A morphological cell atlas of the freshwater sponge Ephydatia muelleri with key insights from targeted single-cell transcriptomes

Abstract

How animal cell types, tissues, and regional body plans arose is a fundamental question in EvoDevo. Many current efforts attempt to link genetic information to the morphology of cells, tissues and regionalization of animal body plans using single-cell sequencing of cell populations. However, a lack of in-depth understanding of the morphology of non-bilaterian animals remains a considerable block to understanding the transitions between bilaterian and non-bilaterian cells and tissues. Sponges (Porifera), one of the earliest diverging animal phyla, pose a particular challenge to this endeavour, because their body plans lack mouths, gut, conventional muscle and nervous systems. With a goal to help bridge this gap, we have studied the morphology, behaviour and transcriptomics of cells and tissue types of an easily accessible and well-studied species of freshwater sponge, Ephydatia muelleri. New features described here include: a polarized external epithelium, a new contractile sieve cell that forms the entry to incurrent canals, motile cilia on apopyle cells at the exit of choanocyte chambers, and non-motile cilia on cells in excurrent canals and oscula. Imaging cells in vivo shows distinct behavioural characteristics of motile cells in the mesohyl. Transcriptomic phenotypes of three cell types (cystencytes, choanocytes and archaeocytes) captured live indicate that cell-type transcriptomes are distinct. Importantly, individual archaeocytes show a range of transcriptomic phenotypes which is supported by the distinct expression of different genes by subsets of this cell type. In contrast, all five choanocyte cells sampled live revealed highly uniform transcriptomes with significantly fewer genes expressed than in other cell types. Our study shows that sponges have tissues whose morphology and cell diversity are both functionally complex, but which together enable the sponge, like other metazoans, to sense and respond to stimuli.

Introduction

Sponges (Porifera) are filter feeding animals that live in a diversity of aquatic habitats from mountain lakes and rivers to the deep sea [117]. As animals they are unusual in lacking polarized bodies with mouths and gut, conventional muscle or nervous systems, but even without these features they carry out the same functions as seen in more complex animals such as feeding and excretion, constructing a skeleton, sexual reproduction, and even active behaviour. Understanding the evolutionary relationships between cells, tissues and regions of sponges and those of other metazoans has been the focus of several decades of research in evolutionary developmental biology [4]. There are now many excellent studies examining the expression in sponges of genes known to be involved in determination of germ cells, polarity, muscle and nervous systems [3, 5, 38, 47, 48, 61, 63, 91, 93]. Nevertheless, our picture for how sponge cells form tissues, how those tissues become distinct body regions, and how those regions function together to enable sensation and active behaviour remains blurry; we are still impeded by an inadequate understanding of sponge morphology.

Molecular clocks suggest that sponges arose some 600–800 million years ago [25, 99]. The events that led to the evolution of the first multicellular animals (metazoans) remain unclear, and may have included increased oxygenation of the atmosphere and oceans [104, 105] or increased predation [15, 16, 64], but it is almost certain that the first animals were very small, and food would largely have been available as a dilute suspension of plankton. Sponges filter water with flagellated pump cells called choanocytes, which strongly resemble a unicellular flagellate, the choanoflagellate. The similarity of these cell types led to the understanding that sponges arose from a filter feeding colonial flagellate [51], and as such were the earliest diverging animal phylum, a view supported by early molecular work confirming that choanoflagellates are the sister group to metazoans [119]. This understanding has for many years reinforced the idea that sponges are a simple animal, composed of a collection of cells that work somewhat like large colonies of choanoflagellates.

In recent years however, phylogenomic data have surprisingly shown that ctenophores—pelagic predatory ciliated gelatinous animals—likely branched off even earlier than sponges [27, 78, 97, 124]. Whereas the traditional hypothesis is that sponges arose first from a unicellular flagellate, prior to the development of complex tissues including muscle and nervous systems [52, 86, 87, 112], this new hypothesis suggests that either ctenophores gained a level of tissue and body complexity, with muscle and neurons, mouths, guts and complex sensory systems, independently of other animals, or that sponges may have lost substantial morphological complexity [79, 96, 113, 132]. Studies show large numbers of gene losses and gains in both ctenophores and sponges [55], but without detailed physiological studies it is still unknown what aspects of morphology and physiology (if any) may have been abandoned for a filter feeding lifestyle.

An important step in understanding the biology of an animal (or any other multicellular organism) is to understand the number of cell types it contains [116]. Single cell RNA sequencing (scRNASeq) has identified 10–18 cell types with different transcriptomic identities in two adult demosponges, and some new cell types have been proposed based on clusters of gene expression [81, 100]. Earlier morphological studies also report 10–15 cell general types (Lévi in [102]), although up to 34 cell types are noted for a freshwater sponge [102]. While single cell gene expression patterns contribute greatly to our understanding of cell type and function, for cell identities to be accurately determined based on scRNASeq a robust knowledge of cell morphology and function is still necessary. For example, a clear understanding of cellular morphology has been insightful in better understanding cell types in placozoans (e.g., [103]) and thus combining this type of data with scRNASeq data will provide a much needed new understanding of sponge tissues.

One of the challenges in describing sponge cells and tissues is the “bilaterian bias” and the misconception that sponges are not “tissue-grade” animals. Tissues are aggregates of cells that together form a structure with a function. But more often than not, tissues are defined by structures known in bilaterians [28] and often this framework prevents the recognition of non-bilaterian-specific structures that clearly fall within the boundaries of this definition. An example of this is the choanocyte chambers, which consist of choanocytes that lie on a mesohyl and pinacoderm that together create the incurrent openings, and whose collars are linked at their apical side to the apopyle cells which form the excurrent opening. The choanoderm and pinacoderms are tissues, and together the whole structure functions as a pump, which in essence is an organ. In addition, the view that sponges are not “tissue-grade” animals has led to the long-held misunderstanding that sponges lack epithelia. Today it is widely agreed that all sponges possess physiologically functional epithelia [2, 66, 90], have spatial organization [5, 44, 129] and with those epithelia comes the ability to sense and respond to stimuli [56, 73].

To date, the best reference on cell biology of sponges remains Simpson’s 1984 book [102], which is out of print and not easy to come by. Although there are many individual papers describing discrete aspects of sponge morphology and biology, Simpson [102] covers all sponge classes, and while this broad and comparative approach has its advantages, it also can be confusing given the substantial differences in cell types and tissue structures among classes and families. Our aim here, therefore, is to provide a guide to understanding the gross morphology, tissues and cells of one well-studied emerging model species Ephydatia muelleri, with a goal to provide an atlas of the morphology and cell types of this species in particular, as a step toward making work on this phylum more accessible to the broader scientific community. We include gene expression of select cells to directly couple gene expression profiles with cell morphology. We chose E. muelleri because much work already exists on the ecology, physiology and genetics of freshwater sponges [39, 41,42,43, 49, 50, 55, 60, 65, 75, 77, 80, 81, 108, 123, 125, 126, 129]. E. muelleri is a pan-continental species, being widely distributed in lakes and rivers, and as these sponges are easily collected and grown in culture, an atlas of morphology is practical for both researchers and teachers. There is also a rich literature on sponges from this genus dating back to the nineteenth century, however many of those studies on morphology and physiology are not in English, and therefore not broadly accessible. We focus on the fully developed young sponge, ‘Stage 5’, that can be grown easily in culture and which has all the characteristics of the adult. Our atlas of morphology of E. muelleri, together with deep sequencing of transcripts from specific cell types of this species, complements any work that uses E. muelleri genome and transcriptome resources [55, 67].

Methods

Sponge collecting and culturing

Gemmules of the freshwater sponge Ephydatia muelleri [69] were collected from logs in Frederick Lake, O’Connor Lake, and the Sooke River, British Columbia, Canada. Depth was 0.5–1 m below the surface. Other samples were scraped from the walls of the Kapoor Tunnel of the Sooke Reservoir during an inspection of the Capital Regional District Head tank, Victoria BC. All gemmules were collected during December or January. Gemmules were stored in distilled water at 3 °C in the dark until use and water was aerated monthly. Gemmules were mechanically separated from the spicule skeleton of the maternal sponge and cleaned of spicule debris, rinsed in 1% hydrogen peroxide and plated in either Strekal’s medium (0.9 mM MgSO₄·7H₂O, 0.5 mM CaCO₃, 0.1 mM Na₂SiO₃·9H₂O, 0.1 mM KCl) [108] or M-medium (0.5 mM MgSO₄·7H₂O, 1 mM CaCl₂·2H₂0, 0.5 mM NaHCO₃, 0.05 mM KCl, 0.25 mM Na₂SiO₃·9H₂O) [88] as described previously [65].

Fluorescence and electron microscopy

For fluorescence microscopy, coverslips with live sponges were placed directly into a mixture of 3.7% paraformaldehyde (PFA) and 0.3% glutaraldehyde in phosphate buffered saline (PBS; 100 mM) for 24 h at 4 °C. Sponges were rinsed and processed as described previously [29]. Actin was labelled with either Bodipy 591 Phalloidin, Alexa 594 Phalloidin, or Bodipy 505 FL Phallacidin (Molecular Probes-Invitrogen, USA) with 10% bovine serum albumin (BSA). Tubulin was labelled with mouse anti-tubulin, either clone E7 from Developmental Studies Hybridoma Bank 1:100 dilution in PBS with PBTX-100 and 10% goat serum, or T6973, Sigma, 1:500 in 10% PBS for 24 h or overnight at room temperature. Secondary antibodies were either Alexa 488 or Alexa 594 goat anti-mouse at a dilution of 1:100 in PBS and 10% goat serum (Invitrogen). DNA was stained with Hoechst (33-342, Sigma), and membranes were stained with FM 1-43FX (Thermo Fisher Scientific). Preparations were viewed with a Zeiss Axioskop2 Plus epifluorescent microscope or Zeiss 710 confocal microscope and a maximum projection image was compiled by 1-µm image stacks.

For electron microscopy the coverslip with the sponge was immersed directly into a cocktail consisting of 1% OsO₄, 2% glutaraldehyde in a 0.45 M sodium acetate buffer (pH 6.4) with 10% sucrose for 24 h at 4 °C [46]. To preserve the glycocalyx mesh around the collars, 10% ruthenium red was added to the fixative. To remove the sponges from the coverslips and also dissolve the glass skeletons, fixed sponges were desilicified in 4% hydrofluoric acid in 70% ethanol and subsequently processed for either scanning or transmission electron microscopy (S/TEM) as described by Ludeman et al. [73]. Thick sections were viewed with a Zeiss Axioskop microscope; images were captured with a QiCam camera with Northern Eclipse software, and were manipulated with Adobe Photoshop CS5. All measurements of structures were performed in ImageJ (NIH, USA v1.52i).

Digital video time-lapse live microscopy

Live Stage 5 sponges were viewed as whole mounts on an Olympus SZX12 stereomicroscope and images were captured with a QiCam camera using Northern Eclipse software (Empix Imaging Inc., Mississauga, CAN). To view live cells in the mesohyl a 20 × or 40 × water immersion lens was used on an Axioskop2 plus or for sponges grown on MatTek dishes (Mattek.com) an Axiovert 200 M was used. Images were captured at a rate of 1 frame/20 s for time-lapse imaging. For high-speed imaging a MegaSpeed5 camera was used at 100–1000 frames per second.

Image analysis

Still images were processed in Adobe Photoshop (CS5). Videos were processed in ImageJ (NIH, USA, v1.52i). Cell perimeter and area measurements were made using ImageJ. Dimensions were averaged from more than 20 measurements and the associated standard error was used. Velocity and path length were calculated with the ImageJ application MTrackJ.

In situ hybridization

Experiments were carried out to determine gene expression in Stage 5 juveniles and in adult individuals of Ephydatia muelleri with germline genes Vasa, Nanos and PL10. To make antisense RNA probes, DNA templates of the genes were amplified by PCR using a specific upstream primer and a downstream primer tagged with a T7 promoter site (5’-TAA TAC GAC TCA CTA TAG GG-3’). Primer sequences are listed in the Supplementary Material (Additional file 1; Table S1). Before preparing the DIG-labelled RNA probes with Megascript® T7 (Ambion, USA), the PCR products were cleaned using MinElute kit (QIAGEN, USA).

Juvenile sponges and adult tissue of E. muelleri were fixed in 4% paraformaldehyde (PF) in 1/4 DEPC-treated Holtfreter’s solution (HS) at 4 ºC overnight. Samples were then rinsed with HS and dehydrated through 25, 50 and 75% EtOH in 1/4 HS and stored in 100% ethanol at -80 ºC. All rinses were 5 min unless otherwise noted. Sponges were rehydrated through 75, 50 and 25% EtOH in PBTw (phosphate buffered saline with 0.1% Tween20), and then washed 4 × in PBTw, followed by treatment with proteinase K (50 μg/μl) in PBTw for 1 min at room temperature. Specimens were post-fixed with 4% PF in PBTw at 4 ºC for 1 h and processed for ISH as described previously [128]. Samples were then post-fixed in 4% paraformaldehyde in PBTw, mounted whole in Entellan for whole-mount observations, or dehydrated through an ethanol series and embedded in Epon resin (TAAB 812, Polysciences, USA) for sectioning. Thick sections (1 µm) were cut on a Leica Ultracut T microtome and stained with Richardson’s [92]. In some post-fixed whole-mount preparations nuclei were post-stained with Hoechst 33-342; preparations were mounted in PBS for fluorescence microscopy. All samples were observed with a Zeiss Axioskop microscope and images captured using a QICam camera using Northern Eclipse software.

Cell dissociation and sequencing

We used targeted Smart-seq single cell sequencing in this study because, unlike a droplet-based single-cell sequencing efforts, it allows us to directly couple gene expression profiles with cell morphology. To dissociate sponge cells, small sections of Stage 5 sponges containing different cell types were mechanically triturated in 400–500 µL of modified (divalent cation-free) Strekal's Medium or M-medium [65]. In some cases, trypsin (1 mg/mL, Sigma-Aldrich) and collagenase (Collagenase type I, 1 mg/mL, Sigma-Aldrich, SCR103) were used to dissociate cells. The dissociated cells were plated on 35-mm Petri dishes filled with M-media and allowed to settle for at least 20–30 min before collection. Petri dishes with cells plated, were kept at room temperature, and used for sampling within about 1 h. The different cell types were identified morphologically and functionally and collected using disposable glass suction pipettes. The pipettes were fabricated from borosilicate capillary glass (BF150-86–10, Sutter Instrument, CA, USA) using a Flaming–Brown micropipette puller (P-87, Sutter Instrument, CA, USA), with a diameter of 2–3 µm. Pipette solution was divalent cation-free M-medium. In most cases single cells were captured, but in the case of choanocytes, which were challenging to separate, some doublets and small clusters of four cells were captured for sequencing.

After capturing a cell in the pipette, it was transferred to 12.5 µl of lysis buffer. The Eppendorf tube was sonicated for ~ 5 s in ice cold water and immediately stored on dry ice. The samples were kept at -80 °C until they were shipped on dry ice to Admera Health (South Plainfield, New Jersey, USA) for amplification, library preparation and sequencing using the SMARTseq V4 with Tagmentation and Illumina sequencing platform with a read length of 2 × 150 bp. Estimated 40 M paired end reads (~ 20 M in each direction) per sample. The lysis buffer was prepared according to the recommendations of the sequencing company and contained: nuclease-free water; dNTP—3.3 mM; RNase inhibitor—1.33 U/µl; Triton X-100—0.133% [109]. The reagents were purchased from Takara Bio Inc., California, USA (10 × Lysis buffer (Cat. No. 634439); RNAse inhibitor (Cat. No. 2313A); 3’SMART-Seq CDS Primer II A (Cat. No. 634439); Nuclease-Free Water (Cat. No. 9012). Slight modifications to the growth media were as follows: M-medium (10x, in mM): CaCl₂·H₂O 1; MgSO₄·7H₂O 0.5; NaHCO₃ 0.5; KCl 0.05; Na₂SiO₃·9H₂O 0.25; Hepes 10; pH7.1 adjusted with Trizma base. Strekal’s medium (10x, in mM): CaCO₃ 0.5; MgSO₄·7H₂O 0.9; KCl 0.1; Na₂SiO₃·9H₂O 0.1; Hepes 10; pH7.1 adjusted with Trizma base. Divalent cation-free M-medium contained (in mM): CaCl₂·H₂O 0.1; MgSO₄·7H₂O 0.05; NaHCO₃ 0.05; KCl 0.005; Na₂SiO₃·9H₂O 0.025; Hepes 1; EGTA 1.

Transcriptome analyses

The transcriptomic analyses are outlined with command lines and information required to replicate the computer environment used in these analyses can be found in our GitHub repository (https://zenodo.org/records/10998268). In summary, we used RSEM and EdgeR as implemented in the Trinity utility scripts align_and_estimtate_abundance.pl, abundance_estimates_to_matrix.pl, run_DE_analysis.pl, PtR, and analyze_diff_expr.pl to align reads to the Ephydatia muelleri gene models (v1), generate counts, normalize counts, conduct principal components analysis, and run differential gene expression analyses.

Statistical comparisons of the number of genes expressed in each sample were conducted. The number of genes expressed was approximated by counting the number of genes non-zero TPMs in the RSEM outputs (multiple isoforms of a single were counted once). We used non-zero TPMs because, unlike approaches using bulk RNA-seq or droplet RNAseq, collecting, washing, and sequencing individual cells means that non-zero TPM represents the real presence of a transcript in a particular cell rather than noise or ambient RNA. To assess the normality of these counts, Shapiro–Wilk tests were conducted for each set of replicates. The assumption of homogeneity of variances was evaluated using Bartlett's test. An analysis of variance (ANOVA) was performed to determine significant differences among the groups. Post hoc Tukey's honest significant difference (HSD) tests were employed to identify specific group differences. All statistical analyses were conducted using R software (version 4.3.2). The Perl script used to count non-zero TPMs from RSEM outputs and the R code and output (included as inline comments) used to run the above tests are included in our GitHub repository (https://zenodo.org/records/10998268).

Results

Overview

Ephydatia muelleri gemmules hatch at room temperature, but different batches of gemmules hatch at slightly different times, such that young sponges are generally referred to by stages (Additional File 1. Figure S1): Stage 1, cells have emerged from the gemmule and attach it to the substrate; Stage 2, the gemmule husk is covered by tissue; an epithelium covers a mass of cells forming choanocytes, spicules and small lacunae; Stage 3, lacunae enlarge and often move slowly in a clockwise direction around the gemmule husk; by Stage 4 lacunae have become narrow canals; at Stage 5 excurrent canals are in their definitive position passing around the side of the now-empty gemmule husk and leading to the now clearly visible osculum.

In Stage 5 sponges (Fig. 1A), the white portions clearly indicate the choanosome, darker channels are large excurrent canals, and the darker periphery marks the edge of the subdermal space. Circular dark regions mark the larger incurrent canals (Additional File 1; Figure S2). Stage 5 sponges have a fully developed aquiferous system with a tent-like tissue raised on spicule bundles above the choanosome (Fig. 1B, C). The tent has an inner and outer epithelium (exo and endopinacoderms) which sandwich a mesohyl with actively wandering cells. The tent sits over the subdermal space (SDS)—a space that acts as a vestibule for incoming water—and lowers momentarily during contractions of the sponge reducing the volume of the SDS. The floor of the subdermal space has depressions that contain openings into the incurrent canals.

Fig. 1

Overview of the morphology of Ephydatia muelleri. A Stage 5, week-old, juvenile sponge viewed by stereo microscopy. B Fracture of a Stage 5 sponge showing the tent held up by spicule shafts and overlying a subdermal space and the choanosome with choanocyte chambers and canals. C Illustration showing the principal features of the sponge. A: Light microscopy, B: Scanning electron microscopy (SEM). osc, osculum; exc, excurrent canals; in, incurrent canals; g, gemmule husk; pd, pinacoderm of the tent; sp, spicule shafts; os, ostia; sds, subdermal space; end, endopinacoderm; cc, choanocyte chambers

The tent: exo and endopinacoderm

The surface of the sponge is formed by an epithelium made of flat pentagonal cells (exopinacocytes) approximately 30 µm across (28.53 ± 5.95, n = 23) and only 3 µm thick (Fig. 2A). The nucleus is central to the cell and while actin labels the periphery of these cells, the boundaries are better marked by a membrane marker FM 1–43 (Fig. 2B). The inner layer is formed by endopinacocytes that are also flat cells (2 µm thick) but are elongate (Fig. 2C). Some amoeboid cells in the ‘tent’ produce collagen as seen by a tail of collagen coming from one end of the cell in SEM images (Fig. 2D). Exopinacocytes overlap each other by less than 1 µm and yet they form a sealed enclosure shown by the fact that ruthenium red included in the fixative does not penetrate beyond the junction (Fig. 2E). Inside is the collagenous middle layer (the mesohyl) with amoeboid cells. The collagen sheet is attached to the basal surface of the exopinacocytes and in this way it both provides support to the exopinacoderm and polarizes the epithelium (Fig. 2E–G).

Fig. 2

Pinacoderm epithelia. A,B. Exopinacocytes in the exopinacoderm, viewed by SEM (A), and by fluorescence using the membrane marker FM 1-43 (B). C Cross section of the ‘tent’ viewed from the underside showing elongate endopinacocytes that have pulled away from one-another during fixation. D View of the collagen layer under the exopinacocytes and being secreted by a lophocyte. E Cross section of a junction between two exopinacocytes fixed with ruthenium red, illustrating that the stain does not penetrate the junction, and that the exopinacocytes are underlain by collagen. F,G Cross section of the ‘tent’ showing exopinacocytes underlain by a thin layer of collagen. Mesohyl cells lie between the exo and endopinacoderms. A,C,D,F: SEM; B: Fluorescence; E: Transmission electron microscopy (TEM). exp, exopinacocytes; end, endopinacocytes; col, collagen; RR, ruthenium red; me, mesohyl

Endopinacocytes have strongly labelled tracts of actin that, at lower magnification, can be seen to cover hundreds of micrometres across the whole sponge and which lead up to and terminate at the tips of spicule posts that form points throughout the ‘tent’ (Fig. 3A-C). Strongly labelled tracts of actin join other tracts in neighbouring cells by distinct junctions that are actin-dense bars lying perpendicular to the tracts (Fig. 3D-F, arrows).

Fig. 3

Actin cytoskeleton in the ‘tent’ of Ephydatia muelleri Stage 5 sponges. A Overview of actin tracts in the whole sponge showing points of attachment at the tips of spicule shafts. B,C Magnification of the spicule shaft tips viewed by scanning electron microscopy (B) and fluorescence of actin (C). D-F. Actin tracts at the edge of the sponge run for hundreds of micrometers across the tent and connect to tracts in neighboring cells at densely staining junctions (arrows). A,C-F: Fluorescence; B: SEM. sp, spicules; g, gemmule husk

Scattered throughout the tent are holes called ostia. Ostia are formed by porocytes, flat cells with a central opening up to 20 µm in diameter that are closely abutted by several pinacocytes (Fig. 4A-B). Ostia occur in arrays with porocytes 10 to 20 µm apart. At the edge of the ‘tent’ where the pinacoderm falls down to the substrate, they can be seen quite clearly in living animals using differential interference microscopy (Fig. 4C). There, ostia open and close in a matter of seconds, while amoebocytes crawl around them in the thin mesohyl space (Fig. 4C). Porocytes are smaller than exopinacocytes and lie between the exo and endopinacoderm layers, anchored by small filopodia to the collagen sheet (Fig. 4D). Porocytes appear to be able to have more than one opening (e.g., Fig. 4E), and the edges of ostia label with phalloidin suggesting that the contractile mechanism is actin/myosin (Fig. 4F). The exopinacoderm and endopinacoderm of E. muelleri lack spicule producing cells (sclerocytes) and the exopinacoderm is not ciliated.

Fig. 4

Ostia and porocytes in the pinacoderm. A Overview showing distribution of ostia with a range of sizes in the surface ‘tent’ of a sponge. B. Higher magnification shows the porocyte tightly adjacent to expinacocytes. C View of two ostia in a live sponge with amoebocytes crawling in the thin mesohyl between the two layers. D. Tearing of the epithelium during tissue preparation for SEM reveals that the porocyte is a flat cell that is anchored by tiny filopodia to the collagen layer just under the exopinacocytes. E. One porocyte can have several ostia. F. Ostia are ringed by an actin cytoskeleton. A,B,D,E: SEM; C: Differential interference microscopy; F, Fluorescence microscopy. Sp, spicule shaft; os, ostia; p, porocyte; exp, exopinacocyte; amb, amoebocyte; co, collagen

Incurrent canals

The floor of the subdermal space is covered by endopinacocytes, flat like those of the tent, but with a range of shapes; endopinacocytes surround circular openings into the choanosome, the incurrent canals (Fig. 5A, B). The openings are formed by one or several cells that together create perforations or holes like ostia. Strands of the cell and their ‘holes’ hang down into the choanosome making contact with choanocyte chambers, and together the cell or cells (for both occur) form a sieve-like structure (Fig. 5C, D). The openings in the sieve cell lead either directly to choanocyte chambers or to further sieve cells, and those together with other endopinacocytes form the lining of incurrent canals that penetrate the whole choanosome. Those canals that go directly down to the bottom of the choanosome are visible under low magnification as dark circular areas in the otherwise white tissue of a relaxed sponge (Additional file 1; Figure S2).

Fig. 5

Incurrent canals and sieve cells. A. Cross section of the choanosome, showing the subdermal space and canals. B. Incurrent canals and sieve cells viewed from the subdermal space. C. Cross section of one sieve cell showing its three dimensional morphology and relationship to the choanocyte chambers. D. Diagram of a sieve cell. A-C: SEM. sds, subdermal space; exc, excurrent canal; ic, incurrent canal; enp, endopinacoderm; sc, sieve cell; cc, choanocyte chamber

Choanocyte chambers

Choanocyte chambers are the main feature in the choanosome (Fig. 6). Chambers are attached to incurrent and excurrent canals as cup-shaped structures, and the slim tissues surrounding the chambers are tightly packed with crawling cells (Fig. 6A). The collars and flagella of choanocytes label strongly for actin and tubulin, respectively; canal endopinacocytes label slightly for actin, and only two cells types label strongly for anti-tubulin: sclerocytes (spicules) and cystencytes (Fig. 6B).

Fig. 6

Choanosome A. Cross section of the sponge showing the relationship of the choanocyte chambers to the incurrent and excurrent canals. B. Actin (red) labelled microvilli of the collars and tubulin (green) labelled flagella show choanocytes in choanocyte chambers surrounding canals; nuclei (blue). Sclerocytes and cystencytes in the mesohyl around chambers also label with antitubulin. A: Light microscopy of an epoxy section; B: Confocal microscopy. cc, choanocyte chambers; cys, cystencytes; exc, excurrent canal; fl, flagellum; in, incurrent canal; n, nuclei; mv, microvilli; scl, sclerocytes; sp, spicules

Fully formed choanocyte chambers are 25–30 µm diameter, and cup or crescent shaped with an opening—the apopyle—at one side that is ~ 13 µm in diameter (Fig. 7A). There are about 50 ± 3.7 (SE, range 24–87, n = 20 chambers) choanocytes in each chamber. Entrances to chambers (prosopyles) are formed by openings in flat cells that directly wrap around the base of choanocytes (Fig. 7B). Choanocytes in Ephydatia muelleri are squat cells 2.5 µm in diameter at the collar and 4–5 µm across the base (Fig. 7C). The choanocyte base extends ‘feet’ that contact neighbouring choanocytes. Collar microvilli are 5–6 µm long and made of an average of 40 ± 2 microvilli. The apopyle is formed by 4 leaf-like cells called cone cells or apopylar cells, whose one edge is attached to the canal wall and the other forms a ruffled margin that is in contact with the tips of collars. Apopyle cells also have a single motile cilium that faces out of the chamber into the canal (Fig. 7D, E) and which in live animals can be seen flicking with a stiff forward stroke–backward stroke action (Additional File 2; Video S1). In typical fixation the collars appear bare as in Fig. 7C, but fixation with ruthenium red highlights the extensive proteoglycan mesh that glues the apical sides of collars together and attaches the collar tips to the apopylar cells (Fig. 7D). Ruthenium red also preserves the mesh that holds adjacent collar microvilli together (Fig. 8A–D). Scanning and transmission electron micrographs show a mesh with holes (40–60 nm) forms a lattice between all collar microvilli. The flagellum, 20 µm long, arises from the centre of the collar and is preserved in the sinusoidal wave of its beat. A glycocalyx sheet forms a wing-like vane that extends out from either side of the flagellum and is connected at the lateral edge to the collar microvilli (Fig. 8A–D). The vane projects beyond the top of the collar for the full length of the flagellum.

Fig. 7

Features of the choanocyte chamber. A. Fracture across a choanocyte chamber showing a view through to the apopyle (exit). B. Prosopyle entrance to the chamber shows choanocyte bases lie on endopinacocytes. C. Choanocytes have a squat cell body and collar of microvilli with characteristic wider spacing at the base of the collar. D. Cross section through a choanocyte chamber fixed with ruthenium red shows attachment of collar tips to one another and to the apopyle cells (arrow heads). Apopyle cells each have a single cilium (arrow). E. A chamber viewed with confocal microscopy shows the wide opening at the apopyle and the cilium associated with the apopyle (arrow). A-D: SEM; E: Confocal microscopy (colours as in figure 6). ap, apopyle; mv, microvilli; n, choanocyte nuclei; fl, flagellum

Fig. 8

Collar-flagella unit of the choanocyte fixed with ruthenium red. A,B. Fracture through the collars of choanocytes showing the tight adhesion of collars to one another, the flagellum and its vane of glycocalyx (A, B, arrows). Microvilli are connected by a lattice of glycocalyx (B, arrowhead). The vane is a cobweb-like structure that is adhered at the edges to the collar. C. Cross section through a collar stained with ruthenium red shows that closest to the flagellum (v, arrow) the vane has two layers. D. Diagram illustrating the interpreted relationship of the vane to the flagellum and collar. A,B: SEM; C: TEM. fl, flagellum; v, vane

Approximately 5% of chambers we examined live using video microscopy had central cells. These cells are oval shaped with pseudopodial extensions that touch the surface of the glycocalyx mesh surrounding collar tips. They are sluggish, not moving very much in the chamber over an hour of time-lapse recordings. The flagella beat continues uninterrupted in the choanocytes below them (Additional file 1; Video S2).

Excurrent canals and osculum

Chambers empty directly into excurrent canals that are lined by endopinacocytes (Fig. 9A). In contrast to those endopinacocytes covering the incurrent surface of the choanosome, these are more pentagonal, and although most are bare, occasionally here and there, a pair of cilia is found on an endopinacocyte in the excurrent canals (Fig. 9A, B). Cilia are short, 4–6 µm long and typically paired. Excurrent canals merge eventually forming the largest diameter canals that meet at the base of the osculum. In stage 5 sponges the osculum varies in length from 400 to 500 µm long and 100–200 µm wide. In older sponges the osculum can be longer, and in contracted sponges the osculum has a sphincter of actin at the base or partially up its length that appears as a denser region of labelled actin, but from the label it is not clear whether the contractile cells are endo- or exopinacocytes (Additional File 1; Figure S4).

Fig. 9

Excurrent canals and osculum A. Excurrent canal showing apopyles of several chambers emptying into the canals. One endopinacocyte has a pair of short cilia (arrow, enlarged in B). C-E. Osculum tip showing cells inside the osculum have a pair of cilia that project out perpendicular to the direction of flow out of the osculum. A,B,D: SEM; C: Light microscopy of a live sponge; E: Phase contrast microscopy of a live sponge

In live animals, the osculum is always moving but it is unclear if this is focused movement towards or seeking a stimulus, or whether the movement is an innate feature of cells moving around its thin middle layer. The opening of the osculum (20–50 µm) is always much narrower than the osculum width at its base or middle, as though it is always slightly contracted (Fig. 9C). The osculum arises from excurrent canals that meet the ‘tent’ and so its outer layer is formed by the exopinacocytes of the tent and its inner layer is made of the endopinacocytes of the canals with a 2–4 µm wide collagenous mesohyl separating the two layers; motile cells are in that thin space. The exopinacocytes have no spicules, nor are they ciliated. However, every endopinacocyte from the base to the tip of the osculum has a pair of 4- to 6-µm-long cilia (Fig. 9D, E). The cilia pair are always perpendicular to the long axis of the osculum, as seen at the top of the osculum in live animals (Fig. 9E inset). There is no structure around the base of the cilia (no microvilli) and no structures are found linking the cilia pair inside the cell. Ciliated endopinacocytes are extremely thin (less than 2 µm thick) and have an ovoid nucleus with heterochromatin; the cytoplasm has endoplasmic reticulum and mitochondria as shown previously [73].

Mesohyl cells and basopinacoderm

Between the inner and outer epithelia of the sponge is a slim collagenous extracellular matrix in which there are a number of actively crawling cell types. These fall into three categories: polyblasts, archaeocytes, and cystencytes (Fig. 10A, Table 1). A fourth category, sclerocytes are not motile themselves but are carried by another amoeboid cell type which we do not describe here as they are rare in Stage 5 sponges because the spicules are already in their terminal positions. The different motile cells have distinct morphological and behavioural characteristics (Additional File 1; Table S2, Figure S5, S6 and S7).

Fig. 10

Cell types in the mesohyl and at the base of the sponge. A. Overview of a region towards the edge of the choanosome of a stage 5 sponge. B. Archaeocyte with a nucleolated nucleolus and birefringent inclusions. C. Polyblast type 1. D. Polyblast type II. E,F. Large (E) and small (F) cystencyte (arrow indicates inclusion). G. Sclerocyte. H. Basopinacocyte. A-H: Phase contrast microscopy; nu, nucleus

Table 1 Cell types in Ephydatia muelleri

Archaeocytes (Fig. 10B): Archaeocytes are spherical cells that can readily be identified by their characteristic large central nucleus with conspicuous dark nucleolus. Archaeocytes change shape, becoming amoeboid with pseudopodia like polyblasts, but always return to a rounder shape. Archaeocytes are distinguished from type I polyblasts in movement, because they move more directly than type I polyblasts, and in morphology because they contain what appear to be birefringent vesicles throughout the cell body (Additional file 1; Figure S5, S6; Video S3). Archaeocytes are found in the mesohyl of all regions of the sponge but are particularly visible in the mesohyl that lies in between the baso- and exopinacoderm near the periphery of the sponge.

Polyblasts (Fig. 10C, D): There are two types of polyblast; both are flat amoeboid cells that move rapidly through the tissue. They can be distinguished by size and speed (Additional File; Figure S5, S6). Type 1 polyblasts are ovoid/circular and extend pseudopodia in leading and trailing directions as they move. They have a larger anterior leading edge than trailing edge; the nucleus is mid-way between front and back. Type 1 polyblasts are the largest cells in the mesohyl; they travel at half the speed of Type 2 polyblasts and cover the least distance of all mesohyl cells tracked.

Type 2 polyblasts are half the size of Type 1 polyblasts. They are the fastest cell in the mesohyl at 0.22 ± 0.02 µm s⁻¹, and cover the most distance (Additional file 1; Figure S5). They move fluidly altering shape around other cells and returning to the basic circular/oval shape (Additional file 2; Video S3). Like Type 1 polyblasts, they are amoeboid and change shape, but they rarely return to a circular-ovoid shape, instead the cell shape constantly changes (Additional file 1; Figure S5; Additional file 2, Video S3). This cell type also changes direction abruptly. Type 2 polyblasts are found in the canal system, in the mesohyl at the periphery of the sponge, and in the mesohyl of the osculum.

Cystencytes (Fig. 10E, F): Cystencytes are a type of storage cell or thesocyte. They have a single central large inclusion that occupies most of the cell while the remainder of the cell extrudes pseudopodia (called a ‘foot’) as it moves. Two sizes of cystencyte are apparent in E. muelleri, and these do not appear to be stages of one cell type (Figure S5, S6). Large cystencytes have an inclusion six times the size of that in small cystencytes (area of inclusion: 665.87 ± 35.67 µm², n = 17 vs 122.5 ± 13.08 µm² n = 17). Cystencytes also have an elaborate microtubule cytoskeleton and often appear the brightest labelled cell type when using anti-tubulin (e.g., Fig. 6B).

In large cystencytes the nucleus is located in the leading edge of the cell in front of the inclusion. Cystencytes are the slowest of the amoeboid cells and both types move at about the same pace at less than 0.1 µm s⁻¹. But unlike the other mesohyl cells, cystencytes can swivel 360 degrees in place when they are unable to move forward, as if seeking a new direction to move toward. When the new track is determined they again extend the single pseudopodium and continue. Unlike other mesohyl cells that move easily and fluidly through the mesohyl, the movement of large cystencytes appears clumsy, and their forward movement is often altered by other cells blocking the way or pushing them off track. For this reason, cystencytes move in all directions and show a more unpredictable track (Additional file 1; Figure S5; Additional file 2; Video S3). This cell type is present in both the canal system and near the periphery of the sponge as well as in the osculum. The cystencyte inclusion, said to be proteoglycans [102], binds fluorescent labels non-specifically.

Sclerocytes (Fig. 10G): Sclerocytes are long spindle-shaped cells that secrete and surround the glass spicules. The sclerocyte cell body takes the form of the elongate spicule and does not extend away from the side of the spicule except to house the nucleus mid-way down the glass shaft. Spicules appear throughout the sponge. Like cystencytes, sclerocytes have a dense network of microtubules that label strongly with anti-tubulin (e.g., Fig. 6B).

Basopinacocytes (Fig. 10H)

The sponge is attached to the substrate by a matrix secreted by the basal epithelium. That epithelium is formed by pentagonal pinacocytes called basopinacocytes, which have a range of adhesion and secretory functions. These include laying down the matrix that attaches the sponge to the substrate, and controlling osmolarity by secreting water. These cells have the same shape as exopinacocytes but differ in labelling strongly for actin and tubulin. Actin tracts branch out from plaques in a cell and sometimes several cells have tracts of actin that continue across several cells but over much shorter distances than those seen in the endopinacoderm. Microtubules labelled with β-tubulin (not alpha tubulin) radiate out from the nucleus (Additional file 1; Figure S7) as described by previous authors [85, 118].

Transcriptomic profiles of select cells

RNA sequencing of isolated cells using Smart-seq V4 libraries provides detailed gene expression profiles of four replicates of isolated archaeocytes (Fig. 11A), five replicates of cystencytes (Fig. 11B), and five replicates of choanocytes (Fig. 11C). Principal components analysis (PCA) of normalized gene expression profiles showed distinct groupings of these choanocytes and cystencytes. Archaeocytes, on the other hand, were sparser and formed a trajectory that intersected with choanocytes (Fig. 11D).

Fig. 11

RNA sequencing of isolated cells using Smart-seq V4 libraries. A-C. All cells were isolated from primary cell cultures of disassociated Stage 5 sponges and individual cell replicates were used for each of the Smart-seq cell transcriptomes. The glass micropipettes used to isolate the cell are visible in the center bottom of these micrographs (out of focus; see Additional File 1; Figure S8 for images of all the cells used to generate the transcriptomes). D. Principal components analysis (PCA) of isolated Smart-seq V4 RNA-Seq data. PCA was performed using PtR utility script included with Trinity. E. Inset shows that the number of expressed genes are significantly higher in archaeocytes and cystencytes relative to choanocytes. F. A heatmap showing all differentially expressed genes with p-value = 2. EphydatiaGene IDs listed alongside human gene IDs corresponding to the top human RefSeq BLASTP hit in parentheses (“no hits” indicates that there was no BLASTP hit to the human RefSeq database). Genes highly expressed in archaeocytes are in white with black background

The number of genes with non-zero TPMs was counted from each sample to approximate the numbers of genes expressed in each cell type. After confirming that the assumption of normality was met for each set of replicates (p-values = 0.7159, 0.3899, 0.3258), and confirming homogeneity of variances among the sets of replicates with the Bartlett test (p-value = 0.3391), analysis of variance (ANOVA) revealed a significant difference in gene counts between the groups (F = 7.555, p = 0.00861). Post hoc Tukey's HSD tests showed that the differences in gene counts between archaeocytes and choanocytes (adjusted p-value = 0.0221719) as well as cystencytes and choanocytes (adjusted p-value = 0.0132390) were significant. Together these results suggest that choanocytes express a far more limited set of genes than the other two cell types (Fig. 11E). Our evidence suggests that the reduced number of genes expressed in choanocytes was not due to sequencing coverage or batch effects (Additional file 1; Data S7).

Differential gene expression analyses between archaeocytes and cystencytes showed 27 genes that exhibited significantly different expression (FDR < 0.001 and log2FC >  ± 2; Fig. 11F). There were three genes that were expressed significantly higher in archaeocytes relative to cystencytes. Only two of these had significant BLAST hits to Uniprot and both of these hits were to CPEB1, a gene involved in cytoplasmic polyadenylation. In addition to these three that were significantly higher in archaeocytes relative to cystencytes, we detected several other key stem cell-related genes expressed in our archaeocyte single-cell transcriptomes, including Nanog and DDX genes (vasa and PL10). Using in situ hybridization, we show that the expression of vasa and PL10 was present in some archaeocytes in Stage 5 juveniles and adults; but while in juveniles they were broadly expressed in the cytoplasm of archaeocytes (Fig. 12A–D), in adult sponges they were expressed consistently around the nucleus of archaeocytes (Fig. 12E–F). In addition, the germline gene nanos was found to be expressed in archaeocytes (Additional file 1; Figure S11A) and choanocytes (Additional file 1; Figure S11B) in juvenile and adult tissues, although it was not found expressed in the transcriptional profiles of single cells. There were 24 genes in our single cell transcriptome data that were expressed significantly higher in cystencytes relative to archaeocytes including vesicle-associated membrane protein (VAMP) and syntaxin (STX). The number of genes expressed in choanocytes in our single cell transcriptomic data was significantly lower than in archaeocytes and cystencytes. This highly imbalanced expression led to very large numbers of genes being designated as highly differentially expressed in archaeocytes and cystencytes when compared to choanocytes (Additional file 1; Figure S9). In contrast, there was only one gene in choanocytes that was consistently expressed above our threshold, Em0011g357a.t1, which BLASTs to a cathepsin L.

Fig. 12

Expression of the germline genes Vasa and PL10 in juveniles and adult tissue of E. muelleri. Arrowheads show labelled archaeocytes, arrows show unlabeled archaeocytes. A,B. A population of archaeocytes show a Vasa positive label in a large area of the cytoplasm (black arrowheads) and around the nucleus in sections (A) and whole mount preparations (B). Labeled archaeocytes are often located close to choanocyte chambers (cc). C. A subset of archaeocytes show a Vasa positive label (black arrowheads) around the nucleus (see insert for enlarged view) and a subset of Vasa-positive choanocytes (black arrowheads). The label appears in small clusters within the cytoplasm of the cells. D,E, Expression of PL10 in archaeocytes (black arrowheads) close to choanocyte chambers (cc) in sections (D) and whole mount preparations (E). F. Expression of EmuPL10 around the nucleus of a subset of archaeocytes (black arrowheads) (inset shows higher magnification). exc, exhalant canal; p, pinacocytes; sc, sclerocyte; cc, choanocyte chamber; sp, spicule; cys, cystencyte; n, nucleus

Discussion

Ephydatia muelleri is a well-studied freshwater sponge and yet our work brings to light some surprising new features, several of which have broader relevance for understanding cell and tissue type evolution in animals. Known features that are represented more clearly here include: a tightly sealed exopinacoderm with extensive actin tracts that form tension lines across the entire endothelium; porocytes are anchored between exo- and endopinacoderm sheets; collar microvilli are linked by a proteoglycan mesh with regular nanometer-wide spacing; a clear proteoglycan ‘vane’ extends on either side of the flagellum to the collar microvilli of choanocytes; and all endopinacocytes in the osculum possess short non-motile cilia.

Our studies also reveal several new important features including: (i) that the outer tissue is distinctly polarized with a basal sheet of collagen under the exopinacoderm; (ii) a ‘new’ cell type—the sieve cell—is described at the entrances to incurrent canals, which suggests a common contractile cell type in the sponge; and (iii) sensory cilia at the exit of choanocyte chamber cells beat in a whip-like manner. We also show that distinct behavioural and transcriptomic characteristics distinguish cell types: choanocytes express a reduced set of genes (relative to cystencytes and archaeocytes) and archaeocytes express a suite of stem cell-related genes.

In our work, we used a morphological approach to describe and understand the cells that make up tissues and regions of the sponge using fluorescent labels, scanning and transmission electron microscopy, live imaging, as well as some single cell sequencing of individual cells. We follow the nomenclature from a rich literature of previous morphological studies [14, 21, 101, 120, 121, 127, 85, 102, 111, 122, 123]. While we were able to map several cell types—choanocytes, apopyle cells, amoebocytes, sclerocytes, and pinacocytes—to those in another freshwater sponge Spongilla lacustris where cells were identified using bulk RNAseq [81], it was not possible to correlate other cells in Ephydatia to the newly proposed nomenclature in the absence of additional morphological and functional data for for S. lacustris. In the following sections, we compare the characteristics of the anatomy, behaviour and transcriptomic profile of the freshwater sponge Ephydatia muelleri with other sponges and discuss their relevance.

The pinacoderm epithelium

Pinacocytes: control of the internal ionic medium

The term pinacocyte comes from the Greek word ‘tablet’, and in fact the outer surface of sponges is so smooth that earlier workers thought they were syncytial. It is this very thin nature of the pinacoderm that makes it difficult to study and therefore to understand. In E. muelleri, exopinacocytes are flat or ‘fusiform’ [10] with edges that overlap very slightly so that the zone of adherence is a very small region that is dovetailed together. Despite the small region of adherence, they seal off the internal mesohyl from the outside, as was shown by using small molecules, like ruthenium red (0.85kDa), which are not able to penetrate the mesohyl [2]. The epithelium of freshwater sponges is also resistant to the passage of electrical current [2], but nevertheless, like freshwater Hydra and Amoeba, freshwater sponges experience passive uptake of water and therefore use contractile vacuoles to maintain osmotic balance. Brauer [13] showed that Spongilla lacustris maintains a very low internal osmotic concentration at 12–15 mOsm, and suggested that sodium is pumped into vacuoles to excrete water and ions that are maintained in low concentration in the tissues.

Origin and differentiation of pinacocytes

Oddly, only two groups of sponges—spongillids (freshwater sponges) and homoscleromorphs—have flat (‘fusiform’) pinacocytes; other sponges have largely T-shaped pinacocytes, with a recessed basal nucleus. According to Boury-Esnault [10], T-shaped pinacocytes can arise from flat pinacocytes during regeneration from wounding. Calcarea seem to have all three stages and there are some thoughts that T-shaped pinacocytes are a contracted form of flat pinacocytes in Calcarea (E. Lanna pers. obs.).

Exopinacocytes in E. muelleri are not motile; as seen in time-lapse videos of FM1-43 labelled E. muelleri, each exopinacocyte holds its position amongst neighbouring cells despite overall movement of the sponge [2], and given the tight sealing and dovetailed adhesion between cells as discussed above, this makes good sense. Simpson ([102], pg. 577) reports that exopinacocytes do not divide, and concludes that they are terminally differentiated cells. The idea that exopinacocytes could be motile comes from studies on regenerating or ‘traumatized’ tissues [8, 9, 18, 24]. During gemmule formation or wounding, various cells including pinacocytes dedifferentiate, although what signals trigger this or what their fates are, is not well-studied, and in any case it has not been documented in the non-regenerating normal epithelium of a sponge. According to Ledger (1976, in [102]) in calcareous sponges, exopinacocytes have been observed during their migration from the exopinacoderm into the mesohyl, thus establishing their ability to become ameboid. However, images of this process are not shown and it is not certain if this was during wounding. Moreover, while it was recently shown that dedifferentiation of choanocytes is triggered during wounding in the asconoid sponge Leucosolenia cf. variabilis [62], exopinacocytes were not found to change position or shape; instead they maintained their connections to other pinacocytes and spread extensions to close the wound surface. The above discussion concerns exopinacocytes and choanocytes, however there is no information as to whether it also applies to endopinacocytes. As there are no reports of pinacocytes undergoing cell division in a fully differentiated sponge, it seems likely that pinacocytes generally are terminally differentiated, in which case new pinacocytes must derive from the division and differentiation of an archaeocyte in the mesohyl.

Polarity of the exopinacoderm

Exopinacocytes in Ephydatia muelleri are underlain by a tightly woven thin sheet of collagen that is secreted by cells referred to by others as collencytes (or lophocytes) that are likely a subset of polyblasts. How the collagen sheet is held onto the exopinacoderm is not yet known, nor whether the proteins secreted by lophocytes are part of the SSC collagen and Type IV collagen families known to be present in freshwater sponges [33, 35], but this definitely needs testing by in situ hybridization.

The presence of Type IV collagen is often thought to be the hallmark of epithelia, but its origin and its role in basement membrane (BM) formation in metazoans is still unclear [36, 37, 68, 90, 115]. Ctenophores have a number of genes encoding for Type IV collagen, but many other proteins considered important for BM formation are missing [37]. Also, while ultrastructure shows a BM in some genera (Beroe and Pleurobrachia), no BM is present in Mnemiopsis leidyi [37]. Type IV collagen is present in homoscleromorph sponges where it forms a basal layer to choanocyte chambers and all pinacocytes [6, 12, 32, 34, 35, 37]. Two Type IV collagen genes were previously identified in the calcareous sponge Sycon coactum [66], although ultrastructure does not show a clear BM below the epithelium. In demosponges, only the homolog of Type IV collagen, SSCC, has been found, as well as other proteins important for BM formation [54], and neither Type IV collagen nor other BM proteins are found in glass sponges [98]. The presence and absence across invertebrates of both Type IV collagen and proteins that link it to the cells (e.g., nidogen and perlecan) is puzzling. Like Mnemiopsis, Trichoplax adhaerens has the gene for Type IV collagen, as well as genes for the BM associated proteins nidogen and perlecan, but there is no extracellular matrix (collagenous or other) in Trichoplax [103, 106]. Likewise, acoels have genes for Type IV collagen but in some species the basement membrane is absent [7]. This is also the case for some nermertodermatids, and Achatz et al. [1] suggest that a network of ciliary rootlets and terminal web provide structural support instead. Overall, a general view is that while Type IV collagen played a large role in the evolution of epithelia for its role in providing structural support, it has been lost multiple times in different taxa where the integrity of the epithelium was sufficient without it.

Contractile cells

The endopinacocytes that lie on the roof of the subdermal space are elongate and are traversed by tracts of actin which connect to neighbouring cells by dense plaques. Due to the strong actin labelling, and the vertical movement of the tent during sneezes, the inference has been that this tissue is contractile [17, 29]. However, the basopinacoderm also has directional actin tracts. Yet, there is no evidence that the basopinacoderm contracts; in contrast damage to the edge of the basopinacoderm causes the tissues to peel back showing they are under tension. Contraction by actin is often thought to compress the sponge, but given the location of actin in the roof of the subdermal space, it is likely that contraction of endopinacocytes would make the ‘tent’ taut thus lifting the tent quickly while relaxation would drop the tent rapidly. The idea that actin tracts in the endopinacoderm are for tension rather than for contraction of the sponge is supported by the finding that high concentrations of glutamate or ATP cause contractions to tear the tent [30, 73], and also with recent suggestions that the sponge maintains itself in an ‘inflated’ state, and relaxes the tent during sneezes [95].

Sponges definitely reduce their size during ‘contractions’, but what cells are involved? Nickel et al. [83] suggested that in Tethya wilhelma contractions by the canal epithelia reduced the water space, so endopinacocytes are implicated as contractile, but perhaps the lacunae in Tethya are also under tension to remain open, and collapse when the rest of the canals contract. Here we describe a sieve-like cell at the entrance to canals, a region that contracts during sponge sneezes. In Tethya wilhelma, a contractile sieve cell (termed a ‘reticuloapopylocyte’) lies at the excurrent side of choanocyte chambers and is implicated in local control of pressure through the chambers [45]. The authors compared that cell to porocytes. Some porocytes have more than one opening. Similarly sieve cells in E. muelleri incurrent canals are perforated by many openings forming a basket-like structure that hangs down into the incurrent canal.

The tips of oscula are also contractile. Weissenfels [120] proposed that oscula are formed by porocytes: at first one porocyte forms the osculum, and then others are added until the entire rim of the osculum is made of adjacent porocytes, all of which are contractile. We observed the same phenomenon. In a stage 2 E. muelleri two openings can be seen. At first we thought these were ostia, but as flow moves into the sponge at a rate that would inflate a closed chamber within hours, there must be an exit, and so one of the openings must be the exit. Our attempts to visualize flow moving into and out of single openings in stage 2 sponges have not been successful as water moves exceedingly slowly through the sponge at that time.

In a number of demosponges there are elaborate ‘sieve plates’ made of groups of porocytes that form ostia ([102] pg. 261). Taken together, the evidence suggests that the contractile cells in sponges may be a family of porocyte-like sieve cells that share a common function. Whether they also share a common origin from porocytes, as suggested by Weissenfels [120] remains to be determined, but one way to do this would be to look for the type of myosin and its expression in porocytes, oscula, sieve plates, and sieve cells as was done in the reticuloapopylocyte cell of Tethya wilhelma [107].

Cilia and flagella

Sponges are again unusual among animals in having both flagella and cilia on cells in the adult, and where these occur is important for understanding their biology. The terms flagellum and cilium are used in different ways by different fields. Generally, for unicells, flagellum is used for organelles that beat in a sinusoidal waves and cilium is used for organelles that beat with a forward and backward whip-like stroke. In metazoans, there is a tendency to refer to both organelles by either flagellum or cilium (or even undulipodia), because, except for the flame cells of flatworms, flagella are only found on sperm in the adults of metazoans. Both organelles consist of the same internal machinery [70] and modelling has shown that the difference in beat depends on how the organelle is anchored [71, 72]. Nevertheless, the type of beat is informative about function, with sinusoidal waves drawing water in at the base of the organelle, and whip-like structures moving water over the surface.

In sponges, flagella are well-known from choanocytes. In contrast, although sponge larvae are typically fully ciliated, in the adult sponge, cilia are less well-known. However, cilia are common on apopyle cells at the exit to choanocyte chambers [22, 23, 45, 57], on some cells in canals, and all cells in the osculum of sponges [73]. Many authors have referred to the organelles on apopyle cells as flagella, and considered them as intermediate cells between pinacocytes and choanocytes (e.g., [22, 23, 58, 59]). However, here we show that the cilia on apopyle cells in E. muelleri beat with an erratic flick that is most similar to that of cilia on the tentacles of the anemone Nematostella vectensis, at 1 Hz, and much slower than the beat of cilia on multiciliated epithelia from a range of animals (6–10 Hz) or flagella (20–30 Hz) (Additional file 2; Video S1) (Leys et al. data not shown). These are clearly not flagella, and they are very likely sensors, possibly of flow, pressure or some fundamental difference in the environment inside and outside the choanocyte chamber. There are four apopyle cells in E. muelleri, which together touch the tips of the choanocyte collars on one side, and on the other side bind the chamber to the pinacocytes. In homoscleromorph sponges all epithelia are fully ciliated, including the apopyle cells [31, 94], but so far cilia have not been reported from the chamber apopyles of either Calcarea or Hexactinellida.

Intriguingly two genes were identified as marking apopyle cells in Spongilla lacustris, C101118_g1 and C85989_g1 (also referred to as C85919_g1) [81]. Following the understanding that these cells must, in some form, sense and provide feedback on flow through each choanocyte chamber, an excellent next step would be to identify the same genes in Ephydatia and use RNAi or equivalent experiments to test their function.

Whether ciliated cells in sponges are derived from choanocytes is a question of long-standing interest. The cone cells in Oscarella lobularis have been proposed as ‘modified’ choanocytes as they appear to be a ‘chimera’ of choanocytes and pinacocytes [11]. There, cone cells are described as having a ruffled side facing the choanocytes with the membrane extensions appearing similar to microvilli, and a smooth side with a cilium facing the canal. Boury-Esnault [11] suggests they either come from pinacocytes or from choanocytes, and comments that as Diaz [24] found evidence that choanocytes could dedifferentiate into pinacocytes, so that apopyle cells were an intermediate state. Another view proposed recently [81], is that apopyle cells are a type of secretory cell together with choanocytes and an amoeboid cell that occurs in the mesohyl. Here, our single-cell transcriptomes of choanocytes showed significant reduction in number of genes expressed in all active choanocytes sampled. Our transcriptome profiles therefore suggest that choanocytes are highly differentiated and unlikely to differentiate into other cell types, in contrast to what has been previously indicated [40].

Mesohyl cells and cell motility

We identified six distinct motile cell types in the mesohyl of E. muelleri, all of which have previously been described in freshwater sponges. These included archaeocytes, two sizes of polyblasts, two sizes of cystencytes, and sclerocytes which are moved by other cells [82]. Here we distinguish these cells based on size classes, differences in labelling with tubulin, behaviour and transcriptomic profiles.

While any motile cell appears ‘amoeboid’ and therefore all (except the sclerocyte) have been termed amoebocytes by other researchers, our observations show that only polyblasts appear classically amoeboid and are constantly changing shape. Archaeocytes have a distinct nucleolus and use lamellipodia to crawl, but at every moment they retain a rounder shape than other amoeboid cells. Archaeocytes are sluggish in comparison to polyblasts, but more agile than cystencytes. The name polyblast comes from Tuzet and Pavans de Ceccatty (1955, in [102]) and is defined as having a low volume of cytoplasm to nucleus, but not being phagocytic, unlike archaeocytes. Many cell types arising during wounding seem to become polyblasts, and some authors have suggested polyblasts are an intermediate stage of choanocytes as they dedifferentiate into archaocytes [18], which is discussed below in the context of choanocytes. Polyblasts in E. muelleri are fast moving cells, but whereas in Type II polyblasts (small) a network of tubulin labelled strongly, there was no such tubulin network in Type I polyblasts (large). The tubulin network suggests secretion of vesicles at the cell perimeter. In some marine sponges polyblasts are also considered to be collencytes due to their high nuclear to cytoplasm ratio ([102], pg. 83). Although some earlier authors believed collencytes were too difficult to define, Borojevic [9] considered collencytes to be pinacocytes capable of secreting collagen. If this were the case, then a cell in E. muelleri seen secreting extracellular matrix under the pinacoderm (e.g., Fig. 2D), would provide a link to the two types of mesocytes found in S. lacustris that were related to pinacocytes under the umbrella category ‘endymocytes’ [81]. Perhaps polyblasts are a category of cell that maintains both the choanoderm and pinacoderm epithelia. Some polyblasts have a nucleolus, and some archaeocytes appeared to transition into polyblasts as described below.

Cystencytes are a cell type with a large spherical inclusion and what is aptly described as a foot, the cytoplasmic base that moves the cell around. The cyst in these cells is said to contain glycoprotein and stain for acid phosphatase, although Pottu-Boumendil (1975, in [102]) found they were not involved in digestion. In our videos the inclusion appeared translucent in phase contrast, it did not change in size, and it was moved around sluggishly behind the rest of the cell as the cystencyte shuffled between choanocyte chambers, along epithelia, and even up and down the mesohyl of the osculum. In cross sections of the choanosome imaged by light and scanning electron microscopy, cystencytes and archaeocytes are both found adjacent to chambers. In both small and large cystencytes a microtubule network labelled strongly in the foot suggesting that these cells have a secretory role, which agrees with Simpson’s [102] conclusion that they deposit ‘ground substance’ (matrix) and is supported by the differential expression of genes associated with vesicular factors like vesicle-associated membrane protein (VAMP; Em0008g145a), syntaxin (STX; Em0020g348a), and charged multivesicular body protein (CHMP; Em0023g656a).

Archaeocytes are the only cell type that we saw dividing in situ. When this happened, the cell stopped crawling, rounded up, and divided into two cells, each of which then continued on its way but now looking much like polyblasts. Some archaeocytes divided once and then a second time and became the ‘nucleus’ for a growing choanocyte chamber, as observed by Tanaka and Watanabe in Ephydatia fluviatilis [111]. These only did two or three such divisions, rarely more, and then subsequent growth of the chamber was by small round cells (from division of a neighbouring archaeocyte) joining the group and inserting themselves in between two choanocytes and differentiating into a choanocyte. This process was described previously in Spongilla lacustris another freshwater sponge [53]. In all the videos we captured we never saw a choanocyte sliding out of a chamber and becoming amoeboid.

A number of genes associated with stem cells are expressed in archaeocytes. For example, the expression of CPEB1 (Em0010g51a) in these cells indicates transcripts may be undergoing cytoplasmic polyadenylation. CPEBs are associated with both germline and non-germline undifferentiated cell types [131]. In the latter case CPEB-mediated cytoplasmic polyadenylation has been proposed to be responsible for quick activation upon extrinsic stimuli [131]. It may be that this activation mechanism and corresponding post-transcriptional regulation is required for the transition of archaeocytes to choanocytes. It will be interesting to investigate whether CPEB proteins are regulating the polyadenylation of transcripts in these cells as cytoplasmic polyadenylation has not been investigated in sponges. Another highly expressed gene in archaeocytes (albeit not significant relative to cystencytes) is an H3.3 gene (Em0020g572a.t1), which is also associated with germline and non-germline undifferentiated cell types [114]. Other stem cell-related genes expressed in archaeocytes include NOG, a number of NOP genes, a GNL gene, two DPP genes, PCNA, pumilio, a number of DDX genes including a vasa homolog, a number of FOX genes, and a PIWI gene (Table 2). Together, these results suggest a highly conserved set of genes involved in preserving “steminess” in these sponge cells.

Table 2 Stem cell-related genes expressed in archaeocytes

Choanocytes

Choanocytes are a type of epithelial cell in that they form an external surface facing the external environment, yet they lie inside the sponge. In comparison to the other cell types we studied, we found all choanocyte transcriptome profiles clustered to express a limited set of genes. The unexpected observation that there are far fewer genes observed in the choanocyte transcriptomes relative to the other cell types is consistent with choanocytes being in a more differentiated state than cystencytes and archaeocytes [133]. However, as we did not anticipate this result, we did not use a spike-in (which are not standard in RNA-Seq analysis) and so quantitative data are needed to definitively determine whether our hypothesis is supported. We also considered the possibility that the low volume of cytoplasm in choanocytes meant that we may have only sampled a random subset of the many transcripts present. However, were that the case, we would not expect our PCA to indicate a highly robust and repeatable expression profile (Fig. 11D). The inclusion of miniature pools of the same cell type as a single sample should not have a major effect on the results because our analysis uses a normalized metric (TPM) to compare the cell types. Furthermore, our PCA, which shows high levels of clustering between the replicates, also suggests that the normalized gene expression profiles of the single-cell transcriptomes are very similar to the paired or miniature pooled samples.

Other single-cell transcriptome work in other sponge species have suggested two types of choanocytes (e.g., [81, 100]). However, from our PCA (Fig. 11D, S9 Figure), which produced a very tightly clustered set of expression profiles, it appears the choanocytes we sampled were likely a single cell type. If there were indeed two choanocyte types in E. muelleri and they were present in similar numbers, which might be presumed based on the size of the two choanocyte clusters in other studies, it would be unlikely to sequence 5 of the same type (for comparison, the probability of five coin flips yielding the same result would be 0.0625).

Choanocytes are very small cells, with high turnover rates [20]. In Stage 5 sponges of Spongilla lacustris, choanocytes originate from archaeocytes when a founder cell first divides into 6–8 cells, after which subsequent choanocytes are formed by archaeocytes dividing and differentiating into choanoblasts, which insert themselves into chambers [53]; we saw similar divisions of archaeocytes to form young chambers in Ephydatia muelleri, and those chambers increased in size by other cells joining from the mesohyl. Choanocytes are lost by sloughing [19] a process that contributes significantly to food for other animals. Some workers suggest that choanocytes can dedifferentiate into other cells (archaeocytes or pinacocytes), as cited in Connes et al. [18] or Diaz [24]. Both of these authors, however, describe regeneration processes: the first study shows that during gemmule formation in the hermit-crab sponge Suberites domuncula, choanocytes dedifferentiate into thesocytes in the process of forming asexual reproduction bodies; the second work describes dedifferentiation and regeneration of wounded tissues of sponges that had been sampled many times for a study of reproduction. Based on those works it seems that this process may only occur during trauma or regeneration. We did not see cells leaving the choanocyte chambers in any of our time-lapse videos of chambers. However, recent work by Melnikov et al. [76] found evidence for dedifferentiation of choanocytes using labels that mark dividing cells (EdU). Most cells first labelled were choanocytes, but once the label was washed out several hours later, by far the greater number of cells that continued to be labelled were in the mesohyl [76]. The authors interpreted this to suggest that either labelled choanocytes had migrated into the mesohyl and continued dividing there, or that a few labelled mesohyl cells had continued dividing while labelled choanocytes were sloughed off [76]. The question as to whether choanocytes migrate into the mesohyl is very interesting and is important to resolve. It may be that cells do this sometimes, usually in sponges, or that it happens differently in different species. One challenge with less transparent models is obtaining a realistic view of cells in their normal non-regenerating state, so developing new methods for this will be important.

Cell nomenclature

A principal goal of this work is to make sponge structure more accessible to readers from a diversity of backgrounds so that more researchers may be encouraged to work on the cell biology of this phylum. The terminology we use reflects the morphological approach, and is supported by a history of functional and morphological studies. Our targeted sequencing of individual cells allows us to map genes onto morphology, and is an approach that could be combined with future droplet or bulk RNAseq as well as functional studies (e.g., feeding) together with single molecule labelling to allow a better comparison of cell type-specific markers across species. Without these data we were not able to distinctly identify cells in the category of myopeptidocytes or metabolocytes as proposed by Musser et al. [81].

One cell type that we do describe and which maps to a newly described cell in Spongilla lacustris is the central cell. In Ephydatia muelleri we found central cells in roughly 5% of chambers, and high-speed imaging showed they lie over the tips of the collars, moving extremely slowly, often sitting for hours in a single chamber (Type D central cell of [89]). The proposed renaming of the central cells as a ‘choano-neuroid cell’ was based on expression of a putative ‘pre-synaptic’ gene set identified from single cell RNAseq analysis [81]. In an extensive review of central cells in sponges, Reiswig and Brown [89] concluded that most likely their function was to control water flow in individual chambers allowing other cells to clean debris that accumulated on the incurrent side of chambers. They were proposed to do this by immobilizing flagella beat, and yet our videos show that flagella beat remains undisturbed while central cells sitting on top of the collars are jiggled by the beat of the collars. While it is not yet understood what the functions of central cells are, the conclusion that these cells are neuroid is not yet supported by functional data. The terms ‘pre-synaptic’ and ‘post-synaptic’ are applied to generic families of genes involved in vesicle trafficking, docking, and secretion (see discussion in [55]). As sponges lack neurons and synapses and the central cells are migratory, it is unclear how these could be involved in rapid information transfer as occurs in other neuroid tissues [74, 110, 130]. Previous studies also found no evidence of phagosomes in central cells, concluding they were not likely feeding on material on the collars or in the chambers [89]. Pavans de Ceccatty [84] concluded that central cells were independent of other cells, and agreed with Dubosq and Tuzet [26] that they were most likely involved in immobilizing flagella and removing degenerating choanocytes. Central cells are fascinating in their ability to exist and function outside of the protection of the sponge’s tissue. Further study of their motility, secretory, or phagocytic activity is highly needed to determine the true function of these unusual cells.

Conclusion

This study provides a detailed morphological and transcriptomic atlas of cell types in the freshwater sponge, Ephydatia muelleri. The discovery of a polarized external epithelium, a new contractile sieve cell, and description on various motile and non-motile cilia types in the excurrent canal system enhances our understanding of sponge biology. The distinct expression profiles of cystencytes, choanocytes, and archaeocytes offer valuable insights into the nature and transcriptomic activity of these cell types. Our findings contribute significantly to the broader goal of understanding the evolutionary origins of animal cell types, tissues, and body plans.