Morphological changes during tissue regeneration in Halichondria panicea explants
We observed extensive tissue regeneration during the transition of explants from ‘no-osculum’(NOE) to ‘water pumping’ stages with an osculum (PE; Fig. 1A-B). In NOE, chambers were disassembled, as seen by individual choanocytes in the mesohyl (Figs. 2A–C and 3A, C–D, F), some of them without a flagellum, which may have been disarranged. The aquiferous system was almost completely disarranged (Figs. 2A–B, 3B), although few canals seemed to remain intact in the mesohyl. Alternatively, these canals might have been newly formed. All 3 NOE individuals lacked subdermal cavities, the exopinacoderm was immediately followed by a high-density cellular organisation of the mesohyl, with a high degree of cell compaction (Fig. 2A), and it completely lacked choanocyte chambers or only few disarranged choanocyte chambers (Fig. 2A–B, 3C–D). The structural organisation of adult sponges (PA), previously described by Reiswig [42] and Barthel [43] and highly similar to PE sponges (Fig. 2D–F), was completely different to the disorganised tissue we found in NOE (Fig. 2A, 3A). During tissue remodelling, NOE sponges had mostly archaeocytes, amoeboid cells in the mesohyl (Fig. 3A–B, E–F), and few choanocytes, both with (Fig. 3C) and without a flagellum (Fig. 3D; Additional file 3: Supplementary Table 4). Choanocytes without flagella still retained similar features as those with the flagellum, including anucleolated nucleus and digestive vesicles, although some showed a larger cell body and a nucleolus (Fig. 3F). Such high numbers of archaeocytes and ameboid cells were not observed in PE and PA sponges (Additional file 3: Supplementary Table 4). In PE explants, choanocytes were observed in clusters or forming new chambers in the mesohyl (Fig. 2C). Also at the PE stage, an osculum and subdermal cavities were present (Fig. 2D), along with complete functional choanocyte chambers (Fig. 2E–F), and an aquiferous system including interconnected in- and excurrent canals (Fig. 2F).
Assembly, mapping and differential gene expression
We obtained 327 million raw reads that were subsequently trimmed (Additional file 1: Supplementary Table 1). The reference assembly of Halichondria panicea was constructed with approximately 245 million clean reads (Additional file 4: Supplementary Fig. 1A) and resulted in 208,269 assembled transcripts (N50 1047 nt, see Additional file 1: Supplementary Table 1), corresponding to 67,844 protein-coding genes. Our BUSCO analysis recovered 99% of the eukaryotic conserved cassette and 96.1% of the metazoan cassette (Additional file 1: Supplementary Table 1). Between 30 and 78% of the clean reads were mapped uniquely to the reference transcriptome per sample (Additional file 1: Supplementary Table 1). Only 37% of the assembled sequences obtained a diamond hit against the refseq database, and 99% of those with a diamond hit had associated GO terms (Additional file 1: Supplementary Table 1).
Two samples (NOE_rep4 and PE_rep1) were discarded because their expression pattern was extremely divergent with respect to the other samples of the same stage (Additional file 2: Supplementary Figs. 2). These disparities were probably due to different timings of sampling and/or developmental stages during the experimental design. Three samples from NOE (NOE_rep1, 2, and 3) were cultured in May, 2016, whilst NOE_rep4 was cultured in September, 2016. It is possible that the sponges sampled for the preparation of explants in May were in a different physiological state than those collected in September, which lies within the reproductive period of this species in the same collection area [44]. Thus, sampling of our H. panicea sponges during the reproductive season (autumn) may have added too much noise to the analysis. Furthermore, we also observed a much faster development of the osculum in explant PE_rep1 (~ 10 days) compared to explants PE_rep2, PE_rep3 and PE_rep4 (17–18 days), which could explain the different expression patterns between these samples (Additional file 5: Supplementary Fig. 2). Removing NOE_rep4 and PE_rep1 resulted in more similar expression levels among samples in each stage (Additional file 4: Supplementary Fig. 1B–C).
A total of 1051 transcripts were upregulated (only 293 with significant diamond hit ID against the databases, hereafter referred as hit ID) and 638 were downregulated (only 261 with hit ID) in PE vs. NOE (Fig. 1D, Additional file 6: Supplementary Table 2). When comparing NOE vs. PA, 1328 transcripts were upregulated (only 243 with hit ID) in PA and 1121 transcripts downregulated (only 488 with hit ID) in PA (Additional file 6: Supplementary Table 2). In contrast, only 91 transcripts were upregulated (only 24 with hit ID) in PE when compared to PA, and 312 downregulated (only 48 with blast ID) (Additional file 6: Supplementary Table 2).
Changes in gene expression during tissue regeneration
Ciliogenesis and ECM remodelling in NOE
In demosponges, transdifferentiation, including both epithelial–to–mesenchymal (EMT) and mesenchymal–to–epithelial (MET) morphogenesis, is the most widespread regenerative processes [6], with pinacocytes and choanocytes (flagellated cells) acting as main cell sources in the EMT morphogenesis and mostly archaeocytes transdifferentiating during MET morphogenesis. Although cell proliferation increases progressively during regeneration in other sponge species [21, 22], such process seems not to be the central mechanism involved in the regeneration of the wounds. Here in the NOE stage of H. panicea, we notably observed a massive decrease in canal structures and almost no choanocyte chambers (Additional file 3: Supplementary Table 4). The few choanocytes we observed lacked flagella and were part of disarranged chambers or free (Figs. 2A–B, 3A–C; Additional file 3: Supplementary Table 4). In addition, in NOE samples, we noticed a high density of archaeocytes (Fig. 2A–B) containing numerous digestive vesicles (Fig. 3E–F). Indeed, there were twice the number of archaeocytes per mm2 in the tissue of NOE individuals compared to both PE and PA (Additional file 3: Supplementary Table 4). In contrast, in PE samples, we found chambers with flagellated choanocytes and canals lined with flattened pinacocytes (Fig. 2C–D), indicating the progressive reconstruction of the aquiferous system. Furthermore, in PE, the formation of an osculum (Fig. 1A) indicated the presence of a fully developed aquiferous system (Fig. 3C-D), confirmed by our observation of active water-pumping in these explants [28, 45]. It remains unresolved if the choanocytes without flagellum observed in our morphological study of NOE (Fig. 3D, F) migrate through the mesohyl until they find chambers to join and then they rebuild their flagellum in the PE stages, like previously shown in explants of Spongilla lacustris [46], or if chambers in PE consist of entirely ‘new’ choanocytes which differentiated from a single or a conundrum of archaeocytes like it occurs during metamorphosis or asexual reproduction in other sponges (e.g., [47, 48]). However, assembly and disassembly of the flagellum is quite a common phenomenon in all organisms, from single-cell algae and sponges to mammals [49]. In Chlamydomonas reinhardtii, for instance, many environmental stimuli can trigger the disassembly of the flagellum [50]. Loss of flagellum could either be achieved by either resorption or deflagellation, which is in fact triggered by stress in many invertebrates [51]. In Aplysina cavernicola and Oscarella lobularis, during metaplasic regeneration, new flagella were formed in old choanocytes using their basal bodies which were not completely resorbed [5, 21]. In our RNASeq results, we found highly enriched functions in NOE compared to PE related to cilium formation, including polymeric cytoskeletal fibre, cilium, motile cilium, ciliary part, microtubule, cilium assembly, cilium organisation, cell projection assembly, and microtubule-based process (Additional file 4: Supplementary Fig. 1D). Many genes responsible for the correct formation and function of the flagellum, as previously observed in proteotranscriptomic studies of ciliogenesis of flagellated unicellular eukaryotes (reviewed in [47]), were upregulated in NOE, including tubulin-α and tubulin-ß, dyneins, outer dense fiber 2, radial spoke head 6, rootletin, cilia- and flagella-associated proteins, and kinesin KIF9 (Fig. 4, Additional file 7: Supplementary Fig. 4). Kinesin and dynein proteins are important for intraflagellar transport (IFT) (Fig. 4 Additional file 6: Supplementary Table 2, Additional file 7: Supplementary Fig. 4), while tubulins are the main structural proteins of the flagellum microtubules; radial spoke heads are connecting proteins between the outer and the inner microtubules, and rootletins are part of the anchoring system and fundamental for maintenance of the flagellum [52]. Interestingly, during the course of flagellar assembly, α- and ß-tubulins are upregulated in the model system Chlamydomonas reinhardtii, but return to basal levels once the flagellum is formed [52]. It is important to note here, that while several transcripts of tubulin-α and tubulin-ß were upregulated in PE and PA, other transcripts with the same annotation were upregulated instead in NOE (Fig. 4, Additional file 6: Supplementary Table 2, Additional file 2: Supplementary Fig. 3C). Other genes related to the structure and maintenance of the flagellum were also found upregulated in NOE, such as MAPK MAK MRK overlapping kinase (MOK) which negatively regulates cilium length [53], and tektin-1, which is involved in the stability and structural complexity of the flagellum [54].
Many other genes traditionally related to the maturation of sperm and capacitation were also upregulated, such as parkin, testis-specific serine threonine- kinases 1 and 2, and cation channel sperm-associated 2 [55]. Overall, these molecular and morphological results indicate that flagella are formed in choanocytes in NOE, regardless of them being newly formed choanocytes (indicating cell proliferation) or pre-existing cells without flagella, as observed in our morphological analysis of NOE.
Besides ciliogenesis, the synthesis and deposition of the extracellular matrix (ECM) are fundamental to maintain the tissue polarity and function. Many markers of ECM deployment were upregulated in NOE (Additional file 4: Supplementary Fig. 1E and Additional file 5: Supplementary Fig. 2), but others were engaged later (in PE) when reorganisation of the tissues and the aquiferous system was completed (Fig. 5). This is similar to the regeneration processes described in other demosponge species, such as O. lobularis and H. caerulea [5, 23], where ECM condensation and reorganisation occur very early in the process (sometimes only 2–3 hr. after injury) but can extend until the wound is completely healed.
Cell proliferation and cell survival
Interestingly, the GO categories (within Biological Process) most frequent among the DE genes in the pumping explant’ stage (PE) were ‘response to ethanol’ and ‘macrophage differentiation’ when compared to the ‘no-osculum explant’ stage (NOE; Fig. 6, Additional file 4: Supplementary Fig. 1E), and ‘apoptotic signaling pathway’ when compared to the intact, actively pumping sponges (PA) collected from the field (Fig. 6, Additional file 4: Supplementary Fig. 1F, J–I); this latter category contains mostly subcategories related to immune responses and cell death or survival. Similarly, during regeneration in the demosponge Halisarca caerulea, apoptotic processes were found to be fundamental for regenerating injured tissues [23, 25]. Within these groups of DE genes, we identified a number of homologs of genes that have been linked to regulated cell death in sponges and other animals.
Despite their position as the possible sister group to all animals, sponges possess many of the same characteristic genes implicated in regulating cell death processes in other animals. These include the apoptotic initiator and executioner proteases known as caspases; pro- and anti-apoptotic Bcl-2 family members, which trigger or block the release of mitochondrial cytochrome C upstream of caspase activation in intrinsic apoptosis; and tumour necrosis factor receptor (TNFR) family members [56], which trigger caspase activation through the action of adapter molecules in extrinsic apoptosis (Fig. 6, Additional file 6: Supplementary Table 2 and Additional file 8: Supplementary Table 3).
In sponges, regeneration often involves cell proliferation and migration of the stem cell pool (archaeocytes and choanocytes), but transdifferentiation has also been described [5, 6, 57, 58]. Here, we have identified several markers of cell proliferation that appear to be upregulated during relatively early phases of regeneration (NOE) (Additional file 6: Supplementary Table 2, Additional file 4: Supplementary Fig. 1G). These include genes involved in nuclear division as well as TNF receptor-associated factor 5 and Nicotinamide phosphoribosyltransferase, which play important roles in cell proliferation in vascular tissues in other metazoans [59, 60]. Several caspases were downregulated in NOE relative to PE and PA, as was the inflammasome component NOD-like receptor family member (NLR), which detects microbe- or cellular stress-associated molecules and initiates inflammatory responses via caspase activation [61]. In addition, we identified a gene downregulated in NOE that exhibited overall sequence similarity to the bilaterian Fas-associated death domain (FADD) genes; FADD proteins interact with members of the Tumour Necrosis Factor Receptor (TNFR) family (which was also downregulated in NOE) and activate downstream apoptotic caspases. Other sponges encode the pro-apoptotic protein DD2, which comprises two death domains and has sequence similarity to both FADDs and the Fas ligand [62]. The FADD-like sequence encoded by the Halichondria panicea transcriptome was partial, such that we could not determine whether it presents the domain architecture of DD2 or of the FADDs found in other animals.
Consistent with these findings, genes upregulated in NOE included several proteins implicated in cell survival, such as the apoptosis-suppressing erythroblast macrophage protein (EMP) [63], a member of the hypoxia-inducible domain-containing protein family, which promotes cell survival under anoxic conditions [64], and the anti-apoptotic Bcl-2-homologous protein (BHP) [65, 66]. Cathepsin L, which, in human cells, may inhibit the BHP by cleaving it [67], was also downregulated in NOE.
Several unknown proteins containing death domains, which in bilaterians mediate protein-protein interactions in a subset of proteins that function in cell death and immunity, were identified in the Halichondria panicea transcriptome. Consistent with the diverse functions of death domain proteins in promoting either cell death or survival [68], many of these proteins were upregulated in PA and PE compared to NOE, while one was upregulated in NOE (Fig. 6, Additional file 6: Supplementary Tables 2 and Additional file 8: Supplementary Tables 3). Similarly, the TNF receptor-associated factor (TRAF) family includes both positive and negative regulators of cell death and innate immunity in bilaterian model organisms [69, 70], with some members functioning in promoting either survival or death depending on cleavage status [71]. Some TRAFs were found upregulated (n = 5) or while others were downregulated (n = 15) in NOE (Fig. 6, Additional file 6: Supplementary Tables 2 and Additional file 8: Supplementary Tables 3). Interestingly, during the later regenerative phase in Halisarca caerulea, TNF receptor factor 3 was found to be upregulated [23].
Taken together, our results suggest that at the point when explants have regenerated their epithelia, i.e. 3–5 days after dissection, they exhibit increased expression of genes linked to cell survival, and repression of genes linked to cell death. This is consistent with earlier reports of apoptosis taking place following dissection in the demosponges Aplysina cavernicola [21] and Halisarca caerulea [23]. In those sponges, two waves of apoptosis take place, 6–12 hours and 48–72 hours post-operation in A. cavernicola and 2 hours and 12 hours post-operation in H. caerulea. Thus, in these examples both waves of apoptosis are complete before the time at which our NOE were sampled, consistent with a shift towards a pro-survival expression profile such as the one we report here.
Remodelling of the aquiferous system: Vasculogenesis- and angiogenesis-like processes in explant regeneration
Sponge morphology and body plan are phylogenetically constrained but they are also modulated to a certain extent by the environment [72,73,74]. For instance, changes in the water flow can induce reorganisation of the aquiferous system to improve filter-feeding and oxygen supply [75]. One of the most dramatic events promoting body remodelling in sponges is the removal of the osculum (where the exhalant current is released), which is considered as the primary organiser of the body plan in sponges [75,76,77,78]. In the absence of an osculum, such as in our NOE stage, the aquiferous system experiences a profound reorganisation until water flow is re-established and the current leaves the sponge body through a newly formed osculum. Interestingly, during the regeneration of the aquiferous system in H. panicea in our experiments, several genes of the molecular toolkit involved in vertebrate blood vessel formation were found upregulated (Fig. 5). As mentioned before, such regeneration of the aquiferous system included choanocyte chamber, canal and osculum formation processes. In vertebrates, blood vessels are formed through two processes called vasculogenesis, or de novo formation of endothelial cells, and angiogenesis, in which new vessels sprout from endothelial cells in pre-existing vessels [79]. Hypoxia is the primary physiological trigger for vessel formation, and this process is regulated first by chemotaxis and recruitment of endothelial cells during vasculogenesis, followed by angiogenesis regulated via both stimulators and inhibitors [80]. In general, vertebrate quiescent endothelial cells become activated by angiogenic signals (VEGF, growth factors, cytokines and chemokines [NO_PRINTED_FORM]), which results in degradation of the basement membrane, proliferation and migration of endothelial cells towards the angiogenic stimulus [81, 82]. These endothelial cells finally form a new basement membrane as well as ECM, and recruit perivascular cells to form a new, functional vessel [82] .
In our experiments, genes with important roles in the formation of new blood vessels (or vasculogenesis), such as vascular cell adhesion molecule 1 and E-selectin [83], were differentially upregulated in NOE (Fig. 5). These chemokine proteins attract endothelial cells to start the formation of new blood vessels [83, 84]. Among the vertebrate stimulators of both vasculogenesis and angiogenesis, the most important are the vascular endothelial growth factor receptor 1, i.e. VEGF-R1 [85], together with ephrin and the ephrin ligands [86]. In turn, both vasohibin-1 and the interplay between semaphorin and plexin-D1 count among the inhibitors of the sprouting process [79, 87]. In marine demosponges, flagellar activity and fluid movement (or pressure differentials in the water canal system) act as stimuli for formation and modelling of the excurrent canals [88]. The aquiferous system of canals in sponges is far from being homologous to the vertebrate vessels, mainly (but not only) because sponges do not transport blood in their aquiferous system but environmental water. However, it is not unreasonable to think that the formation of tubular structures to canalise fluids in metazoans is similar in its molecular regulation and hypoxia–related triggering mechanisms across phyla. In fact, sponges such as our single-osculum H. panicea explants, regularly experience internal hypoxia, and even anoxia [89,90,91,92], due to contractions of the body including the osculum and water canal system [45, 93]. Also, regenerating explants without a functional aquiferous system, are unable to supply their interior with oxygen and nutrients via water-pumping, leaving major parts of the explant depleted of oxygen [89]. Indeed, we found both VEGF-R1 (Additional file 2: Supplementary Fig. 3A-B) as well as the angiogenic factors ephrin type B receptor 1 and ephrin type B receptor 2 more expressed (although not significantly) in NOE and PE explants than in PA, where the canals have been already well-formed (Fig. 5). In addition, the angiogenic inhibitor factors vasohibin-1 and plexin-D1 were differentially upregulated in NOE, when the canals were mostly disassembled (Figs. 2A–B, 5). It is important to note that although ephrin receptors are widely spread within sponges, their ligand, ephrin, has yet to be identified [94].
Other genes encoding proteins involved in signalling vessel formation through the Notch pathway, such as jagged 1 and galectin 3 [95] were differentially overexpressed in PE (Fig. 5). Although their role in the formation of the sponge aquiferous system is completely unknown, perhaps this whole array of stimulators and inhibitors of blood vessel formation have an ancestral role in positioning and forming canalicular systems in metazoans. Interestingly, even though invertebrates lack endothelial cells, the VEGF and VEGF-R molecules have proved to play a role in blood vessel formation in many invertebrate species [81, 82]. Furthermore, the VEGF-like factor of the nematode Caenorhabditis elegans is even able to induce angiogenesis in human cell cultures [96], suggesting the ancestral origin and high conservation of the mechanism. Although we cannot spatially resolve the expression of vascular genes in our experiments, the fact that the molecular machinery for vasculogenesis and angiogenesis is upregulated during remodelling of the aquiferous system in H. panicea explants is suggestive of a role in these processes.
In vertebrates, vasculogenesis and angiogenesis are accompanied by a reorganisation of the extracellular matrix (ECM), cell adhesion processes in the vessels, and formation of the basement membrane lining the vessels [NO_PRINTED_FORM]. During our explant regeneration experiments, we found evidence of a profound reorganisation of the tissue related to the synthesis and deposition of the ECM at both the morphological (Figs. 2–3) and molecular levels (Fig. 5, and Additional file 4: Supplementary Fig. 1E). In fact, collencytes, which are specialised cells secreting collagen, were often observed in NOE (Fig. 5B). Interestingly, genes involved in reorganisation of the ECM during angiogenesis, such as integrin and filamin [97], were upregulated in NOE. Furthermore, we found ubiquitous genes in septate junctions, i.e. neuroglian and contactin, upregulated once regeneration has been mostly completed in (PE), as well as in non-manipulated sponges (PA) (Fig. 5A).
Finally, one of the main aspects related to hemal circulatory systems in invertebrates is the general contractility of the vessels, usually accomplished by myoepithelial cells [81, 82]. In Halichondria panicea, periods of osculum closure and body contractions have been observed in actively pumping explants [93]. The aquiferous system of H. panicea explants seems to be contractile in response to both GABA and L-glutamate stimuli [45], as in other demosponge species [78, 98, 99]. Indeed, in sponges, the Rho/ROCK pathway, its downstream components, and the myosin regulatory light chain (MRLC) of myosin II, which mediate actomyosin contractility [100, 101], regulate the correct formation of the aquiferous system [102]. In our study, myosin genes were highly upregulated in stages with an (already formed) aquiferous system, i.e. PE and PA (Fig. 4) and were found to belong to myosin I and II families (Additional file 2: Supplementary Fig. 3C). Also, the gene twitchin, which, coupled with actin and myosin, is fundamental for the catch phase of contractions in molluscs [103], was highly upregulated in PE (Fig. 4, Additional file 6: Supplementary Table 2). Although we cannot resolve spatially the expression of these myosin genes, they are known to be expressed in pinacocytes lining the water canals and, to some extent, in archaeocytes, as previously shown in single-cell transcriptomic studies [104].
Response to stimulus once the osculum is formed
Sponge contractility and in general, sponge responses, are mostly triggered by amino acid signaling molecules [105]. In pumping explants (PE), GO categories involved in environmental stimuli sensing, including G-protein coupled receptor (GPCR) signalling pathways, response to biotic and external stimulus, response to other organisms, and transmembrane signalling receptor activity, were highly enriched (Additional file 4: Supplementary Fig. 1D, H). In our DE dataset, we found heat shock proteins, GABA receptors, mGluRs, and nitric oxide synthase, upregulated in PE and PA compared to NOE (Fig. 4). Similarly, both GPCRs and GABA receptors were upregulated upon recovery from mechanical damage and grazing in the marine sponge Aplysina aerophoba [25]. In addition, other genes involved in sensing the environment, like several homologs of transient receptor potential cation channel subfamily A member 1 (TRPA), were upregulated in sponges with an aquiferous system, both PE and PA in contrast to NOE (Fig. 4). This family of transient receptor potential ion channels is involved in several processes of environmental sensing, including chemical and temperature signals as well as mechanical stress [106]. In general, environmental receptors including metabotropic glutamate receptors (mGluRs), GABA receptors, and G–protein coupled receptors, are highly diversified in sponges [107]. Interestingly, TRPAs are highly expressed in choanocytes, while mGluRs and GABARs are more expressed in pinacocytes [104]. What is most interesting here, is that our data indicates that the sensory activity usually developed by the aquiferous system in sponges [78, 108, 109] is clearly upregulated when the aquiferous system is fully formed.