Systematics and phylogeography of the Brazilian Atlantic Forest endemic harvestmen Neosadocus Mello-Leitão, 1926 (Arachnida: Opiliones: Gonyleptidae)

Neosadocus harvestmen are endemic to the Southern Brazilian Atlantic Forest. Although they are conspicuous and display great morphological variation, their evolutionary history and the biogeographical events underlying their diversification and distribution are still unknown. This contribution about Neosadocus includes the following: a taxonomic revision; a molecular phylogenetic analysis using mitochondrial and nuclear markers; an investigation of the genetic structure and species’ diversity in a phylogeographical framework. Our results show that Neosadocus is a monophyletic group and comprises four species: N. bufo, N. maximus, N. robustus and N. misandrus (which we did not find on fieldwork and only studied the female holotype). There is astonishing male polymorphism in N. robustus, mostly related to reproductive strategies. The following synonymies have resulted from this work: “Bunoweyhia” variabilis Mello-Leitão, 1935 = Neosadocus bufo (Mello-Leitão, 1926); and “Bunoweyhia” minor Mello-Leitão, 1935 = Neosadocus maximus (Giltay, 1928). Most divergences occurred during the Miocene, a geological epoch marked by intense orogenic and climatic events in the Brazilian Atlantic Forest. Intraspecific analyses indicate strong population structure, a pattern congruent with the general behavior and physiological constraints of Neotropical harvestmen.


Introduction
Harvestmen (Opiliones) constitute one of the largest orders within Arachnida, with more than 6500 species around the world [1]. These organisms exhibit astonishing diversity in the Neotropics (e.g., the family Gonyleptidae, with more than 800 species in over 300 genera [2,3]). The great majority of harvestmen inhabiting the Brazilian coast are endemic to the Brazilian Rainforest and occupy a limited geographical range [4]. Neotropical harvestmen present low vagility (excepting some species of suborder Eupnoi), are very dependent on high levels of humidity and display limited tolerance to temperature variation [3,5,6]. These characteristics have resulted in their generally poor dispersal abilities and narrow species distribution the Brazilian Atlantic Forest) of Neotropical harvestmen make them suitable for biogeographical investigations [28].
In this work, we conducted a taxonomic revision to investigate the validity of Neosadocus species using morphological data and inferred their phylogenetic relationships with molecular data. Additionally, we used a phylogeographic framework to infer the possible historical events that promoted the divergences of the major lineages and shaped the population genetic diversity in the genus.

Sampling
We analyzed 129 Neosadocus specimens collected from 2008 to 2018 in 21 municipalities (32 localities) in the states of São Paulo and Paraná, covering all the known distribution range of the genus (Fig 2, Table 1).  We did not find specimens of N. misandrus to study and consequently did not include it in our phylogenetic analysis. The reference to the type locality of the species is vague (Brazil, according to Kury (2003) [2]), which made it impossible to accurately search for it in a specific region.

Morphological observations and species redescriptions
We analyzed all sequenced male and female specimens of Neosadocus and compared them with the type material for species determination under a Leica MS5 stereomicroscope. The morphological redescriptions were based on the type material of each senior synonym of the valid species. However, we used recently collected specimens to describe the coloration of live animals and the genitalia, and to describe structures that were missing in the type material (see Taxonomy).
All redescriptions follow the same guidelines: the outline of the dorsal scutum is described according to Kury & Medrano (2016) [29]; the topological terms for the appendages are described according to Acosta et al. (2007) [30]; the code for setation of pedipalpal tibia and tarsus are described according to Pinto-da- Rocha (2002) [31]; with respect to tarsi I and II, the numbers within parentheses indicate the number of distitarsi segments; the terminology for the dorsal scutum and leg armature are described according to DaSilva & Gnaspini (2010) [32]; the coloration of 92-96% ethanol-preserved specimens is described according to the NBS/ISCC Color System [http://archive.vn/ufMMn]; the penial macrosetae are described according to Kury & Villarreal M. (2015) [33]; in the descriptions of variation, the numbers within parentheses indicate the number of specimens analyzed; the descriptions of the female only include the characteristics in which they differ from their male counterparts. We followed Pinto-da- Rocha (2002) for the preparation of the male genitalia [31].
The following graphical representation of specimens are provided: photographs of carapace, venter and leg IV of the type material; illustrations of the dorsal scutum, performed using a Leica MZ APO stereomicroscope and a camera lucida; and photographs of the male genitalia, taken with a Zeiss DSM940 scanning electron microscope.

DNA extraction, amplification, sequencing and alignment
All specimens were stored in ethanol 92-96% at -20˚C and the DNA samples were extracted from leg IV muscle tissues using Agencourt DNAdvance kit. We performed polymerase chain reactions (PCRs) for DNA amplification with Thermo Scientific Phusion High-Fidelity DNA Polymerase kit, following manufacturer's protocols. Two genomic regions were selected for our study: the mitochondrial cytochrome C oxidase subunit I (COI), amplified with the primers dgLCO1490 and dgHCO2198 [34]; and the intronic nuclear internal transcribed spacer 2 (ITS2), with the primers 5.8SF and CAS28Sb1d [35] (S1 Table).
We verified the DNA fragments with 1% agarose gel electrophoresis, purified them with Agencourt AMPure XP kit, and measured DNA concentration of each sample (ng/μL) with a Thermo Scientific NanoDrop 2000 spectrophotometer. For DNA sequencing, we used Applied Biosystems BigDye Terminator v3.1 Cycle Sequencing kit. We also performed DNA precipitation using sodium acetate 3M and ethanol 100% and sent the samples for chromatogram production.
The chromatograms were used for consensus sequences assembly with Phred-Phrap-Consed software pack [36][37][38][39]. We manually inspected and edited the sequences in MEGA7 [40] and aligned them with MAFFT software [41], using the default algorithm. Protein-coding COI fragments were examined for stop codons and the sequences of both regions were trimmed to decrease the amount of missing data. Individual ITS2 sequences did not exhibit ambiguous positions (i.e., no heterozygous samples), thus we have not inferred phased nuclear haplotypes. We deposited all sequences in GenBank under accession numbers MT611325 to MT611416 (for COI) and MT630434 to MT630538 (for ITS2; Table 1).

Phylogenetic analysis and molecular dating
We constructed molecular phylogenetic trees for Neosadocus COI and ITS2 sequences using Maximum Likelihood and Bayesian analyses with RAxML [42] and MrBayes (v. 3.2.6) [43], respectively. We first determined the best-fit substitution models for our datasets in jModeltest 2.0 (v. 0.1.1) [44,45] using the corrected Akaike information criterion (AIC). Nine related Gonyleptidae species (Table 1) were chosen as outgroup, the species Acutisoma longipes Roewer, 1913 being the root of the trees. These taxa were chosen according to a previous analysis performed by Pinto-da- Rocha et al. (2014) [46]. In the Maximum Likelihood analyses, we selected each best tree from 100 iterations, and computed the bootstrap values with 1,000 replicates. For the Bayesian inferences, we performed four independent runs with 10,000,000 generations (with chain sampling every 1,000 generations) and determined stationary posterior distribution of parameters by visual inspection in TRACER (v. 1.6). We constructed a 50% majority-rule consensus tree after discarding the first 25% of the trees.
The main lineage divergence events of Neosadocus were dated by performing a calibrated multilocus Bayesian analysis using the � BEAST tool [47], available on BEAST software (v. 1.8.0) [48]. We applied the substitution models selected for each dataset and a Yule tree prior under lognormal relaxed (uncorrelated) molecular clock models, since the likelihood ratio tests conducted in MEGA7 rejected the null hypothesis of strict molecular clock for both loci. We used COI and ITS2 mean substitution rates previously estimated for other Neotropical harvestmen [COI: 0.0055/My (0.003-0.008); ITS2: 0.0004/My (0.0002-0.0006); [49]] in normally distributed priors and performed two independent analyses of 100,000,000 generations (with a 10% burn-in). We verified the convergence of runs and checked for parameters' high effective size values (ESS > 200) in TRACER (v. 1.6) and combined the results in LOGCOM-BINER (v. 1.8). We annotated both maximum clade credibility (MCC) species and gene trees with TREEANNOTATOR (v. 1.8).

Population structure and intraspecific analyses
Haplotype networks were constructed for Neosadocus COI and ITS2 sequences using the Median-Joining algorithm [50] in PopART software (v. 1.7) [51]. We investigated the genetic structure of Neosadocus using Bayesian analyses of population structure (BAPS) in the software BAPS (v. 6.0) [52] and applied the "spatial clustering of groups" option to the COI and ITS2 datasets of each species separately, then inferred the most probable number of clusters (k, allowing a range of one to 20 clusters). We performed five replicates for each dataset and the optimal k values were based on the highest marginal log-likelihood estimates.
The genetic distances among species and populations (considering each municipality as a population) within each species were calculated in ARLEQUIN (v. 3.5) [53]. To verify the possibility of isolation-by-distance among our samples, we analyzed the correlation between the genetic and geographic distances of the different populations through Mantel tests. We also calculated global F ST per species, pairwise F ST values between populations/BAPS groups, and diversity indices [haplotype (h) and nucleotide (π) diversities and number of polymorphic sites] in ARLEQUIN (v. 3.5).

Taxonomy
Our morphological observations confirmed the validity of four Neosadocus species: N. bufo, N. robustus, N. maximus and N. misandrus. This is consistent with the results of the molecular phylogenetic analyses, which recovered the genus as a monophyletic group (the only species absent from the phylogenetic analysis is N. misandrus; see Molecular data, phylogenetic analyses and divergence times).
Herein, we present identification keys, diagnoses, and morphological redescriptions for males and females of N. bufo, N. robustus and N. maximus. We provide a complete diagnosis for the female holotype of N. misandrus, which is sufficient to distinguish it from the other Neosadocus females.

PLOS ONE
rounded (N. bufo and N. maximus) or conical (N. robustus) paremedian elevations on area III, both densely tuberculate; (iv) femur IV prolaterally curved (except in some N. robustus specimens with femur IV straight); (v) basal cluster of tubercles on femur IV; (vi) dorso-basal apophysis of femur IV with apex curved anteriorly, followed posteriorly by a row of long tubercles; (vii) and male genitalia with glans flattened and wrinkled, stylus elongated and curved, ventral process half the stylus length, and ventral plate subrectangular and with three pairs of macrosetae A, one pair of macrosetae B, three pairs of macrosetae C, one pair of macrosetae D and two pairs of macrosetae E. b. females: (i) outline of dorsal scutum γ-type;

PLOS ONE
(ii) strongly granular dorsal scutum with two conical paramedian elevations on area III; (iii) free tergites II and III each with one central spine (with the exception of the female holotype of N. misandrus, which lacks a central spine on free tergites II and III); and (iv) femur IV straight, with blunt (N. bufo, N. robustus and N. misandrus) or conical (N. maximus) tubercles curved posteriorly. In both sexes, the ocularium has two conspicuous paramedian tubercles (with the exception of the female holotype of N. misandrus, instead with two long paramedian spines on the ocularium), and the paramedian tubercles on areas I and II are surrounded each by a circle of smaller tubercles.

Key to the males of Neosadocus
Note: There are two types of outlines of the dorsal scutum in the males of Neosadocus robustus: γT-type (major males) and γ-type (minor males). This is an important diagnostic character for the males of the species and the reason why the species is listed twice on the identification key.  •. Anterior margin of carapace with three tubercles on each side ( Fig 9A); femur IV with conical tubercles (Fig 9D and 9E  Distribution (Fig 2). BRAZIL: Southern coast of the state of São Paulo. There are two questionable records of N. bufo: one record from the municipality of Petrópolis (Southeastern region of Rio de Janeiro state), the locality of the type-species of the genus, Sadocus bufo. We did not not analyze this specimen because it is lost; the other record belongs to Lussanvira (Northwestern region of the state of São Paulo), the type locality of Polybunos tuberculatus. Despite the fact that a great number of Opiliones have been collected in both localities, Neosadocus specimens have not been found among them.
Diagnosis. Neosadocus bufo males differ from the other males of Neosadocus in having dorsal scutum wider than long. Also, N. bufo males possess an apical cluster of tubercles on the anterior surface of the dorso-basal apophysis of femur IV. Otherwise, N. robustus and N. maximus males exhibit small sparse tubercles, not forming a cluster. Neosadocus bufo females differ from the other Neosadocus females in having two lateral clusters of four small tubercles on both sides of the anterior margin of the carapace. Additionally, the paramedian spines on area III of N. bufo females measure over 2.0 mm, longer than in the females of the other species.
Dorsum (Figs 3A and 4A and 4C): Outline of dorsal scutum γT-type. Dorsal scutum wider than long. Anterior margin of carapace with two small tubercles on each side. Ocularium tuberculate, with two conspicuous paramedian tubercles. Carapace area behind the ocularium tuberculate. Lateral margin of prosoma with tubercles concentrated on the posterior half region. Areas I-II tuberculate, each with two prominent blunt paramedian tubercles surrounded by a circle of smaller tubercles. Area III tuberculate, with two rounded paramedian elevations densely tuberculate, each with one central tubercle. Lateral margin of dorsal scutum tuberculate, with an organized external row of tubercles. Free tergites I-III unarmed and with tubercles of different sizes.
Venter ( Fig 4B): Coxa I with a central longitudinal row of four prominent conical tubercles and small sparse tubercles; with a cluster of three prominent tubercles and one isolated prominent tubercle on distal margin. Coxa II with three central prominent conical tubercles and small sparse tubercles; with a cluster of three prominent tubercles and a cluster of two prominent tubercles on distal margin. Coxa III with small sparse tubercles; with an anterior row of eight prominent flattened tubercles and a posterior row of 12 prominent flattened tubercles, the latter almost fused with coxa IV. Coxa IV tuberculate. Stigmatic area tuberculate, with a posterior row of small tubercles. Genital operculum with small sparse tubercles. Sternites I-IV each with a row of small tubercles. Sternite V tuberculate. Anal operculum with sparse tubercles of different sizes.
Chelicera: Segment I with a cluster of three small projections and one small prolateral projection on distal margin. Bulla with small sparse ventral tubercles and nine spiniform ventral tubercles on distal margin. Segment II with 10 teeth. Segment III with six teeth, higher than those of segment II.
Pedipalp: Coxa with a cluster of seven ventro-apical tubercles and one retrodorsal apical tubercle. Trochanter with small sparse dorsal tubercles and two ventro-apical tubercles. Femur elongated, with small sparse dorsal tubercles and a ventral row of six small sparse tubercles. Patella with small setae both prolaterally and retrolaterally on distal margin; with one small prolateral tubercle and one small retrolateral tubercle on proximal margin; with two small ventral tubercles on proximal margin. Tibia with small sparse setae; tibial setation: prolateral IiIi and retrolateral IIi. Tarsus with small sparse setae; tarsal setation: prolateral IiIiiii and retrolateral IIiii.
Legs (Fig 4D and 4E show leg IV): Coxae I-III each with one prodorsal and one retrodorsal tubercle. Coxa III with a retrolateral row of tubercles. Coxa IV tuberculate, with a long proapical apophysis. Proapical apophysis of coxa IV with small dorsal tubercles and sparse ventral setae; with one retroventral proximal tubercle and three distal inflated ventral projections. Trochanters I-III tuberculate, each with one prodorsal and one retroventral tubercle on distal margin. Trochanters I-II longer than wide, each with one prodorsal and one retroventral tubercle on proximal margin. Trochanter III as wide as long. Trochanter IV wider than long and sparsely tuberculate; prolaterally with one conspicuous blunt central tubercle and one proximal cluster of seven small setae; retroventrally with one conspicuous apical tubercle. Femora I-IV with approximately six longitudinal rows (prodorsal, retrodorsal, retrolateral, retroventral, proventral and prolateral). Femora I-II straight, unarmed, with small sparse tubercles. Femur II with tubercles near the distal margin subtly longer. Femur III with sparse conical tubercles of different sizes and two curved ventral tubercles on distal margin. Femur IV prolaterally curved; proximally with small sparse retrolateral tubercles, one dorsal cluster of 32 small tubercles, one ventral cluster of three tubercles and one isolated ventral tubercle; distally with a dorsal cluster of seven small tubercles and two ventral spines, all curved posteriorly; prodorsal and retrodorsal rows of tubercles each decreasing in size distally (the three basal tubercles of prodorsal row longer, the basalmost tubercle almost fused with a dorso-basal apophysis); retrolateral row of tubercles decreasing in size distally (the four basal tubercles large); retroventral and prolateral rows with small tubercles; proventral row with tubercles of different sizes, almost interspersed; the distalmost tubercles of each row curved posteriorly. Dorso-basal apophysis of femur IV inclined distally, with the apex curved anteriorly; 1.5x longer than the longest prodorsal tubercle near it and with small tubercles on the anterior face; with an apical cluster of three tubercles. Patellae I-IV unarmed. Patellae I-II with small sparse tubercles. Patellae III-IV with sparse tubercles curved posteriorly. Tibiae I-IV unarmed. Tibiae I-II smooth. Tibiae III-IV with tubercles curved posteriorly. Tibia III shorter than the other tibiae and with two ventral rows of tubercles increasing in size distally. Tibia IV with one ventral row of tubercles increasing in size distally. Dorsum (Fig 5A and 5C): Outline of dorsal scutum γ-type. Dorsal scutum longer than wide. Frontal hump with two paramedian tubercles. Anterior margin of carapace with four small tubercles on each side. Area III with two long paramedian conical spines encompassed by various tubercles. Free tergites II-III each with one central spine.
Venter (Fig 5B): Coxa I with a proximal cluster of three prominent tubercles and a distal cluster of two prominent tubercles. Coxa II and stigmatic area tuberculate.
Legs (Fig 5D and 5E show leg IV): Coxa III smooth. Coxa IV shorter than in males; proapical apophysis absent. Trochanter IV without a prolateral apophysis; retroventral tubercle and prolateral proximal cluster of small setae both absent. Femora II-III tuberculate. Femur IV straight, with blunt tubercles decreasing in size distally, all curved posteriorly; dorso-basal apophysis absent.  Venter (Fig 6B): Coxa I with a central longitudinal row of three prominent conical tubercles and one isolated prominent central conical tubercle; with small sparse tubercles and two conical tubercles on proximal margin; with a cluster of three and a cluster of two prominent tubercles on distal margin. Coxa II with five central prominent cone-shaped tubercles and small sparse tubercles; with a cluster of two prominent tubercles and one isolated prominent tubercle on distal margin. Coxa III with small sparse tubercles, with an anterior row of five prominent triangular tubercles and a posterior row of 10 prominent triangular tubercles, the latter almost fused with coxa IV. Coxa IV tuberculate. Stigmatic area sparsely tuberculate. Genital operculum tuberculate. Sternites I-IV each with a row of small tubercles. Sternite V and anal operculum tuberculate.
Chelicera (voucher specimen MZSP 76388): Segment I with one small tubercle and one small prolateral projection on distal margin. Bulla tuberculate, with six small ventral tubercles on distal margin. Segment II with nine teeth longer and larger than those of segment III. Segment III with nine teeth, the distal third and fourth greatly reduced.
Pedipalp: Coxa with two ventral tubercles and one retrodorsal tubercle. Trochanter with small sparse dorsal tubercles and two ventro-apical tubercles. Femur elongated, with small sparse dorsal setae and a ventral row of three small sparse setae; with two small setae on the ventroproximal margin. Patella with small sparse dorsal setae; with small prolateral, retrolateral and ventral setae on distal margin; with one small prolateral tubercle and one retrolateral small tubercle on proximal margin. Tibia with small sparse setae; tibial setation: prolateral IiIi and retrolateral IIi. Tarsus with small sparse setae; tarsal setation: prolateral IiIiiiiii and retrolateral IIiiii.
Legs (Fig 6D and 6E show leg IV): Coxae I-III each with one prodorsal and one retrodorsal tubercle. Coxa IV with sparse tubercles of different sizes and a retroventral cluster of four tubercles; with one long proapical apophysis. Proapical apophysis of coxa IV with sparse setae and one retroventral proximal tubercle. Trochanters I-III ventrally tuberculate. Trochanters I-II longer than wide, each with one prodorsal and one retroventral tubercle on both proximal and distal margins. Trochanter III as wide as long, with three prodorsal and three retroventral tubercles on distal margin. Trochanter IV wider than long and sparsely tuberculate; prolaterally with one conspicuous blunt central tubercle and small distal tubercles; retroventrally with one conspicuous apical tubercle and small distal tubercles. Femora I-IV with approximately six longitudinal rows (prodorsal, retrodorsal, retrolateral, retroventral, proventral and prolateral). Femora I-II straight, unarmed, with small sparse tubercles. Femur II with ventral tubercles near the distal margin subtly longer. Femur III with sparse conical tubercles of different sizes and one ventral row of tubercles increasing in size distally. Femur IV prolaterally curved; proximally with one dorsal cluster of 28 small tubercles and small sparse retrolateral and ventral tubercles; distally with a dorsal cluster of nine small tubercles curved posteriorly; prodorsal row of tubercles decreasing in size distally (the three basal tubercles longer); retrodorsal, retroventral, proventral and prolateral rows with small tubercles; retrolateral row with tubercles of different sizes, almost interspersed. Dorso-basal apophysis of femur IV inclined posteriorly, with the apex curved anteriorly; twice longer than the longest prodorsal tubercle near it and with small tubercles on the anterior face; with one small basal tubercle emerging posteriorly. Patellae I-IV unarmed, with small sparse tubercles. Dorsum (Fig 7A and 7C): Outline of dorsal scutum γ-type. Dorsal scutum longer than wide. Frontal hump with two paramedian tubercles. Area III with two long paramedian conical spines encompassed by various tubercles.
Venter (Fig 7B): Coxa I with a proximal cluster of three prominent tubercles and a distal cluster of two prominent tubercles. Coxa II tuberculate, with a proximal cluster of three prominent tubercles and one distal isolated prominent tubercle.
Variation in females not detected. Neosadocus maximus (Giltay, 1928  Distribution (Fig 2). BRAZIL: Northern coast of the state of São Paulo. There are questionable literature records of N. maximus in the Southeastern region of the states of Rio de Janeiro and São Paulo. Neosadocus specimens were never found in any of the various subsequent Opiliones collections conducted in the Southeastern region of Rio de Janeiro state. Moreover, according to our study and as suggested by Kury (1995) [19], only N. bufo and N. robustus occur in the Southeastern region of the state of São Paulo (the latter species also occurs in the coastal region of Paraná state). Given the morphological similarities between N. bufo and N. maximus, the questionable records of the latter species may refer, instead, to records of N. bufo.
Diagnosis. Neosadocus maximus males resemble N. bufo males due to the dorsal scutum coloration pattern and the two rounded paramedian elevations densely tuberculate on area III. However, N. maximus males can be distinguished by the following combination of characters: short dorso-basal apophysis, same size of the longest prodorsal tubercle near it; tubercles of frontal hump, tubercles of ocularium, tubercles of carapace area behind the ocularium, tubercles of lateral margin of prosoma, and the tubercles of opisthosoma near area I pale greenish yellow (centroid 104). N. maximus females differ from the other Neosadocus females by presenting three small lateral tubercles on both sides of the anterior margin of carapace. Also, only N. maximus females have a small proapical apophysis on coxa IV and five long conical basal tubercles on the prodorsal row of tubercles of femur IV. Outline of dorsal scutum γT-type. Dorsal scutum longer than wide. Anterior margin of carapace with three small tubercles on each side. Ocularium tuberculate, with two conspicuous paramedian tubercles. Carapace area behind the ocularium tuberculate. Lateral margin of prosoma with tubercles concentrated on the posterior half. Areas I-II tuberculate, each with two prominent blunt paramedian tubercles surrounded by a circle of smaller tubercles. Area III tuberculate, with two rounded paramedian elevations densely tuberculate, each with one central tubercle. Lateral margin of dorsal scutum tuberculate, with an organized external row of tubercles. Free tergites I-III unarmed and with tubercles of different sizes.
Venter (Fig 8B): Coxa I sparsely tuberculate; with a central longitudinal row of three prominent conical tubercles and one isolated prominent central conical tubercle; with a cluster of three prominent tubercles and one isolated prominent tubercle on distal margin. Coxa II sparsely tuberculate; with a cluster of three prominent tubercles and one isolated prominent tubercle on distal margin. Coxa III with small sparse tubercles, with an anterior row of five prominent flattened tubercles and a posterior row of ten prominent flattened tubercles, the latter almost fused with coxa IV. Coxa IV tuberculate. Stigmatic area sparsely tuberculate, with a posterior row of small tubercles. Genital operculum tuberculate. Sternites I-IV with a row of small tubercles. Sternite V and anal operculum tuberculate.
Chelicera: Segment I with one small tubercle; distally with one small retrolateral projection and one small prolateral projection. Bulla tuberculate; distally with a retrolateral cluster of two small tubercles and three small ventral tubercles. Segment II with seven teeth, the distal first, second and third longest. Segment III with eight teeth, the distal first, second and fourth longest.
Pedipalp: Coxa with one retrodorsal tubercle. Trochanter with two ventro-apical tubercles. Femur elongated, with two small dorsal setae and one small retrolateral seta. Patella with one small dorsal seta; with three small ventral, prolateral and retrolateral setae on distal margin; with two small ventral tubercles on proximal margin. Tibia with small sparse setae; prolateral IiIi and retrolateral IiIi. Tarsus with small sparse setae; prolateral IiIiiiii and retrolateral IIiii.
Legs (Fig 8D and 8E show leg IV): Coxae I-II with one prodorsal and one retrodorsal spiniform tubercle. Coxa III tuberculate. Coxa IV with sparse tubercles of different sizes and one long proapical apophysis. Prolateral apical apophysis of coxa IV with small dorsal tubercles and with sparse ventral setae; with one retroventral proximal tubercle and three distal inflated ventral projections. Trochanters I-III tuberculate, each with one prodorsal and one retroventral tubercle on distal margin. Trochanters I-II longer than wide, with one prodorsal and one retroventral tubercle on proximal margin. Trochanter IV wider than long; prolaterally with one conspicuous blunt central tubercle and small prolateral rounded tubercles on distal margin; retroventrally with one conspicuous apical tubercle and one small distal tubercle; proximally with small sparse tubercles and a prolateral cluster of five small setae. Femora I-IV with approximately six longitudinal rows (prodorsal, retrodorsal, retrolateral, retroventral, proventral and prolateral). Femora I-II straight, unarmed, with small sparse tubercles. Femur III with sparse conical tubercles of different sizes; distally with a dorsal cluster of six conical tubercles and two conical ventral tubercles, all curved posteriorly. Femur IV prolaterally curved; proximally with one dorsal cluster of 11 small tubercles, one ventral cluster of three tubercles and small sparse retrolateral tubercles; distally with a dorsal cluster of five small tubercles curved posteriorly; prodorsal, retrolateral and prolateral rows of tubercles decreasing in size distally (the five basal tubercles of prodorsal and retrolateral rows longer; the distalmost tubercle of prolateral row curved posteriorly); retrodorsal row with small tubercles; retroventral and proventral rows with tubercles of different sizes (the distalmost tubercle of retroventral row curved posteriorly; proventral row with tubercles almost interspersed). Dorso-basal apophysis of femur IV inclined posteriorly, with the apex curved anteriorly; the same size of the longest prodorsal tubercle near it and with small, rounded tubercles on the anterior face. Patellae I-IV unarmed, with small sparse tubercles. Patellae III-IV with the ventral tubercles curved posteriorly. Tibiae I-IV unarmed. Tibiae I-II smooth. Tibiae III-IV with tubercles curved posteriorly. Tibia III shorter than the other tibiae and with two ventral rows of tubercles increasing in size distally. Tibia IV with one ventral row of tubercles increasing in size distally. Dorsum (Fig 9A and 9C): Outline of dorsal scutum γ-type. Dorsal scutum as wide as long. Area III with two long paramedian conical spines surrounded by various tubercles. Free tergites II-III each with one central spine.
Venter (Fig 9B): Coxa I sparsely tuberculate, with a proximal cluster of three prominent tubercles and a distal cluster of two prominent tubercles on distal margin. Stigmatic area without a posterior row of small tubercles.
Legs (Fig 9D and 9E show leg IV): Coxa III smooth. Coxa IV shorter than in males; proapical apophysis with two small tubercles on the anterior face. Trochanter IV without a prolateral apophysis, retroventral tubercle and prolateral proximal cluster of small setae both absent.  Distribution. BRAZIL: There is not a specific locality for the holotype. Diagnosis. Neosadocus misandrus is only known from the female holotype, which can be distinguished from the other females of the genus by the following characters: (i) two long paramedian conical spines on the ocularium (Fig 10A and 10B), whereas all other females have two conspicuous paramedian tubercles on the ocularium; (ii) absence of a central spine on free tergites II and III, whereas they are present in all other females; and (iii) small blunt tubercles of the same size on femur IV, all curved posteriorly (Figs Fig 10C and 10D), whereas the tubercles of all other females decrease in size distally and may be blunt (N. bufo and N. robustus) or conical (N. maximus).

Molecular data, phylogenetic analyses and divergence times
Our dataset comprises 83 COI sequences (42 haplotypes) and 96 ITS2 sequences (32 haplotypes) with 505 base pairs (bp) and 373 bp (with gaps), respectively (Table 1, S2 Table). No stop codons or ambiguous peaks were detected in the COI dataset, suggesting the absence of pseudogenes or numts in the sequences.
The best-fit substitution models for COI and ITS2 datasets were HKY+I+G and HKY+I, respectively. Both Maximum Likelihood and Bayesian phylogenetic analyses revealed gene trees with similar topologies (Fig 12; topologies recovered in ML analyses are shown in S3 Fig). The COI cladogram supported the monophyly of Neosadocus, showing N. bufo, N. maximus and N. robustusas highly supported distinct clades; in the ITS2 tree, however, two outgroup species (Iguapeia melanocephala and Iporangaia pustulosa) were positioned within the Neosadocus clade (which may be a result of the close relationship among these species and lower divergence observed in the nuclear sequences).
The multilocus calibrated tree recovered the monophyly of Neosadocus, corroborated its close relationship with Iguapeia melanocephala and Iporangaia pustulosa, and indicate that

Intraspecific analyses: Population structure and genetic diversity
The COI median-joining network recovers several haplogroups separated by a great number of mutational steps in all three species, while ITS2 haplotypes are separated by only one to three nucleotide differences (Fig 14). Most COI haplotypes were exclusive (i.e., detected in one single location), except for three haplotypes found in populations geographically close (H25, in Morretes and Paranaguá; H26, in Antonina and Morretes; and H32, in Cajati and Barra do Turvo; S2 Table). For ITS2, seven out of the 32 haplotypes were shared among different  Table), which matched the main lineages detected in the phylogenetic analyses (Fig 12). These intraspecific clusters are spatially distinct, although they overlap in few regions (as in Miracatu, Guaraqueçaba and Morretes, populations composed by sequences from different BAPS clusters; Fig 14). Two N. bufo clusters are distributed in the Northern portion of the species range (B1, B4), while the other two are in the Southern region (B2, B3). In the case of N. maximus, samples from the Ubatuba population (at the Northern edge of the species distribution) formed a separate cluster; this result must be treated with caution, since there is an indication of isolation-by-distance in this species (see below). Finally, N. robustus clusters were predominantly distributed to the North (R3, R4, R5) or South (R1, R2, R6) of the Ribeira do Iguape River region and the Northern clusters

PLOS ONE
sympatric with N. bufo (Fig 14). The structure pointed by BAPS for ITS2 sequences was similar but less marked (which is probably due to the higher similarity among sequences), showing the same subdivision for N. maximus, three groups for N. bufo (merging B2+B3 COI clusters), and three for N. robustus (the first combining R1+R2+R6 COI clusters and the other two similar to R3 and R4, except for some samples from Ribeirão Grande and Ibiúna; S2 Table). Given that there was more structure in the COI sequences than in the ITS2 (S2 Table), we extrapolated the geographical distribution of the mitochondrial clusters in the ITS2 haplotype network shown in Fig 14. The genetic distances among samples were considerably greater for the mitochondrial datasets. The mean distances between species varied from 39.59 to 46.91 average pairwise differences for COI sequences, and from 4.02 to 8.01 differences for the ITS2 dataset (S3 and S4 Tables). Within each species, the average number of pairwise COI differences between populations varied from 7. The lowest values were found for N. robustus samples collected between Morretes and Guaraqueçaba (S11-S16 Tables); most values, however, were statistically non-significant due to the small size of the samples.
The haplotype diversity values were high for the three species, varying from 0.88 to 0.97 for COI sequences, and from 0.74 to 0.9 for ITS2 ( Table 2). The nucleotide diversity obtained for COI sequences was substantially higher, varying from 0.034 to 0.041 in each species, while the ITS2 values varied from 0.003 to 0.01 nucleotide differences. Within municipalities, the genetic variation was generally low for both markers, since most populations presented only one or few closely related haplotypes.

Discussion
This is the first taxonomic revision and molecular phylogenetic study focusing on Neosadocus. Herein, we discuss the different aspects of our results, from intraspecific variation in secondary sexual characters in males to species relationships and population genetic structure. We also inferred historical events in the Brazilian Atlantic Forest that possibly influenced the distribution of Neosadocus species and their evolutionary patterns.

Male polymorphism in Neosadocus robustus
Three types of morphological polymorphism are recognized among gonyleptid harvestmen: sexual dimorphism, male polymorphism and variation in non-sexual/non-fixed traits [71]. All Neosadocus species present sexual dimorphism. However, only N. robustus show male polymorphism (S1 Fig), a phenomenon that had been previously documented for other harvestmen but not for this species [69][70][71][72][73]. We observed morphological polymorphism in the following traits of male specimens of N. robustus: outline, length and width of dorsal scutum (γT-type in major males; γT-type in minor males); presence or absence of a central spine on free tergites II and III; length of prolateral apical apophysis of coxa IV (long in major males; reduced in minor males); curvature and armature of femur IV (curved and strongly armored in major males; straight and with reduced armature in minor males); coloration of dorsal scutum and legs. Willemart et al. (2009) [10] reported that the armature of the fourth pair of legs is used by Neosadocus maximus males to fight for resources and nesting sites defense [55] -. We hypothesize that this phenomenon occurs in major males of N. robustus. Nevertheless, the sneaky behavior had been previously documented in minor males of the harvestmen Serracutisoma proximum [74]: minor males invade the territories of major males and furtively mate with egg-guarding females. Guarding eggs had been previously documented for females of N. maximus [11,12]. If N. robustus females also guard their eggs, we conjecture that they also mate with minor males of N. robustus. These cases and hypotheses illustrate the male polymorphism associated with different reproductive strategies [75]. Finally, we also observed an apparent polymorphism of non-sexual/non-fixed traits in males: presence/absence of a central

PLOS ONE
spine on free tergites II and III; coloration of dorsal scutum and legs. Coloration was the only variation we observed in both males and females. We do not believe that the presence/absence of a central spine on free tergites II and III has anything to do with fighting in males, since this trait is variable among major and minor males. We do not have a reasonable hypothesis to explain color variation.

Molecular phylogenetics and divergence times estimation
The molecular analyses indicated a clear differentiation among Neosadocus bufo, N. maximus and N. robustus (Fig 13), corroborating the morphological evidence that these species are valid. The results of the multilocus Bayesian analysis suggests that N. maximus and N. robustus are more closely related to each other than each is to N. bufo; however, this result needs to be looked at with caution, since the posterior probability of this clade is relatively low (0.82), as in all other phylogenetic trees (Fig 12, S3-S5 Figs). The inclusion of more molecular data and a morphological cladistic analysis may assist the elucidation of the relationships among the species of Neosadocus.  [58], also based on molecular data. Caetano & Machado (2013) [76] also suggested that Neosadocus and Progonyleptoidellinae are closely related, based on behavioral, chemical and ecological data. Since this work aimed to uncover the phylogenetic relationships of Neosadocus species, we will not discuss in detail the phylogenetic relationship between Neosadocus and Progonyleptoidellinae. More data are still necessary to further resolve that aspect of the phylogeny.
Most of the genetic divergence in Neosadocus occurred in the Miocene [speciation events at ca. 12-17 Mya (95% HPD = 6-27 Mya); main intraspecific divergences starting at ca. 7-14 Mya (95% HPD = 4-25 Mya; Fig 13, S2 and S3 Figs)]. This period was marked by intense orogenic activity in the Southern Atlantic Forest, which may have driven the evolutionary history of Neosadocus. An important geological structure in the region is known as the Continental Rift of Southeastern Brazil, which includes important present-day relief components, as the Ribeira do Iguape and Paraíba do Sul River Valleys, and the Serra do Mar and Serra da Mantiqueira mountain ranges. Neogene-Quaternary geomorphologic processes along this rift probably increased the topographical differences between mountains and valleys, rearranging rivers and lake systems [77,78]. Other historical events rather than tectonism, for instance changes in climatic conditions (partially imposed by the above-mentioned relief modifications) and marine transgressions [79] may also have played a role in the differentiation of Neosadocus populations, since they could have influenced forest rearrangements or reduction to isolated patches. A phylogeographic study with other Southern Atlantic Forest harvestmen (Acutisoma longipes) corroborates that the drier conditions in valleys (in comparison to Serra do Mar mountain slopes) may have limited the gene flow of populations [26], a consequence of the considerable dependence on humidity and poor dispersal ability of these organisms, promoting lineage diversification.
Studies on the historical relationships among the AoEs of harvestmen in the Atlantic Forest have proposed not only that all the processes mentioned above might be involved in the regionalization of the opiliofauna, but also that they acted as reiterative barriers, which is reflected in the current complex biogeographic patterns with spatial congruence in some areas [7,24]. Using cladistic biogeographic methods, DaSilva et al. (2017) [24] detected an important split between the AoEs mostly at the north and south of Ribeira de Iguape River Valley region (named Southeastern and Southern blocks, respectively). The current distributions of Neosadocus species corroborate this division: N. maximus is restricted to the Northern edge of the range of Neosadocus [being restricted to Serra do Mar of São Paulo (SMSP) and Southern Rio de Janeiro Coast (LSRJ) area, in the Southeastern block], while the distribution of N. bufo and N. robustus partially overlap and occupy the South-Central portion of the genus' distribution, on the AoEs of the Southern block [Southern São Paulo (SSP), Paraná (PR) and Santa Catarina (SC), although many individuals were sampled out of AoEs' congruence cores; Fig  14]. This geographical delimitation, associated with the old (Miocene) divergence times estimated, suggests that ancient historical barriers have influenced the diversification of harvestmen in this region.
Several phylogeographic breaks in the Southern Atlantic Forest had been previously reported for different taxa. A variety of groups exhibit genetic disjunctions that are geographically similar to Neosadocus around the Continental Rift of Southeastern Brazil, as birds [80][81][82], mammals [83], amphibians [84][85][86][87], snakes [88], insects [89,90], and planarians [91]. However, the breaks observed in this study are considerably older than the divergences reported for most taxa, a pattern already observed for other co-distributed harvestmen [25][26][27]. This time-scale heterogeneity reveals that the evolution of the Atlantic Forest biota is complex and cannot be attributed to a single process; rather, it is more reasonable that multiple historical barriers arose reiteratively in a same region. Additionally, spatial and temporal incongruences may result from the different ways species respond to environmental changes, responses that are mediated by taxon-specific traits, dispersal ability, population dynamics, physiological constraints, and/or ecological interactions [92]. Hence, organisms with extremely low vagility, philopatry and humidity-dependence, as harvestmen, tend to retain relatively older phylogeographic signals with respect to other groups, which reinforces the value of these species for comparative biogeographical studies. As different co-distributed taxa are more broadly investigated, it becomes clear that both Quaternary events (as the global climatic oscillations that promoted cycles of forests shifts) and older processes (as the Neogene geomorphological and climatic episodes described above) acted in combination to produce taxon-specific idiosyncratic patterns of diversification in the Atlantic Forest [93].

Population structure and genetic diversity
All three species exhibited great variability and strong population structure. Most haplotypes are geographically restricted to a single location, and the few exceptions are shared among close locations (Fig 14), a result that matches the high genetic distances and F ST values. This pattern is similar to those reported for other Atlantic Forest harvestmen [25][26][27]. Tropical harvestmen, with the exception of some members of the suborder Eupnoi [94], have low vagility -a rare case among arthropods of similar size [3,5,6]. Moreover, they are very humiditydependent and only survive within a narrow temperature zone [5]. All these characteristics limit their chances of dispersal and contribute to the geographical structure commonly found in their populations. The restricted opportunities for relocation may be more intense in regions with valleys (as the Southern Atlantic Forest), since these lower altitude areas are consistently more unstable and are usually drier than the forests inhabited by these harvestmen.
Our BAPS results indicated genetic subdivisions within each species, especially as calculated using COI sequences. Although the genetic groups overlap somewhat, they are geographically separated (Fig 14), which suggests that historical events might have reduced (or interrupted) the gene flow among some populations in the past (as discussed above). This intraspecific differentiation is consistent with the AoEs proposed for harvestmen, corroborating the biotic disjunctions already detected [7,24]: the M1 and M2 groups of N. maximus, for example, occur in SMSP and LSRJ areas, respectively; similarly, the intraspecific groups of N. bufo and N. robustus are spatially congruent with different Southern AoEs blocks (although the association for these two species is less clear; Fig 14). However, as we obtained statistically significant results in Mantel tests for some datasets (especially for ITS2 sequences), we cannot completely discard that these intraspecific groups reflect isolation-by-distance rather than historical barriers.
Although our morphological observations indicate greater variation in N. robustus, the total genetic diversity was similar among the three species (Table 2). Also, we found no correlation between the morphological variation and the genetic groups pointed by BAPS in any species, indicating that such variability is not the result of restricted gene flow. More studies correlating morphological traits and a greater amount of neutral and non-neutral molecular markers could bring more robust results on this issue.

Conclusion
Based on morphological observations and molecular phylogenetic analyses, our study confirmed the taxonomic validity of three Neosadocus species: N. bufo, N. maximus and N. robustus. The latter species presented a high degree of male polymorphism, mostly related to reproductive strategies. Although morphological observations corroborated the validity of N. misandrus, the absence of specimens besides the female holotype did not allow us to include this species in our phylogenetic analysis. Furthermore, our time-calibrated inferences and phylogeographic analyses suggest Neogene speciation events, deep intraspecific lineages and strong population structure in all three species. These results match the patterns commonly found for Neotropical harvestmen and corroborate that historical tectonic and climatic events might have imposed reiterative or longstanding barriers to the gene flow among harvestmen in the Atlantic Forest areas.