Different land-use types equally impoverish but differentially preserve grassland species and functional traits of spider assemblages

Land-use change is one of the major drivers of biodiversity loss by introducing environmental modifications, which excludes native species unable to adapt to the novel conditions. Grasslands are among the most threatened biomes; understanding the influence of different land-use types on native species is crucial to achieving sustainable management policies. We hypothesized that land-use types that partially conserve the original vegetation cover would show higher taxonomic and functional diversity and similarity with native assemblages than land-use types that replace the original vegetation cover. We compared the taxonomic and functional alpha and beta diversity of spider assemblages between soybean crops, eucalypt plantations, and cattle fields with seminatural grasslands. Through null models, we assessed the standardized effect sizes to test differences in the strength of environmental filtering among land-use types. Environmental changes introduced by different land-use types resulted in assemblages differentiated in species and trait composition, taxonomically and functionally impoverished with respect to seminatural grasslands. All land-use types drove species replacement and trait loss and replacement of grassland spiders. Each land-use showed a characteristic species and trait composition. Most of the grassland traits were not lost but were under or over-represented according to the land-use type. Only in soybean crops the formation of spider communities would be mainly driven by environmental filtering. Changes in land-use decreased species diversity and modified the composition of spider species and functional traits leading to differentiated spider assemblages. As spider species and traits varied among land-uses, a mitigation measure against grasslands biodiversity loss could be the development of productive landscapes with a mosaic of land-use types, as each of them would provide microhabitats for species with different requirements. Because land-use types mainly led to the rearrangement of grassland functional trait values, most of spider functions might be conserved in mosaics of land-use types.


Materials and methods
Study area and sampling design. The study was conducted in central-eastern Argentina (31-33°S and 58-59°W, Fig. 1). The climate is humid temperate with precipitations occurring all year round. The dominant vegetation is a grass steppe dominated by Stipa, Piptochaetium, Poa, Eragrostis, and Aristida, wild straws (Melica), wisps (Briza), and bromes (Bromus) 26 . Traditionally, this region was devoted to livestock production. Since the 1990s livestock areas have been replaced by agriculture 27 . The study area is dominated by large extensions of monocultures crops, usually wheat in winter and soybean in summer. Between successive summer soybean crops, the land may remain at rest or livestock may feed on the non-implanted pastures. Eucalypt plantations have also replaced livestock fields 28 . Eucalyptus grandis is the dominant plantation tree in the region, but also pine (Pinus sp.) is planted 29 . Plantations are large-scale monocultures covering hundreds of hectares with stands which have little understory vegetation 5 and large amounts of leaf litter.
We sampled spiders in soybean crops, eucalypt tree plantations, cattle fields, and in protected areas (seminatural grasslands), the latter used as a reference of the native habitat. The sample design consisted of 20 study sites (about 350-800 m 2 ), five sites (i.e., replicates) per habitat type. The study sites were separated by a mean distance of 45 km, from 500 m to 150 km. We selected mature stands (more than 9 year old) of Eucalyptus grandis to avoid heterogeneity among replicates 30 . Spider sample stations were set up at least at 50 m inside each study site to avoid edge effects 31 . Environmental variables. To environmentally characterize the different land-use types, we estimated the percentage of land cover by shrubs, herbs, leaf litter and bare soil. We also measured the mean height, maximum height and density of the herbaceous stratum and the leaf litter depth 32 . At each sampling site, we visually estimated the coverage percentages following Kent and Coker (1992) in three quadrants of 4 m × 4 m. At each quadrant, we measured the maximum vegetation height, the mean vegetation height of the herbaceous stratum from 10 points taken randomly, and the litter depth in the middle of each quadrant. We calculated vegetation density in the lower stratum using a digital camera positioned approximately 1 m from a white cloth of 1 × 1 m used as a background screen and we took a picture capturing the vegetation in the meedle 33 . We estimated the percentage covered by vegetation by transforming the photographs into black for vegetation and white for background screen, and quantifying them with the ImageJ program 34 .
Spider sampling and identification. Spiders were sampled using pitfall traps to collect ground mobile spiders and an entomological vacuum (G-vac) method to catch less mobile ones from the herbaceous strata 35 . At each study site (sample unit) three pitfall traps (diameter = 9 cm, depth = 10 cm) were placed 10 m apart from each other 36 . Traps were half filled with propylene glycol and water (30:70). A plastic roof was placed 10 cm above the trap to prevent flooding and evaporation 37 . Pitfall traps worked for 6 weeks, 3 in spring (November 2015) and 3 in summer (January 2016). The G-vac method was used to suck spiders from the herbaceous strata in a 1 × 1 m square for 1 min using a vacuum with a 110 cm long and 12 cm wide tube. On each site and sampling period we collected five subsamples diurnally and separated at least by 10 m 38 . The material per site was pooled together and conserved in ethanol 70%. Adult spiders were identified to species or morphospecies using taxonomic keys and assistance of arachnologists from the "Museo Argentino de Ciencias Naturales Bernardino Spider traits. Spiders were characterized according to four widely accepted traits that typically reflect spiders adaptation to environmental conditions: body size, hunting mode, dispersal ability and circadian activity 36,40 ( Table 1). For the estimation of body size (i.e., cephalothorax length) of each species two to four specimens were used; measurements were done with a 0.01 mm ocular micrometer mounted on a Carl-Zeiss Discovery V8 stereo-microscope. Hunting mode, dispersal ability and circadian activity states for each species were obtained from literature (Table 1).

Data analysis.
To describe the environmental differences among sites, we performed a principal component analysis (PCA) 41 using R package vegan 42,43 . Since environmental variables were not dimensionally homogenous, they were standardized, and a correlation matrix was used for the analysis.
Taxonomic and functional diversity comparison among land-use types. To analyze spider taxonomic and functional diversity change due to different land-use types, data from pitfall traps and G-vac within each study site were pooled together. For statistical analyses, species richness was expressed as the number of species per sampling site, and functional diversity as the FDis index estimated per sample site. We calculated the alpha functional diversity as the FDis multi-trait index 44 . Life-history trait data for dispersal, circadian activity and hunting mode were coded categorically and body size as a quantitative variable for the index calculation. We also used the community-weighted trait means (CWM onwards) to measure changes in the functional composition of spider assemblages driven by land-use types 45 . To calculate the CWM values for each category, categorical traits were treated as independent binary variables 46 . We ran Generalized Linear Models (GLM) with land-use type as the predictor variable (four levels: the three land-use types and the seminatural grasslands) and the species richness and the FDis index as response variables using R package stats. The Poisson distribution with the logarithm as link function was used for species richness (after meeting for no overdispersion) and the Gaussian distribution with the identity as link function was used for the FDis (after meeting for linearity and homoscedasticity) 47 . Then, Tukey post-hoc comparisons were run to test for differences between land-use types and the seminatural grasslands (used as reference) using emmeans package in R.
Taxonomic and functional similarity among land-use types. To analyze changes in species and functional traits composition among land-use types, we calculated the dissimilarity of spider assemblages between each pair of land-use and seminatural grassland sites. To find out if the dissimilarities were due to replacement or loss of species or traits, we obtained nestedness and turnover components following Baselga (2010) and Villéger et al. (2013). For taxonomic dissimilarity Jaccard index was used. For functional trait dissimilarity, we first obtained the Gower distance between pairs of species based on the table of trait data. Then, we ran a principal coordinate analysis (PCoA) and we used the four resulting axes as continuum functional dimensions to delimit the functional space of each site. Finally, the volumes of multivariate trait space shared by two sites were obtained as the functional similarities between pairs of sites 48 . Taxonomic and functional pairwise beta diversity measures were calculated. Then, to test the differences in the nestedness and turnover components among land-use types we ran permutational multivariate analyses of variance (PERMANOVA) 49 using the vegan package of R (adonis function) and performed post-hoc pairwise comparisons using the RVAideMemoire package of R (pairwise.perm. manova function) (perm = 9999) after checking meeting all assumptions.

Response of individual species and traits to land-use type.
We identified the species involved in the taxonomic dissimilarity patterns by running non-metric multidimensional scaling (NMDS) with Jaccard dissimilarity measures using vegan package of R (metaMDS function). We utilized the scores from axis 1 and 2 of the NMDS ordination to select the three spider species that showed the strongest association with each land-use type. Then, to examine the spider trait composition changes among land-use types, we compared CWM values between land-use types and seminatural grasslands. We ran GLMs in R with land-use type as the predictor variable (four levels: the three land-use types and the seminatural grasslands) and the CWM of each functional trait as the response variable. The Gaussian distribution with the identity as link function was used for the CWM of the spider body size, and the binomial and quasibinomial distribution with the logarithm as link function were used for CWM of categorical variables 47 . Post-hoc comparisons were run to test for differences between land-use types and the seminatural grasslands.
Formation processes of spider communities in land-use types. To assess the environmental filtering process in all land-use types, we compared the observed FDis and the FDis expected by chance for each study site. We ran null models by generating 999 random assemblages per site, using the complete list of species collected 50 . Each random assemblage maintained the same number of species observed per site. To analyze the difference between the observed FDis and that expected by chance for each land-use type, the standardized effect sizes of FDis (SESFDis) was calculated 50 . SES negative values indicated functional convergence (i.e., loss of functional diversity leading to functional redundancy of spiders assemblages) while positive values indicated functional divergence (i.e., gain of functional diversity leading to an increase in spider species functional differentiation) 51 . To estimate whether the SESFDis means were negative or positive, t-tests were performed using stats package of R (t.test function).

Results
We recorded a total of 2085 individuals (823 adults) from 105 spider species or morphospecies (24 families) (Appendix , Table A1) in the 20 sampled sites. Sixty-six species were caught in seminatural habitats while 36 were found in soybean, 40 in cattle fields and 34 in eucalypt plantations. Approximately 29% of species were exclusively from seminatural grasslands. The most abundant families were Lycosidae (20.5%), Theridiidae (17.8%) and Linyphiidae (15.7%) considering only adults. The most frequently observed spider species was Pardosa flammula (Lycosidae, 8% of adults), followed by Erigone sp1 (Linyphiidae, 7% of adults). The PCA based on environmental variables showed that two components explained 71% of the environmental variation (Fig. 2). The first component (43%) differentiated tree plantations and one cattle field from the rest of the sites. The second component (28%) mainly differentiated soybean crops from the other sites. Seminatural  Taxonomic and functional similarity among land-use types. The species composition of the three land-use types was highly dissimilar compared to seminatural grasslands, being the total dissimilarity CI 95 = 0.87 ± 0.09 for soybean crops, CI 95 = 0.84 ± 0.06 for tree plantations, and CI 95 = 0.0.81 ± 0.08 for cattle fields. The PERMANOVA analysis showed that the taxonomic turnover and nestedness components differed among land-use types (F = 3.19, p = 0.04), being turnover higher than nestedness for the three land-use types (Appendix ,  Table A2) (Fig. 4A). In addition, turnover was higher in soybean crops than in cattle fields, but no differences were found in the turnover component between soybean crops and tree plantations, and between tree plantations and cattle fields. Differences among nestedness component were not significant (Appendix , Table A2) (Fig. 4A).
The total functional dissimilarity between each land-use type and seminatural grasslands was high, CI 95 = 0.96 ± 0.03 for soybean crops, CI 95 = 0.93 ± 0.04 for tree plantations, and CI 95 = 0.93 ± 0.03 for cattle fields, with significant differences between the functional turnover and nestedness components among them (F = 4.90, p = 0.01) (Fig. 4B). Turnover was similar among land-use types. For soybean crops, the turnover component was lower than the nestedness component (Appendix , Table A2). Nestedness was higher in soybean crops than in tree plantations and, than in cattle fields (Appendix , Table A2), and no statistical differences were found for the nestedness component between tree plantations and cattle fields (Fig. 4B).
The CWM of spider traits varied among land-use types compared to seminatural grasslands for hunting strategy and circadian rhythm, but not for body size and the ability to disperse by ballooning (Fig. 6). Regarding hunting strategy, active hunters were less represented in soybean crops and tree plantation than in seminatural grasslands (X 2 3,16 = 36.97, p < 0.001; soybean vs grasslands : Z = − 4.08, p < 0.001; plantations vs grasslands: Z = − 4.19, p < 0.001); web-building spiders were more represented in soybean crops and tree plantations than in seminatural grasslands (X 2 3,16 = 32.47, p < 0.001; soybean vs grasslands : Z = 4.21, p < 0.001; plantations vs grasslands: Z = 3.71, p < 0.001); ambush hunters were more represented in tree plantations than in seminatural  Table A1), black circles are tree plantation sites, white circles are seminatural grassland sites, triangles are cattle fields and squares are soybean crops. In bold species that presented the strongest association with the different land-use types based on NMDS scores. Figure 6. Community weighted means (mean ± standard error) of the functional traits of spider assemblages located in tree plantations (plan), soybean crops (soyb), cattle fields (catt), and seminatural grasslands (gras). In each graphic, asterisks represent significant differences between the means of land-use types and seminatural grasslands (p < 0.05). www.nature.com/scientificreports/ grasslands (X 2 3,16 = 16.71, p < 0.001; Z = 2.88, p = 0.01). As regards the circadian rhythm of spider assemblages, cattle fields and soybean crops presented higher proportions of spiders that are diurnal and nocturnal than seminatural grasslands (X 2 3,16 = 16.47, p < 0.001; cattle vs grasslands : Z = 2.64, p = 0.02; soybean vs grasslands: Z = 2.46, p = 0.04); cattle fields presented lower proportions of exclusively nocturnal spiders than seminatural grasslands (X 2 3,16 = 30.47, p < 0.001; Z = − 2.47, p = 0.04), and no exclusively nocturnal spiders were found in tree plantation and soybean crops (Fig. 6); the proportions of exclusively diurnal species did not differ among landuse types (X 2 3,16 = 7.39, p = 0.06).

Formation processes of spider communities in land-use types. Tests performed to assess whether
SESFDis were negative or positive indicated that for cattle fields and tree plantation the SESFDis values did not differ from the null expectations (t 4 = − 0.50, p = 0.64 and t 4 = 2.29, p = 0.08, for cattle fields and tree plantations respectively). For soybean crops the SESFDis was significantly lower than zero (t 4 = − 4.57, p = 0.01) (Fig. 7).

Discussion
Overall, our results showed that land-use types introduced different environmental modifications, however, all land-uses led to assemblages impoverished in grassland spider species and traits irrespective of their environmental similarity with the native habitat. Likely, at the microhabitat scale perceived by spiders, the magnitude of environmental change due to the different land-use types was equivalent. Besides, our results revealed that changes introduced by each land-use type were qualitatively different. Soybean crops retained more grassland spider traits than other land-use types and favored web-building spiders. Tree plantations enhanced ambush hunters while only cattle fields harbored nocturnal species. The three land-use types led to grassland species replacement and loss and replacement of functional traits. However, only in soybean crops, the environmental filtering seemed to be the main process involved in assembly formation. Possibly, the neutral patterns found in tree plantations and cattle fields were due to the compensation of opposite trends driven by environmental filtering and niche differentiation processes acting simultaneously (i.e. functional convergence and divergence, respectively).
Taxonomic and functional diversity comparison among land-use types. We found that soybean crops, eucalypt plantations, and cattle fields equally reduced species richness but not the functional diversity of grassland spider assemblages. As observed in previous studies, environmental changes introduced by human land-uses strongly limit the number of species able to tolerate the new habitat conditions 52,53 and only those species holding functional traits reflecting adaptations to the modified habitats will be able to persist 9 . Except for cattle fields, land-use types did not decrease the functional diversity of spider assemblages compared to seminatural grasslands. The lack of functional diversity response in tree plantation and soybean crops showed that many of the lost species would be functionally redundant and might also be evenly distributed across the traits considered here. However, species loss and replacement could also lead to assemblages with less richness but no decrease in functional diversity due to rearrangements on functional trait values 54 . Taxonomic and functional similarity among land-use types; response of individual species and traits. Land-use types drove species composition shifts. Contrary to our expectations, the spider species composition of the three land-use types was highly dissimilar compared to protected grasslands and was mainly  www.nature.com/scientificreports/ due to species replacement. As environmental conditions differed among land-use types, they might favor different spider species according with their habitat requirements. For example, in tree plantations, Hisukattus transversalis and Asthenoctenus borellii were more abundant, and have been previously found in implanted forests and other habitats closely linked to leaf litter and humidity conditions 55,56 . Also, Alopecosa moesta and Laminacauda montevidensis, highly abundant in cattle fields, have been found in protected areas and urban green areas 57,58 . Tullgrenella morenensis and Steatoda ancorata, which were more abundant in soybean crops, were also found in urban green areas and rice crops 58,59 . These species would be highly tolerant to anthropogenic disturbances and only depend on the herbaceous vegetation layer. Although vegetation cover in cattle fields was more similar to grasslands than other land-use types, at the microhabitat scale perceived by spiders, the magnitude of the environmental change introduced by all land-use types would be equivalent. The land-use types induced changes in functional traits composition due to trait loss and replacement with respect to spider assemblages of seminatural grasslands. Overall, soybean crops were functionally more similar to grasslands than cattle fields and tree plantation. Possibly, the soybean vegetation layer provide a suitable structure for spiders adapted to grassland habitats 21 . While nocturnal spiders were only found in cattle fields and grasslands, most of the changes were due to the under or over-representation of grassland traits in the different land-use types. For example, active hunters decreased in tree plantations and soybean crops compared to grasslands, showing that this guild would be sensitive to changes in the vegetation cover as they depend on habitat architecture for concealment and prey location 60 . Web-building spiders increased in tree plantations and soybean crops compared to grasslands. The litter layer and fallen branches present on the ground of tree plantation and the high vegetation density in soybean crops may offer a variety of physical structures for webs attachment 61,62 . These results suggest that although an equivalent magnitude of the environmental change between land-use types and grasslands was perceived by spiders, the quality of those changes differed. Thus, functional traits modification could be driven by microhabitat loss and replacement 63 , being detrimental for those traits that relied on the lost microhabitats while species holding traits adapted to remaining and novel microhabitats would be favoured. However, some traits would not be affected by land-use habitat change (i.e., body size and ballooning). Likely, body size and ballooning might respond to microhabitat variables such as temperature 19 and wind speed 64 , which would not show high variation among land-use types and natural grasslands.
Formation processes of spider communities in land-use types. Regarding the functional traits considered here, functional convergence, evidence of environmental filtering 50 , only occurred in soybean crops. However, assemblages may also be composed of random subsets of functional traits from the regional pool. Possibly, the results obtained for tree plantations and cattle fields implied different processes that may be acting simultaneously. If niche differentiation and environmental filtering acted simultaneously with a similar strength, because these two processes lead to opposite patterns (i.e., functional divergence and convergence), these would explain a neutral net response for these land-uses. Niche differentiation has been commonly associated with the prevalence of competitive interactions 12 . However, in forest plantations niche differentiation could also occur due to the incorporation of novel forest structures, such as deep leaf mulch or woody structures, that would allow functionally distinct species to occupy novel available microhabitats 65 . While environmental filtering was claimed as the main driver of assemblage formation in anthropogenic habitats 6 , land-use change leading to novel environmental conditions may lead to other assembly processes over environmental filtering.
Environmental changes introduced by each land-use resulted in spider assemblages differentially composed regarding species and functional traits. Besides, environmental filtering may not be the main assembly process for all land-use types. Land-use types impoverished grassland spider assemblages. However, as spider species and traits were highly dissimilar among land-use types and overall retained about 70% of the species registered in grassland sites, the development of mosaics of different land-use types would favour the conservation of grassland spider diversity at the landscape scale. Finally, our results strongly suggest that patches of protected grasslands should always be part of the productive landscapes, as they would act as a refugee for those native species unable to adapt to disturbed habitats (i.e. land-use types). www.nature.com/scientificreports/