1 Introduction

Mesophotic reefs encompass a depth range of approximately 30 to 150 m below the surface of the ocean. They are diverse and can be found in tropical and subtropical regions worldwide, particularly in areas where the ocean floor slopes steeply. Although they receive less light than shallower reefs, mesophotic reefs still support an array of photosynthetic organisms that provide the foundation of the mesophotic food web in addition to receiving energy from productive surface waters (Keesing et al. 2012; Kramer et al. 2022). Mesophotic reefs also often contain unique and diverse marine life adapted to the lower light levels found at these depths (Lesser et al. 2019). In a global study, Lesser et al. (2019) found evidence of a break at 60 m for multiple taxa marking a transition from shallow-water to mesophotic communities. The mesophotic zone also serves as a refuge for species that are threatened in shallower waters, and may contain previously unknown species (Rocha et al. 2018). Despite their potential ecological significance, mesophotic reefs remain understudied and poorly understood. Because they are difficult to access, much of what we know about these reefs comes from surveys using remotely operated vehicles, submersibles, technical divers or dredging (Bongaerts et al. 2015; Glasl et al. 2017; Tamir et al. 2019; Ferrier-Pagès et al. 2022; Cleary et al. 2023a).

Mesophotic reefs have received increasing attention in recent years due to their potential importance as refuges for marine life (Bridge et al. 2013; Rocha et al. 2018; Bongaerts and Smith 2019). They are often thought of as a buffer zone for shallower reefs, acting as a source of coral larvae and other organisms that can help repopulate damaged or destroyed shallow reefs. They also provide habitats for a range of marine species, including corals, crustaceans, and sponges, many of which are unique to these deeper waters (Bongaerts et al. 2013; Estronza et al. 2017; Macartney et al. 2021). For example, in a recent study, 31 sponge species were discovered and thirteen were described from the mesophotic zone of Bonaire and Klein Curaçao using the manned ‘Curasub’ submersible (van Soest et al. 2014). The majority of the listed sponge species appear to be restricted to the mesophotic reef zone.

Species inhabiting both shallow-water and mesophotic zones may also differ in key traits. In the Mediterranean, Idan et al. (2020), for example, showed that shallow-water and mesophotic populations of the sponge species Chondrosia reniformis and Axinella polypoides both adapted their reproduction to changes in temperature and depth. In shallow waters, both species showed seasonality, indicating that temperature affected their reproductive activities. However, in the mesophotic zone where the temperature changes are much smaller, neither species demonstrated seasonality. Chondrosia reniformis had lower fecundity in the mesophotic zone and likely invested more in asexual reproduction by budding. On the other hand, A. polypoides in the mesophotic zone had enhanced and continuous asynchronous sexual reproduction throughout the year. Sponges found in mesophotic zones also tend to have higher proportions of heterotrophic as opposed to phototrophic symbionts, which can absorb dissolved organic matter from the water column rather than relying on light energy (Olson and Kellogg 2010; Olson and Gao 2013; Morrow et al. 2016; Steinert et al. 2016; Macartney et al. 2021; Cleary et al. 2023a).

Sponges and their microbial symbionts have been widely studied as models for host-microbe interactions in the marine environment (Taylor et al. 2007; Hentschel et al. 2012; Cleary et al. 2013, 2015, 2018, 2019a, b; Thomas et al. 2016; Swierts et al. 2018; Busch et al. 2022). Earlier studies showed marked differences in microbial densities among different sponge species and initially referred to sponges with high microbial densities as’bacterial’ sponges and those with low microbial densities as’non-symbiont’ sponges, which later became’high microbial abundance’ (HMA) and’low microbial abundance’ (LMA) sponges. The differences in microbial densities between both groups can be substantial varying from 105—106 microbial cells/g of tissue in LMA sponges to 108—1010 microbial cells/g of tissue in HMA sponges (Hentschel et al. 2006).

Sponge microbiome studies also showed that in addition to differing in abundance, bacterial communities of LMA sponges were less diverse and compositionally more similar to bacterial communities found in seawater (Thacker and Freeman 2012; Blanquer et al. 2013; Moitinho-Silva et al. 2014; Bayer et al. 2014). This led to speculation that these communities were mainly, or exclusively, derived from seawater due to the pumping activities of the sponges and that there was no resident bacterial community to speak of. This was contrasted with the diverse bacterial communities found in HMA species and their marked differences with the bacterioplankton and bacterial communities from LMA sponges (Schmitt et al. 2011; Erwin et al. 2015; Chaib De Mares et al. 2017; Cleary et al. 2019a; Cleary et al. 2020; 2021). Subsequent studies, however, have shown that LMA species may host bacterial communities that are similar to, but not identical to those found in water. Multi-taxon studies have also shown that there can be considerable compositional variation among LMA species (Cleary et al. 2020; 2021; Busch et al. 2022).

HMA species, in contrast to LMA species, tend to house less variable bacterial communities with less variation within and among species than found among LMA species. Several studies have, furthermore, noted the pronounced over-representation of certain phyla in HMA sponges (Bayer et al. 2014; Cleary et al. 2015; de Voogd et al. 2015, 2019; Polónia et al. 2015; Moitinho-Silva et al. 2017). These include the phyla Chloroflexi, Poribacteria, Actinobacteriota, Acidobacteriota, PAUC34f, and Gemmatimonadota (Bayer et al. 2014; Moitinho-Silva et al. 2017). These phyla are much less prevalent, or even absent, in typical LMA species, which instead are usually dominated by Proteobacteria, or cyanobacteria in the case of cyanosponges such as Lamellodysidea herbacea and Terpios hoshinota (Podell et al. 2020; Rützler and Muzik 1993). Finally, it should be noted that various studies have found LMA and HMA sponges to differ in other traits, particularly in relation to morphology and physiology. For example, LMA sponges tend to pump greater volumes of water and have larger aquiferous canals and choanocyte chambers than HMA sponges whereas HMA sponges tend to have higher mesohyl densities. Certain polyketide synthase genes have also only been found in HMA sponges (Vacelet and Donadey 1977; Weisz et al. 2007; Hochmuth et al. 2010). Despite the above, it is not always possible to clearly assign HMA or LMA status based on electron microscopy, microbial composition based on 16S rRNA gene sequencing, or morphological traits.

Although well-studied in shallow waters, fewer studies have assessed the prevalence of HMA and LMA sponges in deeper waters. Most of the sponges, which have been assigned HMA or LMA status (e.g., Gloeckner et al. 2014; Moitinho-Silva et al. 2017), thus far, are shallow-water species with a few extending to the mesophotic zone such as Cinachyrella kuekenthali or true deep-sea species such as Geodia barretti, although its status appears to have been only based on amplicon sequencing data (Radax et al. 2012). Recently, Busch et al. (2022) also provided evidence of the HMA/LMA dichotomy in the deep sea in addition to identifying a novel type of LMA community in glass sponges. However, the deep sea, in contrast to the mesophotic zone, is truly dark and although the deep sea appears even less accessible than the waters up to 150 m, more studies have been carried out on the microbiomes of deep sea sponges (Jackson et al. 2013; Kennedy et al. 2014; Steinert et al. 2016; Tian et al. 2016; Busch et al. 2022; Steffen et al. 2022; Garritano et al. 2023) than those of mesophotic sponges (e.g., Steinert et al. 2016; Indraningrat et al. 2022). Studies of sponge microbiomes in the mesophotic zone have, furthermore, focused on single or limited sets of species and the impact of depth on microbiome composition. For instance, specimens of Giant Barrel sponges (Xestospongia muta) in the mesophotic zone were enriched with Chloroflexi as opposed to Cyanobacteria in shallower waters (Cleary et al. 2023a). To the best of our knowledge, no study has, hitherto, compared the microbiomes of multiple sponge species only from the mesophotic zone. Our study, thus, complements previous studies of the prevalence of the HMA/LMA dichotomy in shallow and deep-sea environments.

In the present study, 26 sponge samples were collected using a submersible at two sites off the islands of Curaçao and Klein Curaçao along depths ranging from 90 to 188 m (note a single sample was collected at 13 m depth). Here, we provide the first in-depth assessment of the prokaryotic communities of 11 sponge species from the mesophotic zone of Curaçao. Most of these sponges have, furthermore, only been recorded from the mesophotic zone (van Soest et al. 2014). The main hypothesis of the present study was to test if the HMA/LMA dichotomy holds for mesophotic zone sponges. HMA or LMA status was assigned to species based on TEM imaging and HMA/LMA status was used as a predictor for variation in sponge-associated prokaryotic composition. In addition to the above, we used an exploratory technique based on machine learning to identify taxonomic indicators of HMA or LMA status.

2 Material and methods

2.1 Sample collection and sites

Curaçao is a small country in the Caribbean, approximately 65 kms from the Venezuelan coast, with a surface area of ~ 444 km2. It consists of the main island of Curaçao and a small, uninhabited island, Klein Curaçao. In the present study, samples were collected off the southern coast of the main island of Curaçao and off of Klein Curaçao with the Curasub submersible from August 29th till October 31st, 2018 (Fig. 1 and Supplementary Data 1). All samples, but one, were collected between 90 and 188 m depth; a single specimen of Xestospongia muta was collected at 13 m depth. The Curasub (Fig. 2) was operated from either the support vessel RV Chapman or directly from the substation at the Curaçao Sea Aquarium. Sponges were visually located and photographed with a GoPro black 7 or Olympus G6 from inside the Curasub. Samples were collected using manipulator arms. The sponges were placed into a basket in front of the Curasub during the dive. Upon surfacing, sponge samples were transferred from the collecting basket to buckets filled with seawater. In the laboratory on board of RV Chapman, the sponges were photographed and a cylindrical section of approximately 1 by 2 cm was cut from each specimen containing both the ectosome and choanosome, and placed in a vial containing 96% alcohol or fixed in a mixture of 0,5% formalin, 1,25% glutaraldehyde in 0.22 uM filtered sea water and transferred to a saline buffer after 48 h. Care was taken to include the surface and interior of each sponge segment. The vials were shipped to the Netherlands and stored at -20 ºC until DNA extraction. The sponges were identified by Nicole de Voogd using classical morphological characters. Voucher specimens have been deposited in the sponge collection of Naturalis Biodiversity Center (as RMNH POR#).

Fig. 1
figure 1

Map of study area showing location of sample sites (blue circles). The inset shows the location of Curaçao with respect to South America

Full size image
Fig. 2
figure 2

A. View of the submersible “Curasub”, B. view from the interior of submarine at a depth of 120 m, C. Calthropella lithistina attached to a steep wall, D. specimen of Neophrissospongia jorgeorum after collection, E. specimen of Acanthella cubensis after collection, and F. fragments of 3 individuals of Petrosia aff. weinbergi after collection

Full size image

The species investigated in the present study belonged to 5 orders and 11 species. The species sampled were Agelas flabelliformis (Carter, 1883), Phakellia folium Schmidt, 1870, Acanthella cubensis (Alcolado, 1984), Neopetrosia ovata van Soest et al. 2014, Petrosia sp., Petrosia aff. weinbergi van Soest, 1980, Xestospongia muta (Schmidt, 1870), Calthropella (Pachataxa) lithistina (Schmidt, 1880), Cinachyrella kuekenthali (Uliczka, 1929), Corallistes typus Schmidt, 1870, and Neophrissospongia jorgeorum Dias, Kelly & Pinheiro, 2023 (Fig. 2). Between 1 and 3 samples were collected per species.

2.2 Transmission electron microscopy

Several pieces of fixed sponge material were cut into approximately 1 × 1 × 2 mm pieces. These were rinsed three times with 0.1 M sodium cacodylate buffer, fixed in 2% osmium tetroxide in the dark for two hours, and rinsed again three times with 0.1 M sodium cacodylate buffer. Subsequently, the material was stained in the dark for an hour in 3% uranyl acetate (aqueous solution). The samples were dehydrated in a series of ethanol concentrations (30%, 50%, 70%, 96%, and twice in absolute ethanol). After dehydration, the material was placed into acetonitrile twice for 15 min and then in a 1 to 1 epon 812 acetonitrile solution overnight. The next day, the epon: acetonitrile solution was refreshed and the mixture was centrifuged slowly for 3 h at a speed of ~ 2 RMP for three hours. Thereafter, the samples were placed twice in a vacuum for 20 min and afterwards in a stove at 60 °C for 48 h. After the embedding process, three separate samples were sectioned into thin slices (80 nm) with a microtome and post-stained for 20 min with uranyl acetate and for 15 min with lead citrate. The samples were imaged using transmission electron microscopy (JEM 1400 Plus, JEOL Ltd., Tokyo, Japan).

2.3 DNA extraction and next-generation sequencing analysis

DNA was extracted using the Qiagen DNeasy Powersoil extraction kit (Qiagen, Venlo, the Netherlands). Approximately 250 mg of tissue was used from each sponge sample; tissue was taken from all sides of the specimen (outside to core, and if applicable top, middle and bottom of the sample). The manufacturer’s protocol was followed with the exception of the initial vortexing step, which was carried out using the Qiagen TissueLyser II (Qiagen NV, Hilden, Germany). Sponge tissue was cut into small pieces using sterilised tweezers and scalpel blades, and transferred to PowerBead Pro tubes containing ceramic and silica beads of different sizes. An extraction blank, in which no tissue was added to the PowerBead Pro tubes, was also included. The library preparation was conducted using a two-step PCR protocol for all samples in addition to two negative controls (mQ water instead of template DNA) and the extraction blank. For the first PCR, the V3-V4 regions of the 16S rRNA gene were targeted and amplified using the primers 314F/ S-D-Bact-0785-a-A-21 (5’-CCTACGGGNGGCWGCAG-3’/5’-GACTACHVGGGTATCTAATCC-3’; Klindworth et al. 2013) with added 5’ Nextera transposase adaptors using the KAPA HiFi HotStart Ready Mix PCR Kit with a T100 Thermal Cycler (Bio-Rad, Hercules, CA, United States). The following PCR conditions were used: initial denaturation at 95 °C for 3 min, 30 cycles of denaturation at 98 °C for 20 s, annealing at 55 °C for 30 s, followed by extension at 72 °C for 30 s. The final extension was carried out at 72 °C for 1 min. PCR success was confirmed on an E-Gel™ (agarose gels at 2%), and the absence of amplification was validated for the negative controls and the extraction blank. PCR products were then cleaned with NucleoMag NGS-Beads (bead volume at 0.9 times the total volume of the sample, Macherey Nagel, Düren, Germany) using the VP 407AM-N 96 Pin Magnetic Bead Extractor stamp (V&P Scientific, San Diego, CA, United States). For the second PCR, the cleaned PCR products (1 µL each) were amplified and labelled using the MiSeq Nextera XT DNA library preparation kit (Illumina, San Diego, CA, United States) with the same thermal cycling scheme limited to 8 cycles. PCR products were then analysed with the Fragment Analyser Agilent 5300 using the DNF-910–33 dsDNA Reagent Kit (35–1,500 bp) protocol (Agilent Technologies, Santa Clara, CA, United States) to confirm successful labelling of the DNA fragments. Negative controls and extraction blanks remained negative after this step. Pooling at equimolar concentration was performed with QIAgility 2 (Qiagen, Hilden, Germany). The final pool was then cleaned with NucleoMag NGSBeads, eluted in Milli-Q and the DNA concentration was verified using Tapestation 4150 (Kit HSD 5000, Agilent Technologies, Santa Clara, CA, United States). Paired-end sequence reads were generated with an Illumina MiSeq v3 PE300 platform at BaseClear B.V. (Leiden, Netherlands). FASTQ read sequence files were generated using bcl2fastq version 2.20 (Illumina). Initial quality assessment was based on data passing the Illumina Chastity filtering. Subsequently, reads containing PhiX control signal were removed using an in-house filtering protocol. In addition to this, reads containing (partial) adapters were clipped (up to a minimum read length of 50 bps). The second quality assessment was based on the remaining reads using the FASTQC quality control tool version 0.11.8. The DNA sequences generated in this study can be downloaded from NCBI BioProject Id: PRJNA715750.

2.4 16S rRNA gene sequencing analysis

The 16S rRNA amplicon libraries were analysed using QIIME2 (version 2019.7; Bolyen et al. 2019). Raw data were imported yielding a demultiplexed ‘qza’ data file (artifact). The DADA2 plugin (Callahan, et al. 2016) in QIIME 2 was subsequently used to trim sequences (final length 400 nt). The DADA2 analysis yielded output archives containing an OTU (at a 100% similarity threshold, also known as amplicon sequence variant or ‘ASV’) table, denoising stats, and a fasta file of representative sequences. The feature-classifier plugin with the extract-reads method was then used with the i-sequences argument set to silva-138–99-seqs.qza. This was followed by the feature-classifier plugin with the fit-classifier-naive-bayes method and the i-reference-taxonomy method set to silva-138–99-tax.qza. Both silva-138 files can be obtained from https://docs.qiime2.org/2020.8/data-resources/?highlight=silva. The feature-classifier plugin was then used with the classify-sklearn method and the i-reads argument was set to the representative sequences file generated by the DADA2 analysis to produce a table with taxonomic assignments for all OTUs. Finally, mitochondria, chloroplasts, and Eukaryota were filtered out using the QIIME 2 taxa plugin with the filter-table method. All OTUs unclassified at Domain and Phylum level were also removed. The final dataset consisted of 333005 sequences and 4094 OTUs. The OTU and taxonomy tables were later merged in R 4.1.2 (R Core Team 2021). Supplementary Data 2 contains all OTU counts per sample and taxonomic assignments of all OTUs. The 50 most abundant OTUs overall, thus including all samples, are presented in Supplementary Data 3.

2.5 BLAST

Accession numbers of closely related organisms to selected OTUs were obtained using NCBI Basic Local Alignment Search Tool (BLAST) (https://www.ncbi.nlm.nih.gov/books/NBK62051/). In the present study, the sequence similarities ranged from 87.82% to 100% (Supplementary data 3).

2.6 Statistical analyses

The OTU counts table was used in R to compare diversity and higher taxon abundance among sponge host species. Diversity indices, namely, rarefied richness, evenness, Shannon's H’ and Fisher’s alpha, were obtained using the rarefy and diversity functions from the vegan 2.6–4 (Oksanen et al. 2022) package in R. Variation in prokaryotic composition among sponge host species was visualised with Principal Coordinates Analysis (PCO). For the PCO, a Bray–Curtis distance matrix was first obtained using the phyloseq 1.44 package (McMurdie and Holmes 2013) whereby the count data were rarefied using the rarefy_even_depth function with the sample.size argument set to the minimum sample size (12206 in the present study) and subsequently log10 transformed. The ordinate function in phyloseq was used to run the PCO analysis and generate data for the ordination. A heatmap was produced showing variation in OTU abundance using the pheatmap function from the pheatmap 1.0.12 package (Kolde 2019).

Sponge species were assigned HMA or LMA status based on a qualitative assessment of TEM images. We tested if HMA/LMA status was a significant predictor of prokaryotic composition using the adonis function from the vegan 2.6–4 package in R. We subsequently used the multipatt function from the indicspecies package in R (De Cáceres and Legendre 2009) and an exploratory technique based on machine learning, Boruta (Kursa et al. 2010), to identify bacterial classes, which distinguished between HMA and LMA sponges. Prior to the Boruta analysis, highly correlated classes (> 0.8) were removed from the data frames. The Boruta function from the Boruta package in R (Kursa and Rudnicki 2010) was then used with the microbial abundance status (HMA versus LMA) as response variable and bacterial class as feature (predictive variable). The doTrace argument in the Boruta function was set to 2 and the maxRuns argument set to 1000; other arguments used default values. A more detailed description of the Boruta method can be found in Cleary et al. (2023b).

3 Results

In the present study, 10 out of the 11 sponge species were investigated using TEM imaging to visualise microbial density (Fig. 3). The species N. ovata, C. lithistina, C. typus, Petrosia aff. weinbergi, Petrosia sp. and N. jorgeorum contained dense and morphologically diverse microorganisms indicative of HMA status (Supplementary Fig. 1). We did not investigate X. muta because multiple specimens have already been investigated using TEM and its HMA status is well established. In contrast to the above, the species A. cubensis, P. folium, and C. kuekenthali, contained relatively low microbial densities indicative of LMA status (Supplementary Fig. 2). Although A. flabelliformis contained relatively large bacterial cells, the density of these cells was low, which also indicates LMA status (Supplementary Fig. 2). As evident in Fig. 3, HMA and LMA sponges separated along the first axis of a PCO ordination and HMA status was a highly significant predictor of prokaryotic composition (adonis: F = 4.529, P < 0.001, R2 = 0.751).

Fig. 3
figure 3

Ordination showing the first two axes of the principal coordinates analysis (PCO) of OTU composition. Symbols are colour-coded and represent samples from different sponge host species as shown in the legend on the right side of the figure. Grey symbols represent weighted averages scores for OTUs. The symbol size is proportional to abundance (number of sequence reads). The legend symbols represent the following sponge species: Af: Agelas flabelliformis, Pf: Phakellia folium, Au: Acanthella cubensis, No: Neopetrosia ovata, Pp: Petrosia sp., Pw: Petrosia aff. weinbergi, Xm: Xestospongia muta, Ca: Calthropella lithistina, Ck: Cinachyrella kuekenthali, Ct: Corallistes typus, and Np: Neophrissospongia jorgeorum. TEM images of all species, with the exception of X. muta, are placed adjacent to the sponge host species samples

Full size image

LMA sponge species were also less diverse in terms of richness and evenness than HMA sponge species although the difference was less pronounced for A. flabelliformis (Fig. 4). The most abundant phyla across all samples were Proteobacteria (99256 sequences, 967 OTUs), Chloroflexi (76186 sequences, 879 OTUs), Actinobacteriota (43541 sequences, 306 OTUs), and Acidobacteriota (27074 sequences, 423 OTUs) (Supplementary Data 2). In terms of bacterial class abundance (Fig. 5), all HMA species, with the exception of N. jorgeorum, were dominated by Dehalococcoidia (Chloroflexi). 98.8% of all sequences classified as Dehalococcoidia, furthermore, belonged to the SAR202 clade. The LMA sponge species were more variable as pertains to Class relative abundance. Agelas flabelliformis, for example, was similar to the HMA sponge species with respect to class representation. Cinachyrella kuekenthali, in contrast, was dominated by Acidimicrobiia members followed by Gammaproteobacteria and Dadabacteriia. Both A. cubensis and P. folium were dominated by Gammaproteobacteria, with this most pronounced in A. cubensis. Phakellia folium, in turn, had the greatest Nitrospiria component.

Fig. 4
figure 4

Variation in richness and evenness among sponge species samples: The legend symbols represent the following sponge species: Af: Agelas flabelliformis, Pf: Phakellia folium, Au: Acanthella cubensis, No: Neopetrosia ovata, Pp: Petrosia sp., Pw: Petrosia aff. weinbergi, Xm: Xestospongia muta, Ca: Calthropella sp., Ck: Cinachyrella kuekenthali, Ct: Corallistes typus, and Np: Neophrissospongia jorgeorum

Full size image
Fig. 5
figure 5

Stacked barplots of the mean relative abundances of the eight most abundant bacterial classes. The x-axis codes represent the following sponge species: Af: Agelas flabelliformis, Pf: Phakellia folium, Au: Acanthella cubensis, No: Neopetrosia ovata, Pp: Petrosia sp., Pw: Petrosia aff. weinbergi, Xm: Xestospongia muta, Ca: Calthropella sp., Ck: Cinachyrella kuekenthali, Ct: Corallistes typus, and Np: Neophrissospongia jorgeorum. The bars are colour-coded according to the bacterial Class as indicated by the legend

Full size image

A heatmap of the 50 most abundant OTUs (representing 37.6% of all sequences) is shown in Fig. 6 including dendrograms of host species and OTUs based on hierarchical clustering. All host sponge species clustered according to HMA/LMA status. Supplementary Data 3 contains a list of these OTUs including taxonomic assignments and the results of the BLAST analyses. Nine OTUs, classified as Acidobacteriota, Spirochaetota, Alphaproteobacteria, SAR202 clade, and Nitrosococcales were enriched in HMA sponge species and formed a cluster in Fig. 6. All of these OTUs had high sequence similarities (> 99%) to organisms detected across a range of sponge species (Supplementary Data 3) including Aplysina cauliformis, Astrosclera willeyana, Ircinia oros, Ircinia strobilina, and Xestospongia testudinaria.

Fig. 6
figure 6

Heatmap of the 50 most abundant OTUs. Dendograms are based on hierarchical clustering of sponge host species and OTUs. The x-axis codes represent the following sponge species: Af: Agelas flabelliformis, Pf: Phakellia folium, Au: Acanthella cubensis, No: Neopetrosia ovata, Pp: Petrosia sp., Pw: Petrosia aff. weinbergi, Xm: Xestospongia muta, Ca: Calthropella Ck: Cinachyrella kuekenthali, Ct: Corallistes typus, and Np: Neophrissospongia jorgeorum. Abundance is shown on a log10 scale as indicated by the legend at the right of the figure

Full size image

Other sets of OTUs were restricted to specific host sponge species as shown in Fig. 6. In the HMA sponge species, four OTUs, for example, were restricted to Petrosia aff. weinbergi, which had high sequence similarities to organisms detected in one coral species (Porites lutea), and three sponge species (X. muta, Ircinia felix, and I. strobilina) (Supplementary Data 3). Five OTUs, classified as Actinobacteriota, Dadabacteria, and Nitrospirota were restricted to the lithistid sponges C. lithistina and C. typus. and had high sequence similarities (> 99%) to organisms detected in the sponges Characella sp., Corallistes sp., Desmacidon sp., and Geodia barretti.

In the LMA sponge species, seven OTUs, classified as Acidobacteriota, Actinobacteriota, Dadabacteria, Entotheonellaeota and Myxococcota, were restricted to A. flabelliformis. These had high sequence similarities to organisms obtained from host sponges identified as Astrosclera willeyana, Geodia pachydermata, Agelas dilatata, Xestospongia exigua, and Isops phlegraei. Three OTUs, classified as Gammaproteobacteria and Nitrospirota, were restricted to P. folium. and were similar to organisms obtained from the sponge species Axinella verrucosa and Axinella corrugata in addition to an organism obtained from a grass carp pond in China, but with a sequence similarity of only 88.3%. Four OTUs, classified as Gammaproteobacteria, genus Nitrosopumilus, and Bdellovibrionota, were restricted to A. cubensis. and were similar to organisms obtained from the sponge Raspailia topsenti and the coral Favia sp. (Fig. 6 and Supplementary Data 3).

The Boruta analysis yielded 14 significant class-level predictors of which eight with the greatest importance values are presented in Fig. 7. The rest of the significant class-level predictors are presented in Supplementary Fig. 3. Results of the Boruta and IndVal (‘indicspecies’) analyses are provided in Supplementary Data 4 and 5. Six of the eight class-level predictors were also significant using the IndVal approach as outlined in the methods section, the exceptions being Dehalococcoidia and the BD2-11 terrestrial group. In contrast, Subgroup 21, Subgroup 11, JG30-KF-CM66, and Thermoanaerobaculia, with the greatest importance values in the Boruta analysis, were all highly significant (P < 0.001) predictors based on the IndVal analysis. All presented class-level predictors, with the exception of Bdellovibrionia, were more abundant in HMA sponge species.

Fig. 7
figure 7

Relative abundances of bacterial classes, which differentiated between HMA and LMA sponge species in order of importance. The x-axis codes represent the following sponge species: Af: Agelas flabelliformis, Pf: Phakellia folium, Au: Acanthella cubensis, No: Neopetrosia ovata, Pp: Petrosia sp., Pw: Petrosia aff. weinbergi, Xm: Xestospongia muta, Ca: Calthropella sp., Ck: Cinachyrella kuekenthali, Ct: Corallistes typus, and Np: Neophrissospongia jorgeorum

Full size image

4 Discussion

The results of the present study using TEM imagery and next-generation sequencing clearly demonstrate the prevalence of the HMA/LMA dichotomy in sponges inhabiting the mesophotic zone of Curaçao. We, furthermore, identified pronounced variation in diversity, and higher taxon abundance, in addition to identifying closely related organisms to the abundant OTUs in selected sponge species. Our results also showed that HMA or LMA status based on TEM imagery is a highly significant predictor of prokaryotic composition. For example, the first axis of the PCO ordination separated sponge species identified as HMA using TEM imagery in addition to the known HMA X. muta (Gloeckner et al. 2014) from species identified as LMA, namely Phakellia folium, A. cubensis, C. kuekenthali, and A. flabelliformis. In addition to the above, the HMA species also had higher richness and evenness than the LMA sponge species although this difference was less pronounced for A. flabelliformis. We previously showed (Cleary et al. 2020, 2021) that evenness was consistently higher in HMA than in LMA species. The HMA species identified in the present study also shared OTUs closely related to organisms detected in known HMA sponge species such as Aplysina cauliformis, Ircinia spp., Astrosclera willeyana, and Xestospongia testudinaria (Karlińska-Batres and Wörheide 2013; Gloeckner et al. 2014; Cleary et al. 2019a). All of these characteristics strongly indicate HMA status, as confirmed using TEM imagery for the species N. ovata, Petrosia sp., Petrosia aff. weinbergi, C. lithistina, C. typus and N. jorgeorum.

Neopetrosia ovata, Petrosia sp. and Petrosia aff. weinbergi belong to the family Petrosiidae (order Haplosclerida) and several other species belonging to this family have been identified as HMA sponges (e.g., Petrosia ficiformis, Petrosia elephantotus, Xestospongia muta, and Xestospongia testudinaria; Gloeckner et al. 2014; Moitinho-Silva et al. 2017). Although sponges belonging to other families (Chalinidae, Callyspongiidae, Niphatidae) within the order Haplosclerida appear to exclusively consist of LMA species, the Petrosiidae appears to only consist of HMA sponges. Sponges within this family are often large, massive and conspicuous as opposed to the soft bodied representatives of other haplosclerid families.

The species C. lithistina (Family Geodiidae), N. jorgeorum, and C. typus (both Family:Corallistidae) all belong to the order Tetractinellida. All three species, furthermore, shared several abundant taxa including OTUs classified as Chloroflexi and Acidobacteriota, which are both HMA indicator taxa (Moitinho-Silva et al. 2017). In comparison to other host species, the tetractinellids, including C. kuekenthali, had relatively high abundances of OTUs classified as Dadabacteria although each species had a distinct abundant dadabacterial OTU (see Fig. 6). Marine pelagic members of the phylum Dadabacteria were shown to have smaller genome sizes, lower nitrogen content and the genomic potential to degrade microbial DOM, particularly peptidoglycans and phospholipids (Graham and Tully 2021). Graham and Tully (2021) also identified distinct clades with the potential for (or lack thereof) photoheterotrophy via proteorhodopsin.

Samples of C. lithistina, C. typus, and N. jorgeorum clustered close to the petrosid N. ovata but were distinct from samples of the other petrosid species. Likewise in Taiwan, tetractinellid species sampled from caves housed prokaryotic communities with high evenness and representation of HMA indicator taxa, but which were compositionally distinct from prokaryotic communities of the petrosid Xestospongia testudinaria sampled from nearby coral reef habitat (Cleary et al. 2024). Recently, three new species belonging to the genus Neopetrosia, including N. ovata, were described from mesophotic reefs of Bonaire and Klein Curaçao (van Soest et al. 2014). These species coexist sympatrically with the giant barrel sponge X. muta, and all belong to the family Petrosiidae differing in spicule size and shape among other traits. We collected our specimens from the same location as the type material.

Dehalococcoidia abundance was particularly high in HMA sponge species and varied from 18.9 ± 3.84% in Petrosia aff. weinbergi to 34.97 ± 10.56% in N. ovata and was much lower in LMA sponge species. Most OTUs classified as Dehalococcoidia belonged to the SAR202 clade (98.8%). Landry et al. (2017) suggested that SAR202 members obtained from the dark ocean were involved in the oxidation of relatively recalcitrant to persistent refractory organic compounds. Likewise, Saw et al. (2020) showed that the genomes of SAR202 MAGs at varying depths encoded for enzymes involved in the oxidation of compounds resistant to degradation due to chiral complexity. In sponges, Bayer et al. (2018) also suggested that sponge-associated Chloroflexi, including SAR202 members, were involved in the degradation of labile and recalcitrant DOM. Taken together, these results, and the high relative abundance of SAR202 in the present study, suggest a potentially important role in the degradation of recalcitrant DOM compounds in mesophotic HMA sponges and possibly a greater dependence on DOM for sponges in the mesophotic zone than in the shallow-water realm, although this requires further validation. It is, furthermore, of interest to note that all sponge species, including the LMA sponges albeit in lower abundances, harboured Dehalococcoidia members in contrast to shallow-water studies (Cleary et al. 2020, 2021). Cyanobacterial abundance, in contrast, was only 0.21 ± 0.52% across all species and six of the seven HMA species had no recorded cyanobacterial sequences.

In contrast to the above-mentioned HMA species, the sponge species P. folium, A. cubensis and C. kuekenthali had relatively low evenness. Phakellia folium and A. cubensis also had very high abundances of small subsets of OTUs, which had relatively low sequence similarities to organisms detected in corals, other sponge species and a grass carp pond. These characteristics suggest LMA status. This same pattern has been previously observed in other LMA sponge species where we showed that they were dominated by small subsets of highly abundant OTUs in addition to large numbers of OTUs shared with seawater (Cleary et al. 2019a, 2020, 2021). This contrasts with typical HMA species, which shared large numbers of OTUs, which were often rare or absent in seawater. The LMA species also hosted abundant taxa with low sequence similarities to sequences in the NCBI nucleotide database with each LMA species hosting its own unique small subset of OTUs.

LMA species have previously been characterised by the dominance of Proteobacteria, Cyanobacteria and/or Bacteroidetes (Moitinho-Silva et al. 2017). In contrast, however, to P. folium and A. cubensis, which were dominated by Gammaproteobacteria, C. kuekenthali was dominated by Acidimicrobiia (Actinobacteriota), which is considered a HMA indicator taxon, and Dadabacteriia. In a screening study of antimicrobial and anticancer activity in marine sponges, Proteobacteria and Actinobacteriota (or Actinomycetota) were shown to be the most prominent producers of bioactive compounds (Gavriilidou et al. 2021). Actinobacterial members have also been shown to be prominent producers of biosurfactants (Freitas-Silva et al. 2022) and potentially play a role in promoting heavy metal tolerance in their host sponge (Bauvais et al. 2015). Importantly, species of Cinachyrella and the related Paratetilla are often found in areas subject to sediment deposition and river plumes as are various HMA species such as X. testudinaria (Cleary and de Voogd 2007; de Voogd et al. 2009; Moura et al. 2016), which should promote a microbiome enriched in genes involved in heavy metal tolerance and DOM utilisation.

The most abundant OTUs in P. folium were classified as the gammaproteobacterial UBA10353 marine group and genus Nitrospira (Nitrospirota). The Nitrospirota phylum includes nitrite-to-nitrate oxidisers and commamox oxidisers, which oxidise ammonia to nitrate (Daims et al. 2016). The abundance of Nitrospira members in P. folium suggests a potentially important role of its sponge-associated prokaryotic community in nitrification. Pajares et al. (2020) previously showed that the tropical Mexican Pacific oxygen minimum zone was dominated by SAR11, SAR406, SAR324, SAR202, Thiomicrospirales, Nitrospinales and the UBA10353 marine group. In contrast, the euphotic zone was dominated by Synechococcales, Flavobacteriales, Verrucomicrobiales, Rhodobacterales, SAR86, and Cellvibrionales. Rust et al. (2020) previously identified a bin in the sponge Mycale hentscheli, affiliated with the UBA10353 marine group. They noted that multiple members of the sponge microbiome, including the UBA10353 bin contributed to the chemical diversity of the sponge. UBA10353 marine group members have also been recorded in a deep sea Fe–Mn crust biofilm (Bergo et al. 2021) and an oxygen minimum zone in the Andaman Sea (Guo et al. 2022).

Despite housing a prokaryotic community similar in certain aspects to HMA sponge species (e.g., in terms of diversity and class abundance), the sponge A. flabelliformis, clustered with the LMA species in the PCO ordination, and did not share any abundant OTUs with the HMA sponge species. Instead, it housed nine abundant OTUs of which only two were shared with another sponge species (A. cubensis). Although Gloeckner et al. (2014) assigned species of Agelas, including A. citrina, A. dilatata, A. dispar, A. wiedenmayeri, A. cervicornis, A. conifera, and A. oroides, HMA status, they did report that these species contained less dense microbial communities compared to other HMA species. The TEM images of this species clearly showed large bacterial cells, but these were not densely packed as shown in the images of the HMA species. Cleary et al. (2019a, b) also previously remarked that species of Agelas sampled in Thailand and Taiwan clustered in between HMA and LMA species in an ordination. The results of the present study, thus, confirm previous studies suggesting that species of Agelas house prokaryotic consortia distinct from both HMA and LMA species.

We were able to identify a number of class-based predictors of HMA/LMA status. The Dehalococcoidia and JG30-KF-CM66 are both classes within the phylum Chloroflexi, which has been identified by various studies as an important indicator phylum of HMA status (Schmitt et al. 2011; Moitinho-Silva et al. 2017). Sequences belonging to the the BD2-11 terrestrial group have previously been recorded across a range of biotopes including marine sediments, sponges, and soils (Hanada and Sekiguchi 2014). The BD2-11 terrestrial group is also a class of the Gemmatimonadota phylum (according to the SILVA database used), which has been recognised as an important indicator phylum of HMA status (Moitinho-Silva et al. 2017). Members of the recently described Thermoanaerobaculia class have previously been implicated in sulfur cycling (Flieder et al. 2021). The Thermoanaerobaculia, Subgroup 11, and Subgroup 21 clades all belong to the Acidobacteriota phylum, which has previously been identified as a HMA indicator phylum (Moitinho-Silva et al. 2017). Acidobacteriota members are important components of the soil bacterial community where they have been shown to be heterotrophic, often oligotrophic and sensitive to shifts in pH (Fierer et al. 2007, 2012; Jones et al. 2009). They have also previously been suggested to be important indicators of wetland restoration and trophic status (Hartman et al. 2008). Sponge-specific clusters were also shown to be ‘particularly well represented’ among certain bacterial taxa including the Acidobacteriota, in addition to Chloroflexi, Cyanobacteria, Poribacteria, and Betaproteobacteria (Simister et al. 2012). In the sponge Rhopaloeides odorabile, Acidobacteriota (and Chloroflexi) members also appeared to be sensitive to elevated temperature under which conditions they were replaced by Proteobacteria, Bacteroidetes and Firmicutes (Webster et al. 2008). Future work should focus on trying to unravel the role that Acidobacteriota members play in marine sponges and the nature of their association with HMA sponges in particular.

5 Conclusion

The results from TEM imagery and compositional analysis align with respect to HMA/LMA status. In addition, this study supports previous research highlighting the prevalence of small subsets of highly abundant OTUs, and concomitantly lower evenness, in LMA sponge species. Taken together, our results demonstrate that the HMA/LMA dichotomy extends to sponges inhabiting the mesophotic zone. Although samples of the LMA sponge species clustered together, suggesting compositional similarity, it should be noted that there were marked differences in the relative abundances of the dominant classes of LMA species with Gammaproteobacteria dominating the prokaryotic communities of P. folium and A. cubensis, whereas Acidimicrobiia (previously identified as a HMA indicator taxon; Moitinho-Silva et al. 2017) dominated the communities of A. flabelliformis and C. kuekenthali.