Expanded sampling of New Zealand glass sponges (Porifera: Hexactinellida) provides new insights into biodiversity, chemodiversity, and phylogeny of the class

Specimen collection and identification

Specimens, seafloor images, and videos were collected as part of Project PoribacNewZ of the Institute for Chemistry and Biology of the Marine Environment (ICBM), Carl von Ossietzky University of Oldenburg and the Department of Earth and Environmental Sciences, Palaeontology & Geobiology (DEES-PG) of the Ludwig-Maximilians-Universität (LMU) Munich, on the new German RV Sonne (voyage SO254) using the GEOMAR Helmholtz Centre for Ocean Research Kiel ROV Kiel 6000 (Simon, 2017). Permit to collect marine organisms was provided by the Ministry of Foreign Affairs and Trade, Wellington, New Zealand.

In total, 101 glass sponge specimens were collected (Table S1). Subsamples were taken on board, stored in appropriate preservatives for morphological and molecular work, and shipped to DEES-PG at LMU Munich. Subsamples were also taken for chemical analysis by cutting 5 cm³ pieces from the specimens where possible, freezing them at −80 °C and shipping them to ICBM at the University of Oldenburg. Specimens were initially assigned to higher taxa by surveying underwater photographs. Subsequently, a molecular phylogenetic survey based on a mitochondrial 16S ribosomal DNA (16S) fragment (cf. Dohrmann et al., 2008) was performed at DEES-PG at the LMU Munich. This allowed assignment of all successfully sequenced specimens (n = 84) to families and, in some cases, to genera with reasonable confidence. Assignment to species was only successful for seven specimens, whose sequences were identical to those published for Walteria leuckarti Ijima, 1896 (Euplectellidae, three specimens) and Aphrocallistes beatrix Gray, 1858 (Aphrocallistidae, four specimens), respectively. Another specimen (NIWA126085) was tentatively identified as Caulophacus (Caulophacus) arcticus (Hansen, 1885) (Rossellidae) by this approach; however, this species’ arctic distribution casted some doubt on this identification, and morphological investigation revealed it as belonging to a new species (see below and Reiswig et al., 2021). The molecular survey further revealed the presence of multiple specimens for several other species (see Table S1 and full 16S phylogeny in Fig. S1).

For further integrative taxonomic and molecular phylogenetic study, 37 specimens (Table 1) were selected to maximize phylogenetic breadth and sequenced for additional mitochondrial (cytochrome oxidase subunit I [COI]) and nuclear (18S and 28S ribosomal DNA [18S, 28S]) markers previously established for the group (see Dohrmann et al., 2008; Dohrmann et al., 2012; Dohrmann et al., 2017 for methods). Preliminary morphological identifications of these specimens were done by MD by analyzing spicule content with light microscopy (LM) of bleach-digested tissue pieces. More thorough investigations and description of species new to science were then started to be performed by HMR using previously described methods (Reiswig & Kelly, 2011; Reiswig & Kelly, 2018). Descriptions of four of these species (as well as two re-descriptions of known species and two new species not selected for further sequencing here; all Rossellidae) were published previously, after the sad passing of Henry Reiswig in July 2020 (Reiswig et al., 2021). Formal descriptions of the remaining 14 species and one subspecies new to science will be provided elsewhere (work in progress).

Molecular phylogenetics

Newly generated sequences (Table 1) and sequences of Rhizophyta yapensis (Shen et al., 2019) were manually added to the alignments provided by Dohrmann (2019) using AliView (Larsson, 2014); duplicates and some problematic sequences (cf. Dohrmann, 2019) were removed. In case of duplicates, i.e., multiple specimens of the same species (mostly from Vargas et al., 2017; Kersken et al., 2018a), the specimen with the least amount of missing data was chosen as that species’ representative. Phylogenetic analyses were performed in maximum likelihood (ML) and Bayesian inference (BI) frameworks (Ronquist et al., 2012; Stamatakis, 2014) using partitioned DNA/RNA substitution models as described in Dohrmann et al. (2017). For BI, chains were run for 10 ×10⁶ generations and the first 10% of samples discarded as burnin. Alignments and trees are deposited in a GitHub repository (https://github.com/PalMuc/SONNE_Hexactinellida).

Chemical fingerprinting

For the chemical extraction, 21 Hexactinellida specimens belonging to eight families with 12 different genera were analyzed, representing a subsample of the specimens included in the molecular phylogenetic study. Specimens were used for chemical fingerprinting if vouchers stored at −80 °C were available, which was not always the case due to limited sponge material. Samples were freeze-dried and ground to a homogeneous powder. For chemical extractions, 1 g of each sample was extracted in 10 mL high performance liquid chromatography (HPLC)-grade MeOH/EtoAc (1:1) on a shaking plate for 30 min and then centrifuged. The supernatant was pipetted, transferred to a second vial, and evaporated to dryness under vacuum. The extraction was repeated two more times, each time with 10 mL MeOH/EtoAc (1:1). Subsequently, the combined extracts were dissolved in HPLC-grade MeOH, filtered (filter size 0.2 µm PTFE membrane filter, VWR international, USA) and brought to a final concentration of 1 mg/L.

For UPLC/MS analysis, an ACQUITY Ultra Performance Liquid Chromatography (UPLC) H-Class system (Waters Co., Milford, MA, USA) coupled with a Synapt G2-Si HDMS high-resolution Q-ToF-MS (Waters Co., Manchester, United Kingdom) was used to obtain the mass spectra. Chromatographic separation was performed using a Waters Acquity BEH C18 column (1.7 µm, 2.1 mm × 50 mm). Analytes were eluted at a flow rate of 0.6 mL min⁻¹ using a linear elution gradient of H₂O (100%, eluent A) to acetonitrile (ACN, 100%, eluent B) both with 0.1% formic acid. The initial condition was 100% A held for 0.5 min, followed by a linear gradient to 100% B in 19.5 min. The column was then washed with 100% B for 9.5 min and subsequently returned and held for 2.9 min to the initial conditions (100% eluent A) to equilibrate the column for the following run. The column temperature was set at 40 °C and the injection volume was 1 µL.

The mass spectrometer was calibrated and handled as described in Kamyab et al. (2020: p. 11). Samples were injected using a Latin Square design to reduce systematic technical errors. At the beginning of the sequence three blank (i.e., methanol) samples were injected, followed by five quality control (QC) samples (i.e., pool of biological samples). QC samples were then equidistantly (every five injections) analyzed throughout the sequence. At the end of the sequence, two QC samples and two blank samples were analyzed.

The open-source tool msConvert from the open-source ProteoWizard library (Chambers et al., 2012) was used to convert the raw data into mzML files. mzML files were processed using the XCMS online tool (version 2.7.2., XCMS version 1.47.3.) (Forsberg et al., 2018) to detect, deconvolute and align features, which provided a matrix containing the retention time, m/z value and integrated peak areas of the identified features. The CentWave algorithm was used for peak detection (maximum m/z deviation = 15 ppm, peak width = c(2,30), mzdiff = 0.01), and obiwarp was used for peak alignment and retention time correction (profStep = 0.5, bw = 5, mzwid = 0.05). Features that had (1) a coefficient of variation higher than 20% among the QC samples, (2) blank intensities that were over 50% of those observed in biological samples (at least one sample), and (3) peak area intensities <10,000 in all sample groups were removed. This resulted in a dataset that contained 4,742 features. A dataset only containing major features (i.e., >5,000,000 integrated peak area in at least one of the sample groups, 112 features) was also generated.

All numerical analyses were performed in R version 4.1.2. (R Core Team, 2021). Integrated peak areas were log transformed prior to multivariate analyses. Non-metric Multidimensional Scaling analysis based on Euclidean distance (NMDS, function metaMDS from the vegan package, without auto-transformation), was used to visualize the similarity in the metabolic fingerprints of the different sponge species. An Analysis of Similarities (ANOSIM) using Euclidean distance was used to test for statistical difference in the metabolic profiles of the different sponge groups (i.e., taxonomic families). A hierarchical cluster analysis (function hclust, Euclidean distance) was used to classify sponge species based on the similarity of their metabolomes. The “Average” algorithm was chosen after analysis of the cophenetic correlation coefficient, and the Kelley-Gardner-Sutcliffe (KGS) penalty function (maptree R package) was used to prune the dendrogram and obtain the optimal number of clusters.

Systematic account

Following the World Porifera Database (De Voogd et al., 2023).

Taxonomic notes

Underwater (in situ) and deck photographs of the new species are provided in Figs. 2–6.

Hyalonema n. sp. (NIWA 126036) (Amphidiscosida: Hyalonematidae), abyssal basin between Three Kings and Colville Ridges, northern New Zealand, 4,158 m. Figures 2A–2C.

NIWA 126036 consists of an oval body, 36 mm high, on a slender, twisted ‘rope’ of six or more long, thick, anchoring spicules, the whole sponge rooted basally in sediment (A). The surface is smooth, felty, and appears faintly cross-hatched; the color in situ is white, and above water, tan (B, C). This specimen cannot presently be assigned to any existing subgenus of Hyalonema, and the material examined was too damaged to erect a new subgenus; for the time being this species is treated morphologically as “subgen. inc. sed”. Furthermore, the spiculation does not match any of the 15 species known from a 1,000 km radius around the collection site. Its most prominent spiculation feature is very long, whip-like (likely atrial) pinules.

Pheronema n. sp. (NIWA 126135, 126138) (Amphidiscosida: Pheronematidae), Southern Kermadec Ridge, northeastern North Island, New Zealand, 1,169 m. Figures 2D–2F.

NIWA 126135 and NIWA 126138 are barrel-shaped sponges, 105 mm, and 165 mm diameter, respectively (D), with a wide, smooth opening at the apex leading to a deep atrium that extends almost to the base of the sponge (E, F). The exterior is covered in a beard of very fine, hair-like spicules that trap sediment (F), the color in situ is pinkish brown, and above water, dark tan. These specimens clearly differ in spiculation from all 20 described species of the genus.

Poliopogon n. sp. (NIWA 126337) (Amphidiscosida: Pheronematidae), Southern Kermadec Ridge, northeastern North Island, New Zealand, 1,150 m. Figures 2G, 2H.

NIWA 126337 is a large pillar, about 400 mm high and 230 mm wide, with large concave inhalant and exhalent cones, attached to the substrate by a long, spreading fringe (G). The exterior surfaces are a smooth crosshatch and the margins of the aquiferous cones, thin. Color in situ is pale lemon tan, peach out of water (H). This specimen shows a unique shape and size range of microuncinates and shape and size of macramphidiscs, not known from any described species of this genus.

Eurete n. sp. 1 (NIWA 126028) (Sceptrulophora: Euretidae), abyssal basin between Three Kings and Colville Ridges, northern New Zealand, 4,160 m. Figures 3A–3C.

NIWA 126028 is a pedunculate, cavernous, honeycombed body, about 12 cm long and eight cm wide, arising from a tough, solid stem (A). The body is bulb-shaped with a restricted apex opening into a wide tubular, perforated atrium (B, C). Texture is crunchy, brittle, color in situ white, tan out of water. This specimen was identified as Homoieurete n. sp. 1 by HMR due to the presence of some discohexasters; however, this contradicts the molecular phylogenetic results (see below) and needs further investigation.

Eurete n. sp. 2 (NIWA 126276) (Sceptrulophora: Euretidae), Wairarapa Slope, southeastern North Island, New Zealand, 1,663 m. Figures 3D, 3E.

NIWA 126276 is a shallow chalice with corrugations visible as translucent and opaque lines radiating from the base of the sponge body in life (D), about eight cm diameter at the apex. The external sculpted corrugations are readily visible in the deck image (E). Color in situ is a faint pinkish white, tan out of water. This specimen was identified as Homoieurete n. sp. 2 by HMR due to presence of some discohexasters; however, this contradicts the molecular phylogenetic results (see below) and needs further investigation.

Farrea occa n. ssp. (NIWA 126307) (Sceptrulophora: Farreidae), Seamount 986, off Hawke Bay, eastern North Island, New Zealand, 893 m. Figures 3F, 3G.

NIWA 126307 is a thick, curved, cavernous, honeycombed fan, about seven cm wide, attached along the base (F). Texture is crunchy, brittle, color in situ white, tan out of water (G). This specimen has hooks on most of the anchorate clavules, which is a character not known from the ten described subspecies of F. occa (cf. Lopes, Hajdu & Reiswig, 2011: Table 2).

Farrea n. sp. (NIWA 126004) (Sceptrulophora: Farreidae), Seamount 986, off Hawke Bay, eastern North Island, New Zealand, 893 m. Figures 3H, 3I.

NIWA 126004 is a pedunculate, tree-shaped sponge with numerous, hollow, tubular branches arising from the primary stem, about 12 cm high, attached to hard substrate by a tough, solid base (H). The ends of the branches are flared and irregularly extended in places (I). Texture tough, crunchy, brittle, color in situ and out of water, white. This specimen has a combination of clavule types (pileate and anchorate with hooks), microscleres (oxyhexasters only), and morphology of surface pentactins (microtuberculated) that does not match with diagnoses of the 37 described accepted species given in Lopes, Hajdu & Reiswig (2011): Table 2) and later additions (Reiswig & Kelly, 2011; Reiswig & Stone, 2013; Reiswig, 2014; Reiswig, 2018; Reiswig, 2020; Boury-Esnault et al., 2017; Tabachnick et al., 2019).

Corbitella n. sp. (NIWA 126122, 126123, 126350) (Lyssacinosida: Euplectellidae), Southern Kermadec Ridge, northeastern North Island, New Zealand, 1,164–1,205 m. Figures 4A–4G.

NIWA 126122 (A, D), NIWA 126123 (B, E) and NIWA 126350 (C, F, G) have long, hispid, tubular bodies, up to about 400 mm long and about 100 mm wide, with a delicate corona of marginal spicules projecting in a broad, thin, wavy cuff on the upper body wall, with eyelash-like hexactine parietal prostalia projecting from the rim. These sponges usually harbour a pair of resident Venus shrimps. Reiswig & Kelly (2018: 84) assigned these three specimens to Regadrella hispida (Reiswig & Kelly, 2018), due to their apparent similarity to the holotype of R. hispida, illustrated in Reiswig & Kelly (2018): fig 31, F): all appeared to have an extremely hispid body wall and long marginal spicules projecting from the margin of the upper body wall. However, Reiswig & Kelly (2018) based their assignments to R. hispida on images only (NIWA 126122: seafloor image 68). Genetic differences in this study suggested these three specimens were a different species, and a new species of Corbitella rather than Regadrella. It is obvious to us now that Corbitella n. sp. has a thin, wavy, marginal cuff, that projects from the upper rim of the sponge body, whereas this appears to be absent in the, albeit damaged, holotype of R. hispida. Furthermore, the body of R. hispida is barrel-shaped, not tubular, and tall as in Corbitella n. sp.

Trychella n. sp. (NIWA 126306) (Lyssacinosida: Euplectellidae), Seamount 986, off Hawke Bay, eastern North Island, New Zealand, 893 m. Figures 4H, 4I.

NIWA 126306 is a thin-walled, bulbous, sack-shaped, sponge, 68 mm high (H), with a wide, globular, curved base, tapering to a broad, open apex surrounded by a thick margin from which project sparse, long, spine-like, marginal spicules (I). The wall is lumpy and punctate, texture slightly compressible, color in situ and out of water icy white. This specimen differs from the only described species, T. kermadecensis—among other spiculation characters—by having floricomes and lacking tetradiscs and pentadiscs.

Aulocalyx n. sp. (NIWA 126318, 126176, 126301) (Lyssacinosida: Aulocalycidae), Seamount 986, off Hawke Bay, eastern North Island, New Zealand, 875 m. Figures 5A–5C.

NIWA 126318 (A, B), NIWA 126176, and NIWA 126301, are cup-shaped sponges, about 85 mm wide, with a smooth, perforated interior (B) and heavily pocketed exterior (A, C). The tubular wall extensions are thin, delicate, and translucent, some are blind, and some are open-ended. The overall appearance is of a fragile, translucent flower. There are currently three described species in this genus: A. irregularis Schulze, 1886, A. serialis Dendy, 1916, and A. australis Reiswig & Kelly, 2011. The new specimens clearly differ in microsclere combination to these species: They have big spheric discohexasters, which are missing in A. australis and A. irregularis; have small spheric discohexasters, which are missing in A. serialis; lack big stellate discohexasters, which are present in A. australis and A. serialis; lack oxyhexactins, which are present in A. australis and A. irregularis; and have smaller rhopalasters than the described species.

Rhabdodictyum n. sp. (NIWA 126083) (Lyssacinosida: Aulocalycidae), Kermadec Trench slope, northeastern North Island, New Zealand, 4,833 m. Figures 5D, 5E.

NIWA 126083 is a moderately thin, beautifully regular, arching fan, about 15 cm wide, attached at the center of the base of the fan (D). The surface is punctate, the texture is compressible, appearing fibrous in close-up. Color in situ is translucent white and dark tan out of water (E). This specimen has a very different body shape than the only other known species, R. delicatum Schmidt, 1880 (cf. Reiswig, 2002b), and differs by having no other microscleres than spirodiscohexasters.

Chaunoplectella n. sp. (NIWA 126325) (Lyssacinosida: Leucopsacidae), Seamount 1247, off East Cape, eastern North Island, New Zealand, 1,389 m. Figures 5F, 5G.

NIWA 126325 is low, elongate to ovoid in overall form, about 85 mm diameter, with a large, deep, perforated atrium. The skeleton is very loose and airy with surface pentactins visible sitting on the surface. Color in life translucent icy white, tan out of water. This specimen lacks the large anchorate discohexasters typical for species of Chaunoplectella and could therefore be assigned to a new genus; however, given that it otherwise matches well with the diagnosis of Chaunoplectella, amendment of that diagnosis seems more appropriate.

Leucopsacus n. sp. (NIWA 126063) (Lyssacinosida: Leucopsacidae), West of Raoul Island, Kermadec Ridge, northeastern North Island, New Zealand, 335 m. Figures 5H–5J.

NIWA 126063 is cup-shaped with very thick walls and a deep atrium, about 45 mm high and 35 mm diameter (H, I). Surface relatively smooth with a relatively compact skeleton. Color in life translucent icy white, white out of water (I, J). This specimen clearly differs in spicule composition and dimensions from the five described species of the genus.

Lophocalyx n. sp. (NIWA 126005, 126014) (Lyssacinosida: Rossellidae: Lanuginellinae), Kiwi Seamount, Three Kings Ridge, northern New Zealand, 760 m. Figures 6A–6C.

NIWA 126005 and NIWA 126014 are sack-like sponges with a soft, felty, fibrous body that expands from a solid restricted attachment base, about 400 mm high (A), collapsing inwards slightly out of water (B). The upper body wall margin is thin and easily torn (C). Color in situ is white, whitish tan (attached detritus) out of water. This specimen clearly differs in spicule composition from the 12 described species of the genus.

Lanuginellinae n. gen. n. sp. (NIWA 126169, 126302, 126310, 126315, 126319) (Lyssacinosida: Rossellidae: Lanuginellinae), Seamount 986, off Hawke Bay, eastern North Island, New Zealand, 802–893 m. Figures 6D–6I.

NIWA 126169 (D, E) and other specimens examined are soft-bodied sponges up to 50 cm high, arising from solid, irregular, cylindrical, or laterally compressed stems of highly variable morphology. Stems are curved, anastomosing, or diverging to form branches which give rise to one or more soft, smooth, billowy, tulip to mushroom-shaped bodies, soft pinkish white in situ and out of water (D–I). This species has some similarities in spiculation to species of Sympagella Schmidt, 1870, and was initially suggested by HMR as a new species of that genus. However, the body shape is highly unusual and very distinct from the typical stalked, mushroom-like appearance of described Sympagella species. Furthermore, assignment to Sympagella would contradict the molecular evidence (see below). We therefore conclude that it is best classified in a new genus.

Molecular phylogeny

Deep phylogenetic relationships of Hexactinellida (Fig. 7, Figs. S2–S4) are largely congruent with previous analyses (Dohrmann et al., 2017; Dohrmann, 2019). However, within Lyssacinosida, the sister group of Rossellidae is here Leucopsacidae with moderate bootstrap support (BS) and posterior probability (PP), instead of Aulocalycidae (Dohrmann, 2019) or a clade of Aulocalycidae + Leucopsacidae (Dohrmann et al., 2017: Fig. 7), in line with a genus-level total-evidence (DNA + morphology) analysis (Dohrmann et al., 2017: Fig. 10). Thus, the hypothesis that Aulocalycidae evolved their fused (dictyonal) skeleton from an ancestral leucopsacid-like choanosomal skeletal organization (Dohrmann et al., 2017: p. 17) could not be corroborated further by increased taxon sampling of these two small families.

Another discrepancy with Dohrmann (2019) concerns the low BS for monophyly of Lyssacinosida (36%), which was 82% in the previous study. However, we observed that after exclusion of the problematic Heterorete sp. (Hexasterophora inc. sed.), BS for Lyssacinosida rose to 75%, and when also removing the dactylocalycids Dactylocalyx and Iphiteon, even to 96% (Figs. S5–S6). Similarly, in the phylogeny including all taxa (Fig. 7, Fig. S2) BS for monophyly of Euplectellidae is only 43% but rose to 61% and 96%, respectively, under the reduced sampling schemes (Figs. S5–S6). These results indicate that Heterorete and the dactylocalycids behave as “rogue taxa” in the bootstrap analysis (see also Shen et al., 2019 for discussion of potential long-branch attraction issues). In contrast, BI was robust to these issues and yielded strong support for these well-established groups with the full dataset (Fig. 7, Figs. S2, S4). However, both methods were unable to resolve the exact positions of Dactylocalyx, Iphiteon, and Heterorete—clearly, more data are needed to determine their relationships (Dohrmann et al., 2017).

Below we briefly discuss the positions of the additional species for major hexactinellid subgroups (Amphidiscophora, Sceptrulophora, and the families of Lyssacinosida), and some implications for taxonomy and evolution.

Amphidiscophora (Fig. 8).—Among Pheronematidae, monophyly of the type genus Pheronema is corroborated by inclusion of an additional species (Pheronema n. sp.). Inclusion of a second species of Poliopogon (Poliopogon n. sp.) reveals an apparent diphyly of this genus. However, the position of Poliopogon n. sp. is congruent with predictions of a total-evidence analysis (Dohrmann et al., 2017: Fig. 8), whereas the position of P. microuncinatus Kersken et al., 2018 close to Schulzeviella n. sp. might indicate generic misassignment by the describing authors (i.e., Kersken, Janussen & Martínez Arbizu, 2018b). This gains some support from the presence in this species of the eponymous dagger-shaped microuncinates, which have also been described from Schulzeviella gigas (Schulze, 1886), originally placed in Poliopogon (see Tabachnick & Menshenina, 2002). A similar issue has been suggested for the apparent diphyly of Semperella (Dohrmann, 2019); additional investigations of the morphology of P. microuncinatus and S. jialongae Gong, Li & Qiu, 2015 will be necessary to resolve their taxonomy.

Among Hyalonematidae, Hyalonema n. sp. is very closely related to H. (Cyliconemaoida) ovuliferum Schulze, 1899, suggesting it should be placed in that subgenus. The placement of Lophophysema and Tabachnickia as ingroups of Hyalonema suggests that they might be better classified as subgenera of Hyalonema. As pointed out by Dohrmann et al. (2017) and Dohrmann (2019), major integrative taxonomic revisions of the family will be necessary to solve the phylogenetic status of subtaxa and their assignment to genera or subgenera. For further discussion of hyalonematid phylogeny, see Dohrmann (2019).

Sceptrulophora (Fig. 9)—Addition of five further species of Farrea corroborates the artificial nature of the type genus of Farreidae (see discussion in Dohrmann, 2019) but also reveals that the genus Aspidoscopulia is not monophyletic. This implies that aspidoscopules (the diagnostic spicule type of Aspidoscopulia) are a poor phylogenetic character that can be expressed in multiple unrelated species. Therefore, the two genera—as well as Lonchiphora (two species), which is also nested within Farrea (Fig. 9; Dohrmann, 2019)—should be considered to be synonymized in future revisions of the family.

Regarding the “waste-bin” family Euretidae (Reiswig & Dohrmann, 2014; Dohrmann et al., 2017), we here provide the first sequences of the type genus Eurete (Eurete n. sp. 1 and 2). Both are firmly placed in the Homoieurete-Sarostegia-Chonelasma-Bathyxiphus clade (Kersken et al., 2018a), which should therefore become the new core for a revised scope of Euretidae (cf. Dohrmann, 2019). Consequently, we move Homoieurete and Sarostegia from Sceptrulophora inc. sed. (Reiswig & Dohrmann, 2014; Dohrmann et al., 2017) back to Euretidae, and instead transfer Conorete, Lefroyella, and Verrucocoeloidea to Sceptrulophora inc. sed. The position of Eurete within Sceptrulophora recovered here contrasts with a previous total-evidence analysis (Dohrmann et al., 2017: Fig. 9), although analysis of morphological data alone indicated a closer relationship with Chonelasma (Dohrmann et al., 2017: Fig. 5). Eurete itself is here recovered paraphyletic, with Eurete n. sp. 2 more closely related to Bathyxiphus subtilis Schulze, 1899 and Chonelasma bispinula Kersken, Janussen & Martínez Arbizu, 2019 (although with poor support). However, this result must be taken with caution since the latter species are only represented by a 16S sequence (B. subtilis) and a partial 18S sequence (C. bispinula), respectively (Kersken et al., 2018a), which could have disturbed accurate reconstruction of this quartet due to excessive missing data; when the incompletely sampled B. subtilis and C. bispinula are excluded, Eurete is recovered as monophyletic (e.g., Fig. S1). Despite the poor resolution of the relationships of Bathyxiphus, Chonelasma, and NIWA 126028 and 126276, they nonetheless group together in a strongly supported clade as sister to Homoieurete macquariense Reiswig & Kelly, 2011, separated by a relatively long branch (Fig. 9, Fig. S3). Thus, assignment of the new species to Homoieurete due to presence of discohexasters (see above) seems inappropriate; instead, generic diagnoses of Homoieurete and Eurete should be emended to better distinguish them morphologically.

Lyssacinosida: Euplectellidae (Fig. 10)—With 13 so far unsequenced species, most of them recently described or new to science (Reiswig & Kelly, 2018; this study), in eight genera (four sequenced here for the first time), Euplectellidae was the best-represented family in the Sonne SO254 collection (Table 1; Table S1). As in Shen et al. (2019), Rhizophyta yapensis and the discoplumicome-bearing clade (DPC, here represented by Hertwigia and Saccocalyx; cf. Dohrmann et al., 2017) form successive sister groups to the remaining euplectellids in both the ML (Fig. 10A) and BI (Fig. 10B) trees. Notably, regarding the deeper relationships of the latter, there were marked discrepancies between results of the two methods.

In general, monophyly of Trychella, Amphidiscella, Saccocalyx, and Bolosoma could be (further) corroborated here. In contrast, Regadrella and Corbitella are clearly recovered as polyphyletic, indicating that these relatively poorly defined genera are in need of revision. Amphidiscella and the two recently described new genera of Bolosominae (Amphoreus and Trychella; Reiswig & Kelly, 2018) expectedly group with the remainder of that subfamily (except for Saccocalyx and Rhizophyta; see Dohrmann et al., 2017; Shen et al., 2019). However, this clade also includes the newly sequenced Dictyaulus hydrangeaformis, which has a typical venus-flower basket (VFB) body shape (Reiswig & Kelly, 2018), in contrast to the mushroom- or wine glass-shaped bolosomins. Further, in the BI tree, a clade containing members of the VFB genera Corbitella and Regadrella is strongly supported as the sister group to a clade containing the majority of VFB and bolosomine species. Thus, the “Bolosominae sensu stricto” and “VFB clade” concepts proposed by Dohrmann et al. (2017) as a starting point for a new subfamily classification of Euplectellidae could not be further corroborated here (see also Shen et al., 2019). A natural subgrouping of euplectellid genera therefore seems out of reach and should best be abandoned.

Lyssacinosida: Leucopsacidae and Aulocalycidae (Fig. 11)—Monophyly of these two small and thus far poorly sampled families is further supported here by inclusion of Chaunoplectella and Leucopsacus (Leucopsacidae), and Rhabdodictyum and Aulocalyx (Aulocalycidae), respectively. Genus-level taxon sampling is thus complete for Leucopsacidae; Aulocalycidae contains four further genera, but it is unclear if the elusive lyssacinosid specimen SMF12068 (cf. Dohrmann, 2019) belongs to any of those or represents a new genus. Among Aulocalycidae, monophyly of Aulocalyx is confirmed, whereas among Leucopsacidae the type genus appears paraphyletic with strong support, as it includes Oopsacas. This latter result is somewhat surprising as Leucopsacus is morphologically well defined (Tabachnick, 2002b); clearly this needs to be investigated further before any taxonomic action can be taken. A possible reason could be gene-tree-species-tree conflict as the genus was resolved as monophyletic (81% BS) in the 18S single-gene tree (not shown).

Lyssacinosida: Rossellidae (Fig. 12)—Monophyly of Caulophacus and non-monophyly of its subgenera Caulophacus and Caulodiscus (Kersken et al., 2018a; Dohrmann, 2019) is further supported here by inclusion of C. (Caulophacus) serpens (described in Reiswig et al., 2021) and C. (Caulophacus) discohexaster (re-described in Reiswig et al., 2021), which seem to be closely related. Also, according to the 16S tree (Fig. S1; see comment in section “Specimen collection and identification”), C. (Caulophacus) ramosus (NIWA126085; described in Reiswig et al., 2021) appears to be very closely related to C. (Caulophacus) arcticus, C. (Caulodiscus) valdiviae Schulze, 1904, and C. (Oxydiscus) weddelli Janussen, Tabachnick & Tendal, 2004, although it shows no obvious morphological similarities to these species, beyond the shared characteristics of the genus. Inclusion of another species of Lophocalyx, Lophocalyx n. sp., did not corroborate the monophyly of this genus, as the species seems to be close to Caulophacella. More taxonomic work on the morphology of the poorly known Caulophacella and its relationship to Lophocalyx, as well as better taxon sampling of the latter (currently 12 described species) will be necessary to settle this issue. In any case, Caulophacella clearly groups outside Caulophacus, confirming its status as a separate genus (Gong et al., 2023; contra Boury-Esnault et al., 2015). The other species (and new genus) of Lanuginellinae sampled during Sonne Cruise SO254 with several specimens (NIWA 126169 and four other samples; Table 1), despite having some similarities in spiculation to Sympagella (see above), the sister taxon to all remaining lanuginellines, clearly diverges after that genus in the molecular phylogeny and forms a distinct branch in the Lanuginellinae clade (Fig. 12, Fig. S3).

Among the remainder of Rossellidae, the recently described genus Nubes (Reiswig et al., 2021) is monophyletic and appears closely related to Crateromorpha. Although the two genera are overall similar in spiculation, no striking potential synapomorphies are apparent to us, except perhaps the presence of a peduncle in Crateromorpha and N. poculiformis, which is however a homoplasy-prone character. Monophyly of Bathydorus, reconstructed here as the sister group of Crateromorpha + Nubes, is further confirmed by inclusion of B. poculum (see Reiswig et al., 2021 for description). Scyphidium australiense is strongly supported as the sister group of Rossella. Although the taxonomic history of the two genera is somewhat intermingled (see Tabachnick, 2002c), again no strong support from morphological characters is apparent to us. It is very closely related to S. variospinosum (see Reiswig et al., 2021 for description) according to the 16S analysis (Fig. S1). In fact, the sequences of the three investigated Scyphidium specimens (Reiswig et al., 2021) are so similar that, based on the molecular results only, we assumed they were conspecific (hence we did not attempt to sequence the other markers for all of them). This high genetic similarity likely points to a very recent divergence or incipient speciation of S. variospinosum.

Chemical fingerprinting

The metabolic profiles of different glass sponge families were significantly different (ANOSIM R = 0.75, Fig. 13A); however, these differences were stronger when only the major metabolites were analyzed (ANOSIM R = 0.87, p = 0.001, Fig. 13B). When all metabolites were analyzed, five significantly different clusters were obtained (Fig. 14A). The cluster analysis was congruent with the NMDS (separation along NMDS1) and showed a first separation into two major clusters: cluster 1, which contained the families Farreidae, Euretidae, Pheronematidae and Hyalonematidae, and cluster 2, which contained the families Euplectellidae, Aulocalycidae, Leucopsacidae, and Rossellidae. While cluster 2 is congruent with monophyly of Lyssacinosida, cluster 1 contradicts monophyly of Hexasterophora (i.e., Sceptrulophora [here Farreidae + Euretidae] closer to Amphidiscophora [Pheronematidae + Hyalonematidae] than Lyssacinosida), suggesting limitations of chemical fingerprinting at this deep phylogenetic level. The first cluster split into two subclusters separating the two Amphidiscophora families (which formed by themselves unique clusters) from the Farreidae and Euretidae, which were also clearly separated into statistically different clusters (Fig. 14A). Metabolic fingerprinting of all metabolites did not allow a clear resolution of cluster 2, i.e., of the four families of Lyssacinosida, only Aulocalycidae was recovered. Analysis of only major metabolites allowed identification of 7 statistically different clusters (Fig. 14B), with Euplectellidae now also being clearly separated. However, genera of Rossellidae and Leucopsacidae were still intermingled. One reason for the unclear taxonomic resolution when all metabolites were included could be the time span between actual collection at depth and preservation of the sponge pieces on board. This could have led to the production of some additional minor metabolites by the sponge as a result of the physical injury during collection and transport to the surface and ship (which could have been up to 6 h for certain specimens). Despite this, in general a good clustering by family and genus can be observed, suggesting a higher replication could allow a better chemotaxonomic discrimination (possibly to genus or species level).