Phylogenomics resolves ambiguous relationships within Aciculata (Errantia, Annelida)

Aciculata (Eunicida + Phyllodocida) is among the largest clades of annelids, comprising almost half of the known diversity of all marine annelids. Despite the group ’ s large size and biological importance, most phylogenomic studies on Annelida to date have had a limited sampling of this clade. The phylogenetic placement of many clades within Phyllodocida in particular has remained poorly understood. To resolve the relationships within Aciculata we conducted a large-scale phylogenomic analysis based on 24 transcriptomes (13 new), chosen to represent many family-ranked taxa that have never been included in a broad phylogenomic study. Our sampling also includes several enigmatic taxa with challenging phylogenetic placement, such as Histriobdella , Struwela , Lacydonia , Pilargis and the holopelagic worms Lopadorrhynchus , Travisiopsis and Tomopteris . Our robust phylogeny allows us to name and place some of these problematic clades and has significant implications on the systematics of the group. Within Eunicida we reinstate the names Eunicoidea and Oenonoidea. Within Phyllodocida we delineate Phyllodociformia, Glyceriformia, Nereidiformia, Nephtyiformia and Aphroditiformia. Phyllodoci- formia now includes: Lacydonia , Typhloscolecidae, Lopadorrhynchidae and Phyllodocidae. Nephtyiformia includes Nephtyidae and Pilargidae. We also broaden the delineation of Glyceriformia to include Sphaerodoridae, Tomopteridae and Glyceroidea (Glyceridae + Goniadidae). Furthermore, our study demonstrates and explores how conflicting, yet highly supported topologies can result from confounding signals in gene trees. in this study. A Amphiduros pacificus (Hesionidae) B Chrysopetalum occidentale (Chrysopetalidae) , C Pholoe baltica (Sigalionidae), D Tomopteris sp. (Tomopteridae), E Histriobdella sp. (Histriobdellidae), F Sphaerodorum gracilis (Sphaerodoridae), G Nephtys hombergii (Nephtyidae), H (Typhloscolecidae), (Microphthalmidae), iocasica (Aphroditidae), Lacydonia Lopadorrhynchus (Lopadorrhynchidae).


Introduction
With about 5900 named species Errantia contains almost half of the diversity of marine annelids (Pamungkas et al., 2019;Rouse et al., 2022). This includes some of the most iconic and well-known polychaete species such as the emerging model organism Platynereis dumerilii (Nereididae), the magnificent sand-striker worm, Eunice cf. aphroditois (Eunicidae), the furry sea mouse, Aphrodita (Aphroditidae), and the venomous "blood worms'' Glyceridae. However, despite the diversity and ecological importance of the group, most broad phylogenomic studies on annelids have had only limited sampling within Errantia (Andrade et al., 2015;Helm et al., 2018;Martín-Durán et al., 2021;Struck et al., 2015;Weigert et al., 2014). With the broad pattern of annelid relationships now becoming somewhat settled, more recently phylotranscriptomic datasets have also been employed to resolve phylogenies within subclades of Annelida and it is evident that some relationships cannot be confidently resolved by small numbers of loci Tilic et al., 2020).
Errantia is an old grouping that was originally erected by Audouin and Milne-Edwards (1832) and the membership included Amphinomida, Eunicida and Phyllodocida. However, both Errantia and the complementary Sedentaria were largely abandoned in the later 20th century (Dales, 1962;Fauchald, 1977). The name was recently resurrected by Struck et al. (2011) and currently consists of three main clades: Eunicida, Phyllodocida and Protodriliformia. The name Aciculata was erected by Rouse and Fauchald (1997), and it was essentially synonymous with the older taxon Errantia. However, with the placement of Protodriliformia within Errantia (Andrade et al., 2015;Struck et al., 2015), the clade Eunicida + Phyllodocida needed a name and Andrade et al. (2015) reinstated the name Aciculata for this grouping. The focus of this study is on Aciculata as delineated in Andrade et al. (2015).
For this study we have analyzed twenty-four Aciculata transcriptomes, including thirteen that are new. The main aim of our study was to resolve the phylogeny of Phyllodocida, focusing on taxa that have been hard to place and often referred to as incertae sedis. We present a well-supported phylogeny of Aciculata, that includes representatives of seven family-ranked clades that have never been included in a broad phylogenomic analysis. These encompass Hesionidae (Fig. 1A), Chrysopetalidae (Fig. 1B), Sphaerodoridae (Fig. 1F), Aphroditidae (Fig. 1J), Pilargidae, Lacydonia (Fig. 1K), the enigmatic Struwela (Fig. 1I), the holopelagic Lopadorrhynchidae (Fig. 1L) and Typhloscolecidae (Fig. 1H) and from Eunicida the tiny "Charlie-Chaplin-worms" Histriobdellidae ( Fig. 1E) that live as commensals with crustaceans. Our results have significant implications for the systematics of the group, allowing us to name and place some of these problematic clades that no longer need to be referred to as incertae sedis. Furthermore, we also show how conflicting, yet highly supported topologies can result from confounding signals in gene trees and discuss the challenges this creates in resolving phylogenetic uncertainties within Annelida.

Taxon sampling and transcriptome sequencing
Twelve species of Phyllodocida and Histriobdella sp. were sampled for transcriptome sequencing (Fig. 1). In addition to these, eleven previously published transcriptomes were included in the analyses. Specimen details, voucher information and accession numbers for the sequence data are summarized in Table 1. COI Barcode sequences from the assembled transcriptomes were uploaded to GenBank (Table S1) and voucher specimens are deposited to the Benthic Invertebrate Collection of Scripps Institution of Oceanography (SIO-BIC), to facilitate future identification of taxa sequenced as part of this study.
Tissue samples were preserved in RNAlater quickly after the animals were collected and stored at − 80 • C until RNA extraction. For larger specimens, only the anterior region of the animals was sampled, for smaller specimens the whole individual was used for RNA extraction. For Histriobdella sp. multiple individuals, collected from a single host, were pooled during extraction. RNA extractions were performed from Trizol, using Direct-zol RNA Miniprep Kit with an in-column DNase treatment (Zymo Research). mRNA was isolated with Dynabeads mRNA Direct Micro Kit (Invitrogen).
RNA concentration was estimated using Qubit RNA broad range assay kit, and quality was assessed using RNA ScreenTape with an Agilent 4200 TapeStation on an Agilent Bioanalyzer 2100. Values were used to customize library preparation protocols following manufacturer's instructions. Library preparation was performed with the KAPA-Stranded RNA-Seq kit, targeting an insert size in the range of 200-300 bp. Quality, concentration and molecular weight distribution of libraries were assessed using a DNA ScreenTape, a Bioanalyzer 2100. Libraries were sequenced in multiplexed pair-end runs using 150 bp paired end Illumina HiSeq 4000, with 8 libraries per lane, resulting in an average sequencing depth of 48 million reads (range: 16 million -108 million). To minimize read crossover, we used 10 bp sequence tags designed to be robust to indel and substitution errors (Faircloth and Glenn, 2012). All sequence data have been deposited in NCBI's sequence read archive (SRA) ( Table 1) with Bioproject accession number PRJNA743301.

Transcriptome Assembly and phylogenetic analyses
Sequence adapters and low-quality regions were removed from the raw reads of each species using Trimmomatic v.0.36 (Bolger et al., 2014) with default parameters. All transcriptomes were de novo assembled and orthologous gene sequences were identified using the automated pipeline Agalma v.1.0.1 (Dunn et al., 2013;Guang et al., 2017). In brief, Agalma assembles transcripts with Trinity (Grabherr et al., 2011), maps reads with Bowtie2 (Langmead and Salzberg, 2012) and identifies orthologs using the topology of gene phylogenies and DendroPy (Sukumaran and Holder, 2010). Assembly statistics and the number of genes identified by Agalma are summarized in Table S2.
To decide on the placement of the root, we conducted a preliminary analysis with an extended taxon sampling including 19 additional taxa spanning the diversity of annelids. We conducted a concatenated ML analysis with IQTREE with the same parameters as outlined below on a supermatrix with 70% occupancy (817 genes).
In addition to ML analyses on the concatenated supermatrix, coalescent-based species tree was inferred using ASTRAL-III v.5.7.3 (Zhang et al., 2018b). Individual gene trees were estimated with IQTREE under the best-fit model for each alignment. Gene alignments were tested for model violation using the -symtest option in IQTREE (Naser-Khdour et al., 2019). Out of 9657 orthologous gene alignments identified by Agalma, 518 rejected SRH (stationary, reversible and homogeneous) assumptions and were excluded from the ASTRAL analysis. All the remaining 9139 gene trees were included as ASTRAL benefits from more data Nute et al., 2018). Poorly supported branches with less than 10% bootstrap support were contracted in the input trees for ASTRAL as this can improve species tree inference (Zhang et al., 2018a). Input trees for ASTRAL were also filtered using TreeShrink (Mai and Mirarab, 2018), which implements an algorithm to identify and remove unexpectedly long branches caused by e.g contamination, mistaken orthology, and misalignment. The ASTRAL tree inferred from this curated dataset revealed no effect on the topology and only some very minor differences in branch lengths and support values when compared to the uncurated dataset.
To assess if the phylogenetic signal differed between slow-and fastevolving proteins, two submatrices were constructed from the 80% occupancy matrix with 854 genes. Individual gene trees were analyzed to rank each gene by evolutionary rate (total tree length/n terminals). The ranked genes were divided into the fast-evolving 50% and the slowevolving 50%. The alignments were then concatenated and used to construct a maximum likelihood tree using IQTREE under the best-fit model chosen according to BIC (LG + F + R5). Support was assessed with 10,000 bootstrap pseudo-replicates.

Topology testing
To explore the strength of support for the five nodes that differed between the ASTRAL and ML trees, we used two approaches. To assess how much less likely alternative topologies were compared to the best ML tree from the concatenated dataset, we used topology tests as implemented in IQ-TREE using expected likelihood weights (Strimmer and Rambaut, 2002) and with the approximately unbiased (AU) test (Shimodaira, 2002). All tests performed 10,000 resamplings using the RELL method (Kishino et al., 1990). In each assessment, we changed only one branch to assess whether the alternative topology was statistically less supported than the best ML tree.
For the tests in a coalescent-based framework, we used ASTRAL as above but constrained the analysis to either the ML topology or the ASTRAL topology (Rabiee and Mirarab, 2020). This way, we could assess whether there is support for the alternative hypotheses proposed by the ML trees that were not recovered in the main ASTRAL analysis. We tested how support (measured as local posterior probability PP (Sayyari and Mirarab, 2016)) for the two topologies changed with additional data. A true relationship is expected to gain support with additional gene trees, while a spuriously recovered signal is more likely to remain supported with low support values. We randomly split the 9657 gene trees into subsets of 250, 500, 1000, to 9000 gene trees, and replicated each subset 20 times. From each subset, we built a species tree using ASTRAL v.5.6.9, while constraining to the topology of the ML or ASTRAL analysis. For the resulting species tree, we recorded PP support for each of the five nodes that differed between ASTRAL and ML trees.

Data analyses and matrix Assembly
Assembly statistics and values to assess the quality of each transcriptome (number of assembled contigs, mean contig length, N50, BUSCO values, and number of detected loci) are summarized in Table S2. The smallest number of reads in the total data set was the 16 million reads sequenced for Lacydonia sp. (assembled into 91,326 contigs), whereas the largest one was sequenced for Histriobdella sp. with 108 million reads (assembled into 45,904 contigs).

Phylogeny reconstruction based on supermatrices
The unpartitioned RAxML analysis, as well as the partitioned IQTREE analysis using integrated model selection and the separate analyses of slow-evolving and fast-evolving genes recovered the same topology (Fig. 2). All nodes had > 94% bootstrap support.
The tree in Fig. 2 was rooted with Eunicida forming a clade including Histriobdella. This decision was based on the tree topology recovered in the extended analyses with broader taxon sampling of annelids (Fig. S1-S4), which matched the tree shown in Fig. 2, apart from the placement of Histriobdella. In the expanded analyses Histriobdella grouped with a clade of Diurodrilidae, Dinophilidae and Myzostomida, or with Sabellidae which all have notably long terminal branches and may be a result of artifactual long branch attraction (see Andrade et al. 2015). The numerous morphological similarities that Histriobdellidae share with members of Eunicida make this a reasonable assumption.
Within Phyllodocida several clades were recovered that we apply names to, based on support and congruence with the ASTRAL result: Phyllodociformia, consisting here of Lacydonia sp., Phyllodoce medipapillata Moore, 1909 and two holopelagic taxa; Travisiopsis sp. and Lopadorrhynchus sp.
Nephtyiformia is a new name we use to include Pilargidae and Nephtyidae, based on Pilargis verrucosa Saint-Joseph, 1899 and Nephtys sp. grouping together with full support.
Hesionoidea. We could not consistently recover a clade that could be referred to as Nereidiformia. In the ML analysis Platynereis dumerilii (Audouin & Milne Edwards, 1833) grouped together with Struwela camposi Salazar-Vallejo et al. 2019 and this clade was the sister to Glyceriformia + Phyllodociformia. We do name the well-supported clade comprising of Chrysopetalidae and Hesionidae as Hesionoidea, which has an ending signifying superfamily rank, and the authority is Grube (1850).

Coalescent-based species tree analysis with ASTRAL
The ASTRAL topology remained stable when potentially extraneous sequences were filtered with TreeShrink. Most of the nodes had full support (Fig. 2). Though the topologies recovered from concatenation based supermatrix analyses and the coalescent-based species tree from ASTRAL were broadly similar, there were few significant differences indicating a conflicting phylogenetic signal in the gene trees.
In the ASTRAL tree Histriobdella and Ophryotrocha did not group together. Histriobdella was the sister to the remaining Eunicida, which formed a clade that had an ASTRAL local posterior probability of 0.89. Within Glyceriformia the position of Tomopteris sp. (Fig. 1D) shifted Fig. 2. Phylogeny of Aciculata based on transcriptomic data. The tree on the left shows the topology recovered from maximum likelihood inference on the supermatrix with 80% taxon occupancy (242,492 AA sites; 854 genes). The black bars are proportional to the number of genes present for each terminal. The same topology was also found when slow-evolving and fast-evolving genes were analyzed separately. The tree on the right shows the results of the ASTRAL analysis summarizing all 9139 gene trees. Bootstrap support values and ASTRAL local posterior probabilities are only shown on the tree if a node was not recovered with full support. Note that the delineation of Glyceriformia is expanded to include Tomopteridae and Sphaerodoridae, former Glyceriformia (Glyceridae + Goniadidae) therefore becomes Glyceroidea. We also bring back the names Eunicoidea (Eunicidae + Onuphidae) and Oenonoidea (Lumbrineridae + Oenonidae) both originally assigned by Orensanz (1990). Note the dashed line of Histriobdella indicating the conflicting topologies shown in Figs. S1-4, likely a result of long-branch artefacts.
In contrast to the supermatrix analysis the ASTRAL result showed Platynereis forming a clade with the hesionid Amphiduros pacificus Hartman, 1961 and the chrysopetalid Chrysopetalum occidentale Johnson, 1897, which could be regarded as the taxon Nereidiformia (ASTRAL support 0.62), but further assessment of this is required given the conflict with the ML result (Fig. 2). Instead of grouping with Platynereis, the microphthalmid Struwela camposi was the sister group to a Glyceriformia + Phyllodociformia clade. Lastly, Syllis gracilis became the sister of all Phyllodocida except Aphroditiformia. The remaining topology was identical to that recovered by the maximum-likelihood analyses.

Topology testing
Two out of the five nodes that differed between the ML and the ASTRAL trees were found to result in significantly worse trees using the AU test (p-value less than 0.05). These involved the placement of Nereididae and Microphthalmidae (Fig. 3D, E). However, alternative topologies for three nodes, involving Syllidae, Sphaerodoridae and Histriobdellidae ( Fig. 3A-C) were not significantly different and so cannot be ruled out based on the data. The ASTRAL subsetting analyses showed strong support in favor of the ASTRAL topology for four out the five nodes (Fig. 3). It is notable that large amounts of data were needed to gain high PP support (>0.90) for most of these nodes. The position of Syllidae as the sister of Aphroditiformia was strongly supported when more than ~ 5000 gene trees were analyzed, while the ML hypothesis was only poorly supported (Fig. 3A). On the other hand, the AU test could not reject the alternative topology, with Syllidae as sister to the remaining Phyllodocida, over the ML topology (p-value 0.137). The ASTRAL position for Sphaerodoridae and for Histriobdellidae was strongly favored and support levels increased with additional data, while the support for the alternative topology decreased (Fig. 3B-C). However, the trajectory of the curves suggest that more than 9000 loci could be needed to obtain full support on those nodes. Both topological changes did not result in statistically different trees according to the AU test. The position of Struwela (Microphthalmidae) as the sister group to Nereididae as in the ML topology was strongly supported but a lower-level support for the alternative topology as the sister to Glyceriformia + Phyllodociformia remained relatively constant (Fig. 3D). This contrasted with the results from the AU test, which indicated that the alternative position of Struwela led to strong statistical differences compared to the best ML tree (p-value 2.15 e -33 ). The position of Nereididae was the only node where both alternative topologies had consistent and equal support and no obvious effect of increasing the number of analyzed gene trees existed (Fig. 3E). Interestingly, the AU test did detect statistically significant differences between the two topologies (p-value 0.0055).

Discussion
We presented a phylogenetic analyses of Aciculata based on transcriptomic data from 24 species that resulted in an overall wellsupported and congruent topology. Using subsetting analyses, we show that some difficult nodes require large numbers of loci to obtain strong support, consistent with what was observed in other clades within annelids . Difficulty remains in some branches, which appear to be impacted by long branches and conflicts between analysis types . Nevertheless, the stability and support for a number of clades allows for a series of nomenclatural recommendations that are outlined below.

Relationships within Eunicida and implications on the systematics of the group
Eunicida (Dales, 1962) is a well-defined clade of annelids whose members show a notable synapomorphy in the ventral muscular proboscis with a complex jaw apparatus that is mineralized or sclerotized (Tzetlin and Purschke, 2005). These jaws can often be found in the fossil record (scolecodonts) and date the group back to the late Cambrian (Paxton, 2009  Previous phylogenetic analyses based on Sanger sequencing and combined morphological data have been largely congruent in recovering the monophyly of most these clades (Budaeva et al., 2016;Struck et al., 2015Struck et al., , 2006Tilic et al., 2016;Zanol et al., 2014). Studies that included Pettiboneia (Struck et al., 2006(Struck et al., , 2002 failed to recover a monophyletic Dorvilleidae, but this was only based on limited sequence data (16S rRNA, 18S rRNA genes). Given the long branches and low support in these analyses further investigation was warranted.
Inferred relationships within Eunicida have become more stable with the introduction of phylogenomic datasets. Our results presented herein are congruent with Struck et al. (2015) in recovering a clade consisting of Eunicidae + Onuphidae, another clade with Lumbrineridae + Oenonidae and then Dorvilleidae (possibly together with Histriobdellidae) as sister to these two clades. The sister group relationship of Eunicidae and Onuphidae is well supported both on morphology (Budaeva et al., 2016;Fauchald, 1992;Orensanz, 1990;Paxton, 2009;Zanol et al., 2014) and on sequence data (Struck et al., 2015(Struck et al., , 2006Tilic et al., 2016). In contrast, Rouse and Fauchald (1997) had recovered Eunicidae as sister to Lumbrineridae + Dorvilleidae and Onuphidae as sister to this clade. This topology, however, is no longer supported and the monophyly of Eunicidae and Onuphidae has become undisputed. Orensanz (1990) had named this clade (ranked superfamily) Eunicoidea, which we adopt here. The synapomorphies that unite Eunicoidea are: eulabidognath maxillae, five prostomial appendages, peristomial cirri, and subacicular hooks in median and posterior parapodia (Budaeva and Zanol, 2020).
The second well-supported clade within Eunicida contains Lumbrineridae and Oenonidae. In previous phylogenetic analyses based on few Sanger-sequenced markers (Struck et al., 2006;Tilic et al., 2016) Lumbrineridae was recovered as the sister to all other Eunicida, however phylogenomic studies based on transcriptome data, both in Struck et al. (2015) and in this paper, now support the sister group relationship of Lumbrineridae and Oenonidae. This was also favored by earlier morphological hypotheses and was given the name Oenonoidea by Orensanz (1990) with a superfamily ranking and we also adopt it here.
The placement of Dorvilleidae within Eunicida using molecular phylogenies has been more problematic and challenging owing to long branch artefacts (Struck et al., 2006). In Struck et al. (2006) Dorvilleidae did not form a clade. Tilic et al. (2016) recovered Oenonidae and Dorvilleidae as sister taxa, though this had very low support. Struck et al. (2015) included transcriptomic data for several Dorvilleidae terminals, which were recovered as monophyletic and as the sister to remaining Eunicida. Our analyses only included a single Dorvilleidae terminal (Ophryotrocha globopalpata), which was the sister taxon to Eunicoidea + Oenonoidea either on its own (ASTRAL) or together with Histriobdella sp. (RAxML/IQTREE). Even though the ASTRAL position was favored based on our topology testing, this placement requires further investigation, with better taxon sampling within Dorvilleidae which might help resolve the incongruencies between the two analyses.
Though we were not able to unambiguously resolve the placement of Histriobdella, our phylogenomic analysis is the first attempt to place Histriobdellidae within Eunicida using molecular data. In all ML analyses with the concatenated supermatrix Histriobdellidae was recovered as the sister taxon to Dorvilleidae. However, there was some confounding signal in the gene trees, as the ASTRAL analysis supported a different topology (Histriobdella as sister to the remaining Eunicida), which was strongly favored in subsetting analyses over the ML topology (Fig. 3C) also with full support. An affinity of Histriobdellidae with Dorvilleidae has also been suggested based on some morphological similarities of their jaws (Tzetlin et al., 2020;Tzetlin, 1980). In the extended analyses of annelids (Fig. S1-S4), Histriobdella grouped together with other long-branched taxa outside Aciculata. This indicates that confounding long-branch effects may be causing the placement of Histriobdella among taxa with long terminal branches. This matches what was observed for other annelid taxa with symbiotic or parasitic lifestyles (Andrade et al. 2015). Removing long-branched taxa was also not sufficient to place Histriobdella within Eunicida (Fig. S2, S4), indicating that there seems to be little signal in the Histriobdella data combined with many mutations that are not shared with any other Eunicida. Alas, given the relatively long branches of both Ophryotrocha globopalpata and Histriobdella, and the conflicting topologies, we cannot confidently resolve the phylogenetic position of Histriobdellidae, and this still warrants further investigation.
The placement of Hartmaniella within Eunicida still remains a mystery. Orensanz (1990) suggested that Hartmaniella has its sister group within Eunicoidea, then again Fauchald and Rouse (1997) interpreted the jaws differently and considered them to be closer to Oenonoidea. With only three named species and very limited records, and a complete lack of molecular data nothing is resolved about their placement within Eunicida.
Our phylogenomic analyses of transcriptome data has the largest taxon sampling within Phyllodocida to date, representing many of the above-mentioned major lineages and resolves, with high support, many of the relationships within this diverse clade. Our results have significant implications on the systematics of the group. We reinstate and change the delineations of some existing names and name new clades when no preexisting ones were available. All new and amended taxon names are above family rank, therefore the rules and regulations of the ICZN do not apply.
One of the largest clades within Phyllodocida is Aphroditiformia, the scale-worms, the relationships of several family-ranked taxa within this clade is addressed in recent studies by Gonzalez et al., (2018);Norlinder et al., (2012);Zhang et al., (2018b). Our dataset includes four Aphroditiformia terminals representing Aphroditidae (Laetmonice cf. iocasica), Sigalionidae (Pholoe baltica) and Polynoidae (Lepidonotopodium sp. and Harmothoe imbricata). Though limited in sampling, our topology within Aphroditiformia is congruent with previous studies (Gonzalez et al., 2018;Zhang et al., 2018b) in recovering Aphroditidae as sister to the latter two taxa. Our concatenated supermatrix analysis placed Aphroditiformia as sister to Syllidae, a placement that has also been found in phylogenomic analyses of Weigert et al. (2014) and Struck et al. (2015). In contrast to this, Syllidae was recovered as the sister to all Phyllodocida excluding Aphroditiformia in our ASTRAL analysis, and this alternative topology could not be rejected with the AU test, which indicates that future work is required to fully resolve the placement of Syllidae within Phyllodocida.
Phyllodocida excluding Syllidae and Aphroditiformia form a wellsupported clade, which we choose not to name here pending resolution of the placement of Syllidae. Nephtys sp. and Pilargis verrucosa together form the sister taxon to the rest of this clade. The position of Pilargidae has long been problematic. Several studies based on both morphology (Fitzhugh and Wolf, 1990;Glasby, 1993;Licher and Westheide, 1994;Pleijel and Dahlgren, 1998) and molecular data (Rousset et al., 2007) have often placed them within Nereidiformia (or Nereidoidea). Rouse and Fauchald (1997) placed Pilargidae as sister to Sphaerodoridae. In Struck et al.'s (2007) phylogeny of annelids based on Sanger sequenced data, Pilargidae were either in a trichotomy together with Syllidae and Nephtyidae, or as sister to the two others. The position of Pilargis verrucosa is highly supported in all analyses we performed, and we therefore use the new name Nephtyiformia for the Pilargidae + Nephtyidae clade.
As with Pilargidae, Chrysopetalidae have never been included in a broad scale molecular analysis prior to this study. Molecular phylogenies assessing the relationships within Chrysopetalidae were not able to resolve the sister taxon of the group (Aguado et al., 2013;Jimi et al., 2019;Ravara et al., 2007). Glasby (1993) and Pleijel and Dahlgren (1998), both based on morphology, identified a sister-group relationship between Chrysopetalidae and Hesionidae. All our analyses also place Chrysopetalum occidentale in a well-supported clade with Amphiduros pacificus, therefore we name the clade consisting of Hesionidae and Chrysopetalidae, Hesionoidea (Grube 1850). The position of Nereididae differed depending on the analysis we performed; the supermatrix analyses grouped Platynereis together with Struwela, whereas ASTRAL recovered Nereidiformia (Nereididae + Hesionoidea). The latter relationship was also obtained by Glasby (1993) and Pleijel and Dahlgren (1998).
Struwela camposi is one of only two known Struwela species and was described only recently (Salazar-Vallejo et al., 2019). These animals are morphologically somewhat aberrant, with large hooks on the second segment, and both live in association with sand dollars. Salazar-Vallejo et al. (2019) have redelineated Microphthalmidae (and raised it to family rank) to include Struwela as well as Microphthalmus, Hesionides, Uncopodarke, Westheideius, Hesionella and Fridericiella. Furthermore, molecular sequence data also support the close affinity of Struwela and Microphthalmus (Rouse, unpubl.). Though we cannot resolve the placement of Struwela with certainty in our analyses, the transcriptome data we publish herein for this aberrant genus, makes it clear that it and Microphthalmidae are not closely related to Hesionidae and so not likely to be a subgroup of this clade as seen also in Salazar-Vallejo et al. (2019).
Glyceridae (Glycera dibranchiata), Sphaerodoridae (Sphaerodorum gracilis) and Tomopteridae (Tomopteris sp.) always formed a wellsupported clade in our analyses. To avoid introducing a new name, we here expand the delineation of Glyceriformia to include both Sphaerodoridae and Tomopteridae and propose the superfamily-ranked taxon name Glyceroidea for Goniadidae + Glyceridae.
The last clade we recovered within Phyllodocida was Phyllodociformia, which was well-supported in all our analyses. We included Lacydonia, as well as the two holopelagic taxa; Typhloscolecidae (Travisiopsis sp.) and Lopadorrhynchidae (Lopadorrhynchus sp.) always grouped together with the Phyllodocidae terminal (Phyllodoce medipapillata). The placement of Lacydonia within Phyllodociformia is in accordance with previous morphology-based hypotheses, where they have been inferred as the sister group to Phyllodocidae (Pleijel and Dahlgren, 1998;Rouse and Fauchald, 1997). There has been some confusion regarding the placement of the holopelagic clades Typhloscolecidae and Lopadorrhynchidae resulting from analyses published using chimeric sequence data (Struck and Halanych, 2010), which was demonstrated and further discussed in Nygren and Pleijel (2011). The topology we recover here using a much larger and comprehensive phylogenomic dataset also corroborates the results of Nygren and Pleijel (2011).
Though we were able to recover highly supported relationships for most of the clades within Phyllodocida, few taxa where no transcriptome data was available remain as Phyllodocida incertae sedis. These are Paralacydonia, and two holopelagic taxa, Pontodora and Yndolaciidae.

Aciculata incertae sedis that still need assessment
Based on morphology several potential members of Aciculata, or at least Errantia, can be identified, which need to be assessed in future studies. This includes Spinther Johnston, 1845, Myzostomida Graff, 1877, Aberranta Wolf, 1987 and Nerillidae Levinsen, 1883 have always been hard to place within Annelida. No transcriptome data is available for Aberranta and previous phylotranscriptomic studies have been unsuccessful in confidently placing Myzostomida, Nerillidae and Spinther; that all have extremely long branches (Andrade et al., 2015;Struck et al., 2015;Weigert et al., 2014). This also appears to be the case for Histriobdellidae (Fig. S2). While our data shows that large numbers of loci can confidently resolve most relationships within Aciculata, the long branch of Histriobdella remains a problem. To break the long branch of Histriobdella and other Aciculata incertae sedis, future analyses with increased and broader taxon sampling could be useful, in addition to the analysis of non-coding data that could be less prone to confounding effects compared to protein coding data Reddy et al., 2017).

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.