Phylogeny of the millipede order Spirobolida (Arthropoda: Diplopoda: Helminthomorpha)

Taxonomic and systematic history of Spirobolida

The taxonomic history of Spirobolida up through the middle of the last century was well illustrated by Keeton (1960a) and Hoffman (1980). It is not our goal to reproduce their excellent scholarship, but only to bolster this and to supplement this work where we feel further discussion is warranted. We provide reference to the major advancements in Spirobolida taxonomy as they relate to the order, suborder, and family levels within this lineage, followed by a focus on advances made since Keeton’s (1960a) and Hoffman’s (1980) publications.

The history of the name Spirobolida begins with the description of the genus Spirobolus by Brandt, including two species, Spirobolus bungii Brandt and Spirobolus olfersiiBrandt (1833), with no type designation made at that time. The first type designation in the genus was by Pocock (1894), although he provides no discussion as to why S. bungii was selected as the type species over S. olfersii. In the original publication, S. olfersii was the first species listed after the diagnosis of the genus followed by S. bungii so order of appearance in the original manuscript was not the criterion used; it is possible that the decision was simply made based on alphabetical order, but this remains unclear and may never be resolved. The first suprageneric taxon based on Spirobolus was the subfamily Spirobolinae within the family Julidae, proposed by Bollman (1893) in a work published after his death. Based on Bollman’s newly established subfamily, other workers began recognizing the lineage as family Spirobolidae soon thereafter (Verhoeff, 1893; Pocock, 1894). Most notably was the introduction of the order Anocheta by Cook (1895) to accommodate Spirobolidae, representing the first time ordinal status was given to what is now order Spirobolida. The first use of the name Spirobolida to define the order was by Chamberlin (1943), although this change was not explained until Chamberlin and Hoffman’s (1958) publication. Hoffman (1980) applied Cook’s (1895) ordinal name Anocheta to the superorder housing Spirobolida. Recent classifications omit the superordinal name Anocheta (Enghoff, 1984; Shelley, 2003).

There are 11 families housing 184 nominal genera in Spirobolida, 115 of which are valid (see Table 1 for a breakdown of genera by family); herein we only trace the history of family names currently recognized as being valid. We also trace the history of Trigoniulinae due to the uncertainty of its taxonomic status. Spirobolidae was the first family name associated with the modern order Spirobolida (Verhoeff, 1893); Hoffman (1980) attributes the transition from subfamily Spirobolinae to family Spirobolidae to Pocock (1894), but it was actually Verhoeff (1893) who first made the change. Cook (1897) introduced the second family, Pachybolidae, and his differentiation between families was based on the shape of the lateroventral portions of the collum, pointed in Pachybolidae and rounded in Spirobolidae. Attems (1909, 1910) split this lineage into Euspirobolidae (much the same as Spirobolidae of Cook) and Trigoniulidae (much the same as Pachybolidae of Cook) based on the presence or absence of a sternite connecting the posterior gonopods, respectively; this established the major morphological differentiation that has been used to diagnose the suborders Spirobolidea and Trigoniulidea to this day. The seminal work in the early history of family-level taxonomy was by Brölemann (1914); he proposed the families Rhinocricidae, Spirobolellidae, and Pseudospirobolellidae, defining them exclusively on characters of the male gonopods. He did not provide illustration of the character states used to define these taxa, making it somewhat cumbersome to interpret his definition and description without visual representation. The family Atopetholidae was proposed by Chamberlin (1918) to include North American spirobolidans with an acutely narrowed collum that projects ventrally beyond the second body ring.

The 1950s and 1960s brought new life to the taxonomic landscape of Spirobolida. Floridobolidae, originally described as a genus within Spirobolidae (Causey, 1957), was elevated to family status by Keeton (1959) based mainly on the presence of a small sclerotization in the membrane connecting the anterior and posterior gonopods. Keeton (1960b) treated a small collection of orphaned North and Central American genera and placed them in a new family, Allopocockiidae. This family was placed in Spirobolidea and distinguished from other families by the following two characters: presence of a distinct coxite in the posterior gonopods and presence of large apodemes associated with the coxite of the anterior gonopods. The family Messicobolidae was proposed by Loomis (1968) to accommodate three genera that had previously been left unassigned to families (Chamberlin, 1922; Hoffman and Orcutt, 1960). The family was defined based on the presence of distinct medioventral elongation of the coxites, the sternite being short and broad, and the posterior gonopods being stout and lacking complexity. Hoffman (1969) first recognized a unique subfamily of spirobolellid millipedes, which lacked ocelli and had ozopores beginning on the third body ring, a character unique in Diplopoda; the lineage was later (Hoffman, 1980) elevated to family status as Typhlobolellidae. More recently, Shelley (2001) described the monotypic family Hoffmanobolidae from a series of specimens collected from southern Mexico. Although other characters are listed in the diagnosis, the only diagnostic character unique to this taxon is the enlarged collum, suggesting that further study of this species is necessary to determine if it represents a distinct family or is simply a derived member of an established family.

It was in the 1960s that the first modern revisions were completed within Spirobolida. Both Spirobolidae (Keeton, 1960a) and Atopetholidae (Hoffman and Orcutt, 1960) were revised, but neither of the works addressed the position of these families in relation to other families in the order. The first treatment that provided hypotheses of relationship was by Hoffman (1980). Hoffman’s insight into gonopod evolution within Spirobolida revealed one of the major obstacles in conducting morphological phylogenetic analyses within this order and in millipedes in general. He discussed the trend towards reduction of the posterior gonopods and how this may lead to implied relationship based on similarity due to reduction, not due to shared evolutionary history. Hoffman also discussed the family Floridobolidae and questioned whether this monotypic family should be recognized or if it belongs as a subfamily of Spirobolidae. In his work, the suborder Trigoniulidea comprised a single family, Pachybolidae, with Trigoniulinae (Trigoniulidae) recognized as a subfamily. The tree produced using information from Hoffman’s (1980) taxonomy resolves the suborders as monophyletic, with each being a polytomy beyond that point (Fig. 1). Hoffman (1982) provides a unified treatment of Spirobolida and presents insights into relationships among families. In this work, Hoffman discusses affinities between Floridobolidae and Spirobolidae, Allopocockiidae and Atopetholidae, Atopetholidae and Spirobolellidae, and Spirobolelllidae and Typhlobolellidae. Although phylogenetic hypothesis were beyond the scope of Hoffman’s (1982) work, it is possible to infer a tree for Spirobolidea based on his classification and his discussion of interfamilial affinities (Fig. 2).

Wesener et al. (2008) has presented the first phylogenetic analysis of the order Spirobolida based on morphological characters. The main purpose of this work was to investigate internal relationships of Pachybolidae, but because the authors included a sampling of spirobolidan taxa that extended beyond this family, it is possible to infer trends in relationships from their study. They recovered Spirobolida and Spirobolidea (sensuHoffman, 1982) as monophyletic (their fig. 2, p. 42). As they only included four species of Spirobolidea in their analysis, there is little that can be gleaned about interfamilial relationships. The authors recognized and articulated this limitation and only discussed the monophyly of the suborder.

Present study

It is due to the strong taxonomic foundations established by workers before us that we are now capable of moving forward to investigate millipede relationships and taxonomies in a phylogenetic context. Our work serves to test the family limits within Spirobolida as established by these previous taxonomic studies. We take a total-evidence and an exemplar approach (for a detailed discussion of the exemplar approach to phylogenetic reconstruction see Marek and Bond, 2006). Using this methodology, our study is established on the largest reliable sample of characters and character systems as possible, with terminals sampled from as wide a geographical and morphological spread as possible within a family given specimen availability, while maintaining a rigorous test of inter- and intra-family relationships. As it would be prohibitive in cost, time, and resources to sample and analyse both morphological and molecular data from all nominal spirobolidan species, we feel this approach provides a strong initial test for monophyly of families and will serve as a foundation to future studies at lower taxonomic levels by establishing a phylogenetic framework within which taxonomic hypotheses can be constructed and evolutionary questions can be posed.

Specimens

We collected morphological and molecular data from 33 ingroup and three outgroup taxa. The specimens used in this work originated from collections at the Field Museum of Natural History and the California Academy of Sciences and from field collecting expeditions to California and Australia by K.M.P. When choosing terminals, we attempted to maximize the morphological and geographical diversity of taxa sampled from each family. Some families, such as Typhlobolellidae and Pseudospirobolellidae, have few members, making such sampling difficult. We were unable to sample members of Spirobolellidae throughout their geographical and morphological range due to a lack of available material suitable for molecular analyses, and this is emphasized in the discussion of results pertaining to this family. Males were used as voucher specimens from which morphological and molecular data were collected. This allowed us to provide accurate identifications of taxa, as most diagnostic characters at lower levels are based on male gonopod characters.

Specimens were identified to family level using Hoffman et al.’s (1996) key to Neotropical families and Hoffman’s (1982) descriptions of families. Generic and species-level identifications were done by comparing collection data from specimens with records from a catalogue of original descriptions of spirobolidan millipede species (P. Sierwald, unpublished data). Once candidate taxon identities were established at this level, we compared male gonopods against images of type material in the literature in order to make a positive identification at the generic and species levels. This was possible for the majority of species used in our analyses. Four terminals are identified only to the family level, two of which are species of Pachybolidae, a large family that has never been revised, and two are undescribed species of Pseudospirobolellidae from China that represent new species and at least one new genus. Information pertaining to voucher specimens used in this work can be found in Table 2.

The most recent ordinal-level phylogenetic analysis of Diplopoda by Sierwald and Bond (2007) recovered Spirobolida sister to (Spirostreptida + Julida); together these three orders form the superorder Juliformia. The Juliformia along with Nematophora and Merocheta comprise the subterclass Eugnatha. Given this phylogenetic framework, we chose members of the orders Spirostreptida (Juliformia), Julida (Juliformia), and Polydesmida (Merocheta) as outgroup taxa. All cladograms are rooted on Polydesmida, as this is the outgroup taxon most distantly related to Spirobolida (Sierwald and Bond, 2007). Information on outgroup taxa can be found in Table 2.

Tissues from two specimens used in the molecular portion of this study were sampled, preserved, provided, and identified by Thomas Wesener; these specimens will be deposited in the collections of the Field Museum of Natural History in the future. For all other specimens either legs or muscle tissues (both from body rings and from gonopods) were collected from individual specimens and placed in 95% ethanol (previously preserved specimens) or in RNAlater^® (Ambion, Austin, TX, USA) (live specimens).

Isolation of genetic material

Our study is the first molecular phylogenetic work within Diplopoda that relies heavily on museum specimens as sources of genetic material. When explicitly discussed, previous studies have used DNA extracted from muscle tissues isolated from body rings (Lavrov et al., 2002) and from legs (Marek and Bond, 2006); Marek and Bond (2006) specify that legs were pulled from live specimens and immediately preserved, whereas the condition or length of preservation of specimens used in other studies was not presented (Bond and Sierwald, 2002; Lavrov et al., 2002).

While dissecting specimens, we noted that muscle tissue associated with male gonopods appeared to be the best preserved soft tissue within the body. As the gonopods evert and articulate during the act of mating, there are areas where muscle in this region of the body is only covered by membranous tissue. Also, the muscle tissue associated with the seventh body ring of males appeared to be better preserved than tissue associated with body rings further from this membranous tissue, in that it remains fibrous whereas further away the tissue begins to soften. Because of this, we hypothesize that, when specimens are preserved, this membranous tissue serves as an entry point for ethanol which results in well-preserved tissue in these areas, and this tissue remains viable for molecular studies. Using muscle tissue associated with the gonopods and body ring seven, we successfully amplified genes from tissues extracted from specimens collected as early as 1991 (Messicobolus hoplomerus); it should be noted that this represents the oldest specimen from which we attempted to isolate DNA, meaning that it may be possible that future studies will find yet older specimens that are viable for molecular analyses. This demonstrates that museum collections are viable sources of molecular data for diplopod systematics studies.

Gene choice and molecular protocols

We chose to utilize molecular data from complete 18S and partial 28S ribosomal RNA genes, as these genes have been used in proposing phylogenetic hypotheses within Arthropoda at a variety of taxonomic levels and covering a range of the diversity within the phylum (Giribet et al., 1996; Edgecombe et al., 1999; Mallatt et al., 2004; Gillespie et al., 2005a,b). DNA was extracted using DNeasy© extraction kits (Qiagen Inc., Valencia, CA, USA) from legs or muscle tissues following standard protocols. Specimens used in this study were given unique identification numbers such that they could be recognized later and are housed in the collections of either the Field Museum of Natural History or the California Academy of Sciences.

Target gene regions were amplified using polymerase chain reactions (PCR); all amplifications were performed to obtain double-stranded products. Primers used in PCR were also used in sequencing reactions; primer sequences for 18S and partial (D3–D5) 28S fragments and their sources are detailed in Table 3. The following reaction protocol was implemented: 15.8 μL ultra-pure water, 2 μL 25 mm MgCl₂, 2.5 μL 10× buffer (Roche Diagnostics Co., Indianapolis, IN, USA), 1 μL 8 mm each primer, 1.5 μL 2 mm each dNTP, 0.2 μL AmpliTaq Gold Taq polymerase (1 U) (Roche Diagnostics Co.), and 1 μL DNA. Double stranded products were amplified for all reactions. Reaction conditions were typically 94°C for 12 min, 35 cycles of 94°C for 1 min, 50.5°C for 1 min, and 72°C for 1 min, and 72°C for 10 min, although the annealing temperature was varied between 45 and 52°C when amplifying from specimens that failed in initial reactions.

PCR products were purified using ExoSAP-IT^® (USB Co., Cleveland, OH, USA) employing a modified protocol. To each PCR tube, 0.8 μL of a 20× reaction buffer [400 mm Tris–HCL (pH 8), 200 mm MgCl₂] and 0.2 μL ExoSAP-IT were added. Samples were incubated at room temperature overnight; after the overnight incubation, samples were incubated at 80°C for 15 min to deactivate enzymatic activity. We used the BigDye^® Terminator v3.1 Cycle Sequencing Kits (Applied Biosystems, Foster City, CA, USA) to label PCR products following standard protocols and using standard reaction conditions. Cycle sequencing reactions were purified and sequencing was conducted using an Applied Biosystems 3730 DNA Analyzer (Applied Biosystems).

The software package Sequencher 4.8 (Gene Codes Co., Ann Arbor, MI, USA) was used to produce contigs of each gene fragment and to check and correct consensus sequences for missed calls. In rare instances where only a single strand for a target gene region was sequenced cleanly, the single strand was included in analyses if correcting base calls was negligible due to high-quality sequence.

Sequence alignment

A variety of multiple sequence alignment protocols are currently employed to propose homology statements within molecular datasets (De Laet and Wheeler, 2003; Gillespie et al., 2005a; Löytynoja and Goldman, 2005; Zaldivar-Riverón et al., 2006). We utilize two methods to align 18S and 28S data, progressive alignment using ClustalX (Thompsón et al., 1997) using default gap opening and extension parameters with no manual refinement, and alignment guided by secondary structure models (Kjer, 1995; Gillespie et al., 2005a,b). Use of a progressive alignment provides an implicitly repeatable and objective methodology (Giribet and Wheeler, 1999), although it has been suggested that manual refinement of alignments from automated methodologies, such as ClustalX, produce superior alignments than from automated methods alone (Edgar and Batzoglou, 2006). We employ the default parameters in ClustalX, as they have been optimized by the program’s authors to work for a majority of datasets (Chenna et al., 2003). If we were to alter the optimization criteria the only justifiable test would be to include all permutations of optimality criteria so all could be considered when choosing which set is optimal for our dataset. This sort of sensitivity analysis is beyond the scope of this paper. By conducting a second alignment methodology, we test the sensitivity of results based on our dataset to differing alignment criteria.

In order to form and function properly, ribosomal RNA subunits have a specific secondary structure, and this basic structure is conserved between divergent taxa (Hillis and Dixon, 1991). Alignment of rRNA genes (rDNA), such as 18S and 28S, among taxa is therefore constrained by their secondary structure. As a result, proposing multiple sequence alignments based on secondary structures proposed for the RNAs coded for by these genes is justifiable (Kjer, 1995; Hickson et al., 1996). Because stem regions are composed of complementary strands that base-pair with each other (including uracil pairing with guanine) forming the secondary structure, and because nucleotides in loop regions are not constrained by complementary base changes, stem and loop regions in a secondary structure model can be treated as different divisions in terms of nucleotide site change (Vawter and Brown, 1993; Muse, 1995). This can lead to higher mutation rates and higher prevalence of indels in loop regions compared with stems, leading to obscuring of phylogenetic signal. Gillespie et al. (2005a,b) identify and exclude from analyses portions of 18S and 28S where, due to high levels of base change and increased insertion/deletion events, positional homology assignments are ambiguous across the entire dataset; as in other studies (Deans et al., 2006; Pitz et al., 2007) these hypervariable areas are excluded from our analyses of secondary structure aligned data.

As there have been few studies of interfamilial relationships within Diplopoda using molecular character systems, no existing secondary structure models explicitly proposed for millipede ribosomal RNAs were available. Therefore, we aligned our dataset against a secondary structure model as proposed for Arthropoda for our 18S alignment (Gillespie et al., 2005a) and a secondary structure for Ichneumonoidea (Hexapoda: Hymenoptera) (Gillespie et al., 2005b) for our 28S alignment. Areas of variable length that could not be clearly reconciled using existing models were folded using the program Mfold (Zuker, 2003) to predict secondary structure for these spans.

Morphological data

To date there have been no analyses of higher-level relationships within Spirobolida employing morphological character suites in a phylogenetic context. Previously unrecorded morphological characters and character states were investigated using light microscopy (Leica MZ 12s stereoscope; Leica Microsystems Inc., Bannockburn, IL), digital imaging (Microptics™ imaging system; Microptics Inc., Ashland, VA), and scanning electron microscopy (SEM) (Leo Scanning Electron Microscope; Carl Zeiss SMT, Peabody, MA) (see character list and description in Appendix 1). All figures derived from SEM studies are provided with scale bars.

New character systems were combined with characters and character states interpreted from descriptions of higher-level taxa (Hoffman, 1982), from the few family-level revisions available within Spirobolida (Hoffman and Orcutt, 1960; Keeton, 1960a), and from a recently published phylogenetic analysis within Trigoniulidea (Wesener et al., 2008) to create a morphological data matrix using DEscription Language for TAxonomy (DELTA) (Dallwitz, 1980; Dallwitz et al., 1999). In our character definition, character state definition, and character coding, we have attempted to eliminate what we interpret as improper character and character state definitions and miscodings that were found in Wesener et al.’s (2008) character list, character definition, and data matrix.

Morphological data were recorded from the same specimen as was used to collect molecular data in all cases, with the exception of Narceus americanus (Palisot de Beauvois), where a specimen from the same collection event and locality as the one sequenced was used to code morphological data. Images resulting from our investigation into morphological systems were edited using Adobe^® Photoshop^® CS software (Adobe Systems Inc., San Jose, CA, USA) to clean up backgrounds and to format those chosen for use in this publication.

Analyses

Multiple methodologies were employed towards elucidating relationships within Spirobolida, including Bayesian inference (BI) and maximum parsimony (MP) analyses. Both MP and BI analyses were conducted on datasets resulting from progressive and secondary structure alignment methods. An analysis of morphological data alone was conducted using MP.

Maximum parsimony analyses

The program PAUP*4b10 (Swofford, 2002) was used to run unweighted MP analyses on datasets consisting of morphological data alone, combined ClustalX aligned molecular data and morphological data, and combined secondary structure aligned data and morphology. In all parsimony analyses containing molecular data each single base gap was treated as a fifth state. Each dataset was analysed using 10 000 random addition sequence replicates using TBR swapping (PAUP commands: hsearch nreps = 10000 addseq = random). Bootstrap support values (Felsenstein, 1985) and Bremer support values (Bremer, 1988) were calculated for consensus topologies resulting from MP analyses of both datasets. Bootstrap analyses consisted of 10 000 pseudoreplicates each with five independent random heuristic searches (PAUP commands: bootstrap nreps = 10000/hsearch nreps = 5 addseq = random). Bremer support values were calculated using default commands in the program Autodecay (Eriksson, 1998) in conjunction with PAUP*. In all instances where more than one most-parsimonious reconstruction was recovered, strict consensus trees were calculated.

Bayesian analyses

Using MrBayes 3.1.2 (Huelsenbeck and Ronquist, 2001; Ronquist and Huelsenbeck, 2003), a mixed model approach was taken in analysing datasets consisting of both alignments of 18S and 28S combined with morphological data. The program MrModeltest 2.2 (Nylander, 2004) was used to select appropriate models of evolution for both 18S and 28S fragments; the Akaike information criterion (AIC) was chosen as the favoured model for use in analyses (Posada and Buckley, 2004). The Markov k model (Lewis, 2001) both with and without gamma was applied to the morphological partition in both Bayesian analyses. Analyses consisted of two simultaneous analyses employing one heated and three cold Markov chain Monte Carlo (MCMC) chains that were run for 1 × 10⁶ generations sampling trees every 1000 generations. Analyses were concluded when average standard deviation of split frequencies dropped below 0.01 (Ronquist et al., 2005); if this threshold was not met during the initial 1 × 10⁶ generations, an additional 1 × 10⁶ generations were run until the average standard deviation of split frequencies fell below this value. Plots of generation versus log likelihood were graphed for both runs to determine when each run had stabilized; the higher of the two values for generation at which stabilization was reached was used to dictate burn-in. All parameters were analysed using the program Tracer 1.4 (Rambaut and Drummond, 2003) to determine that they had also reached stationarity. Using the sump command, post burn-in parameter values were averaged and, using the sumt command, post burn-in posterior probabilities (pp) for each clade were calculated. A 50% majority rules consensus tree was calculated with pp values recorded on this topology in all instances.

Morphological characters

We identified and coded 54 discrete and unordered characters for our analyses (Table 4). Of these characters, 20 (37%) are directly related to or coded from male reproductive structures, including structures related to both gonopore and gonopods, with the remaining 34 (63%) being somatic characters. The character list and discussion can be found in Appendix 1 and information about character systems not included in these analyses and why these were not coded is presented in the Discussion.

Table 4.
Matrix of 54 characters for 30 spirobolidan millipede taxa and three outgroup taxa

Sequencing, alignment statistics, and Bayesian models

Multiple sequence alignment using ClustalX resulted in an aligned sequence of 2386 bp, with 18S and 28S being 1933 and 453 bp, respectively. Secondary structure alignment resulted in an aligned sequence of 2020 bp, with 18S and 28S being 1754 and 265 bp, respectively. For ClustalX aligned data, mean base composition was A = 0.25785, C = 0.23116, G = 0.28212, T = 0.22887 with a homogenous nucleotide frequency among taxa (χ² = 99.1574, d.f. = 105, P > 0.64). For secondary structure aligned data, mean base composition was A = 0.26734, C = 0.22070, G = 0.28637, T = 0.22559 with a homogenous nucleotide frequency among taxa (χ² = 60.1254, d.f. = 105, P > 0.99). Using MrModeltest, the substitution models selected for both alignments of 18S and 28S data were GTR + I + G and GTR + G, respectively. Topologies resulting from all analyses conducted in this study are available at Treebase (accession no. S2549).

Maximum parsimony

For ease of discussion, we have divided our support values into artificial discrete divisions of strongly, moderately, and poorly supported. For all MP analyses, we consider clades with bootstrap values of 69 and below to be poorly supported, between 70 and 89 to be moderately supported, and 90 and above to be strongly supported. Clades with Bremer support values above five are considered strongly supported, those with values between two and four are considered moderate supported, and those with one and below are considered poorly supported. Values are reported as (Bremer/bootstrap) throughout the remainder of the Results.

Analyses of morphological data alone recovered 18 most-parsimonious trees of length 141 (CI = 0.482; RI = 0.722), from which a strict consensus was calculated (Fig. 3). Of the 54 characters in total, 47 were parsimony-informative. The order Spirobolida was recovered as monophyletic with moderate to strong support (5/81). With the exception of the families Spirobolidae and Spirobolellidae, all families included in the analyses were recovered as monophyletic. The three included species of Spirobolellidae were recovered scattered throughout the topology, whereas Spirobolidae is rendered paraphyletic by Floridobolidae. There is moderate to strong support for the monophyly of Messicobolidae (3/92); otherwise, support values are generally poor to moderate for the monophyly of families in this analysis (1–3/< 50–82). This demonstrates that we have yet to sample deeply enough within morphological character systems to provide strong support for familial groupings.

Analyses of both molecular datasets combined with morphological data recovered similar topologies (4, 5). The monophyly of Rhinocricidae was strongly supported (16–30/100) and this family was recovered as sister to all other included spirobolidan taxa. Pachybolidae, and therefore Trigoniulidea, was recovered as monophyletic (32–44/100). A monophyletic clade consisting of Typhlobolellidae + Atopetholidae + Spirobolidae + Messicobolidae + Floridobolidae was recovered with moderate to strong support by both total evidence analyses (6–8/89–92).

Combined analysis of molecular data aligned using ClustalX with morphological data resulted in two most-parsimonious trees of length 4107 (CI = 0.555; RI = 0.622) from which a strict consensus was calculated (Fig. 4). Of the 2440 total characters, 1134 were constant, 367 were parsimony-uninformative, and 939 were parsimony-informative. Spirobolida was recovered as monophyletic with strong support (58/100). All families, with the exception of Spirobolidae, were recovered as monophletic and with strong support (10–82/100); Spirobolidae was rendered paraphyletic by Messicobolidae and Floridobolidae in this analysis. Spirobolellidae was recovered as sister to Trigoniulidea, but the relationship (Spirobolellidae + Pachybolidae) lacked strong support (4/55).

Combined analysis of molecular data aligned using secondary structure and morphological data resulted in 13 most-parsimonious trees of length 2440 (CI = 0.556; RI = 0.620) from which a strict consensus was calculated (Fig. 5). Of the 2074 characters included in the analysis, 1225 were constant, 256 were parsimony-uninformative, and 593 were parsimony-informative. Spirobolida was recovered as monophyletic with strong support (41/100). All families with the exception of Spirobolidae were recovered as monophyletic with strong support (5–54/94–100). Spirobolidae, Messicobolidae, Atopetholidae, and Floridobolidae were recovered as a polytomy, with Spirobolidae being recovered as three monophyletic clades: Tylobolus, Hiltonius, and (Spirobolus + (Narceus + Chicobolus)). Spirobolellidae was recovered as a member of a polytomy with Pseudospirobolellidae and (Spirobolidea excluding Rhinocricidae).

Bayesian

For all Bayesian analyses, we consider clades with posterior probabilities of 0.69 and below to be poorly supported, of 0.70–0.89 to be moderately supported, and of 0.90–1.00 to be strongly supported. In all analyses except that of the secondary structure aligned molecular data with gamma, the average standard deviation of split frequencies fell below 0.01 within the first 1 × 10⁶ generations. In this analysis, it took 8 × 10⁶ generations to fall below 0.01. Topologies resulting from analyses with and without gamma-distribution rates applied to morphological partitions were identical at the familial level. Results are reported for analyses where gamma-distribution rates were not applied. A single difference in topology (below the family level) was found when comparing result of analyses with and without gamma from datasets aligned by ClustalX, and this is discussed below.

Because the same general topology was recovered by both BI analyses, we discuss results from both combined analyses (6, 7) together in the following section. All families, with the exception of Spirobolidae, were recovered as monophyletic and with strong support (1.00). Rhinocricidae was recovered as sister to all remaining species of Spirobolida. Pachybolidae was recovered with strong support for its monophyly (1.00). There was strong support for the monophyly of Spirobolidea (excluding Rhinocricidae) (1.00). Pseudospirobolellidae and Spirobolellidae were recovered as sister taxa with strong support (1.00). Typhlobolellidae was recovered as sister to (Atopetholidae + (Spirobolidae + Messicobolidae + Florodibolidae)) (1.00) and the clade (Spirobolidae + Messicobolidae + Floridobolidae) was recovered as monophyletic with moderate to strong support (0.81–1.00). In the analysis of the Clustal + morphology dataset excluding gamma-distribution rates, Floridobolidae was recovered as sister to (Hiltonius + Tylobolus). In the analysis of the same dataset employing gamma-distribution rates (tree not shown), Hiltonius was recovered as sister to (Floridobolidae + Tylobolus). Both analyses suggest that Spirobolidae is not a monophyletic lineage, while in the analysis of the secondary structure alignment + morphology, Floridobolidae was recovered as sister to Messicobolidae, with this clade originating from a polytomy; whether Spirobolidae may or may not be monophyletic is thus ambiguous.

Subordinal classification

Our study supports monophyly of the suborder Trigoniulidea composed of Pachybolidae. Previous taxonomic work grouped all remaining families not placed in Trigoniulidea into suborder Spirobolidea. This study demonstrates that suborder Spirobolidea (as previously defined) represents a paraphyletic lineage, with Rhinocricidae recovered as sister to all other spirobolidans. With the remaining spirobolidans, Trigoniulidea was recovered as sister to a monophyletic lineage composed of Atopetholidae, Floridobolidae, Messicobolidae, Pesudospirobolellidae, Spirobolellidae, Spirobolidae, and Typhlobolellidae; because Hoffmanobolidae and Allopocockiidae were not available for study we cannot assess their phylogenetic position based on our analyses, but based on morphological characteristics and geographical data we suggest that they are allied most closely to species recovered in Spirobolidea excluding Rhinocricidae (a detailed treatment of these families is given in the Discussion). Based on the results as detailed in 5-7, we propose the suborder Rhinocricidea composed of family Rhinocricidae and we redefine Spirobolidea to exclude Rhinocricidae. Trigoniulidea remains as previously defined (Hoffman, 1982; Shelley, 2003; Wesener et al., 2008).

The newly proposed Rhinocricidea contains the single family Rhinocricidae. Phylogenetically, it is sister to all remaining taxa within Spirobolida. The lineage is defined by a number of morphological characters. In both sexes, the clypeal foveolae are stabilized at 2 + 2 (Hoffman, 1982). All rhinocricid taxa examined in this study lacked a prominent lateral groove on the anteriodorsal margin of the collum; within Spirobolida, the only other taxon examined that exhibited this state was Ergene sp. nov. (Typhlobolellidae). In males, the anterior gonopod is distinct in shape; the sternite is developed into a large triangular plate such that there are no lateral extensions of the sternite (Fig. 8A) as are found in other spirobolidans (Fig. 8B). The coxite of the anterior gonopod forms a slight groove in which the posterior gonopod rests (Fig. 8C), but does not wrap around to create a cavity in which the posterior gonopods are housed (Fig. 8D). The telopodite of the posterior gonopod is simple, elongate, and thin and often with a bifurcation distally (Fig. 9A).

Trigoniulidea and Spirobolidea sensu this study are united by the form of the coxite of the anterior gonopod, which wraps around the posterior gonopod creating a distinct cavity in which the posterior gonopods reside (Fig. 8B, D). Trigoniulidea is characterized by the presence of a sclerotized sternite dorsally connecting the two posterior gonopods in males and by the tracheal apodemes of the posterior gonopods being nearly at a right angle to the coxite + telopodite (Hoffman, 1982). Spirobolidea sensu this study accommodates Allopocockiidae, Atopetholidae, Floridobolidae, Hoffmanobolidae, Messicobolidae, Pseudospirobolellidae, Spirobolellidae, Spirobolidae, and Typhlobolellidae. None of the characters examined in this study optimize as apomorphies for Spirobolidea, but Spirobolidea can be characterized by the following combination of characters: the absence of a sternite connecting the posterior gonopods (shared with Rhinocricidea), the tracheal apodeme running subparallel to the remainder of the posterior gonopod (shared with Rhinocricidea), and the anterior gonopods forming a cavity in which the posterior gonopods reside (apomorphic for Spirobolidea + Trigoniulidea). No somatic characters that unite either Trigoniulidea or Spirobolidea have been proposed previously or were uncovered by this study.

Familial classification and relationship

With the exceptions of Spirobolellidae and Spirobolidae, family limits as established prior to this study are supported by morphological character systems (Fig. 3); with the addition of molecular markers, the monophyly of Spirobolellidae was supported, with monophyly of Spirobolidae still being ambiguous (Fig. 7). The following section discusses all families included in our study with the exception of Rhinocricidae; for discussion of Rhinocricidae, refer to the previous section on suborder relationships.

It is necessary to stress that in the case of Spirobolellidae, taxon sampling may be a driving factor in this assemblage being recovered as monophyletic in our analyses. We also do not include any members of Trigoniulinae in this study. While assembling specimens for this study, we were unable to locate viable material from this group. The three species of Spirobolellidae selected as exemplars for this study are native to eastern Australia (two) and New Caledonia (one). We demonstrate the monophyly of these three species, thus implying monophyly for the family using the exemplar approach. Because of the proximity of the native habitats of these species, this may again be an artefact of taxon sampling, which is in turn a result of the availability of specimens. We maintain Spirobolellidae as a valid family because we recovered it as monophyletic in our analyses, with the understanding that this is a tentative hypothesis based on limited samples that will need to be assessed by future phylogenetic studies that include larger taxon samplings.

Spirobolidae was never recovered as a monophyletic lineage, regardless of data type, alignment criterion, or analytical technique implemented. In all cases, Spirobolidae was rendered paraphyletic by Floridobolidae (3, 6) or Floridobolidae and Messicobolidae (Fig. 4) or as part of a polytomy such that relationships within Spirobolidae and between Spirobolidae, Floridobolidae, Messicobolidae (Fig. 7), and Atopetholidae (Fig. 5) remain ambiguous. Analyses of sequence data aligned using ClustalX show higher levels of resolution within this complex of families than those aligned using secondary structure. This suggests that the regions of ambiguous alignment identified across the entire dataset are supplying characters that support the internal relationships within this assemblage. By employing a ClustalX alignment over sequences for all exemplar taxa from families Typhlobolellidae, Atopetholidae, Spirobolidae, Floridobolidae, and Messicobolidae and a subsequent parsimony analysis over this dataset, we were still not able to recover a topology with fully resolved relationships between Spirobolidae, Messicobolidae, and Floridobolidae that were well supported; within this truncated dataset only 98 of 2140 aligned base pairs are parsimony-informative (our unpublished data). This, along with the very low levels of genetic divergence within this assemblage (Fig. 10), demonstrates that the molecular markers sampled in this study are too conserved to address relationships within this clade. From its phylogenetic position, this clade is inferred to be young; the speciation rate may be elevated compared with others in Spirobolida, leading to faster divergence and less genetic difference between taxa.

Our results demonstrate that the monophyly of Spirobolidae was not supported, and the relationships among genera of Spirobolidae and between these genera and Floridobolidae and Messicobolidae will remain unclear until future study of this assemblage is undertaken. This problem would benefit from a detailed study at the generic level across Spirobolidae, Messicobolidae, and Floridobolidae including new molecular and morphological data; we know of an ongoing study that will help us better understand some lower-level relationships within this group (M. Walker, East Carolina University, personal communication) and we feel it prudent to wait for more data before making a taxonomic change in this regard. Based on our results, it is probable that such a future study will result in synonomy of Floridobolidae and Messicobolidae with Spirobolidae.

The family Atopetholidae was supported by a large number of morphological characters, including: the collum coming to blunt points ventrally below the level of the second body ring, the presence of a fringe of closely spaced setae on the paraprocts, having the mesal edges of paraprocts inturned and meeting at a reentrant angle, and having hypertrophied tarsal claws on leg pair 1 of males. As members of this family are known mainly from arid regions of North America and few other spirobolidan millipedes are known from this habitat type, it is not surprising that atopetholids would be morphologically unique; it is unclear whether any of the previously mentioned characteristics can be correlated with life in this habitat, but it is certain that future detailed studies of this family will identify morphological adaptations to this environment.

Two undescribed species of Pseudospirobolellidae are included in this study. These taxa differ from described members of the family in lacking semicircular depressions dorsally on the mesozona of body rings (Hoffman, 1982); these specimens do possess membrane-like pads on the prefemora, with this being the morphological characteristic that unites these species in our analyses and identifies these two species as members of Pseudospirobolellidae. With the exception of the species Pseudospirobolellidae sp., China st (Fig. 9B), all members of Pseudospirobolellidae lack a sternite on the anterior gonopods of males (Fig. 9C). Hoffman (1982) interprets this character as the sternite being fused to the coxites. Because there are very short lateral extensions present on the sternite of this new species, it appears that the sternite is not fused to the coxite, but it is in fact reduced in this species and absent completely in other species. The discovery of this new morphological variation within the family means the morphological boundaries of the group must be redefined to include presence of a median sternite of the anterior gonopod. A detailed study of these two species is planned in which a formal taxonomic treatment of the family will be made.

Messicobolidae is recovered as a monophyletic group in all analyses. Male specimens exhibit prominent ventrally produced endites on the coxites of the anterior gonopod (Fig. 9D), which represent a morphological autapomorphy for this family. Unlike in members of Floridobolidae and Spirobolidae, the tracheal apodeme of the posterior gonopods of the messicobolid exemplars used in this study are not fused to the coxite, but freely articulating; this condition is similar to what is found in all members of the order outside of Spirobolidae and Floridobolidae.

There are few other interfamilial relationships that are supported by morphological character systems. The relationship (Atopetholidae + (Spirobolidae + Floridobolidae + Messicobolidae)) was recovered in all analyses (as a polytomy in MP analysis of a total-evidence dataset aligned using secondary structure). In this group, the coxite of the anterior gonopod is elongated along its lateral axis (Fig. 8B), with all other exemplars (Pachybolus ligulatus being the exception) having the coxite elongated along the longitudinal axis (Fig. 11A). The group (Ergene + (Atopetholidae + (Spirobolidae + Floridobolidae + Messicobolidae))) is supported by having the coxite on the third and fourth leg pairs of males produced ventrally (Fig. 11B) (this character state is not present in Atopetholus sp.); this character state is present in two species of Rhinocricidae and two species of Spirobolellidae, suggesting that although it is a uniting characteristic for this lineage, it is not one that can be used to solely define it.

Coding morphological data

The branch support recovered from analysis of the morphological data alone (Fig. 3) reflects the current size of the morphological data matrix (54 morphological characters for 33 taxa). Across millipedes and Spirobolida, morphological character suites are insufficiently developed to propose relationships; for example, few ultrastructual studies exist examining mouthparts and the distribution of sensilla on the gnathochilarium and the antennae. Millipede systematics is hampered by our inability to accurately homologize across many character systems, especially male gonopods. This is a result of two distinct issues facing systematists: strong divergence among orders and a lack of knowledge in relation to variation. Divergence in form among orders is often so great that it is impossible to homologize morphological variation beyond assigning primary homology to gross morphological structures, eliminating the ability to propose homology of finer-level variation. It will be through the implementation of large-scale studies of specific character systems (discussed below) that we will be able to build a foundation for examination of morphological variation within Diplopoda.

Sense cones, sense fields, setae, and hair-like sensillae found on the labral region, gnathochilarium, and antennae are character systems that may highlight variation between taxa at a variety of levels (species, genera, families, and above). These features were not examined in this study, as the only way to assess details of these structures is through scanning electron microscopy (SEM). Spirobolidan (and diplopod) morphology has, as yet, not been subjected to large-scale SEM work across a broad range of taxa, and such a project is well beyond the scope of this study. Therefore, weak branch support recovered in the MP analysis of morphology alone is probably strongly affected by a lack of data from such studies. A detailed examination of variation of these structures outside the context of a phylogenetic revision would be most beneficial, and we hope to pursue such studies in the future with the goal of identifying phylogenetically informative character systems using novel morphological models.

In defining and coding morphological characters, we recognized that there are not enough landmarks and the currently available taxon sampling is not dense enough to allow us to homologize structures of the telopodite of the posterior gonopod among ingroup and outgroup taxa due to strongly divergent gonopod morphologies. We hope to study gonopodal homology in the future, specifically relating to intergeneric relationships within families. Many gonopod characters are coded as inapplicable in outgroup taxa because we are not able to assess homology beyond that of the gross structure of the telopodite and coxite themselves, placing homology of the fine structures on the telopodite and coxite in question.

No vulva characters were included in this study due to the need to develop this character system from the ground up, which was outside the scope of this work. Also, it is often not possible to determine the identity of female specimens unless they are collected in association with males. Because our dataset relies on both museum specimens and recently collected individuals it was not possible to identify, with the level of accuracy needed for a study of this kind, viable female specimens for all terminals.

Historical biogeography

It has been well established that, with the exception of the genus Spirobolus, all members of the families Atopetholidae, Floridobolidae, Messicobolidae, Spirobolidae, Allopocockiidae, Hoffmanobolidae, and Typhlobolellidae are known only from Central and North America. In all topologies, we recovered a monophyletic lineage containing all exemplars from these families (3-7). All members of these families share a common ancestor, and the diversity of this lineage is probably a result of a single common ancestor and subsequent radiation into North and Central America. The exception to this is the genus Spirobolus, which is known only from China (Keeton, 1960a). Spirobolus was recovered as sister to (Chicobolus + Narceus) in all analyses in which it was included (4-7).

Chicobolus and Narceus are found only in eastern North America, with all remaining genera of Spirobolidae known only from northern Mexico and the western USA. This begs the question of how Spirobolus or its ancestors became established in China. Millipedes are limited in their dispersal ability by their habitat requirements, their lack of dispersal mechanism (neither ballooning, flying, nor swimming) and their slow locomotion (Hopkin and Read, 1992). Based on our knowledge of spirobolidan millipedes, the hypothesized food source of the lineage that gave rise to Spirobolus is decaying vegetation. Because of this, colonization of intervening areas across land with subsequent extinction seems the most likely scenario, but whether this was due to island hoping, colonization over Beringia, or by other means is uncertain. Forms in either of these areas, if present, have yet to be discovered. Future studies involving calibration points in the phylogeny may lead to establishing a timeframe for the split between Spirobolus and Chicobolus + Narceus, which would help us begin to understand when this divergence happened and possibly provide clues as to how it happened as well.

Allopocockiidae and Hoffmanobolidae

These two families represent some of the least studied and least sampled taxa within Spirobolida. Allopocockiidae contains five genera and a total of eight valid species (Hoffman, 1999), and Hoffmanobolidae is represented by a single species known only from the type series collected in 1966 (Shelley, 2001). Both families are restricted to Mexico and Central America, with a single species of Allopocockiidae being known from south-eastern Texas in the USA. Until a detailed study of both families is possible, where both molecular and morphological data can be collected, it will not be possible to understand the true relationships between these taxa and the remaining families of Spirobolida.

Previous work has proposed relationship between Allopocockiidae and Atopetholidae (Hoffman, 1982), but without proposing characters to support this relationship. Shelley (2001) suggests that the structure of the coxites of the anterior gonopods, being medially elongated, allies Hoffmanobolidae with Atopetholidae, but he did not recognize that members of Messicobolidae also exhibit this trait. Examining images of gonopods from all families of Atopetholidae and Messicobolidae found that this characteristic is present in both families. A cursory comparison of gonopod structure between members of the genera Atopetholus, Watichelus, Centrelus, Comanchelus, Eurelus, Toltecolus, Onychelus, Arinolus, Piedolus, Scobinomus, Tarascolus, Cyclothyrophorus and Hoffmanobolus mexicanusShelley, 2001 illustrate some of the problems in hypothesizing affiliations between Hoffmanobolidae and other spirobolids.

The anterior gonopods of Hoffmanobolidae and a number of genera of Atopetholidae are similar in multiple ways. They both have medioventral elongations of the coxites, the sternite is produced as a thin, dorsal band with little to no medial ventral projection, and the telopodites are usually deeply uncinate to bilobed in appearance. These are characteristics also found in members of Messicobolidae, and a comparison of the anterior gonopods of Hoffmanobolidae (Fig. 11C) to those of a species of Messicobolidae (Fig. 9D) demonstrates this similarity, and finds them to be more similar to the gonopods of Messicobolidae than to gonopods of Atopetholidae. This is in contrast to the suggestion by Shelley (2001) that Hoffmanobolidae may be affiliated with Atopetholidae based on the coxite structure (Fig. 11D). It is apparent that without a detailed study of specimens using both morphological and molecular data, the relationships between these and other families will remain unresolved. Although it appears that Hoffmanobolidae’s fate may be a synonymy in the future based on empirical evidence, we chose not to burden the literature with conjecture before evidence so we maintain the family Hoffmanobolidae until adequate data are collected to support our supposition. Based on the geographical origins and ranges of Hoffmanobolidae and Allopocockiidae, together with the morphological similarities, this suggests these lineages are related to the other families of Spirobolidea restricted to North and Central America (Atopetholidae, Floridobolidae, Messicobolidae, Spirobolidae-Spirobolus, and Typhlobolellidae). Therefore, these two families are maintained as members of Spirobolidea, probably related to Atopetholidae or Messicobolidae, with the anticipation of testing this hypothesis with the addition of data in the future.

Comparison against previous works

As outlined in the Introduction, there has been little phylogenetic work done in Spirobolida and there have been few hypotheses of interfamilial relationships previously proposed. Hoffman (1980) did not present explicit hypotheses of relationship between families of Spirobolidea, although these can be inferred from his work (Fig. 2), but he did make predictions that we can examine in light of our results. The most relevant of his observations were with regard to the trend towards reduction in the posterior gonopods and loss of the sternite of the posterior gonopods.

If we assume, as Hoffman (1980) did, that the presence of a sternite on the posterior gonopods is the pleisomorphic condition, then our results support the hypothesis that there have been multiple independent losses of this character: once in the common ancestor of Rhinocricidae and again in the ancestor of the Spirobolidea sensu this study. Hoffman (1981) questioned the suborder classification within Spirobolida and suggested that the lack of a sternite connecting the posterior gonopods might be a convergent character, and that Rhinocricidae may not be closely related to other families lacking such a sternite. Hoffman’s (1981) hypothesis is clearly supported by this study, as Rhinocricidae was recovered as sister to all remaining spirobolids and not associated with Spirobolidea.

The families Rhinocricidae, Spirobolellidae, and Typhlobolellidae exhibit reduction/simplification of the posterior gonopods. Our results did not find these lineages to share immediate common ancersors as all were sister to families whose gonopods are not reduced in form. This supports Hoffman’s (1980) notion that the reduction of posterior gonopods has multiple evolutionary origins.

Hoffman (1980) also suggests that Floridobolidae may represent a subfamily of Spirobolidae and not a true family. In the majority of trees, including that for morphology alone, Floridobolidae was recovered as a sister to members of Spirobolidae. This suggests that Floridobolidae renders Spirobolidae paraphyletic and should be formally transferred. We have chosen to take a conservative approach and have not formally transferred Floridobolidae because we have not fully resolved relationships among Floridobolidae, Messicobolidae, and Spirobolidae; future work addressing these relationships is planned and will address this question.

It should be noted that Floridobolidae (a monotypic family) and Typhlobolellidae are each represented by a single terminal in this study; therefore, recovering these families as monophyletic was implicit. Our results are exciting because, in most instances, they affirm the previously proposed taxonomic composition of families through the use of multiple sources of data; they also provide fertile ground for the development of future research questions.

Wesener et al. (2008) has published phylogenetic work aimed at elucidating relationships within Pachybolidae, but including a sampling of non-pachybolid spirobolids. Their results showed support for the monophyly of Spirobolida, Spirobolidea (sensuHoffman, 1982), and monophyly for subgroups within Trigoniulidea. Within Spirobolidea they included specimens of Atopetholidae, Rhinocricidae, and Spirobolidae. Their subordinal relationships differ strikingly from those that we recovered in our analyses. We supported the monophyly of Rhinocricidae and, unlike their study, found this lineage to be independent of Spirobolidea. We also recovered Trigoniulidea as monophyletic, which was not found in their work. Our inclusion of a number of taxa from Spirobolidea excluding Rhinocicidae allowed us to robustly test the monophyly of this group, which could not be done in their study.

As the scope of Wesener et al.’s (2008) paper varies markedly from ours, it is not surprising that they include a number of characters that are not included in our study. A number of these characters were originally proposed and coded within our data matrix but later excluded for reasons such as invariability in our taxon sampling or because they represented autapomorphies; certain autapomorphic characters were maintained in our dataset in instances where they defined monotypic families. In our character list (Appendix 1), all characters shared between the two datasets are referenced to the corresponding character in Wesener et al. (2008) and any alternative interpretations of characters between their and our analyses are discussed to make it possible for future researchers to marry the two datasets if it is so desired (Appendix 1).

Conclusions and future directions in spirobolidan research

This study provides the first phylogenetic test of inter- and intra-familial relationships within the order Spiroboilda, and it sets forth a phylogenetic framework in which future studies of lower-level relationships among spirobolidan millipedes can be developed. We have proposed testable hypotheses of relationship based on multiple data sources, and we hope this will be built upon by future work in this group. The majority of families as proposed prior to this work have been recovered as monophyletic lineages, validating much of the historical taxonomy within this group.

A major advance in this study is uncovering the relationships among families and identifying morphological synapomorphies that define the order, suborders, families, and interfamilial groupings as recovered in our analyses. Even more intriguing is finding that certain lineages of dubious monophyly when examined using only morphological data, such as Spirobolellidae, are recovered as monophyletic with the addition of molecular data. This suggests that the currently defined morphological character systems are not able to accurately reconstruct the evolutionary history of Spirobolida. The limited external morphological variation combined with strong divergence in male gonopod structure at higher levels suggest that molecular data may serve as the key to resolving questions of higher-level relationships within Diplopoda.

Much of our future work within Spirobolida will focus on resolving intrafamilial relationships through the addition of a new molecular marker and new morphological character systems. Of interest will be an assessment of the clade containing Floridobolidae, Messicobolidae, and Spirobolidae. Not only will this provide insight into the evolutionary history of the lineage, but it will allow researchers to address questions pertaining to colonization of and dispersal within the USA by this ecologically important group of millipedes.