Capitella (Annelida: Capitellidae) species in the Gulf of Mexico: Delimitation, phylogeography and phylogeny

Capitella spp. are found in marine and estuarine benthos world‐wide. They are often treated as environmental sentinels because they occur in high densities in areas of anthropogenic disturbance and organic pollution. However, there are many cryptic species around the world, often grouped into the Capitella capitata complex. Historically, these have been differentiated using allozymes and developmental studies. Recently, single‐gene sequencing, especially cytochrome c oxidase subunit I, and phylogenetic methods have been used to characterise populations around the world. We present molecular support and provide descriptions for five new Capitella spp. from the Gulf of Mexico: Capitella gomexa sp. n., Capitella mandingensis sp. n., Capitella tampe sp. n., Capitella maculosa sp. n. and Capitella loucindae sp. n. Two were recently differentiated on the basis of larval development and morphology. Additionally, we present an updated look at global Capitella phylogeny. While no apparent patterns of phylogeography were recovered, we did find support for a single origin of thoracic acicular spines.

Two morphotypes of Capitella have been reported from the northern Gulf of Mexico (GoM) (Hilliard et al., 2016).Based on the presence of acicular spines, one of these matches the description of Capitella aciculata (Hartman, 1959), while the second one, without acicular spines, fits the general description of C. capitata.However, COI sequence data could not confirm the separation of the two morphotypes as separate clades (Hilliard et al., 2016); instead, they apparently constitute a single, morphologically plastic species.In addition, the sequence data suggest a genetic break between the eastern and western populations in the northern GoM.To document developmental differences between these populations, we described differences in the early development and adult morphology between Capitella spp.from Tampa Bay, FL and Tamiahua Lagoon, Veracruz, Mexico (Méndez et al., 2019).Populations with acicular spines also occur along the Brazilian coast, but were genetically distinct from the GoM populations and described as a new species, Capitella neoaciculata (Silva et al., 2017).
For the present study, we examined additional samples from the US and Mexican coasts of the GoM.Our objectives were to (1) determine molecular support for the Tampa Bay and Tamiahua Lagoon populations (Méndez et al., 2019) and identify possible geographical patterns, (2) delimit species boundaries, describe novel species, and re-evaluate the significance of acicular spines as diagnostic characters and (3) analyse the phylogeny of the genus Capitella using the newly generated data in conjunction with publicly available data.

| Specimen collection and storage
Capitella specimens were collected from shallow mud/ sand flats, canals and the edge of red mangrove swamps along the GoM coast and in Miami, Florida (Figure 1 and Table 1).Sediment was collected with a shovel or Ekman grab at a depth not exceeding 15 cm and sieved through a 0.5 mm or 1 mm sieve.Some specimens were hand picked off the mesh and stored in a centrifuge tube with ambient water while others were later picked from retained material (stored in a plastic storage container with ambient water).Most adult Capitella spp.(identified by the presence of nine thoracic chaetigers and/or genital spines in the 8th and/or 9th chaetigers) were processed within several hours of collection by relaxing them with a 7% magnesium chloride (or magnesium sulphate) solution and then fixed and stored in 95% ethanol.Some specimens from Apollo Beach Preserve in Tampa Bay, Florida and from Tamiahua Lagoon, Veracruz were maintained in culture to document their early development (Méndez et al., 2019).Some of the cultured specimens were sequenced for this study as well.
F I G U R E 1 Sampling locations throughout the Gulf of Mexico.The shapefiles used to make this map are from version 2.3.7 of the Global Self-consistent, Hierarchical, High-resolution Geography (GSHHG) Database, accessible at http:// www.ngdc.noaa.gov/ mgg/ shore lines/ gshhs.html.The map was produced using QGIS.
Cycle sequencing was conducted with the BigDye® Direct Cycle Sequencing Kit (Life Technologies, now Thermo Fisher Scientific) using 1 μL Big Dye® Terminator v3.1 Cycle Sequencing RR-100, 2 μL Big Dye® Terminator v1.1/3.1 Sequencing Buffer (5X), 1 μL primer (10 μM), 1 μL template DNA and 5 μL autoclaved Milli-Q® water.The thermocycler protocol consisted of 15 cycles at 96°C for 10 s, 50°C for 5 s and 60°C for 75 s; five cycles with the same settings except an extension time of 80 s; and five cycles with the same initial settings except an extension time of 120 s (Hilliard et al., 2016).Reaction products were cleaned with a ZR DNA Sequencing Clean-Up Kit™ (Zymo Research) following the manufacturer's protocols using half the designated volume of all reagents with final elution in 15 μL of Hidi formamide (Thermo Fisher Scientific).Products were analysed with an ABI 3130 Genetic Analyzer (Applied Biosystems, now Thermo Fisher Scientific).Resulting electrochromatograms and base calls were edited in Sequencher™ 4.8 (Gene Codes Corporation) by assembling the forward and reverse reads of each specimen.Final sequences were exported as FASTA files and are accessible in GenBank (Table 1).
ML analysis of H3 was completed with the Jukes Cantor substitution model (Jukes & Cantor, 1969), rapid bootstopping (Stamatakis et al., 2008) with a random seed of 333 and automatic bootstopping (Pattengale et al., 2009), a final search for the best-scoring tree, and Heteromastus sp. as the outgroup.The BI analysis was completed with the Jukes Cantor substitution model (Jukes & Cantor, 1969), 9,000,000 total generations, a 1000-generation sampling frequency, a 5000-generation diagnostic frequency and a 25% burn-in fraction.
Bayesian Poisson Tree Process (bPTP) (Zhang et al., 2013) was used to delimit species using the COI ML T A B L E 1 All sampling localities, collection dates and GenBank accession numbers for samples collected by the authors in this study (italicised) and Hilliard et al. (2016).

COI H3
Texas tree.A seed of 333 was set and Markov Chain Monte Carlo was run for 6,000,000 generations.The input tree was rerooted on the longest branch and the H. filiformis and N. profondus were specified as outgroups.

| Morphological observations
All specimens used for the sequence analyses were characterised with light microscopy.Some additional specimens from the Apollo Beach culture were relaxed with a 7% magnesium sulphate solution and then fixed with 95% ethanol.The Methyl Green staining pattern (MGSP) was assessed for all specimens of the five novel species.This was accomplished by bathing them in a saturated Methyl Green and 70% ethanol solution for at least 60 s, transferring them to clean 70% ethanol until there was no obvious destaining, and a final transfer into 70% ethanol for imaging and characterisation (Blake, 1996).Imaging was accomplished with a Leica S6D stereoscope, Olympus BH2 compound microscope and Leica EC3 camera.
All specimens collected for this study and used for species descriptions and/or phylogenetics are deposited in the Fields Museum of Natural History (FMNH) in Chicago, Illinois under Catalogue Numbers 15996-16115.FMNH 15998-16010 and FMNH 16108-16110 were collected during the Benthic Invertebrate Taxonomy, Metagenomics and Bioinformatics (BITMaB) Workshop 2017, part of the Gulf of Mexico Research Initiative (GoMRI) project 16-052 (Reuscher, 2017).
This manuscript and the species described within are registered in the Official Register of Zoological Nomenclature (ZooBank) in compliance with the International Code of Zoological Nomenclature.The registration number for this manuscript is urn:lsid:zooba nk.org:pub:BADB0ADD-C625-4161-B88B-9893F356D2BA and it can be viewed using this URL: https:// zooba nk.org/ badb0 add-c625-4161-b88b-9893f 356d2ba.

| COI analysis
A total of 112 specimens were successfully sequenced for COI.Analysis of COI sequences revealed a total of 17 Capitella clades (Figure 2).While six of the clades (1,2,8,9,11,12) are limited to a single location, the remaining ones encompass specimens from at least two locations.Clades 3,5,6,13,14,16 and 17 have geographical origins from multiple locations within one country, so only that country is listed in Figure 2 (see Table 2 for additional details).
The GoM specimens fall into seven different clades in the phylogenetic tree (clades 2, 8, 9, 10, 11, 12 and 15) (Figure 2).Only one of them, clade 15, can be assigned to a named species, Capitella nonatoi (Silva et al., 2017), originally described from Araçá Bay, São Paulo, Brazil and here reported from Ciudad del Carmen and Texas (Galveston and Suter Park).Specimens from Tamiahua, Veracruz form a clade with some specimens from Ciudad del Carmen, Texas, and Florida (Clade 10).This clade was referred to as Capitella sp.TV in Méndez et al. (2019) and is hereafter referred to as Capitella gomexa sp.n.
Some specimens from Apollo Beach Preserve in Tampa Bay form a distinct clade (Clade 2).This clade was referred to as Capitella sp.TF in Méndez et al. (2019) and is hereafter referred to as Capitella maculosa sp.n.The remaining four clades with samples from the GoM cannot be linked to any previously reported species and three are described as new species (Capitella loucindae sp.n., Capitella tampe sp.n., Capitella mandingensis sp.n.).Clade 8 is not formally described as there was only a single specimen (Figure 2). Figure 3 shows that C. gomexa sp.n. is split into Western and Eastern regions, with all Texas and Mexico samples in one group and all Florida samples in the other, respectively.The two groups are separated by three mutations.Specimens with acicular spines were found throughout (Figure 3).Within the Eastern Region are all of the samples from Miami and nine of the samples from Apollo Beach Preserve, including field collected and cultured specimens.The remaining eight field-collected specimens from Apollo Beach Preserve are described here as C. loucindae sp.n. and form a distinct clade sister to C. gomexa sp.n. (Figure 2).Two cultured specimens form a clade near the base of Capitella (Figure 2).This clade has been described as C. maculosa sp.n. and morphological and developmental differences from C. gomexa sp.n. have been described (Méndez et al., 2019).
Capitella nonatoi (Figure 2) includes specimens from Corpus Christi, Galveston and Ciudad del Carmen, in addition to the sequences from Brazil.Reconstruction of a haplotype network revealed geographical separation between the GoM and Brazil, with some haplotypes shared between Ciudad del Carmen and Corpus Christi (Figure 4).K2P genetic distances for COI between all populations are reported in Figure S1.Multiple samples were collected from Suter Park, Corpus Christi Bay (11 January 2017 and 29 May 2017) and Galveston Bay (10 July 2015 and 23 May 2018) Texas, but C. nonatoi specimens were only found during May sampling events.Capitella gomexa sp.n. was present during the other samplings as well as the other locations throughout both bays.
Species delimitation analysis using bPTP clearly separated the 17 Capitella clades (Figure 2).While 13 of the clades were confirmed as species under both maximum likelihood and Bayesian posterior probability, four clades (Clades 11, 13, 14 and 17) were additionally split up into several species, depending on the criterion used.For Capitella tampe sp.n. (Clade 11), both solutions delimited two species (each represented by a single specimen).Capitella neoaciculata (Clade 13) contained a single species under the maximum likelihood criterion but eight species under Bayesian posterior probability.Clade 14 includes two species under maximum likelihood and forms a single species under posterior probability.Capitella biota (Clade 17) (Silva et al., 2017) was delimited as six species under maximum likelihood and as eight under posterior probability.

| H3 analysis
Analysis of H3 did not reveal much, if any, geographical structure.The two Ciudad del Carmen specimens that clustered within C. nonatoi by COI analysis formed a well-supported clade with two specimens from Fish Pass, Corpus Christi, Texas (Figure S2).The average K2P distance between C. nonatoi H3 and all other Capitella is 8.54%.Average K2P distances among Clades 9-12 range 0.517%-11.5%.We did not successfully sequence COI for the Fish Pass specimens.Several of the samples from Apollo Beach Preserve formed supported clades, but the clade membership does not align with that of COI.

| Morphological analysis
The newly collected and sequenced specimens ( 85 and S10).The species were characterised using the prostomium shape, thoracic chaetal formulae, length of the thoracic region, morphology of the pygidium, and grooves (ventral and lateral).Species descriptions can be found in the Appendix S1.
Capitella gomexa sp.n. exhibits variable Methyl Green staining depending on sex and geographical origin and

| DISCUSSION
Our results show that Capitella has high diversity along the GoM coast.2).Capitella species are generally not isolated from each other to particular locations in the GoM and most locations host several species.Furthermore, the proportional contributions of different species at a particular location may be temporally variable, depending on season, environmental conditions, or random factors (Gamenick et al., 1998;Grassle & Grassle, 1976).For example, C. nonatoi was only collected in May 2017 and 2018 in Suter Park, Corpus Christi Bay and Galveston Bay, respectively.In comparison, C. gomexa sp.n. was the only species found at these locations at other times of the year and they were found in Suter Park in May 2017.
The acicular spine morphology occurs in a wellsupported derived clade (Clades 8-13, Figure 2), suggesting a single evolutionary origin.Within this acicular clade, C. neoaciculata (Clade 13) was reported from Brazil and it is monophyletic with species from southern Veracruz and Campeche, Mexico.No Capitella species with acicular spines are known outside of the Western Atlantic and its adjacent basins.We suggest that the aciculate morphology could have evolved after the opening of the GoM and Caribbean Sea during the Jurassic Period  (Ross & Scotese, 1988).This also indicates that aciculate species could be present, but unknown, in the Caribbean Sea.However, there is currently not enough evidence to establish a molecular clock for Capitellidae.Some recent phylogenomic analyses place the Capitellidae and Echiura as the sister group to the Clitellata (Helm et al., 2018) while others find a sister group relationship between Clitellata and Terebellomorpha (Weigert et al., 2014).Regardless of which of these hypotheses is favoured, the divergence of the Capitellidae from clitellates predates the radiation of the Clitellata, which dates back to the Triassic Period (Manum et al., 1991).Further work using the fossil record of clitellates will be necessary to calibrate a molecular clock for Capitellidae and related taxa.
Capitella capitata was described from Greenland and has been redescribed with material from there (Blake, 2009).Blake (2009) hypothesised that C. capitata is widely distributed in Arctic and Subarctic waters and this is supported by the current analysis (Figure 2) and the original report of the C. capitata COI sequence from Greenland (Silva et al., 2017).Interestingly, there is support for samples from the Indo-Pacific region being true C. capitata (Figure 2).This is likely the result of an introduction by anthropogenic activity.Capitella teleta is also widely distributed and has been detected in the Mediterranean Sea and waters of South Korea and Japan (Figure 2).It was hypothesised that this, too, is due to anthropogenic activity (Tomioka et al., 2016).Continued sampling around the world will shed light on these patterns.
Three species occurred sympatrically at Apollo Beach Preserve in Tampa Bay, Florida.At the time of collection, we did not detect any obvious differences among the specimens and preserved some of the larger individuals for morphological and molecular studies.The smaller specimens and the remaining larger specimens were kept in culture, assuming they all belonged to the same species.Our COI sequence analysis revealed that (1) C. gomexa sp.n. was present both in the culture and the field, (2) C. maculosa sp.n., a smaller species, was only detected in the culture and (3) C. loucindae sp.n., was only sequenced from field material (although it is possible that some are present in the culture but have not been sequenced).2).Locations marked by a triangle in the legends had some specimens with acicular spines.One hash represents a single mutation.Small black squares within the network represent a predicted haplotype that is not present in the data set.
(referred to as Capitella sp.TF) by their larval development, based on samples collected from Tamiahua, Veracruz and Tampa Bay, FL, respectively.These three species may have different ecological characteristics and/or occupy different sediment depths (Gamenick et al., 1998) or they may maintain sympatry through differing seasonal dynamics (Grassle & Grassle, 1976).However, our sampling design did not address this.Additional work is needed to F I G U R E 4 Median joining haplotype network of COI for Capitella nonatoi (Clade 15, Figure 2).One hash represents a single mutation.Small black squares within the network represent a predicted haplotype that is not present in the data set.
further understand developmental and ecological differences of these three species.
Capitella gomexa sp.n. from Miami were collected at Matheson Hammock Park, the type locality for Capitella caribaeorum (Warren & George, 1986).They are different sizes with C. caribaeorum having a thorax about 1.29 mm long (average of measurements made using Figures 1b and 2a in Warren & George, 1986) and C. gomexa sp.n. having an average thorax length of 3.01 mm (Taxa Description, Appendix S1).Additionally, our specimens from this location were collected differently.While Warren and George (1986) described C. caribaeorum from a culture stemming from intertidally collected decaying mangrove leaves, our specimens were retrieved by sieving sediment from sand flats with interspersed algal mats.
We further considered the possibility that C. maculosa sp.n. (Clade 2) represents C. caribaeorum, but differences in size (C.caribaeorum: mid-thorax width 0.7 mm and thorax length 1.29 mm; C. maculosa sp.n. holotype: midthorax width 0.22 mm and thorax length 1.95 mm) and life cycle make this unlikely as well.Capitella caribaeorum larvae metamorphose in the brood tube and emerge as juvenile worms (George, 1975) whereas C. maculosa sp.n. larvae emerge from brood tubes as swimming metatrochophores, settling to metamorphose after 3 days (Méndez et al., 2019).
H3 had limited utility for species delimitation in Capitella, as the phylogenetic analysis did not support most of the clades revealed in the COI analysis.The only species that was supported as monophyletic in the H3 analysis was C. nonatoi.Overall, the K2P distances between the GoM Capitella likely reflect more recent speciation.Although H3 has been successfully used to delimit three cryptic Heteromastus species from Korea (Man-Ki et al., 2019), the marker is likely too conserved to resolve relationships among closely related Capitella species.
We found MGSP useful overall, but additional studies focused on this feature are needed.For example, C. gomexa sp.n. exhibited sexual and geographical variation (Taxa Description, Appendix S1) and the degree of staining varies among individuals.Without the complementary COI analysis, it is not unreasonable that this species may be split if only considering the morphological data.We also found that of the four specimens of C. nonatoi checked for MGSP (FMNH 16086-16,087, FMNH 16100-16,101), none developed the staining pattern described for this species (Figure S3) (Silva et al., 2017), indicating geographical variation.Continued documentation of MGSP, histology work in particular, may reveal more details about what causes these patterns, leading to more taxonomic utility.
In conclusion, at least five undescribed species of Capitella were present in our samples as revealed by COI analysis.Two of them, C. gomexa sp.n. and C. maculosa sp.n., have documented differences in larval development.C. gomexa sp.n. is widespread throughout the GoM, ranging to at least the southern Atlantic coast of Florida, while the range of C. maculosa sp.n. still remains to be determined.Our phylogenetic tree supports a single origin of acicular spines.Our analyses included our own sequences in combination with publicly available sequences from multiple locations worldwide.Many parts of the world still remain unstudied with regard to Capitella diversity.Increased sampling efforts would lead to a more global picture of Capitella phylogeny and biogeography.Finally, multiple lines of evidence are useful when delimiting Capitella species and should be used when possible.
Acicular spines are found in C. gomexa sp.n., C. mandingensis sp.n., and C. neoaciculata (Figure 2, shaded boxes).Capitella gomexa sp.n. and C. mandingensis sp.n. include specimens with and without acicular spines.The description of Capitella neoaciculata indicates all specimens have acicular spines.

T A B L E 2
All clade names, geographical locations and GenBank accession numbers for samples downloaded from GenBank.found in the MGSP of the species description (Taxa Description, Appendix S1).Capitella gomexa sp.n. and C. mandingensis sp.n. both have acicular spines and discussion of this and other species with acicular spines can be found in the Remarks of the C. gomexa sp.n. description (Taxa Description, Appendix S1).Three of the species occur sympatrically (C.gomexa sp.n., C. maculosa sp.n., and C. loucindae sp.n.) and differences between them can be found in the Remarks of the C. maculosa sp.n. and C. loucindae sp.n. descriptions (Taxa Description, Appendix S1).Capitella gomexa sp.n. also occurs sympatrically with C. tampe sp.n. and discussion of this can be found in the Remarks of the C. tampe sp.n. description (Taxa Description, Appendix S1).

F
Maximum likelihood phylogenetic tree based on COI sequences.Branch labels are bootstrap support values|posterior probabilities.Known details of each clade's geographical origin and species name are listed.Novel species are bolded.The clades with species where acicular spines have been documented are indicated shading.