Microhabitat change drives diversification in pholcid spiders
- 1k Downloads
Microhabitat changes are thought to be among the main drivers of diversification. However, this conclusion is mostly based on studies on vertebrates. Here, we investigate the influence of microhabitat on diversification rates in pholcid spiders (Araneae, Pholcidae). Diversification analyses were conducted in the framework of the largest molecular phylogeny of pholcid spiders to date based on three nuclear and three mitochondrial loci from 600 species representing more than 85% of the currently described pholcid genera.
Assessments of ancestral microhabitat revealed frequent evolutionary change. In particular, within the largest subfamily Pholcinae, numerous changes from near-ground habitats towards leaves and back were found. In general, taxa occupying leaves and large sheltered spaces had higher diversification rates than ground-dwelling taxa. Shifts in speciation rate were found in leaf- and space-dwelling taxa.
Our analyses result in one of the most comprehensive phylogenies available for a major spider family and provide a framework for any subsequent studies of pholcid spider biology. Diversification analyses strongly suggest that microhabitat is an important factor influencing diversification patterns in pholcid spiders.
KeywordsMicrohabitat Diversification rates Speciation Leaf dwelling Pholcidae Phylogeny
16S, 18S, 28S: 12S, 16S, 18S, and 28S ribosomal RNA genes, respectively
Akaike information criterion
Bayesian information criterion
Cytochrome oxidase c subunit 1 gene
- GTRCAT, GTR-Γ
General time reversible models of nucleotide substitution with CAT and gamma approximations of rate heterogeneity, respectively
Histone 3 gene
Polymerase chain reaction
- QC, QD, QF, QU
Support measures of the quartet sampling method (quartet concordance, quartet differential, quartet fidelity, and quartet uncertainty, respectively) 
Rapid bootstrap support 
Standard bootstrap support
- SH-like aLRT
Shimodaira-Hasegawa-like approximate likelihood ratio test 
Species numbers differ vastly among groups of organisms – a phenomenon observed at any taxonomic level. Differences in species richness of clades of different age are sometimes explained by the longer time that older clades had to accumulate species (e.g. [1, 2]). However, sister clades which are of the same age per definition often differ substantially in species richness. Therefore, net diversification rates (speciation minus extinction rates) must differ even among closely related groups. Indeed, it was recently suggested that diversification rates may explain most variation in species richness among organisms .
A range of factors that potentially affect rates of diversification are known. Climate and in particular changes of climatic niches among species are thought to be among the main causes of diversification rate differences [4, 5, 6, 7, 8, 9]. On the macro-ecological level, invasions into new adaptive zones play a major role and have promoted some of the largest radiations. So is the diversity of many phytophagous insect lineages likely triggered by the rise of angiosperms in the Cretaceous [10, 11, 12]. Further factors that may affect rates of diversification are differences in body size and size dimorphism [13, 14], sexual selection [15, 16, 17, 18], diet , habitat [20, 21], and parasitism . The total rate of species production is highest in tropical biomes – either caused by increased speciation rates  or simply by the vast number of species that are already present there . Higher rates in the tropics may be caused by increased opportunities for the evolution of reproductive isolation, faster molecular evolution, or the increased importance of biotic interactions .
Recently, microhabitat has been suggested as one of the most important factors that drive variation in diversification rates among vertebrates [20, 25, 26, 27]. Its effect may even supersede that of climatic niche , often changing several times within evolutionary young taxa . It has been proposed that traits like microhabitat that are involved in local-scale resource use (alpha niche) may be more important in explaining patterns of diversification than those related to the broad-scale distribution of species (beta niche), as suggested in analyses across vertebrates and oribatid mites [25, 29, 30]. This might be because alpha-niche traits primarily change over deeper time scales while beta-niche traits (e.g., climate preferences) frequently change on lower time scales, which was shown for amphibians, reptiles, and birds [6, 29, 31, 32, 33].
In the present study we inferred the evolutionary history and plasticity of pholcid spiders’ microhabitats using a newly developed molecular phylogeny based on three nuclear and three mitochondrial DNA markers. Compared to previous studies, we extended the taxon sampling to include 600 species representing more than 85% of described pholcid genera. We also collected microhabitat information first hand for 88% of the examined species. Separate analyses of leg proportions as a proxy for microhabitat allowed a near-complete species coverage. We investigated the evolutionary plasticity of microhabitats by ancestral state reconstructions. Using current species numbers and estimates of extant diversity, we analyzed diversification rates in pholcids and tested the effect of microhabitat on diversification dynamics.
Sampling and molecular lab procedures
The taxon sampling for the phylogenetic analyses is based on previously published phylogenies of the group [9, 36, 39, 44, 45, 46, 47, 50, 51, 52, 53] and aims to include as many species as possible to reduce phylogenetic error [54, 55] and minimize biases of macroevolutionary inferences [56, 57, 58]. The final dataset included 635 pholcid terminals representing 600 pholcid species from all major lineages, covering more than 85% of the described genera and 38% of the described species. Of these, 391 species (423 specimens) were collected and sequenced as part of this study and data for additional 229 species were downloaded from GenBank (Additional file 1: Figure S1, Tables S1, S2). Thirty two outgroup species from GenBank were included based on Dimitrov et al. . Previously missing loci were sequenced for 17 species.
Total genomic DNA was extracted from one to three legs, depending on the size of the specimen, and rarely from whole specimens using Qiagen® DNeasy Blood & Tissue Kit. The Qiagen® Multiplex PCR Kit was used to amplify partial sequences of three mitochondrial (12S rRNA, 16S rRNA, and cytochrome c oxidase subunit 1 [CO1]) and nuclear (18S rRNA, 28S rRNA, and histone 3 [H3]) loci each. 1.6 μl of each primer (Additional file 1, Table S3) and 1.2 to 2.5 μl undiluted DNA were used in 20 μl reaction mixes. The following protocols were used: hot start Taq activation: 15 min at 95 °C; 35 cycles á 35 s denaturation at 95 °C, 60 s annealing at 49 °C (12S, 16S) or 51 °C (18S, 28S, H3) and 60 s elongation at 72 °C; 10 min final elongation at 72 °C. A touch down program was applied for CO1, reducing the annealing temperature by 1° per cycle during the first 15 cycles, starting at 55 °C, and subsequent 25 cycles at 50 °C annealing temperature and 90 s elongation time. PCR products were purified using Qiagen® QIAquick PCR purification kits or 3 M sodium acetate precipitation. Samples were sent to Macrogen (Amsterdam, Netherlands) for forward and reverse Sanger sequencing and edited manually in Geneious v. 7.1.8 (Biomatters; available from ). Primers were cut from the sequences prior to multiple sequence alignment. Contaminations were identified by BLAST searches against the GenBank nucleotide data collection and by the help of preliminary gene trees. Since repeated amplification of single genomic DNA extracts for 28S yielded varying products depending on the PCR program used, suspicious assemblies of species in the corresponding gene trees were evaluated for potential paralog copies of the locus. If such assemblies split apart species from several morphologically well-supported species groups and were not recovered by other loci, the sequences with the conflicting signal were discarded.
We applied the divide-and-conquer realignment technique implemented in SATé-II 2.2.7  which improves multiple sequence alignment particularly when highly variable regions are included. In several iterations, the data are deconstructed to smaller subsets of related specimens (subproblems) which are subsequently merged. A phylogenetic tree based on all loci is simultaneously inferred guiding the alignment of each locus. The break strategy was set to ‘centroid’ to create subproblems with a maximum size of 100 taxa which were aligned with MAFFT-linsi v. 7.299b [61, 62] and subsequently merged with MUSCLE v. 3.7 [63, 64]. Searches for alignment guide trees were done with RAxML v. 8.2.9  on the partitioned supermatrix. Five more iterations were done after SATé failed to find a tree/alignment pair with a higher likelihood score than in the previous iteration. Alignments were manually checked for reverse complement sequences, stop codons, and obvious errors in Aliview v. 1.18.1 .
In order to reduce the amount of missing data, we included 52 chimera taxa (Additional file 1, Table S4). Most of these (48) originated from specimens from the same sampling event (same vial). In four cases, specimens originated from geographically close localities and preliminary analyses indicated a very close relationship. Although some loci had large amounts of missing data, they were included in the analyses since their exclusion may reduce phylogenetic accuracy . This applies in particular to conservative genes like 18S and 28S that might bear information on deeper nodes. In addition to the complete dataset, a dataset with reduced missing data was compiled that included specimens having at least 4 markers successfully sequenced. A third dataset was created by the exclusion of rogue taxa. Rogue taxa can affect phylogenetic inference by having an unstable position in the tree due to ambiguous or insufficient phylogenetic signal [68, 69, 70]. We ran multiple iterations of RogueNaRok  using the web service  until no more rogue taxa were found. In each iteration, rapid bootstrap supports  from 1000 iterations were maximized for a maximum likelihood tree inferred by RAxML v. 8.2.9  based on reduced data from the previous iteration (GTRCAT model; data partitioned by loci). Optimal partition schemes and substitution models for subsequent thorough tree searches were inferred with PartitionFinder v. 2.1.1 [73, 74, 75] for all three datasets separately using a greedy search  and evaluating all available models of evolution.
Searches for the maximum likelihood tree were done multiple times [76, 77] with two different algorithms: RAxML v. 8.2.8  and IQ-TREE v. 1.5.4 . In RAxML, we conducted 100 replicates, each starting from a distinct parsimony tree using partitions based on PartitionFinder and using the GTRCAT model of sequence evolution. We refrained from estimating invariant sites since their inference may conflict with gamma categories inference . We used 25 CAT-gamma categories which sufficiently cover sites with low variation, making an extra parameter superfluous . IQ-TREE implements (partially) terrace-aware algorithm  which efficiently handles gappy alignments and may lead to trees with improved likelihood . Models and partition schemes were again chosen based on PartitionFinder and 100 searches for the maximum likelihood tree were conducted, finally choosing the one with the highest likelihood.
Branch support was assessed with (i) 100 standard bootstrap replicates (SBS), (ii) 1000 rapid bootstrap replicates (RBS) , (iii) Shimodaira-Hasegawa-like approximate likelihood ratio test (SH-like aLRT)  supports, and (iv) quartet sampling . Requirements of SBS (e.g., site independence) are rarely met by empirical data and may be particularly problematic with many missing data [80, 81, 82]. SBS, RBS, and SH-like aLRT supports are known to underestimate the true probability of a clade to be correct, although RBS seems to have a tendency to be less conservative . SBS ≥ 80, RBS ≥ 95, and SH-like aLRT supports ≥80 roughly correspond to a 95% probability for the clade to be correct and are thus considered to present reasonably good support; SH-like aLRT supports < 50 are not representative for true clade support . SH-like aLRT supports are fast to compute but only evaluate alternative topologies around the branch of interest [74, 84] and can thus be interpreted as local supports. Recently published measures of branch support based on quartet sampling  are less affected by missing data. Four statistics, i.e., quartet concordance (QC), quartet differential (QD), quartet uncertainty (QU), and quartet fidelity (QF) are calculated, which measure overall branch support (QC), the potential presence of alternative evolutionary histories (QD), data information content (QU), and individual taxa tendency to produce alternative topologies (similar to rogue taxa; QF). Its ability to distinguish between lack of information and conflicting signal as causes for low branch support offers more comprehensive and specific information on branch support.
All RAxML, IQ-TREE, PartitionFinder, and quartet sampling analyses were conducted on the computing cluster of the Zoological Research Museum A. Koenig.
Diversification analyses depend on the branching pattern inferred by time calibration of the phylogeny. We therefore applied three different dating approaches: non-parametric rate smoothing implemented in treePL v. 1.0 , Bayesian relaxed-clock dating using MCMCtree v. 4.9e , and RelTime, a fast ad hoc approach implemented in MEGA v. 7.0.20 [87, 88, 89]. All methods were applied to the best maximum likelihood (ML)-tree for the complete dataset without changing the topology.
Calibration points were adopted from Dimitrov et al.  without using the fossils for Quamtana and Nephilidae, since their identity or phylogenetic position has been contested [90, 91]. Stem ages were calibrated with minimum fossil ages (Additional file 1, Table S5). The Macaronesian clade of Pholcus was calibrated with a maximum age of 14 My [45, 92]. Fossil age uncertainty was implemented in MCMCtree using heavy tailed Cauchy distributions.
For treePL, a ‘prime’ analysis with smoothing = 1 was done to assess the best optimization method using ‘thorough’ estimation. Cross-validation was used to estimate the best fitting smoothing value (by 10 iterations in a range from 1000 to 0.000001). The smoothing value affects how strong rate variation among branches is penalized. Final analyses were done with ‘thorough’ optimization and increased number of penalized likelihood total optimization iterations (5, default = 2) and increased number of penalized likelihood simulated annealing iterations (10,000, default = 5000).
RelTime  is a very fast method which was originally intended for the estimation of relative divergence times in large phylogenies, but can also assess absolute times. It was shown to outperform other non-Bayesian methods when high rate increases in specific clades are present . However, a recent study revealed shortcomings of RelTime in relaxing the clock among internal branches of specific datasets, arguably because it essentially does infer divergence times under a strict clock . Results were thus checked for loss in variation of relative branch rates at deeper node ages. We estimated ‘all clocks’ under the GTR-Γ model, using all sites.
MCMCtree uses an approximation to speed up likelihood calculations and thus outperforms BEAST in terms of speed. We used the independent rate model to avoid unrealistic rates [94, 95, 96]. The birth-death tree prior was set to a uniform distribution of nodes (BDparas = 1 1 0). The locus rate prior (rgene_gamma) was set to a Dirichlet, hence posterior time estimates are insensitive to the rate prior . We set a diffuse gamma distribution G(1, 7) with mean 0.14, which was derived from the average pairwise genetic distances between the six loci of two distant species (Gertschiola macrostyla (S434) and Holocnemus caudatus (S435)), assuming a divergence time of about 210 Mya . The prior for σ2 was set to G(1, 10), indicating serious violation of the strict clock . The Markov Chain Monte Carlo (MCMC) analysis was run on the ZFMK computing cluster for 2e5 generations, sampling every 20 generations after a burnin phase of 2e4 generations.
Data and trees were plotted with ggplot2  and ggtree v. 1.9.2 , respectively. Efforts to infer divergence times with the widely used BEAST software  were unsuccessful due to lack of convergence of the MCMC chain.
To reduce the bias introduced by unequal sampling of clades, diversification analyses (Additional file 1, Figure S2) were conducted based (1) on the number of currently described species and (2) on an estimate of actual species richness. For the latter, species numbers of 102 taxonomic entities (Additional file 1, Table S6; mostly species groups, genera, or groups of genera) were updated by adding undescribed species available in collections and by accounting for obviously misplaced species and then multiplied by 2 or 3 depending on their assignment to one of two categories: (i) taxa from temperate regions, with limited distribution, focused collection, low endemism, easy to find (multiplied by 2); (ii) taxa from tropical regions, with wide distribution, large sampling gaps, high local endemism, difficult to collect (multiplied by 3).
To evaluate the influence of microhabitat on diversification (speciation + extinction) rates, each species was assigned to one main habitat type, i.e., “ground”, “leaf”, or “space” (represented by 206, 174, and 178 specimens, respectively) based on direct field observations. Since this information was available for only 88% of the species, the usage of the metatarsus to tibia ratio of the first leg as a proxy of microhabitat was evaluated. The correlation of this ratio with microhabitat was tested using a phylogenetic generalized least-squares analysis  as implemented in the R  package ape v. 4.1  in conjunction with nlme v. 3.1–128 . Fits of Ornstein-Uhlenbeck [106, 107] and Brownian Motion [108, 109] models of trait evolution were evaluated. Ancestral microhabitats were estimated by maximum likelihood using the ape-function ace and the underlying expm-package  and by maximum parsimony (function MPR in ape). Blomberg’s K  and Pagel’s lambda  were calculated as measures of phylogenetic signal of the metatarsus to tibia ratio of the first leg. All diversification analyses were done on the dated trees inferred with different methods. Outgroups and duplicate species were pruned from the trees.
Models of diversification rates that were used with MuSSE. All models were evaluated for all dated trees (MCMCtree, treePL, and RelTime), choosing the best fitting one by the AIC value
no difference in speciation or extinction rates between microhabitats
λ1, λ2, λ3
speciation rates differ among microhabitats
no difference in speciation rate of leaf and space, but difference to ground
μ1, μ2, μ3
extinction rates differ among microhabitats
λ1, λ2, λ3
μ1, μ2, μ3
speciation and extinction rates differ among microhabitats
Data for the calculation of the leg ratio were available for 91% of the species. Using the leg ratio as a proxy for habitat preference, speciation rates were estimated as a function of this ratio using QuaSSE . Linear, sigmoidal, and hump shaped speciation functions were evaluated with constant extinction rate. All models were estimated with and without the drift parameter, which describes the directional component of character evolution due to selection or any other within-lineage process that has a directional tendency .
Additionally, shifts in speciation rates were inferred with BAMM v. 2.5.0 [14, 119, 120, 121], using the currently described and the estimated species numbers for the calculation of clade-specific sampling fractions. Because of the tree size, 50 speciation rate shifts were expected a priori; other prior values were set using BAMMtools v. 2.1.6 . Rate shifts were allowed to occur in clades with a minimum size of two taxa. The Metropolis- coupled MCMC chain was run for 10 Mio generations, sampling every 1000 generations after a burnin of 20%. BAMMtools was used to visualize the results.
Cross validation of the smoothing parameter in treePL favored very small values (10− 6), indicating strong rate heterogeneity across the tree. Absolute time estimates conspicuously differed among the applied methods (Additional file 1: Figures S10 – S13).
Phylogenetic relationships inferred in the present study largely confirm previous findings based on morphological and molecular data [36, 38, 39, 40, 44, 45, 46, 47, 49, 53, 125, 126, 127, 128]. Species groups that were previously identified based on morphological apomorphies were mostly recovered (Additional file 1: Table S6). However, several low support values and the presence of unstable taxa (whole clades or single rogue-species) lead to uncertainties, in particular in deeper relationships. Quartet sampling scores  suggest the presence of both data with low phylogenetic information content and conflicting signal (Additional file 1: Figure S9). The existence of paralog copies of the nuclear ribosomal array (including 18S and 28S rRNAs) may also act as a confounding factor. Paralogs of these gene arrays are also known from other arachnid groups [129, 130, 131], emphasizing the need for future phylogenomic scale datasets and approaches that explicitly address confounding factors and processes [132, 133]. A detailed systematic discussion will be published in a standalone article .
Due to the limited fossil evidence for the group and deviating estimates of divergence times across methods, estimates of lineage ages could not be proposed in the present study. A potential inability of RelTime to relax the molecular clock between internal branches  was not evident in our analysis.
Evolutionary shifts of microhabitat
The present analyses with strongly increased species sampling corroborate indications from previous phylogenetic studies [45, 46, 47] that microhabitat frequently changes even among closely related species (Fig. 3, Additional file 1: Figures S10 – S12). Also the preference of the Brownian Motion model and low alpha parameter values of the Ohrnstein-Uhlenbeck model for microhabitat (PGLS regressions of microhabitat and leg ratio) indicate evolutionary instability of microhabitat use in pholcid spiders.
High phylogenetic signal of a trait might be interpreted as indication for the trait to change at deeper timescales . Thus, the high phylogenetic signal in the leg-ratio approximation for microhabitat (Additional file 1: Table S9) might be interpreted as indication for less frequent change of microhabitat. Given the correlation of microhabitat and diversification rates, this would be in concordance with the idea that traits that differ at deep timescales may be more important for diversification . However, regression analysis and plots of the distribution of leg-ratios clearly reveal increased values in leaf and space living taxa (Additional file 1: Figure S14, Table S8). Thus, the similarity in the ratio among leaf dwellers and space living species likely biases the phylogenetic signal calculations towards higher values since changes from space living to leaf dwelling and vice versa are not captured; i.e., the similarity of leaf dwellers and space living species artificially increase phylogenetic signal by increasing the probability of closely related species to resemble each other. Additionally, bimodalities were present in the ratio distributions of each microhabitat (Additional file 1: Figure S14). These were likely caused by different leg proportions among species with equal microhabitat preference in different phylogenetic lineages. Thus, similarity within a phylogenetic lineage is increased and phylogenetic signal further increases. The phylogenetic signal of the leg-ratio might thus overestimate phylogenetic signal of microhabitat preference.
Increased diversification in leaf and space microhabitats
The present study suggests that microhabitat influences rates of diversification in pholcid spiders (Fig. 4). Despite the variance in absolute divergence times that we observed among methods (Additional file 1: Figure S13), relative estimates of diversification rates were largely concordant (Fig. 4, Additional file 1: Figure S16). Thus, their comparison among microhabitats is justified. In the context of microhabitat, accelerated diversification in pholcid spiders seems to be related to two factors: (i) frequent microhabitat change in a phylogenetic sense and (ii) leaf or space dwelling. Microhabitat change might facilitate the coexistence of many species on a local scale (e.g. by resource partitioning or intraguild predation escape [26, 134]) and thus explain its relation to diversification rates (Additional file 1: Figure S15). The causality of the observed relation between species numbers and microhabitat, however, remains subject to future studies. A leaf dwelling or space living lifestyle is associated with several factors that differentiate it from ground living conditions. Among those are prey availability and protection from predators which is also reflected in body color (leaf dwellers are pale whitish to green while ground dwellers tend to be brown). Leaf dwelling implies varying sizes and shapes of leaves that may require different webs  and consequently further differences in vibratory signal conduction, humidity, etc. Such factors might drive increased rates of speciation or reduce extinction [135, 136], e.g. by sexual selection, predator-prey interactions, or competition. Given the frequent change of microhabitat in the evolutionary history of pholcids, we do not expect that minor topological changes of the tree will alter the general conclusions of the present study.
Current methods to infer shifts in diversification rates [119, 137, 138, 139] are known to underestimate the number of speciation rate shifts on a phylogeny [119, 138]. The consistent inference of a speciation rate shift by BAMM in Pholcinae, where most shifts to leaf dwelling were observed, thus underlines the impact of leaf microhabitat on speciation rates. Inferences of ancestral microhabitat actually located a shift to leaf dwelling in close phylogenetic vicinity of the respective branch (Fig. 3, Additional file 1: Figures S10 – S12).
The inclusion of a world-wide sampling might confound estimates of speciation rates in microhabitats by potentially increased diversification rates in tropical areas . Our data did not support higher diversification rates in tropical biomes since high speciation rates were for instance also found in the Mediterranean biome (Additional file 1: Figure S17). A confounding effect on the inference of diversification rates in microhabitats was excluded by analyzing species from tropical broadleaf forest only, from where the vast majority of species originated (Additional file 1: Figure S16). Exceedingly high extinction rates that were inferred for some biomes (Additional file 1: Figure S17) were likely caused by lacking statistical power since they were only covered by less than three species  and by the general difficulty of extinction rate inference from phylogenies of only extant taxa [58, 140, 141, 142].
The present study reveals frequent evolutionary changes among pholcid spider microhabitats and explains the remarkable variation of the associated morphology (such as body size, leg proportions and color). While additional factors are likely to play a role in the diversification of pholcid spiders, the increase in net diversification rate in leaf dwelling but also in space living species emphasizes the importance of microhabitat for the evolution of high species richness. This is further supported by the observed shift in speciation rate in the subfamily Pholcinae that includes a large percentage of leaf dwelling taxa. In addition, our analysis of six molecular loci resulted in one of the most comprehensive phylogenies available for a major spider family and provide a framework for any subsequent studies of pholcid spider biology. Given the problems likely encountered due to multiple independently evolving nuclear ribosomal arrays in lycosid, jumping, and pholcid spiders, future phylogenetic studies should rely on genomic scale data, which allows to specifically address gene orthology. The general conclusions of the present study, however, are unlikely to be affected by minor topological changes in the presented phylogeny, and provide a strong argument favoring microhabitat as a major diversifying factor in pholcid spiders.
We thank S. Benjamin, I. Agnarsson, F. Lopez Osorio, N. Lopez, and C. Ribera who allowed us to use their unpublished pholcid DNA sequences and countless other colleagues and collaborators who either provided specimens or helped during expeditions, in particular N.M. Abdul Aziz, S. Aharon, J. Altmann, M. Alves Dias, S. Benjamin, A. Brescovit, S. Bumrungsri, L.S. Carvalho, D. Court, R. Duncan, G. Eilu, H. El-Hennawy, E. Gavish-Regev, A.R.M. Ghazali, A. Giupponi, C. Grismado, C. Griswold, C. Haddad, D. Harms, S. Huber, P. Jäger, R. Joqué, J.K.H. Koh, M. Komnenov, G. Kovacs, J. Kral, H.-J. Krammer, A. Kury, P. Kwapong, C.-W. Lai, P. Le Gall, C. Leh Moi Ung, J. Malumbres-Olarte, Y. Marusik, J.F. Mavoungou, P. Michalik, E. Monsanto,O.M. Nuñeza, A. Pérez-González, B. Petcharad, M. Ramirez, R. Raven, C. Rheims, J. Ricetti, M. Sidibe, M. Siyam, S. Sutono, I-M. Tso, R. Victor, O. Villarreal, C. Warui, Z. Yao. We also thank two anonymous reviewers for their valuable comments on the manuscript.
This research was mainly supported by the German Research Foundation (DFG project HU 980/11–1). DD received further support as Mercator Fellow under DFG project HU 980/11–1 and from the Danish National Research Foundation grant DNRF96 to the Center for Macroecology Evolution and Climate. AVM received further support from Catedras Consejo Nacional de Ciencia y Tecnología (CONACyT, project No. 59). None of the funding organizations interfered with the design of the study; the collection, analysis, or interpretation of data; nor with the writing of the manuscript.
Availability of data and materials
All newly generated sequences are deposited in GenBank (Accession Numbers MG267426 – MG269191; Additional file 1: Table S1). Newly extracted genomic DNA is stored in the Biobank at the Zoological Research Museum Alexander Koenig, Bonn (ZFMK) (Additional file 1: Table S1) and voucher specimens are deposited in the collections of the ZFMK. Leg ratio measurements and microhabitat assignments of all specimens are supplied as Additional file 2.
Study concept and design: JE, BAH, DD. Field work: BAH, AVM. Molecular lab work: JE, AVM. Phylogenetic and diversification analyses: JE, DD. Manuscript writing and figures: JE, BAH, DD. Final approval of the manuscript: JE, DD, AVM, BAH. All authors read and approved the manuscript
We obtained permission from the Zoological Research Museum A. Koenig, Bonn (ZFMK) to access, loan and dissect the material in the collections that was used in this study.
Consent for publication
The authors declare that they have no competing interests.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
- 9.Valdez-Mondragón A, Francke OF. Phylogeny of the spider genus Ixchela Huber, 2000 (Araneae: Pholcidae) based on morphological and molecular evidence (CO1 and 16S), with a hypothesized diversification in the Pleistocene. Zool J Linnean Soc. 2015;175:20–58. https://doi.org/10.1111/zoj.12265.CrossRefGoogle Scholar
- 15.Ritchie MG. Sexual selection and speciation. Annu Rev Ecol Evol Syst. 2007;38:79–102. https://doi.org/10.1146/annurev.ecolsys.38.091206.095733.CrossRefGoogle Scholar
- 34.Huber BA. Life on leaves: leaf-dwelling pholcids of Guinea, with emphasis on Crossopriza cylindrogaster Simon, a spider with inverted resting position, pseudo-eyes, lampshade web, and tetrahedral egg-sac (Araneae: Pholcidae). J Nat Hist. 2009;43:2491–523. https://doi.org/10.1080/00222930903207876.CrossRefGoogle Scholar
- 35.Huber BA, Schütte A. Preliminary notes on leaf-dwelling Metagonia spiders (Araneae: Pholcidae) in the Esquinas rainforest near La Gamba, Costa Rica: leaf preference, mimesis, and web structure. Contrib to Nat Hist. 2009;12:681–97.Google Scholar
- 36.Huber BA. Revision and cladistic analysis of Pholcus and closely related taxa (Araneae, Pholcidae). Bonner Zool Beiträge. 2011;58:1–509.Google Scholar
- 37.Huber BA. Pholcidae. In: Roig-Juñent S, Claps LE, Morrone JJ, editors. Biodivers. Artrópodos Argentinos, Vol. 3. 3rd ed. Sociedad Entomológica Argentina; 2014; p. 131–140.Google Scholar
- 38.Huber BA. Revision and cladistic analysis of the Afrotropical endemic genus Smeringopus Simon, 1890 (Araneae: Pholcidae). Zootaxa. 2012;3461:1–138.Google Scholar
- 39.Huber BA. New World pholcid spiders (Araneae: Pholcidae): a revision at generic level. Bull Am Museum Nat Hist. 2000:1–341. https://doi.org/10.1206/0003-0090.
- 43.Natural History Museum Bern. World Spider Catalog version 18.5. http://wsc.nmbe.ch. Accessed 12 Sept 2017.
- 46.Huber BA, Carvalho LS, Benjamin SP. On the New World spiders previously misplaced in Leptopholcus: molecular and morphological analyses and descriptions of four new species (Araneae:Pholcidae). Invertebr Syst. 2014;28:432–50. https://doi.org/10.1071/is13050.
- 50.Astrin JJ, Misof B, Huber BA. The pitfalls of exageration: molecular and morphological evidence suggests Kaliana is a synonym of Mesabolivar (Araneae: Pholcidae). Zootaxa. 2007;1646:17–30.Google Scholar
- 59.Biomatters. Geneious - Molecular Biology and NGS Analysis Tool. http://www.geneious.com. Accessed 3 Dec 2017.
- 69.Sanderson MJ, Shaffer HB. Troubleshooting molecular phylogenetic analyses. Annu Rev Ecol Syst. 2002;33:49–72. https://doi.org/10.1146/annurev.ecolsys.33.010802.15059. CrossRefGoogle Scholar
- 70.Wilkinson M. Majority-rule reduced consensus trees and their use in bootstrapping. Mol Biol Evol. 1996;13:437–44. https://doi.org/10.1093/oxfordjournals.molbev.a025604. CrossRefPubMedGoogle Scholar
- 71.Exelixis Lab. RogueNaRok. 2015. http://rnr.h-its.org/. Accessed 10 July 2017.
- 78.Yang Z. Computational molecular evolution (Oxford series in ecology and evolution). Oxford, UK: Oxford University Press; 2006.Google Scholar
- 80.Pease JB, Brown JW, Walker JF, Hinchliff CE, Smith SA. Quartet Sampling distinguishes lack of support from conflicting support in the plant tree of life. bioRxiv. 2017; https://doi.org/10.1101/148536.
- 92.Dimitrov D, Arnedo MA, Ribera C. Colonization and diversification of the spider genus Pholcus Walckenaer, 1805 (Araneae, Pholcidae) in the Macaronesian archipelagos: evidence for long-term occupancy yet rapid recent speciation. Mol Phylogenet Evol. 2008;48:596–614. https://doi.org/10.1016/j.ympev.2008.04.027.CrossRefPubMedGoogle Scholar
- 103.R Core Team. R: A Language and Environment for statistical computing. Vienna: R Foundation for Statistical Computing; 2016.Google Scholar
- 105.Pinheiro J, Bates D, DebRoy S, Sarkar D, R Core Team. nlme: Linear and Nonlinear Mixed Effects Models. 2016;Google Scholar
- 106.Phylogenies FJ, Characters Q. Annu Rev Ecol Syst. 1988;19:445–71. https://doi.org/10.1146/annurev.es.19.110188.002305 CrossRefGoogle Scholar
- 109.Edwards AWF, Cavalli-Sforza LL. Reconstruction of evolutionary trees. In: Heywood VH, McNeill J, editors. Phenetic phylogenetic Classif. London: Systematics Assocaition Publications; 1964. p. 67–76.Google Scholar
- 110.Goulet V, Dutang C, Maechler M, Firth D, Shapira M, Stadelmann M. expm: Matrix Exponential. 2015; p. R package version 0.999–990.Google Scholar
- 115.QGIS - A Free and Open Source Geographic Information System.Google Scholar
- 116.Tveite H. The QGIS NNJoin Plugin. https://plugins.qgis.org/plugins/NNJoin/. Accessed 3 Dec 2017.
- 123.Huber BA, Eberle J, Dimitrov D. The phylogeny of pholcid spiders (Araneae, Pholcidae): a critical evaluation of relationships suggested by molecular data. ZooKeys. In press.Google Scholar
- 130.Murphy NP, Framenau VW, Donnellan SC, Harvey MS, Park Y-C, Austin AD. Phylogenetic reconstruction of the wolf spiders (Araneae: Lycosidae) using sequences from the 12S rRNA, 28S rRNA, and NADH1 genes: implications for classification, biogeography, and the evolution of web building behavior. Mol Phylogenet Evol. 2006;38:583–602. https://doi.org/10.1016/j.ympev.2005.09.004.CrossRefPubMedGoogle Scholar
- 131.Vink C, Dupérré N, McQuillan B. The black-headed jumping spider, Trite planiceps Simon, 1899 (Araneae: Salticidae): redescription including cytochrome c oxidase subunit 1 and paralogous 28S sequences. New Zeal J Zool. 2011;38:317–31. https://doi.org/10.1080/03014223.2011.613939.CrossRefGoogle Scholar
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.