Phylogenetic analysis of Micrathena and Chaetacis spiders (Araneae: Araneidae) reveals multiple origins of extreme sexual size dimorphism and long abdominal spines

Choice of terminal taxa

At least two species of each Micrathena species group and four species of Chaetacis were chosen so that the monophyly of these could be tested. The M. militaris and M. kirbyi groups were represented by more species as they are morphologically more diverse. Genera of Araneidae that are closely related to Micrathena and Chaetacis, according to Levi (1985, 1996) and Scharff & Coddington (1997), were chosen as outgroups. Terminal taxa were chosen to cover the morphological and geographical variation of the genera and restricted to species known from both sexes. The following species were selected: outgroups: Actinosoma pentacanthum (Walckenaer, 1842), Argiope argentata (Fabricius, 1775), Araneus venatrix (C.L. Koch, 1838), Aspidolasius branicki (Taczanowski, 1879), Cyclosa fililineata Hingston, 1932, Enacrosoma anomalum (Taczanoswki, 1873), Gasteracantha cancriformis (Linnaeus, 1758), Hypognatha belemLevi, 1996, Wagneriana dimastophora (Mello-Leitão, 1940), and Xylethrus superbusSimon, 1895. Chaetacis: Chaetacis aureola (C.L. Koch, 1836), Chaetacis cornuta (Taczanowski, 1873), Chaetacis picta (C.L. Koch, 1836), Chaetacis bandeiranteMagalhães & Santos, 2011. Micrathena species groups: M. gracilis group: M. gracilis, Micrathena horrida (Taczanoswki, 1873); M. guerini group: M. bifida (Taczanowski, 1879), M. guerini (Keyserling, 1864), Micrathena nigrichelis Strand, 1908; M. kirbyi group: Micrathena clypeata (Walckenaer, 1805), Micrathena digitata (C.L. Koch, 1839), Micrathena excavata (C.L. Koch, 1836), Micrathena fissispina (C.L. Koch, 1836), Micrathena furcula (O.P.-Cambridge, 1890), Micrathena gaujoni Simon, 1897, Micrathena guanabara Levi, 1895, M. kirbyi (Perty, 1833), Micrathena patruelis (C.L. Koch, 1839), Micrathena plana (C.L. Koch, 1836), Micrathena triserrataF.O.P.-Cambridge, 1904; M. lepidoptera group: Micrathena decorata Chickering, 1960, Micrathena lepidoptera Mello-Leitão, 1941; M. militaris group: Micrathena cyanospina (Lucas, 1835), Micrathena furcata (Hahn, 1822), Micrathena lata Chickering, 1960, Micrathena sagittata (Walckenaer, 1842), Micrathena swainsoni (Perty, 1833); Micrathena schreibersi group: Micrathena balzapambaLevi, 1985, M. schreibersi (Perty, 1833), Micrathena spitziMello-Leitão, 1932, Micrathena vigorsiPerty, 1833; Micrathena spinosa group: Micrathena brevipes (O.P.-Cambridge, 1890), Micrathena pichinchaLevi, 1985, M. spinosa; M. triangularispinosa group: Micrathena acuta (Walckenaer, 1842), Micrathena evansi Chickering, 1960, Micrathena jundiaiLevi, 1985, Micrathena schenkeli Mello-Leitão, 1939; incertae sedis: M. funebris, M. pungens.

Argiope argentata was chosen to root the tree as this genus was considered a basal araneid by Levi (1983) and a close relative of the spiny orb-weavers by Scharff & Coddington (1997).

Studied specimens (Appendix 2) come from the following institutions (names of curators in parentheses): AMNH, American Museum of Natural History, New York (N.I. Platnick); IBSP, Instituto Butantan, São Paulo (I. Knysak); INPA, Instituto Nacional de Pesquisas da Amazônia, Manaus (A.L. Henriques); MCZ, Museum of Comparative Zoology, Harvard University, Cambridge (G. Giribet); MNHN, Muséum National d'Histoire Naturelle, Paris (C. Rollard); MPEG, Museu Paraense Emílio Goeldi, Belém (A.B. Bonaldo); MNRJ, Museu Nacional do Rio de Janeiro, Rio de Janeiro (A.B. Kury); MZSP, Museu de Zoologia da Universidade de São Paulo, São Paulo (R. Pinto-da-Rocha); SINMNH, Smithsonian Institution, National Museum of Natural History, Washington, (J.A. Coddington); NRM, Naturhistoriska Riksmuseet, Stockholm (G. Lindberg); UFMG, Coleções Taxonômicas da Universidade Federal de Minas Gerais, Belo Horizonte (A.J. Santos).

Character survey

Many characters were taken from Levi (1985), especially the ones used to define Micrathena species groups, and Scharff & Coddington (1997). Besides this, several new characters are here proposed (see character list in Appendix 1). Autapomorphic characters were included only in the Bayesian analyses because they are not decisive in parsimony analyses (Bryant, 1995), and all characters were treated as unordered. Coloration character scoring follows the Pantone (http://www.pantone.com) colour palette and colour codes are cited every first time a colour is mentioned in a character description. Besides discrete characters, continuous characters were also included in the analysis using the technique described in Goloboff, Mattoni & Quinteros (2006). Continuous characters are especially suitable for studying traits that cannot be sorted into discrete states, such as sexual size dimorphism and length of abdominal spines. They can also be used to infer the phylogeny itself (Wiens, 2001; Goloboff et al., 2006; de Bivort, Clouse & Giribet, 2010 and references therein), and have been used in chelicerate phylogenetic analysis recently (see Hendrixson & Bond, 2009; de Bivort et al., 2010; Lopardo, Giribet & Hormiga, 2011). We took measurements from one to 18 males (mean = 4.43) and from one to 19 females (mean = 7.94) per species depending on material available. Carapace length was expressed in millimetres and all other continuous characters were expressed as ratios to carapace length (see de Bivort et al., 2010; Lopardo et al., 2011). This correction is necessary to remove the effect of size, avoiding spurious groupings containing similar sized species, which would be a violation of the principle of independence amongst characters. Whenever a structure was absent in the taxon being measured, the character was scored as 0 for the continuous characters associated with it. We took this approach because we believe that in cases where a species has a structure very reduced in length (e.g. a very short spine or carapace rim), this character state is more closely related to the absence of the structure than to the fully developed structure. Nevertheless, we also tested data matrices with this character state scored as inapplicable, with no changes in the topologies obtained.

Only adult specimens were considered in the analysis. They were examined and measured immersed in 75% ethanol under an Olympus SZ40 stereoscopic microscope. For examination of male genitalia, especially in the cases where the terminal apophysis covers part of the palpus in mesal view, palpal bulbs of some species were expanded after a few minutes of immersion in a saturated KOH solution followed by a wash in distilled water. For observation of internal structures of females and booklung covers, these parts were removed with entomological pins and treated with a pancreatin solution (prepared as described in Álvarez-Padilla & Hormiga, 2008). Booklung covers were mounted in temporary slides, in 75% ethanol and without coverslip, and observed under an Olympus CX40 light microscope in search of stridulatory files. Genitalia were examined in the standard position for Araneidae (Levi, 1985, 2002) except where stated otherwise. Male specimens of M. decorata could not be obtained and the scoring of masculine characters for this terminal was carried out from Levi's (1985) measurements and illustrations. However, because of the limitations in this approach, some structures could not be observed and the corresponding characters were scored as missing entries in the matrix.

Illustrations were created either on a Motic K400 or on a Leica M205c stereoscopic microscopes, each equipped with a camera lucida. Specimen photographs were taken on a DFC295 digital camera in a Leica M205c stereoscopic microscope and composed as a single, multifocal image using Leica Application Suite. Specimens for scanning electron microscopy (SEM) were dried under a bulb lamp, fixed to aluminium stubs with conductive adhesive copper tape, sputter coated with 10 nm of gold, and examined and photographed in a Quanta 2000 SEM microscope in Centro de Microscopia da UFMG.

Parsimony analysis

The discrete character matrix was assembled in the Character Matrix Editor module of MESQUITE 2.74 (Maddison & Maddison, 2010) and exported in the .tnt format. Continuous characters were inserted manually in this file using Microsoft NOTEPAD. Searches for most parsimonious trees (MPTs) were carried out in TNT 1.1 (Goloboff, Farris & Nixon, 2008), using heuristic searches with 10 000 random addition sequences (RAS) followed by rounds of tree bisection and reconnection (TBR), retaining 100 trees per replication. Searches for maximum fit trees (MFTs) were carried out using the implied weighing method (Goloboff, 1993, 1995). As the k constant affects the resulting trees and the choice of this value is arbitrary, we used a script described in Mirande (2009) to search for MFTs in 11 different weighting schemes. Then, a majority-rule consensus was inferred from all the obtained trees and the lowest entire k-value that yielded the most similar tree to this consensus tree was chosen to perform the character optimization. Retention and consistency indexes were calculated using the wstats.run script, which is part of the TNT package. Branch supports were calculated using the Bremer decay index (Bremer, 1994) and symmetric resampling (P = 33; Goloboff et al., 2003). For the latter, calculations were carried out based on 1000 pseudoreplicates with 1000 RAS followed by TBR and saving 1000 trees per replication, and support values are given in frequency differences (GC; Goloboff et al., 2003).

Bayesian analysis

Apart from the parsimony analyses, we analysed our matrix using Bayesian inference of phylogeny using a likelihood model suitable for morphological data (Lewis, 2001). We decided to do so for two main reasons. The first is that this approach permits branch lengths to be estimated, an outcome not possible with parsimony analyses alone. The second is that the likelihood approach takes these branch lengths in consideration and helps in the resolution of clades in the presence of high levels of homoplasy (Lewis, 2001). Thus, by combining both approaches we could test whether our parsimony analysis could be under the influence of long-branch attraction. The likelihood model for morphology only handles discrete data (Lewis, 2001). Thus, eight of the 18 continuous characters were recoded as discrete (see details in the character list, Appendix 1). The remaining continuous characters were not considered in this analysis because they could not be coded in unambiguous discrete states. Two analyses were performed in MrBayes 3.1.2 (Ronquist & Huelsenbeck, 2003), one under the Mkv model (Lewis, 2001) and one under the MkvΓ model (Nylander et al., 2004). The Γ parameter allows rate variation amongst characters to follow a gamma distribution, which may lead to homoplasy-penalizing effects that are analogous to those of implied-weighted parsimony (Nylander et al., 2004). For each of these analyses, two independent runs with four chains (one cold and three heated) each were run for five million generations, sampling trees every 250 generations. We checked for stationarity of the chains using the standard deviation of the split frequencies and plotting the log-likelihood of the trees against generation using TRACER 1.5 (Rambaut & Drummond, 2009). Twenty-five per cent of the trees were discarded as burn-in and the remaining were used to calculate the posterior probabilities. The consensus trees containing branch lengths and node posterior probabilities were plotted in FigTree 1.3.1 (Rambaut, 2009).

Character optimization and ancestral states reconstruction

Character optimization was carried out in TNT and in WinClada 1.00.08 (Nixon 1999–2002). We reconstructed ancestral states of the continuous characters corresponding to posterior spine lengths and male and female carapace lengths using the Trace Character History and Reconstruct Ancestral States modules from MESQUITE 2.74 (Maddison & Maddison, 2010) under a squared-change parsimony model.

Spine and apodeme homology

The term ‘spine’ in this paper will usually refer to the pointed, sclerotized structures that project from the abdomen of females. The pointed, rigid, movable setae in the legs will be termed ‘macrosetae’. We grouped Micrathena spines in anterior, lateral, and posterior spines [a classification similar to the one used by Simon, (1895: 849–850), which included épines antérieures, dorsales, angulares, and postérieures]. The anterior pair of spines lies in the anterior edge of the abdomen and is anteriorly directed (Figs 1, 2, AS). All pairs of lateral spines are usually laterally directed. The first pair of lateral spines lies in the same line or anteriorly in relation to the first pair of primary apodemes (Figs 1, 2, LS1), the second pair of lateral spines is located between the first and second pairs of primary apodemes (Figs 1, 2, LS2), and the third pair of lateral spines is located at the same line or posterior to the second pair of primary apodemes (Figs 1, 2, LS3). The fourth pair of lateral spines is located at the base of the first pair of posterior spines. The first pair of posterior spines (Figs 1, 2, PS1) is usually dorsolaterally directed and always the largest, except in the M. spinosa group, in which the second pair of posterior spines is the largest. The second and third (respectively PS2 and PS3 in Fig. 2) pairs of posterior spines are posteriorly directed.

The primary apodemes (indicated by arrows in Fig. 1) lie in the centre of the abdominal dorsum and consist of two or three pairs, being present in most araneid spiders examined. Secondary apodemes are here defined as the sclerotized disks that surround the primary apodemes [they correspond to what Scharff & Coddington (1997) called ‘sigillae’, char. 56].

Phylogenetic analyses

For each of the 51 taxa, up to 146 characters were scored, of which 18 are continuous, 111 are binary, and 17 are multistate. The character matrix covers mostly the female body (67 characters) and male genitalia (42), but also includes 14 female genitalic and 23 male somatic characters. All characters are described in Appendix 1. Additionally, the matrixes in .tnt and .nex files, used as inputs to TNT and MrBayes, respectively, are available as supporting information (Appendices S1, S2) or upon request to the first author.

A single MPT [length = 575.3; consistency index (CI) = 0.305; retention index (RI) = 0.693; Fig. 3] was obtained through the equal weights parsimony analysis. Following the procedure described above, the value of k for the implied weighted parsimony analysis of this data set was 5. This analysis yielded a single MFT (fit = 105; length = 579.718; CI = 0.303; RI = 0.689; Fig. 4). The Bayesian analyses yielded trees that are very similar to the ones obtained through parsimony analyses (Figs 5, 6). The only differences are the position of M. kirbyi in the Mkv run (Fig. 5), C. bandeirante being sister to C. picta + C. aureola in the MkvΓ run (Fig. 6) and Micrathena ruschii being sister to M. excavata in both runs (Figs 5, 6). Overall nodal support was also similar between Bayesian and parsimony analyses. The Mkv model yielded a tree that is slightly better resolved within the outgroups although it has some polytomies within Micrathena (Fig. 5), whereas the MkvΓ model shows the inverse pattern (Fig. 6).

As all the topologies obtained were very similar, character optimizations were performed on the trees obtained through weighted parsimony analysis only, except when stated otherwise. In all of the four trees, Micrathena plus Chaetacis form a monophyletic, well-supported group. In the weighted analysis, this clade is supported by 12 non-ambiguous synapomorphies (Fig. 7): second posterior pair of spines (char. 10), a projecting cone surrounding the spinnerets (char. 11), projected femoral setal bases (char. 42: 1), third pair of lateral spines (char. 53: 1), sclerotized ring around the spinnerets (char. 65: 1), two median ventral apodemes in the abdomen [a small, anterior (char. 67: 1), and a large, median one (char. 68: 1)], oval male carapace (char. 89: 1), rectangular male abdomen (char. 101: 1), conductor lobe (char. 125: 1), basal projection of the median apophysis (char. 135: 1) and a well-developed paracymbium (char. 139: 1). Micrathena, however, is paraphyletic in relation to Chaetacis, which was recovered as the sister-group of M. pungens. Chaetacis is monophyletic and supported by four non-ambiguous synapomorphies (Fig. 7): curved post-ocular macrosetae absent (char. 22: 1), cephalic tubercles or spines (char. 23: 1), carapace dimples modified into sulci (char. 34: 1), and basal projection of the median apophysis tooth-shaped (char. 136: 3). Based on these results, we here propose that Chaetacis be considered a junior synonym of Micrathena (see Discussion and Taxonomy below).

The two analyses yield very different results when the outgroups are considered. In both of them, Gasteracantha, Xylethrus, Aspidolasius, and Hypognatha form a monophyletic group, hereafter denominated the ‘gasteracanthoid clade’. In the equal weights analysis, Cyclosa is sister to Micrathena + Chaetacis, supported by the high thoracic region (char. 27: 1), glabrous pleura (char. 35: 1), male palpus with a tegular projection (char. 108: 1), and a conductor lobe (char. 125: 1). These three genera are closer to Enacrosoma and Wagneriana + Actinosoma than to the gasteracanthoid clade. In the implied weights analysis, Micrathena + Chaetacis was recovered as sister to the gasteracanthoid clade. The group is supported by three non-ambiguous synapomorphies (Fig. 7): sclerotized female abdomen (char. 44: 1), wide and modified setal bases in the abdomen (char. 62: 1), and loss of the male endite tooth (char. 90: 0).

Micrathenaspecies groups

Considering the internal phylogeny of Micrathena, two great subdivisions were recovered (Fig. 8). The first group, hereafter named clade M. swainsoni, is formed by the M. triangularispinosa, M. militaris, and M. spinosa groups (sensuLevi, 1985) and is supported by the loss of the tapetum in the posterior median eyes (char. 25 : 1), a sclerotized plate anterior to the epigynum (char. 73: 1), and by a laterally projected paracymbium (char. 140: 1) (Fig. 8). The second, hereafter denominated the M. clypeata clade, is formed by Chaetacis and the remaining Micrathena species groups (M. gracilis, M. funebris, M. pungens, M. lepidoptera, M. kirbyi, and M. guerini, sensuLevi, 1985) and is supported by four non-ambiguous synapomorphies (Fig. 8): wrinkled booklung covers (which are also modified into stridulatory files; char. 70: 0; char. 71: 1), epigynum with a pair of anterior apodemes (char. 75: 1), copulatory openings small and concealed beneath lateral plates (char. 84: 1), and terminal apophysis fused to the embolus (char 114: 1). Most species groups proposed by Levi (1985) are monophyletic, excluding the M. militaris group (paraphyletic in relation to the M. spinosa group) and the M. kirbyi and M. guerini groups (which are not mutually monophyletic). Therefore, to increase agreement between taxonomy and phylogeny, we here propose a new, natural classification scheme for Micrathena groups (7-9; see Discussion and Taxonomy below).

Abdominal spines

The abdominal spines of females have a complex evolution within Micrathena. The anterior pair of spines appears at the base of the M. clypeata clade and is lost at least three times. The first pair of lateral spines is highly homoplastic, and appears at least four times in the outgroups. In Micrathena, it is lost three times, and one of its losses is synapomorphic for the M. furcula group. The second pair of lateral spines appears three times in Micrathena, supporting the M. cornuta, M. gracilis, and M. plana groups (although these are three independent gains). The third pair of lateral spines is a synapomorphy of Micrathena + Chaetacis, being lost at least five times. The fourth pair of lateral spines is a synapomorphy of the M. plana group, appearing independently in M. pungens. The first pair of posterior spines appears at the base of the cladogram (Fig. 10) and is lost twice, in Cyclosa and Hypognatha, being present in all Micrathena and Chaetacis species. Additionally, there are at least eight independent events of extreme elongation of this pair, in which the spines are longer than the carapace. Extreme elongated spines appear in the M. triangularispinosa group, in M. cyanospina, M. pungens, M. kirbyi, M. schreibersi, M. balzapamba, M. guanabara, and M. guerini (Fig. 10). The second pair of posterior spines is one of the synapomorphies of Micrathena + Chaetacis (appearing convergently in Gasteracantha + Xylethrus) and is lost at least four times (Fig. 10). The extreme elongation of the second pair of spines occurred a single time and is a synapomorphy of the M. spinosa group. The third posterior spine is gained at the base of the M. clypeata clade and is lost at least eight times.

Sexual dimorphism

The extreme sexual size dimorphism in the group is here defined as when the female carapace is at least two times larger than in males (following Scharff & Coddington, 1997; Hormiga et al., 2000). Sexually dimorphic lineages appear seven times in the phylogeny (Fig. 11): in Aspidolasius + (Gasteracantha + Xylethrus), in M. cyanospina, M. spinosa, M. schreibersi, M. vigorsi, M. lepidoptera, and M. horrida + M. gracilis. In four of these cases, sexual size dimorphism was attained through enlargement of females in relation to the ancestral size, rather than male diminishment. However, in two lineages males are smaller than the estimated ancestral size (M. lepidoptera and the M. gracilis group; Fig. 11), showing that in these lineages sexual dimorphism was attained through male reduction. In some other cases (e.g. M. spitzi + M. balzapamba, M. fissispina, M. swainsoni), enlargement of females was also observed, but the male became larger as well, and hence these lineages have remained monomorphic with respect to size.

Phylogenetic analyses

The highly supported monophyly of Micrathena + Chaetacis confirms previous ideas of a close relationship between these two genera (Simon, 1895; Levi, 1985; Scharff & Coddington, 1997). The analysis, however, indicates that the first is paraphyletic in relation to the latter, which was recovered as sister group to M. pungens. Micrathena pungens males are astonishingly similar to those of Chaetacis, whereas females do not have most of the diagnostic characters of this genus. This led Levi (1985) to suggest the possibility of a mismatch between males and females of M. pungens. Specifically, he thought the males could belong to Chaetacis abrahami Mello-Leitão, 1948. However, males and females attributable to M. pungens, based on Levi's (1985) diagnosis, have broadly overlapping distributions and are frequently found together in museum collections. Additionally, a male has been collected in a supporting thread of a female's web in Belém, Pará, Brazil by the first author. Moreover, there are derived characters shared between Chaetacis and M. pungens females, such as spiny fourth coxae and reniform spermathecae. Hence, the present evidence supports the male−female match in this species and thus, there are no strong reasons to doubt the sister-group relationship between M. pungens and Chaetacis. As Micrathena is an older name than Chaetacis, we here propose that the latter be considered a junior synonym of the first (see Taxonomy section below).

The relationships amongst the outgroups differ between this study and the topologies obtained by Scharff & Coddington (1997), where (1) Enacrosoma and Cyclosa (as well as Alpaida, closely related to Wagneriana and Actinosoma; Levi, 1988) were recovered closer to Araneus than to Gasteracanthinae genera; ( 2) Xylethrus is recovered as sister group to Micrathena + Chaetacis; and (3) these latter three genera are closer to Hypognatha than to Gasteracantha. These differences probably reflect the lower sampling of genera in the present study. However, it should be noted that Scharff & Coddington (1997) did not provide supporting measures of any kind for the clades recovered in their analysis and thus it is not immediately possible to know how well supported are their groupings. Additionally, the authors themselves stated that their topology is very unstable. Our results are congruent with other studies (Scharff & Coddington, 1997; Álvarez-Padilla et al., 2009; Sensenig, Agnarsson & Blackledge, 2010) in refuting the monophyly of the spiny orb-weavers (Micrathena + Chaetacis and Gasteracantha + related Palaeotropical genera).

Micrathenaspecies groups

Most Micrathena species groups proposed by Levi (1985) are monophyletic in our analyses, with some important exceptions. Levi (1985) suggested that the M. spinosa group might be a subgroup of the M. militaris group, a view confirmed by our results. The M. guerini and M. kirbyi groups were not clearly defined by Levi (1985). As a matter of fact, the author did not provide a single characteristic to tell males from these two groups apart, providing a single key for the males of both groups. This is reflected in the fact that, in our analyses, they were recovered as not mutually monophyletic. We propose some nomenclatural changes, especially concerning Chaetacis and Micrathena of the M. kirbyi, M. spinosa, and M. guerini groups. These are discussed in detail in the Taxonomy section (see below).

Abdominal spines

The results of our analyses suggest multiple gains and losses of abdominal spines in Micrathena, and multiple events of spine elongation throughout the phylogeny. Abdominal spines are large, sclerotized structures, probably energetically costly. Their convergent evolution in different Micrathena species is intriguing and may suggest that they play a functional role. An explanation based on sexual selection seems unlikely, as abdominal spines are almost exclusively feminine structures; males are the competing sex and courtship in araneid spiders is primarily tactile, not visual (Robinson & Robinson, 1980; Bukowski & Christenson, 1997b). Several authors have advocated that they may represent antipredator devices (Peckham, 1889; Edmunds & Edmunds, 1986; Cloudsley-Thompson, 1995; Gonzaga, 2007) because spines play this role in several animal and plant groups (e.g. Milewski, Young & Madden, 1991; Arnqvist & Johansson, 1998; Gilbert, 2001; Bollache et al., 2006). Most orb-weavers are nocturnal and hide themselves during the day or, if diurnal, build retreats outside the web, where they remain most of the time (Blackledge & Wenzel, 2000). Spiders that remain in the hub of their webs during the day, such as Micrathena, frequently make web stabilimenta, which are possibly defensive structures (e.g. Argiope, Blackledge & Wenzel, 2000; Cyclosa, Chou et al., 2005; Gonzaga & Vasconcellos-Neto, 2005a; see Herberstein et al., 2000 and Blackledge, Kuntner & Agnarsson, 2011 for reviews on web stabilimenta, with suggestions of other possible functions). Despite some records of stabilimenta in Micrathena webs (Robinson & Robinson, 1980; Nentwig & Heimer, 1987; Herberstein et al., 2000), this behaviour does not seem to be widespread in the genus and the structure is usually inconspicuous (I. L. F. Magalhães, pers. observ.). Furthermore, there is evidence that the stabilimenta built by Micrathena sexspinosa are prey-attracting structures (Gálvez, 2011; but see Blackledge et al., 2011), although they have not been tested as a defensive strategy. Hence, in the absence of other obvious defensive structures, we suggest that the abdominal spines may be antipredator devices.

There is evidence both that Micrathena may be avoided by some kinds of orb-weaver predators and that other genera of spiny orb-weavers may use their spines as defences. Wasps of the families Sphecidae and Crabronidae are important predators of orb-weaving spiders, which are captured to serve as food to their developing larvae inside tubular mud nests built specially for this purpose (Coville, 1987; Gonzaga, 2007; see references below). Several studies show that wasps capture spiders of many different araneid genera, but not Micrathena, despite the availability of species of the genus in the areas studied (Genaro & Alayón, 1994; Genaro, 1996; Camillo & Brescovit, 1999a, b, 2000; Gonzaga & Vasconcellos-Neto, 2005b; Araújo & Gonzaga, 2007; Buschini, Borba & Brescovit, 2008; Buschini et al., 2010). Other studies, however, show that Micrathena are occasionally captured, and at least some indicate that these spiders are the preferred prey in some cases (Obin, 1982; Landes et al., 1987; Gonzales-Bustamante, 1994; Camillo, 2002). In fact, Levi (1985) stated that large numbers of individuals of some species [Micrathena mitrata (Hentz, 1850), M. furcula, M. swainsoni, M. sexspinosa (Hahn, 1822)] were obtained from wasp nests. However, none of these species would be classified in the present study as having extremely long spines, except M. sexspinosa, which has the second posterior pair elongated (but running in parallel to the main body axis). In fact, M. mitrata and M. furcula belong to the M. furcula group (see below), whose species frequently possess only one or two pairs of very short spines. Other important spider predators are hummingbirds (Aves: Trochilidae) (Peckham, 1889; Hoffmaster, 1982). It is possible that long spines are a hindrance to predation by wasps and hummingbirds, as they would make it more difficult to insert the spider through a narrow tubular nest or a bird's beak because of the mechanical impediment. Additionally, long abdominal spines could reduce wasp attacks on Micrathena species, as they would occupy too much space inside mud nests, making the nest construction and provisioning more costly for female wasps. If this is confirmed, the emergence of several lineages with long spines should reflect similar selective pressures exerted by predators on different species. There is some evidence that wasps and other predators avoid spiny orb-weavers. Edmunds & Edmunds (1986) observed that birds attack individuals of Nephilengys cruentata (Fabricius, 1775), but not of the spiny spider Gasteracantha curvispina (Guérin, 1837). Additionally, only small individuals of the latter have been found in the nests of predatory wasps, possibly because mature females are too large to fit into the nests (Edmunds & Edmunds, 1986). The absence of spines in males supports the hypothesis of defence against predators because adult males are less likely to be captured by wasps (Buschini et al., 2008, 2010). Males possibly suffer selective pressures against the evolution of long spines as they would make their movement outside webs, in search for females, more difficult. Another explanation for the apparent bias against capture of Micrathena by predatory wasps could be the sclerotized abdomen of these spiders. Elgar & Jebb (1999) observed that wasps do not provision their first-instar larvae with Thelacantha brevispina (Doleschall, 1857) and suggested that this could be because of the hard integument of the spiders. However, the possibility of spines functioning as antipredator devices against wasps, birds, and other predators is not incompatible with this hypothesis and should not be ruled out.

It is important to stress that our suggestions that Micrathena spines are adaptations against predators are tentative suppositions based on their multiple independent origins, and should be addressed by specific experiments in the future. Studies of adaptation certainly are more accurate when carried out in a phylogenetic context (Coddington, 1988, 1994). However, our study only accounts for the pattern of spine evolution in Micrathena, and we can say nothing directly about the process that leads to it (see de Pinna & Salles, 1990; Grandcolas & D'Haese, 2003). Nevertheless, it is important to note that the current study indicates that there are several clades in which one of the species is short-spined whereas its sister group is long-spined. These should be of great value in comparative studies of spine functions, and this field surely deserves more empirical attention in the future.

Sexual dimorphism

The current study indicates that Micrathena is primitively monomorphic in respect to size and that at least six of its lineages became secondarily dimorphic independently. The primitively monomorphic condition suggested here conflicts with Scharff & Coddington (1997), who suggested that a return to a monomorphic condition is a synapomorphy for Micrathena + Chaetacis. Despite receiving attention from biologists at least since Darwin (1871: 314), sexual dimorphism in spiders still does not have its causes entirely understood. According to Hormiga et al. (2000), dimorphism is subject to two relatively independent traits: male size and female size. Some studies (Prenter, Elwood & Montgomery, 1998, Prenter et al., 1999; Hormiga et al., 2000) have suggested that, in orb-weavers, dimorphism usually arises when females are larger in relation to the ancestral size – the ‘giant females’ hypothesis. One of the possible explanations for this pattern is that larger females can produce larger clutches of eggs, and thus are to be positively selected given that all other conditions are equal (Blackledge, Coddington & Agnarsson, 2009). There is some empirical support for larger sizes leading to increased fecundity (Prenter et al., 1999; Higgins, 2002). The alternative hypothesis is that, in dimorphic lineages, males are reduced in relation to ancestral size – the ‘dwarf males’ hypothesis, suggested by Vollrath & Parker (1992). In this case the mechanisms are much more unclear and could include differential mortality, protandry because of scramble competition, the ‘gravity hypothesis’, and copulatory and postcopulatory processes (Foellmer & Moya-Laraño, 2007 and references therein). Our study suggests that, at least in Micrathena, extreme sexual size dimorphism evolves primarily because of enlargement of females. Nevertheless, male diminishment has also occurred in at least two lineages. We demonstrate that new sexually dimorphic lineages can arise within genera, something that could not be detected by Hormiga et al. (2000) because they used genera as terminal taxa in their analysis. More importantly, our results show that dimorphism can evolve by either female enlargement or male reduction even amongst closely related species. The fact that Micrathena exhibits several dimorphic lineages that have arisen independently, and by different mechanisms, makes it an excellent model for studies on the evolution of sexual dimorphism in spiders.

Biogeography

The results of the phylogenetic analyses suggest that Micrathena originated in South America. All Micrathena lineages include at least one South American species, and the M. guerini, M. schreibersi, M. kirbyi, M. triangularispinosa, and M. lepidoptera groups are endemic to this subcontinent. Nevertheless, several clades include species from Central or North America, including many species endemic to Caribbean islands, suggesting that these areas have been repeatedly colonized from South America. For example, Micrathena osa (Levi, 1985) (M. cornuta group), M. brevipes and M. sexspinosa (M. spinosa group), Micrathena molesta Chickering, 1961, and M. triserrata (M. plana group), and M. furcula and Micrathena bimucronata (O.P.-Cambridge, 1899) (M. furcula group) occur in Central America and Mexico. Three lineages (M. militaris, M. furcula, and M. gracilis groups) reach as far as the north-eastern USA (M. sagittata, M. mitrata, and M. gracilis) and some Caribbean islands, such as Cuba [Micrathena banksiLevi, 1985, and M. militaris (Fabricius, 1775), Micrathena cubana (Banks, 1909), M. horrida, and Micrathena forcipata (Thorell, 1859)], Hispaniola (M. militaris, Micrathena similis Bryant, 1945, M. forcipata) and Jamaica [Micrathena rufopunctata (Butler, 1873), M. horrida]. The M. gracilis group is of special interest because five out of seven described species are from Central America. In this case, the group main diversification probably took place in Central America with later invasions of South America (M. horrida) and North America (M. gracilis). All of these distribution patterns suggest that Micrathena originated in South America and subsequently invaded the continents to the north. As each of the groups is thought to have invaded Central America independently, our results suggest at least six different invasions of this subcontinent. Following the same line of thought, both the Caribbean islands and North America have been independently colonized three times each, probably by species from Central America. The routes used in these invasion are unclear, however, as there is no fossil record for the genus (see Dunlop, Penney & Jekel, 2011), but include the Panama Isthmus, the Caribbean islands, and overseas dispersal before closure of the former.

Although most Micrathena species and species groups have wide distribution areas, some clades are endemic to certain regions, such as the M. lepidoptera group, which occurs in the Sierras Nevadas of northern Colombia; M. gaujoni + (M. bifida + M. guerini), native to Andean cloud forests of eastern Colombia and Ecuador; and M. cyanospina + the M. spinosa group, which occur predominantly in the Amazon and southern Central America. Interestingly, some clades have disjunct geographical distributions, such as M. spitzi (south-eastern Brazil) + M. balzapamba (Ecuador) and M. guanabara (south-eastern Brazil) + (M. gaujoni + (M. bifida + M. guerini) (Ecuador and Colombia). Additionally, several species also have disjunct distributions and occur in both the Atlantic forest of eastern Brazil and the Amazon forest of northern South America, such as M. fissispina (Levi, 1985: map 3), Micrathena macfarlanei Chickering, 1961 (Levi, 1985: map 5), M. excavata (Levi, 1985: map 6), Micrathena triangularis (C.L. Koch, 1836) (Levi, 1985: map 6), Micrathena brevispina (Keyserling, 1864) (Levi, 1985: map 6), M. schreibersi (Levi, 1985: map 9), and M. lata (Levi, 1985: map 10). This pattern could be the result of vicariance (in this case, this would support the hypothesis that the Amazon and Atlantic forest have been continuous in the past) or long-range dispersal through the open, dry biomes of central South America.

The results of this study raise several interesting questions regarding Micrathena evolution and natural history. The availability of a phylogenetic hypothesis for the group opens new opportunities for the use of this genus as a model for biogeography and evolutionary biology. For instance, the existence of lineages that independently acquired long spines makes it possible to test hypotheses about the functions of these spines using phylogenetically independent replicates. Additionally, the genus would be an interesting model for studies of sexual size dimorphism evolution, as the extreme size difference between the sexes has evolved several times independently in the genus, both through female enlargement and male diminishment. Finally, the geographical distribution is restricted to the New World, with localized cases of endemism, making this genus an interesting model to study the evolution of the American Biota.

Family Araneidae Clerck, 1757

GenusMicrathena Sundevall, 1833

Micrathena Sundevall, 1833: 14. Read in April 1833 (F.O.P.-Cambridge, 1904: 525); publication date uncertain [F.O.P.-Cambridge (1904) states that the work was publicly read in April 1833, therefore the actual publication date must be earlier. Thus, the name Micrathena Sundevall, 1833 has priority over Acrosoma Perty, 1833.]. Type species Epeira clypeata, only species listed in section one of the genus, designated by Simon (1895: 848).

Acrosoma Perty, 1833: 193. Published in December 1833 (F.O.P.-Cambridge, 1904: 525). Type species Acrosoma swainsoni, designated by F.O.P.-Cambridge, 1904: 525. First synonymized with Micrathena by F.O.P.-Cambridge (1904).

Meganopla Simon, 1864: 292. Type species Meganopla cyanospina, designated by Bonnet, 1957: 2752. First synonymized with Acrosoma by Butler (1873).

Keyserlingia O.P.-Cambridge, 1890. Type species by monotypy Keyserlingia cornigeraO.P.-Cambridge, 1890 (= M. sexspinosa); preoccupied by Pander, 1861 (Brachiopoda) (Levi, 1985). First synonymized with Micrathena by Simon (1895).

Ildibaha Keyserling, 1892: 31. Type species by monotypy I. albomaculataKeyserling, 1892 (= M. flaveola (C.L. Koch, 1839)). First synonymized with Micrathena by Levi (1985).

Chaetacis Simon, 1895: 863. Type species by original designation Acrosoma affinis C. L. Koch, 1839 (= C. aureola). syn. nov.

Thaumastobella Mello-Leitão, 1945. Type species by monotypy Thaumastobella moureiMello-Leitão, 1945[= M. saccata (C.L. Koch, 1836)]. First synonymized with Micrathena by Scharff (1991).

Diagnosis: Micrathena females have a circular, excavated thoracic fovea (Magalhães & Santos, 2011: figs 15, 18, 21, 33) and a ventral, median line of up to three apodemes in the abdomen (Fig. 12, LMA, SpA). The carapace is glabrous or has reduced setae (but is hirsute in the M. plana and M. furcula groups) and is modified in shape, having a high thoracic region and often a pair of lateral rims and up to three pairs of dimples. The abdomen is usually longer than wide, heavily sclerotized with a ring around the spinnerets (which is found on the tip of a ventrally projected cone; Fig. 2, SCL), often has the dorsum glabrous (hirsute in the M. plana group) and has from one to nine pairs of spines (usually three to seven), but no median posterior tubercle (Figs 1, 2).

Micrathena males also present a circular thoracic fovea. They have an elongated abdomen that is longer than wide and rectangular to subtrapezoidal in form (Levi, 1985: figs 15, 487, 561; Magalhães & Santos, 2011: figs 15, 18, 21). The paracymbium of the palpus is large (Fig. 13, P) and often has extra lobes or is modified otherwise (Levi, 1985: figs 541, 711; Magalhães & Santos, 2011: fig. 40). The median apophysis has a basal projection that is partially fused to the frontal base of the radix (Figs 14–16, BP; Magalhães & Santos, 2011: figs 16, 22, 39, BP; frequently small and not fused to the radix in the M. plana group, Fig. 17; Magalhães & Santos, 2011: fig. 19).

Micrathena cornuta group

Diagnosis: Females can be diagnosed by the presence of a pair of tubercles or spines in the carapace (Magalhães & Santos, 2011: figs 33, 34; except M. pungens), spiny fourth coxae (Fig. 18), reniform spermathecae (Levi, 1985: figs 805, 823, 855; Magalhães & Santos, 2011: fig. 38; except M. osa, in which it is rounded), epigynal lateral keels and epigynum without lobe, frequently conical and ventrally projected (Magalhães & Santos, 2011: figs 36, 37). Males can be diagnosed by the presence of spiny setal bases at the margins of the carapace (except M. pungens) and by the palpus lacking a terminal apophysis, with a long, distal tegular projection, a large conductor that projects beyond the margin of the tegulum (Magalhães & Santos, 2011: fig. 39), and a large and modified paracymbium with a dorsal lobe (Magalhães & Santos, 2011: fig. 40, DL).

Composition: This group corresponds to species previously allocated in Chaetacis, plus M. pungens. Eleven species are included: M. abrahami comb. nov., M. aureola comb. nov., M. bandeirante comb. nov., M. carimagua (Levi, 1985) comb. nov., M. cornuta comb. nov., M. cucharas (Levi, 1985) comb. nov., M. necopinata Chickering, 1960 comb. nov., M. osa comb. nov., M. picta comb. nov., M. pungens, M. woytkowsii (Levi, 1985) comb. nov.