The Amazon River mouth drains an enormous sedimentary basin carrying a massive load of siliciclastic sediment of Andean origin, suspended in 20% of the freshwater input in the oceans (Lentz and Limeburner, 1995). A large, low salinity plume (up to 2 × 10⁶ km²) moves northwest influencing the insular environment in the Southeast Caribbean (Lumpkin and Garzoli, 2005). The seabed in this area presents a regionalization. The central and southern areas are predominantly carbonate, while the northern area suffers more direct influence of the Amazon plume and its terrigenous sedimentation (Collette and Rützler, 1977; Moura et al., 2016). The confluence between the Amazon River and the Atlantic Ocean generates high fisheries production in this relatively shallow and wide shelf, including important demersal resources (fishes and crustaceans), associated to reefal environments (Moura et al., 2016).

An oceanographic expedition was carried in the Amazon mouth with the Brazilian Navy ship NHo “Cruzeiro do Sul” in 24–29 September 2014. A total of 90 specimens were collected by bottom trawls and dredges. The specimens were sorted, photographed and labeled on ship before fixation in 92% ethanol. After that, they were deposited in the Porifera Collection of Museu Nacional—UFRJ (MNRJ). A fragment of each species sorted on board was preserved in liquid N₂, but one of the Arenosclera spp. later recognized as new (namely A. klausi sp. nov. described below), was not cryopreserved because both were thought conspecific.

Samples were identified based on microscopic preparations of dissociated spicules, and thick anatomical sections obtained from fragments. These procedures are described in detail in Hajdu et al. (2011). For each specimen 100 oxeas were randomly selected and measured, unless stated otherwise. Dissociated spicules were also analyzed in a JEOL JSM6390LV Scanning Electron Microscope (SEM). Spicule dimensions are presented in micrometers as minimum–mean–maximum (±standard deviation). New species were compared with the specimens listed in Table 1.

The new species described in this paper are registered online at ZooBank (http://zoobank.org/) as requested by the International Code of Zoological Nomenclature. Each species received a Life Science Identifier (LSID) discriminated above.

Fragments of specimens were washed in filtered sea water to remove all microorganisms attached to the sponge's outer surface. Afterwards, samples were macerated and embedded in Solution A [CTBA—2% + 100 mM EDTA + 1.4 M NaCl + 100 mM Tris-HCL (pH 8.0) + 2.0 μL Beta-mercaptanol + 0.5 μL proteinase K a 20 mg/mL]. Three heat-shock cycles were made (65°C by 3 min and −80°C by 30 min). After that, 1 ml of phenol:chloroform:isoamilic Alcohol (25:24:1) was added, and centrifuged in 13,400 rpm for 10 min at 4°C. The supernatant was then transferred to a new tube, and the spin filters from PowerSoil DNA Isolation Kit (MoBio, USA) were used to further purify the solution containing nucleic acids (Garcia et al., 2013).

Next generation sequences were obtained through HiSeq (Illumina, USA) by the staff of Laboratório de Microbiologia (at Departamento de Biologia Marinha/UFRJ). DNA libraries were prepared using Nextera XT DNA Sample Preparation Kit (Illumina, USA), that uses a tagmentation reaction for transposon cleaving and tagging of the double-stranded DNA with a universal adapter, followed by a limited cycle PCR to add primer sequences and indices. The size distribution of reads in each library was evaluated with a 2100 Bioanalyzer and its High Sensitivity DNA Kit (Agilent, USA). The accurate quantification of the libraries was accomplished using the 7500 Real Time PCR (Applied Biosystems, USA) and the KAPA Library Quantification Kit (Kapa Biosystems, USA). Paired-end sequencing (2 × 150 bp) was performed using the kits TruSeq® Rapid SBS Kit–HS (50 cycles and 200 cycles) and TruSeq® Rapid PE Cluster Kit–HS. Quality control of the metagenomes was undertaken with the PRINSEQ software (Standalone Lite Version 0.20.4; Schmieder and Edwards, 2011) by removal of sequences presenting Phred score lower than 30 (Q < 30), and with duplicate reads. Paired-end reads were merged using PEAR (Zhang et al., 2014) with a base Phred quality score of 20.

Besides Brazilian Arenosclera, we used the metagenomes of Amphimedon compressa Duchassaing and Michelotti, 1864 and Callyspongia vaginalis (Lamarck, 1814) collected in the same expedition, as controls to verify if there is influence of the environmental microbiome in this analysis. In addition, these species are important for the recovery of the phylogenetic affinities of Arenosclera as shown below. The biological materials available for each of these species are illustrated as a Figure S1. Sequences of the 28S rRNA from Arenosclera spp. nov., Arenosclera brasiliensis, A. compressa, and C. vaginalis were submitted to Genbank (Table 2). Arenosclera brasiliensis materials used here came from the Cabo Frio region (Rio de Janeiro, Brazil), and had been reported upon by Trindade-Silva et al. (2012). This metagenome originated from SANGER sequencing which generates larger reads (>500 bp). All other metagenomes were generated by HiSeq (Illumina, USA), and are thus composed of smaller reads.

Karlin's signatures compare the relative abundances of dinucleotides in different sequences (Karlin and Burge, 1995), and were used in the comparison among Amazon reef sponges' metagenomes and those obtained by Trindade-Silva et al. (2012) for A. brasiliensis. Frequency tabulation of the sequence data was performed using homemade Perl scripts according to Willner et al. (2009).

We chose as genetic marker the 28S because it is a well-established gene for evolutionary studies with sponges, and this is the only sequence available from the type species of Arenosclera. The 28S sequences used were recovered from metagenomes compared to a referential database compiling all Genbank sequences showing over 80% similarity to the available nearly complete 28S sequence of Arenosclera heroni (2,488 bp; Thacker et al., 2013). Sequences compiled for this database were KC869461, KC869526, KC869599, KC869567, KC869609, KC869497, KC869562, KC869607, KC869455, AB511881, KC869527, KC869473, AY561893, KC869622, KC869553, KC869626, KC869626, KC869620, KC869460 from Genbank. Contigs were aligned against the 28S referential database created using BLASTN algorithm with at least 70% query coverage and an E-value cut-off of 10⁻⁵ (Altschul et al., 1990). We compared the metagenomes of A. brasiliensis, C. vaginalis, A. compressa, and the two new Arenosclera sponges from Amazon reefs. SPAdes software (v. 3.5; Bankevich et al., 2012) was used as a first approach to assemble the high quality short reads (<300 bp), followed by a second approach using CAP3 software (Huang and Madan, 1999).

Additional sequences reported in Thacker et al. (2013) were obtained from GenBank to settle our comparative nearly complete 28S rRNA dataset. We refrained from using further 28S GenBank sequences for the Haplosclerida because in general these were much smaller than the sequences we were working with, and spurious preliminary results suggested that extensive discussion of phylogenetic affinities within this order might be needed, going way beyond the scope of this contribution. A global alignment was performed using MAFFT software (version 7; Katoh and Standley, 2013) with FFT-NS-2 parameter. Gaps were only eliminated when occurring on 20% or more of analyzed sequences. Maximum Likelihood analysis was run in RAxML (Stamatakis, 2014) in CIPRES platform (Miller et al., 2010) using the substitutions model GTR+GAMMA+I selected by Model Generator (Keane et al., 2006) with 1,000 bootstrap.

Class Demospongiae Sollas, 1885

Sub-class Heteroscleromorpha Cárdenas, Pérez and Boury-Esnault, 2012

Order Haplosclerida Topsent, 1928

Family Callyspongiidae de Laubenfels, 1936

Genus Arenosclera Pulitzer-Finali, 1982

Diagnosis: Callyspongiidae with ectosomal skeleton of sand and foreign debris cemented by scarce spongin, rounded meshes. Choanosomal fibers irregular, discontinuous, with foreign debris, proper spicules or both (Desqueyroux-Faúndez and Valentine, 2002).

LSID: urn:lsid:zoobank.org:act:28C5BD00-0AA4-4903-82ACBC1AAD666EA5

LSID: urn:lsid:zoobank.org:act:16CBBD72-49FF-41A0-82E9-A2FA9D6C6930

The state of uncertainty as regards higher taxa diagnoses in the Haplosclerida renders the assignment of new species to currently accepted generic names a taxonomic roulette. On the basis of the 18S marker, Redmond et al. (2013) retrieved not only polyphyletic families and genera, but also species such as Callyspongia fallax, C. vaginalis, H. curacaoensis, and H. tubifera. How can higher taxa boundaries in the Haplosclerida be discussed if species-limits remain ambiguous? Currently, one has to resort to morphology to recognize a new species' genus, regardless of the certainty of forthcoming major changes in higher taxa classification in the order; and to molecules for its clade A–E (Redmond et al., 2011, 2013), alas frequently not containing the genus' type species.

Accordingly, we observed the single available 28S sequence of A. heroni to cluster with two species of Haliclona, as originally retrieved by Thacker et al. (2013), instead of with the other Arenosclera spp. dealt with here. We had the opportunity to revise the voucher of Thacker et al.'s (Op. cit.) A. heroni (NCI 198), and found it to bear a clearly callyspongiid architecture, with a neat reticulation of spicule- and debris-cored fibers with abundant cementing spongin. However, our included C. vaginalis is inserted nowhere close to A. heroni. Rather, it groups with Cladocroce sp. with high support, and then, with moderate support to Haliclona implexiformis. Several clade A species reported in Redmond et al. (2013), namely C. molitba, Haliclona fascigera, H. manglaris, H. oculata, and H. tubifera integrate a larger clade including A. heroni with 87% bootstrap support. Still, the same two species of Neopetrosia (N. rosariensis and N. subtriangularis) and two species of Petrosia (P. strongylata and P. weinbergi), present in Redmond et al.'s (Op. cit.) clade A, are here placed in this same clade, and highly supported. Unfortunately, these authors were unsuccessful in retrieving an 18S sequence for A. heroni, which prevents further discussion on this genus' best phylogenetic assignment.

As pointed out above, all three Brazilian Arenosclera spp. clustered elsewhere, in a clade containing the type species of the niphatid genus Amphimedon. The same clade includes H. curacaoensis, a sponge of uncertain phylogenetic affinity (see above), and D, varia, allegedly a callyspongiid. The latter, given its lack of spicules, is at best taken as a possible callyspongiid. However, it is interesting to note that Redmond et al.'s (2013) clade C, including both niphatid species clustering with Brazilian Arenosclera spp., also includes Niphates erecta, the type species of Niphates. Thus, albeit indirectly so, reinforcing the niphatid affinity of these Arenosclera spp. Obviously, Arenosclera needs a thorough revision, considering not only the species above, but also other additional little-known species from the Indo-Pacific.

Table 1 lists the comparative haplosclerid materials used here. Specimens from Heron Island referred to, as A. heroni, by Muricy and Ribeiro (1999) were found to be better assigned to Dactylia, as not a single spicule could be found in them. Thus, aside from Pulitzer-Finali's (1982) original description, our comprehension of A. heroni's morphospace derives from hands-on study of Desqueyroux-Faúndez (1984) and Thacker et al.'s (2013) materials, alas, not topotypical. Both new species proposed appear quite distinct from this West Pacific species (Table 5) in overall habit (erect, somewhat lamellate), ectosomal architecture (commonly bearing uni- or pauci- to multispicular tracts), and usually considerably larger spicule dimensions (up to 112, 130, or 180 μm in the new species, against a maximum of 90 μm in A. heroni).

As simple as it is, these larger oxeas in the Amazon reefs' species set them apart from every other species previously assigned to Arenosclera (Table 5), where the maximum reported length for the oxeas was 108 μm in A. brasiliensis. The remaining Arenosclera spp. all possess oxeas smaller than 90 μm. Nevertheless, the paratype of A. amazonensis sp. nov. has comparatively smaller oxeas, thus coming quite close to the spectrum observable in A. brasiliensis. We are confident that both species are distinct as not a single (among hundreds observed by CVL and EH) A. brasiliensis possessed a lamellate-pedunculate body. Moreover, extensive surveys conducted along the Brazilian coast and its oceanic islands in the last couple decades, failed to spot any A. brasiliensis further north than 20°S, thus suggesting this species has a geographic range determined by the influence of SE Brazilian upwelling. This is in marked contrast to the Amazon environment were both new Arenosclera spp. were retrieved. While both new species proposed have flat, erect, somewhat lamellate morphologies, all other species considered bear massive to tubular habit (Table 5). In this way, we do not see risk of mistaken identifications, and the new species appear easy to recognize among other congeners.

Given the insertions of A. heroni, and the Brazilian Arenosclera spp. in our 28S phylogeny, monophyly of the remaining Arenosclera spp. is unlikely. This is an obvious assumption, consequence of the findings by McCormack et al. (2002), Raleigh et al. (2007), and Redmond et al. (2007, 2011, 2013), who have shown rampant polyphyletism of haplosclerid genera classified in every family. For instance, distribution of alkaloids such as arenosclerines and haliclonaciclamines cuts right through current genera and families. These compounds are present in A. brasiliensis (currently in the Callyspongiidae; Torres et al., 2000), Haliclona (Chalinidae; Charana et al., 1996) and Pachychalina (Niphatidae; Oliveira et al., 2007). It is thus clear that a major taxonomic revolution is necessary, unlikely to spare any such tool as detailed morphology, multi locus sponge genetics, metagenomics and metabolomics.

The use of metagenomes as a data source for phylogenetic analyses in place of the traditional approach of isolation, amplification and sequencing of individual markers, proved a good strategy. Despite its currently larger cost, the advantages of producing a database for the genes of a given species, to be assessed at any time, without the need for further primers, isolation, amplification and sequencing, far outweighs any drawbacks. The characterization aimed at for the biome off the mouth of the Amazon river included metagenome analyses of the water column, sediments and biota (Moura et al., 2016), so that several sponge metagenomes had already been generated independently of any foreseen taxonomic necessity.

A curious outcome of our metagenome dinucleotide dissimilarity analyses is the observed 10% dissimilarity in the six available metagenomes of A. brasiliensis (Trindade-Silva et al., 2012), which is of the same magnitude as the value retrieved for A. compressa and C. vaginalis, species currently assigned to distinct families. These three species exhibit 15% dissimilarity, while A. amazonensis sp. nov. appears as most dissimilar. Given that all species, aside A. brasiliensis, were collected in the Amazon reefs area, it appears that neither a phylogenetic, nor an ecological signal were determinants for the observed dissimilarity.

One way to look at dissimilarity focuses on its translation into the uniqueness of the associated microbiota. As such, the higher the dissimilarity, the greater the potential for the finding of novelties, which suggests good avenues for the investigation of the chemo-pharmacological applicability of the associated microbiota in A. amazonensis sp. nov.

On the other hand, it is remarkable that despite originating from the same short stretch of SE Brazilian rocky coast, the six specimens of A. brasiliensis yielded 10% dissimilar metagenomes. This reminds of the importance to consider multiple samples when characterizing metagenomes, which has unfortunately not been possible here for many species considered, due to the limited number of duplicate and triplicate specimens obtained in the Amazon cruises.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. The handling Editor declared a shared affiliation, though no other collaboration, with several of the authors, and the handling Editor states that the process met the standards of a fair and objective review.