Systematics of ‘lithistid’ tetractinellid demosponges from the Tropical Western Atlantic—implications for phylodiversity and bathymetric distribution

Background Among all present demosponges, lithistids represent a polyphyletic group with exceptionally well-preserved fossils dating back to the Cambrian. Knowledge of their recent diversity, particularly in the Tropical Western Atlantic Ocean (TWA) where they are common in deep waters, is scarce making any comparison between present and past major ‘lithistid’ faunas difficult. In addition, the lack of sufficient molecular and morphological data hamper any predictions on phylogenetic relationships or phylodiversity from this region. The Harbor Branch Oceanographic Institute (HBOI, Fort Pierce, Florida) holds the largest collection of TWA lithistid sponges worldwide, however, the majority remain to be taxonomically identified and revised. Principal Findings In this study we provide sequences of 249 lithistid demosponges using two independent molecular markers (28S rDNA (C1-D2) and cox1 mtDNA). In addition, a morphological documentation of 70 lithistid specimens is provided in the database of the Sponge Barcoding Project (SBP). This integrated dataset represents the largest and most comprehensive of the TWA lithistids to date. The phylogenetic diversity of ‘lithistid’ demosponges in the Bahamas and Jamaica are high in comparison to other TWA regions; Theonellidae and Corallistidae dominate the fauna, while Neopeltidae and Macandrewiidae are rare. A proposed tetractinellid suborder, one undescribed genus and several undescribed species are recognized and the Pacific ‘lithistid’ genera, Herengeria and Awhiowhio, are reported from the TWA for the first time. The higher-taxa relationships of desma-bearing tetractinellids are discussed and topics for revision suggested. Conclusion This first integrative approach of TWA ‘lithistid’ demosponges contributes to a better understanding of their phylogenetic affinities, diversity and bathymetric distribution patterns within the TWA. As in the Pacific, the TWA ‘lithistid’ demosponges dominate deep-water habitats. Deeper taxonomic investigations will undoubtedly contribute to a better comparison between present major ‘lithistid’ faunas and their fossil record in the Mesozoic.


Figure 1 Distribution map of investigated HBOI and other desma-bearing tetractinellids and Vetulin
. These expeditions aimed to conduct a biodiversity inventory and collect samples for biomedical research focused particularly on sponges, octocorals and algae. Various habitats from the fore reef slopes and escarpments to the deep shelf slopes were sampled using either a claw, suction tube or scoop in depths from 0-1000 m. Sponge samples from this collection, were pre-identified by S.P. and M.K., and frozen and/or stored in 70% ethanol. For comparison, additional material from the Southwest Pacific (New Caledonia and New Zealand), and Indo-Pacific region, in the National Institute of Water and Atmospheric Research (NIWA) collection in Auckland and its invertebrate collection (NIWA Invertebrate Collection, NIC)  Undetermined samples from the TWA (all HBOI subsamples) were identified to the genus level according to their phylogenetic position relative to known species. Based on this, we selected 71 samples with distinct genotypes for a deeper morphological investigation. For those taxa we examined collection pictures, prepared thick sections and spicule and skeleton stubs for Scanning Electron Microscopy (SEM). We used the methodology outlined in Pisera & Pomponi (2015) to illustrate and evaluate morphological characters. Based on this, 249 specimens were identified to genus and/or species level. Morphological documentation for the 71 representative specimens are provided in the SBP (http://www.spongebarcoding.org/). SEM stubs and spicule slides including thick sections are deposited at the Bavarian State Collection for Paleontology and Geology (BSPG) Munich, Germany.

Molecular investigations
Genomic DNA was isolated from small pieces of sponge tissue preserved in 70% ethanol using a modified protocol of the DNeasy (Qiagen) Blood and Tissue Kit, which included an additional centrifugation step just before transferring the lysate to the spin column. A Nano-Drop 1000 Spectrophotometer (Thermo Scientific) was used to quantify the isolated genomic DNA. Amplification of a fragment of the mitochondrial cytochrome c oxidase subunit 1 (cox1, partial ≈ 659 bp) was performed using the primers dgLCO1490 and dgHCO2198 (Meyer, Geller & Paulay, 2005). Additionally, a fragment of an unlinked nuclear ribosomal gene (28S; partition C1-D2, 768-832 bp) was amplified using the forward C1 ASTR (Cárdenas et al., 2010) and the reverse universal D2 (Lê, Lecointre & Perasso, 1993) primers. Both amplifications follow the PCR protocol and settings outlined in Schuster et al. (2015). Amplification success was checked on a 1.5% agarose gel. For the majority of the 28S fragments we observed an additional non-specific shorter band at ≈ 650 bp, which was subsequently identified as originating from a bacterial template. Therefore, separation of double bands and PCR clean-up was performed using a modified freeze-squeeze method (Tautz & Renz, 1983), as described in Schuster et al. (2015). For sequencing of the 28S fragment, 6 µl of the remaining supernatant from the clean-up was used with the PCR primers and BigDye Terminator v3.1 (Applied Biosystems, Forster City, CA, USA) chemicals. For sequencing of cox1 we used a 1:10 dilution of the PCR products together with the PCR primers and BigDye Terminator v3.1 chemicals. Sequencing was carried out on an ABI 3730 Genetic Analyzer at the Sequencing Service of the Department of Biology (LMU München). Sponge origin of novel sequences were tested by BLAST searches against NCBI GenBank (https://blast.ncbi.nlm.nih.gov/Blast.cgi). Raw trace files were post-processed by base-calling using CodonCode Aligner v.3.7.1.1 (CodonCode Corporation). Geneious v.8. 1.8 (http://www.geneious.com, Kearse et al., 2012) was used for the assembly of forward and reverse reads. Sequences will be deposited at the European Nucleotide Archive (ENA) and the Sponge Barcoding Database (SBD) of the SBP under accession numbers SBD#1794 to SBD#2108.

Phylogenetic reconstructions
Alignments were generated separately for cox1 and 28S using MAFFT v.7 under the L-INS-I algorithm (Katoh & Standley, 2013) because of heterogeneous taxon sampling and moderate sequencing success of cox1. Saturation of both markers was evaluated using Xia's test (Xia et al., 2003) as implemented in DAMBE v5.1.5 (Xia, 2013) which compares an estimated substitution saturation index (Iss) to a critical substitution saturation index (Iss.c). For the cox1 dataset, sequences of Halichondria panicea (Pallas, 1766) (subclass: Heteroscleromorpha Cárdenas, Pérez & Boury-Esnault, 2012, order Suberitida Schmidt, 1870 and Aplysina aerophoba (Nardo, 1833) (subclass: Verongimorpha Erpenbeck et al., 2012, order Verongiida Bergquist, 1978 were chosen as outgroups. For the 28S dataset sequences of the order Sphaerocladina were chosen as outgroup. All outgroups have been used in earlier phylogenetic studies on tetractinellids (see e.g., Schuster et al., 2015;Kelly & Cárdenas, 2016). The final cox1 alignment comprised 307 sequences of which 122 are newly generated sequences for this study. The alignment was 635 bp long, of which 295 bp were constant, 40 bp were parsimony uninformative and 300 bp were parsimony informative. The final 28S alignment comprised 474 sequences of which 305 are newly generated sequences for this study. In total this alignment was 905 bp long, of which 325 bp were constant, 66 bp parsimony uninformative and 514 bp parsimony informative. Both alignments from this study are freely available at OpenDataLMU 10.5282/ubm/data.221. Phylogenetic tree reconstructions for both datasets were performed on a parallel version of MrBayes v3.2.4 (Ronquist et al., 2012) on a Linux cluster. The most generalized GTR+G+I evolutionary model, indicated as the most suitable by jModelTest v.2.1.7 (Darriba et al., 2012), was used. Analyses were run in two concurrent runs of four Metropolis-coupled Markov-chains (MCMCMC) for 100,000,000 generations and stopped when the average standard deviation of split frequencies dropped below 0.01. The first 20% of the sampled trees were removed as Burn-in from further analyses.

Inclusive molecular phylodiversity and abundance analyses
The Inclusive Phylogenetic Diversity (PD I ) is the sum of all branch lengths of a gene tree connecting a set of taxa from the root of the tree to the tips of all phylogenetic branches spanned by this set of taxa (see e.g., Lewis & Lewis, 2005). To evaluate the PD I , a Maximum Likelihood (ML) tree was first calculated from the most comprehensive dataset (28S, C1-D2 partition) using RAxML 7.2.8 (Stamatakis, 2014). The GTRGAMMA nucleotide evolutionary model selected by jModelTest v.2.1.7 (Darriba et al., 2012) was taken with 1000 fast pseudo-replicated bootstraps. The resulting tree topology was used to calculate the PD I for several areas in the TWA using a modified python script from Vargas et al. (2015). All non-TWA genera and all TWA genera less than five were excluded from this analysis. In total, the PD I of Bonaire, Curaçao, Florida, Honduras, Jamaica, Puerto Rico and Turks and Caicos was calculated. In order to compensate for different sampling efforts across the seven regions, rarefaction curves (Sanders, 1968) were used for each location. The rarefaction curves were generated in RStudio (R Studio Team, 2014). Both scripts are available at https://bitbucket.org/molpalmuc/.
The family Theonellidae consists of the genera Discodermia Du Bocage, 1869, Manihinea Pulitzer-Finali, 1993, Racodiscula Zittel, 1878, Siliquariaspongia Hoshino, 1981 and Theonella Gray, 1868. Theonellidae possesses tetraclone desmas and phyllotriaenes to discotriaenes as characteristic megascleres. Typical microscleres are acanthorhabds, spirasters and amphiasters (Pisera & Lévi, 2002a). Until now, only Theonella and Discodermia species as well as one Manihinea sp. were sequenced in different phylogenetic studies using 18S, 28S and cox1 (see e.g., Redmond et al., 2013;Hall, Ekins & Hooper, 2014;Schuster et al., 2015). By providing sequences for all known genera, our 28S phylogeny (Fig.  3) recovers Theonellidae as monophyletic (PP = 1.0), thus conclusively support earlier findings of Schuster et al. (2015), while the cox1 phylogeny (Fig. 13) lacks support in this respect. The 28S phylogeny indicates the monophyly of the genera Discodermia, Manihinea, Racodiscula and a potential new taxon, here denoted as Theonellidae gen. sp., a potential new genus mainly distinct by the layered network of tetraclone desmas with smooth rays and strongly tuberculated tips and the less abundant microscleres on the ectosome (SBD#2102-2106). The sister relationship of Manihinea conferta to Theonella sp. is highly supported (PP = 0.99) by 28S (Fig. 3), whereas it is not supported by cox1 (Fig. 13). A close relationship of Theonella and Manihinea was observed in an earlier study by Redmond et al. (2013) using a nearly complete 18S gene fragment, but unsupported. The genus Discodermia is sister to a clade consisting of Manihinea +Theonella +Siliquariaspongia, which is sister to Racodiscula +Theonellidae gen. sp. The genus Racodiscula is highly supported (PP=1.0) as sister to Theonellidae gen. sp. Although the outer morphology of Theonellidae gen. sp. (SBD#2106 A-D) is very similar to that of Racodiscula, it differs in spicule composition, desma and skeleton structure: the usually abundant spinose microacanthorhabds, covering the surface of Racodiscula species (SBD#2065) building a dense crust on the surface, are rarer or even absent in Theonellidae gen. sp. Instead of microacanthorabds, phyllo-to discotriaenes are the main components of the dense surface crust. In addition, Theonellidae gen. sp. possesses desmas with smooth rays and strongly tuberculated tips (SBD#2105 and SBD#2106) building a layered network (SBD#2102), which clearly differs from Racodiscula (Schuster et al., 2018).
The key size and shape differences between the 'tetralophs' of T. mirabilis and other Placinolopha species were noted by Muricy & Diaz (2002), who suggested that the species mirabilis had a more likely affinity with species in family Theonellidae. Sequences of 28S (Fig. 3) unite specimens identified as Theonella mirabilis in a single clade with a specimen identified as T. conica (Kieschnick, 1896) which also has tetraloph-like desmas, suggesting that species with non-articulated 'tetraloph' desmas may be monophyletic and separate from other Theonella spp. However, cox1 sequences (Fig. 13) separate T. mirabilis into two groups, nesting them within diverse species of Theonella. Theonella mirabilis is very similar in spicule complement to the type species of the genus Siliquariaspongia, S. japonica Hoshino, 1981 (Family Theonellidae), although the latter lacks the strongyles and possesses frilly discotrianes, the latter occasionally recorded in T. mirabilis. Our phylogenies clearly place all of the sequenced Theonella mirabilis species within the Theonella +Manihinea clade (Figs. 3 and 13), confirming that this species belongs to the family Theonellidae. This result is supported by the discovery of potent new depsipeptides mirabamides A-D, that inhibit HIV-1 infection, adding to a small class first exemplified by the papuamides from various Theonella spp. (Plaza et al., 2007).
The family Macandrewiidae is monogeneric with currently seven valid species (Van Soest et al., 2018a). Until know, only Macandrewia rigida Lévi & Lévi, 1989 from the Solomon Islands has been sequenced (28S C1-D2 region, LN624160, G317931) (Schuster et al., 2015). The present study includes a further sequence of an undescribed Macandrewia sp. from the Bahamas (909 m depth), which clearly differs from M. rigida (Fig. 4). Morphological differences in desmas (SBD#2004) corroborate the genetic difference to M. rigida and provide further evidence of a possible new species, which would be the first record in the TWA. Nevertheless, further morphological observations and comparison with the type material of M. rigida as well as its sequences are needed to conclusively describe and distinguish this potential new species. Both Macandrewia species group within the Geodiidae, close to the Erylinae Sollas, 1888, within a clade of non-desma bearing astrophorins (Calthropella Sollas, 1888, Caminella Von Lendenfeld, 1894 (Fig. 4). This relationship is currently not supported by morphology  and in any case suggests a distinct evolutionary history of Macandrewia to other 'lithistid' families Corallistidae and Neopeltidae (Schuster et al., 2015), where Macandrewia was previously allocated (Kelly, 2000;Pomponi et al., 2001).
A clade of six as yet unidentified specimens from the HBOI collection (SBD#1814) is sister to the monophyletic Neophrissospongia (PP = 1.0, Fig. 6). We assume that this clade consists of species from the as yet unsequenced genus Awhiowhio Kelly, 2007 from the Pacific based on morphological evidence. These show similar mega-and microsclere types to Awhiowhio such as dicranoclone desmas and smooth dichotriaenes in Awhiowhio sp. 1 from the Bahamas (SBD#1815), most similar to the Awhiowhio osheai from New Zealand (Kelly, 2007), but slightly different in terms of desma ornamentation. Streptaster microscleres and acanthose microrhabds in Awhiowhio sp. 1 (SBD#1814, 1815) differ from those in Awhiowhio osheai in sizes and shapes (SBD#1814). The cox1 phylogeny (Fig. 9) indicates the sister group relationship of Awhiowhio osheai Kelly, 2007 to Neophrissospongia. A close relationship of Awhiowhio to Herengeria as suggested by Kelly (2007) based on morphological features is not supported by any of our phylogenies. Instead, both markers independently suggest a close relationship (strongly supported by PP = 1.0) to Neophrissospongia. The genus Pleroma Sollas, 1888 (family Pleromidae) is recovered as paraphyletic in both phylogenies (Figs. 5,6,9 and 13). Pleroma menoui (Sollas, 1888) is distant to other Pleroma spp. (including the type species P. turbinatum) in a close relationship to Corallistidae (Figs. 5 and 9). cox1, 18S and 28S markers showed the separation of all rhizomorine-bearing sponges from Spirophorina (=Tetillidae). The present enlarged dataset corroborates again the absence of desma-bearing sponges in Spirophorina and their grouping in a well supported clade along with the Stupendidae and the Thrombidae (Fig. 2). In order to establish this clade as a new taxa, we await further molecular data from the latter two families (work in progress, MK and PC).

Subordinal structure of Tetractinellida
Kelly & Cárdenas (2016) provided strong support for families Azoricidae, Scleritodermidae and Siphonidiidae within a proposed suborder, supported in part by the common possession of rhizomorine desmas.
Within the polyphyletic rhizomorine family Scleritodermidae, its genera Aciculites is polyphyletic and Scleritoderma is paraphyletic, while Microscleroderma, Amphibleptula Schmidt, 1879 andSetidium Schmidt, 1879 are monophyletic (Fig. 8). The genus Amphibleptula is currently monospecific with A. madrepora Schmidt, 1879 from the Caribbean (Pisera & Lévi, 2002c). Morphologically, A. madrepora is very similar and easy to confuse with Microscleroderma spirophora Lévi, 1960as discussed in Van Soest & Stentoft (1988. Amphibleptula is here sequenced for the first time and our 28S phylogeny shows Microscleroderma and Amphibleptula sp. 1 as sister groups, although unsupported in the 28S phylogeny (PP = 0.63). Morphological observations (SBD#1802,1803) provide conclusive evidence that our three samples are Amphibleptula species, due to their dense tuberculated/blunt spinose rhizoclones, the protruding bundles of oxeas in the oscula area (SBD#1802,1803) as well as the presence of sigmaspires (SBD#1802). Differences to A. madrepora are the diactine spicules present in all three Amphibleptula sp. 1. In addition, fusiform spined microxeas and acanthorhabds are found in the specimen from Jamaica (HBOI 1-IX-93-1-006, SBD#1802). To conclude, we sequenced two potential new species of Amphibleptula with clear unique morphological characters, different from A. madrepora. & Lévi, 2002a, Pisera & Lévi, 2002d, it is difficult, to draw any conclusion of different depth zones or habitats influencing growth form patterns in these two families. However, further 'lithistid' families with a similar bathymetric trend are observed and growth forms are suggested to play a role in the bathymetric distributions of 'lithistids'. For instance Leiodermatium spp. (Azoricidae) are abundant (27 specimens) in depth zones 301-1,000 m. Similar to Corallistidae Leiodermatium possess a dense heavily articulated skeleton, but of strongly spinose rhizoclone desmas. The growth form of Leiodermatium species are described as being foliated or vase to ear-shaped (Pisera & Lévi, 2002b). Such growth forms are suggested to improve the water circulation in sponges, in particular of those in the deep-sea habitats, and to be more resistant to higher water viscosity and scarcity of particles (Levinton, 1982;Gage & Tyler, 1991). Many vase to cup or ear-shaped sponges have their inhalant pores facing the outer side and exhalant openings on the upper side separating incoming and processed water (Sará & Vacelet, 1973;Pronzato, Bavestrello & Cerrano, 1998), which may reduce any negative effect on filtering due to a sedimentation. This is in contrast to Siphonidiidae, a family represented in this analysis by the genera Gastrophanella and Siphonidium, which are rather encrusting or irregular cylindrical, thus more abundant in the depth zone of 61-150 m.
Scleritodermidae occurred more often on vertical walls in depth 301-600 m, but was also not observed to be a major component of the 'lithistid' fauna in the TWA. The greatest number of desma-bearing demosponges were found in depth zone 301-600 m (87 specimens), with Corallistidae as the dominant family (34 specimens) followed by Azoricidae with 27 specimens. Diverse habitats from fine mud and sand slopes to rock pinnacles, boulders and vertical walls in this depth zone (Fig. 18) could be a possible explanation. The families Neopeltidae and Macandrewiidae are rare in our study with only one species discovered at 909 m depth on a vertical wall in the Bahamas (Macandrewia sp.), Figure 18 Bathymetric distribution and relative abundance (%) of TWA desma-bearing demosponges based on 234 samples of eight families. Numbers in each bar represent the number of samples investigated. The following genera for each family were included: Leiodermatium (Azoricidae); Corallistes, Herengeria, Neophrissospongia and Awhiowhio (Corallistidae); Macandrewia (Macandrewiidae); Daedalopelta and Neopelta (Neopeltidae); Aciculites, Amphibleptula, Microscleroderma, Scleritoderma and Setidium (Scleritodermidae); Gastrophanella and Siphonidium (Siphonidiidae); Discodermia, Racodiscula and Theonella (Theonellidae); Vetulina (Vetulinidae). Geomorphological characterizations of depth zones are given below the graph and follows Pomponi et al. (2001) and Reed & Pomponi (1997 and two Daedalopelta spp. collected from the Bahamas at 301-600 m. This corroborates the findings of Pomponi et al. (2001), because they found one species of Daedalopelta nodosa Schmidt, 1879 at 452 m in the Bahamas, one Neopelta perfecta in 116 m depth from Grenada and one Macandrewia clavatella in the southwest coast of Florida. These families and species are also found to be rare in the Southwest Pacific (Lévi, 1991;Kelly, 2000). Besides the tetractinellid 'lithistid' sponges, we noted that other desma-bearing sponge lineages, such as family Vetulinidae (Order Sphaerocladina) constitute only a minor component in any depth-zone in the TWA. Further testing is required to assess whether geomorphological conditions resulting of a variety of complex tectonic interactions (e.g., strike-slip faults, thrust fault, subduction and seafloor spreading in Cayman Trough, see Fig. 1), directly affect diversity and bathymetric distribution of 'lithistids' in the TWA (Fig. 18).

CONCLUSION
In summary, this is the first integrative approach using molecular and morphological data on TWA 'lithistid' demosponges, thus contributing to a better understanding of their phylogenetic affinities, diversity and bathymetric distribution patterns. The present study points to specimens/groups in need of deeper taxonomic investigations and revision, however, additional morphological as well as other independent markers are needed. With recent evidence (Pomponi et al., 2001) that 'lithistids' are dominant components among all investigated TWA regions, we suggest a comparable diversity to the Pacific 'lithistids' as well as to the Mesozoic fauna. Furthermore, there is a clear shift of lithistids with a rigid and heavily articulated desma towards deeper habitats (Corallistidae and Azoricidae), whereas 'lithistids' with a less articulated skeleton tend to occur in more shallow habitats (Theonellidae and Siphoniidae). A major effect causing this shift is the availability of silica in the ocean throughout time. Our robust phylogeny enables relaxed molecular clock analyses in conjunction with the rich fossil record of lithistids to better correlate such shifts to geological/geochemical events in the past.
his help with the modification of the phylodiversity script. Nicole Enghuber is thanked for her help in spicule preparations and Simone Schätzle is thanked for sequencing assistance (all at the Dept. of Earth and Environmental Sciences, LMU München, Munich, Germany).

ADDITIONAL INFORMATION AND DECLARATIONS Funding
Financial support for this study was provided by the German Science Foundation to Dirk Erpenbeck and Gert Wörheide (DFG ER 611/3-1, DFG Wo869/15-1, respectively). The

Grant Disclosures
The following grant information was disclosed by the authors: The German Science Foundation: DFG ER 611/3-1, DFG Wo869/15-1. • Shirley A. Pomponi and Michelle Kelly analyzed the data, authored or reviewed drafts of the paper, identified genera/species and contributed to the sampling, and approved the final draft.
• Andrzej Pisera and Paco Cárdenas analyzed the data, authored or reviewed drafts of the paper, identified genera/species, and approved the final draft.
• Gert Wörheide analyzed the data, authored or reviewed drafts of the paper, and approved the final draft.
• Dirk Erpenbeck conceived and designed the experiments, authored or reviewed drafts of the paper, and approved the final draft.