtropical origin, considered relics of the wetter past climatic conditions (e.g. Z. lotus, Periploca angustifolia Labill., Lycium intricatum Boiss., Maytenus senegalensis, Withania frutescens (L.) Pauquy. and Pistacia lentiscus L.) (Sánchez-Gómez & Guerra, 2003). Their fruits are key trophic resources for frugivorous vertebrate animals as they provide both nutrients and hydration (Robbins, 1983).

In this sampling area, we selected three sites situated more than one km apart from each other in a region stretching from the coast Southeast of Cartagena to Cabo de Palos and bordering the Valley of Escombreras to the north: Zone 1: Escombreras; Zone 2: Gorguel and Zone 3: Cola de Caballo (Fig. 1). From the second half of the last century, this valley has undergone drastic physiognomic changes that have led to its current occupation by a large petrochemical refinery complex and some local industries. Moreover, there are non-endemic plant species associated to gardens in anthropogenic areas and crop plant species from oldfieds in the area.

Plant community composition and sampling of frugivore-dispersed seeds

We sampled the local plant community to determine which plant species could potentially be dispersed by animals at the three study sites. Vegetation sampling was carried out on circular plots of 100 m² (nine plots at each sampling site, 27 in total), in which all fleshy-fruited woody plant species were counted to determine their frequency of occurrence following the protocol developed previously (Zapata & Robledano, 2014).

The three sampling sites were visited fortnightly from September to December 2016 (six field trips) to collect animal-dispersed seeds (defecated in the case on mammals and defecated or regurgitated in the case of birds). Sampling procedures were performed during the peak abundance for fleshy-fruited plants in Mediterranean shrublands (Jordano, 1988). We looked for animal faeces (mammals and birds) and regurgitated seeds from birds in five fixed 100 m transects at each sampling site that were distributed in order to cover the landscape heterogeneity of the study areas. One transect in each study area included a dirt road, since many mammalian species have a preference for defecating in such anthropogenic structures (Suárez-Esteban, Delibes & Fedriani, 2013). We made a preliminary classification of the mammalian and bird faeces in the field, then each faecal sample was photographed next to an object reference for posterior identification. All samples were labeled with the date and site and stored in an airtight bag. Mammalian faeces were examined under a laminar flow chamber on plastic trays in the laboratory. These seeds were separated, washed with distilled water, dried at room temperature, and then stored at 5 °C for taxonomic classification. In the case of bird-dispersed seeds, we collected the faecal and regurgitated samples in sterile tubes and stored them at −20 °C for identification through molecular techniques. The plants dispersed by both animal classes were identified by the seed’s external morphology, comparing this with a reference collection of seeds of fleshy-fruited shrubs.

Determination of seed dispersers

The taxonomic classification of mammal species dispersing seeds in Sierra de la Fausilla was achieved by analyzing photographs and the characteristics of the faeces collected, and by comparing our material with information available in trail and footprint guides (Rubines, 2015). These classifications were confirmed by a researcher with experience in the collection and processing of mammalian faecal samples. Seeds in bird faeces were separated from faecal material to be used in DNA extraction in this way reducing PCR inhibitors and increasing PCR yield (see Marrero et al., 2009 for rationale). Dispersers from bird-dispersed seeds were identified by DNA barcoding achieved by amplifying the animal DNA that remained on the seed surface after passage through the gut or regurgitation (González-Varo, Arroyo & Jordano, 2014). This molecular technique is based on amplification of a fragment of mitochondrial DNA (cox1: cytochrome c oxidase subunit I, Hebert, Ratnasingham & De Waard, 2003), which allows closely related species to be distinguished. Genomic DNA was extracted from seeds with the NucleoSpin Tissue extraction Kit (Macherey-Nagel, Düren, Germany). For PCR reactions, COI-fsdF and COI-fsdR primers were used following the PCR profile described previously (González-Varo, Arroyo & Jordano, 2014). PCR reactions were performed in a total volume of 12.5 µl, containing 2.5 µl of PCR buffer (5×), 0.5 µl of each primer (0.04 µM), 0.312 µl of BSA (0.5 mg/ml bovine serum albumin; Roche Diagnostics, Barcelona, Spain), 0.2 µl of Taq polymerase (1 U/Tube; Bioline, London, UK), and 6.48 µl of ultrapure water. As the study sites were close to the sea, implying high humidity, the DNA of many samples was degraded and hence, we used specifically designed primers for such cases: COI-fsd-degF and COI-fsd-degR (González-Varo et al., 2017b). Sequence edition and alignment was performed with Mega7 (Kumar, Stecher & Tamura, 2016), and “BOLD: The Barcode Life of Data System” (Ratnasingham & Hebert, 2007) was used to identify the bird species to which each sequence belonged.

Data analysis

The density of fleshy-fruited species and the seed rain were analyzed in the three sampling sites using a non-parametric Kruskal–Wallis test. To determine whether the plant communities dispersed by mammals and birds were significantly different to those established in the sites studied, we assessed the data of the species dispersed by each taxon in relation to the density of the plants species found in vegetation samples, performing a Permutational Multivariate Analysis of Variance (PERMANOVA) with a Bonferroni correction, and using Bray–Curtis distances and 9,999 permutations. We graphically represent the three plant communities—plant species recorded in vegetation samples and plant species dispersed by birds and mammals—with non-metric multidimensional scaling (NMDS) (Anderson, 2014). Data of species dispersed by each taxon were considered as the number of interactions between each plant and each animal species, that is, the number of times that viable seeds of a plant species were found in animal-dispersed samples. Moreover, we calculated both Chao1 and Chao2 estimators in order to appraise the diversity of plant species that was observed for the sampling plots of native plants, and identified in seed dispersal of birds and mammals (Sarmah, 2017). Chao1 estimates the total diversity in a sample of sites using abundance data, whereas Chao2 estimates species richness based in presence and absence information (Escribano-Avila, 2019). We considered each sampling plot of native plants and the mammal and bird dispersed samples as the sampling effort units for the samplings of native plants and plant species dispersed by both animal classes, respectively, and the different plant species as the abundance or presence data.

In addition, a quantitative network approach has been used to evaluate the structure of the interactions between fleshy-fruited plants, and mammalian and avian seed dispersers (Bascompte & Jordano, 2007). We pooled the number of interactions (i.e., the total number of samples in which at least a viable seed was found; if seeds of two species were found in the same sample it counted as two interactions; Vizentin-Bugoni et al., 2019), for each plant–frugivore pair (mammalian or avian) from all the animal-dispersed samples in the sampling transects and across the three sites throughout the season. The network structure was depicted as a bipartite graph and evaluated using basic parameters that represent complementary aspects of the structure of a mutualistic network (Peredo et al., 2013): connectance (C), the proportion of interactions detected relative to all the potential interactions in the network; linkage density to assess the average linkage level; nestedness (WNODF), the degree to which the interactions of the poorly connected species are a subset of the highly connected species; complementary specialization (H′₂), a measure of niche complementarity between species; generality (G), a weighted average of the number of plants dispersed per frugivore; and vulnerability (V), a weighted average of the number of frugivores per plant. We calculated species–specific parameters to identify the number of partners of each plant and disperser species (degree), which species were the most selective (by means of d′) and which species were the most important of each trophic level (species strength). To test whether the parameters estimated from the empirical network differed significantly from networks with randomly interacting species, we compared the observed values with those of 1,000 random networks based upon a Patefield null model (Dormann, Fründ & Gruber, 2014). We also calculated network modularity (Q), which indicates the extent to which resistance or infectivity interactions can be partitioned into distinct groups, each of which has many internal interactions but few with other groups (Dormann & Strauss, 2014), identifying the main modules related to the roles of frugivorous guilds (mammals vs. birds). Moreover, null model comparison was calculated by means of z-scores, where z-scores values above two indicate a significant modularity.

All analyses were carried out using R software version 3.4.0 (R Development Core Team, 2016). Statistical tests and Chao estimators were performed with the vegan R package (Oksanen et al., 2007) and the graphs were obtained with ggplot2 (Wickham, 2010). Specifically, the network analyses and their graphs were performed using the bipartite R package (Dormann, Fründ & Gruber, 2014; Dormann & Strauss, 2014).

Plant composition and seed disperser species

There were 12 fleshy-fruit species identified on the plots of vegetation studied here (Table S1). Mean density values of the fleshy-fruit species and of their dispersed seeds were similar among the three sampling sites (Density of fleshy-fruit species—Kruskal–Wallis χ² = 0.104, df = 2, p = 0.949; dispersed seed density—Kruskal–Wallis χ² = 3.046, df = 2, p = 0.218). Overall 650 animal-dispersed samples were collected from the SCI—Sierra de la Fausilla, of which 33.2% were mammalian and 63.1% were avian, with 3.7% of undetermined taxonomy. Escombreras was the sampling site with most seeds, followed by Cola de Caballo and Gorguel (Fig. 2; Table S2). Of the 216 mammalian faecal samples, only 29 contained seeds and these faeces were deposited by five species of mammals: beech marten Martes foina Erxleben.; badger Meles meles L.; fox V. vulpes; rabbit Oryctolagus cuniculus L. and hedgehog Erinaceus europaeus L. Significantly, most seeds were dispersed by carnivorous mammals like marten (43.44%) and fox (30.05%). Approximately, 38% of the samples collected from birds contained seeds (155 of 410 samples). Mammals dispersed a larger number of seeds than birds, 762 and 325 seeds, respectively, although this quantity was clearly biased by the consumption of figs (Ficus carica L.) by mammals. This human cultivated fruit tree usually has 100 seeds per dispersal unit and consequently, it is probable that the seeds observed in one faecal sample came from just one fig. In addition, we detected seed dispersal of Pygmy date palm (Phoenix roebelenii O’Brien), an exotic plant, by V. vulpes and Martes foina (Table S2).

Regarding birds, the disperser species was successfully identified in 92% of the seeds sampled (n = 299). We sequenced fragments amplified with different combinations of the four primers mentioned above (see “Data accessibility” for details about the sequenced strand), although only one strand was sequenced for most samples (COI-fsd-degF) as the length obtained was sufficiently long for species identification (Length range = 159–465 bp; mean = 233.34 ± 4.08 bp). DNA barcoding allowed us to identify eight birds dispersing seeds in our study area based on a similarity threshold >99%: Erithacus rubecula L.; Phoenicurus ochruros S. G. Gmelin; Sylvia atricapilla L.; S. melanocephala Gmelin; S. undata Boddaert.; Turdus merula L.; T. philomelos C. L. Brehm.; and T. viscivorus L.

Concerning the relationship between the composition of the plant community and the dispersed seeds, the density of fleshy-fruit shrubs did not show any relationship to the dispersed seed density (R² = 0.073, p = 0.331), although we registered the highest values of fleshy-fruited shrubs density and dispersed seeds in Escombreras. The plant community registered in each plot of vegetation was dissimilar to the plant species dispersed by birds (PERMANOVA, F = 3.377, R² = 0.082, adjusted p-value = 0.006) and by mammals (PERMANOVA, F = 10.142, R² = 0.229, adjusted p-value = 0.003), and there were also significant differences in the plant species dispersed by each frugivorous guild (F = 5.355, R² = 0.196, adjusted p-value = 0.003) (Table S3). However, NMDS showed a higher contribution of birds to disperse plant species occurring in the study sites (Fig. 3).

Seed dispersal networks: identifying the key species in seed dispersal

Chao1 estimated true diversity as 12.25, 7, and 13 species for sampling of native plants, plant species dispersed by mammals and birds, respectively, while Chao2 estimated the richness of the sample as 14, 7.25, and 15.5 for sampling of native plants, plant species dispersed by mammals and birds, respectively (Fig. 4). These results highlight the role of birds as main seed dispersers in the study area, at least in terms of the diversity of plant species dispersed.

The bipartite network between fleshy-fruit plants and vertebrates (mammals and birds) included a total of 181 interactions (with 1,061 dispersed seeds), showing strong biases in the distribution of these interactions (Fig. 5). From the perspective of the 15 plant species dispersed, Pistacia lentiscus was the most dominant species, with Asparagus albus L., O. lanceolata, L. intricatum, and W. frutescens among the most dispersed plants in a second order. Pistacia lentiscus was the species associated with the largest number of frugivores (n = 7). Regarding mammals, most of the dispersal events were done by V. vulpes, being also that dispersed the widest variety of seeds. In terms of the avian species, T. viscivorus was responsible for the 76.3% of the dispersal interactions by birds. Moreover, while Pistacia lentiscus was dispersed by a wide range of bird species, L. intricatum and W. frutescens were only dispersed by two species. From all the bird species identified, T. viscivorus was the dominant disperser and it interacted with the largest number of plant species. The species level parameters (Table S4) also showed that the most important fleshy-fruited plant of the priority habitat 5220* was Pistacia lentiscus and the most important seed disperser was T. viscivorus since they showed the highest values of degree and species strength. On the other hand, Rhamnus lycioides L. and O. cuniculus were the most selective species from the fleshy-fruited plants and seed dispersers, respectively, supported by the highest values of d’. This selectivity could be explained by the scarcity of R. lycioides seeds and the occasional frugivorous role of O. cuniculus.

Network parameters showed that there were a lower connectance (C) and linkage density in the seed dispersal network than expected under a null model (Table 1), since more than half of all the potential plant–frugivore links were not observed (Fig. 5). Nestedness (WNODF) was higher than expected in the network (Table 1), given that the interactions of poorly connected species (S. undata, S. atricapilla and Erinaceus europaeus) were highly nested within those of highly connected species (like T. viscivorus and V. vulpes Fig. 5). The degree of specialization within the network (H’₂) was higher than the anticipated by the null model, probably because rarer plant species were almost exclusively dispersed by occasional frugivores, for example O. cuniculus only dispersed R. lycioides (represented in one module, Fig. S2). Modularity (Q) was 0.412, being significantly modular (z-scores = 15.942; Qnull = 0.161) and identifying five modules in the network (Fig. S2), where frugivorous taxa had a complementary role: whereas mammals dispersed especially exotic and human-cultivated plant species (represented in two modules), birds dispersed most representative plant species from habitat 5220*.

It is also relevant to note the high degree of complementarity among mammals and birds in the seed dispersal network, as F. carica, Phoenix dactylifera L. and R. alaternus L. were the only plants dispersed by members of both taxa (mammals and birds; Fig 5). This may be another factor to explain the high degree of specialization within the network. According to the generality (G), each frugivore dispersed an average of six plant species (Table 1). Vulnerability (V) was lower than generality (with an average of two dispersers per plant), since the average number of frugivores per plant was lower than the number of plants per frugivore, although both parameters were lower than in the null model (Table 1).