Single-nucleus profiling highlights the all-brain echinoderm nervous system

Abstract

Metazoans comprise diverse tissues and cell types, each essential for the organismal survival. Most of these types are established early in embryogenesis and persist into adulthood. In indirectly developing sea urchins, however, the continuity between embryonic and adult stages is interrupted by a planktonic larval stage that undergoes complete metamorphosis. While gene regulatory networks controlling embryonic and larval lineages are well studied, the molecular and morphological identities of postmetamorphic cell types remain poorly understood. Here, we reconstructed the cell atlas of postmetamorphic Paracentrotus lividus juveniles using single-nucleus transcriptomics, revealing conservation of regulatory mechanisms. We identified signatures of eight distinct cell type groups and analyzed 29 neuronal families, including 15 unique photoreceptor types. By combining transcriptomics, spatial analysis, and ultrastructure, we identified vertebrate neuronal and opsin homologs expressed across the juvenile. These findings show that the echinoderm body plan is predominantly head-like and exhibits an “all-brain” organization.

A sea urchin is a head with a brain-like organization and a vertebrate-type retinal signature.

INTRODUCTION

During the past decade, single-cell (scRNA-seq) and single-nuclei (snRNA-seq) RNA sequencing has proven to be an indispensable tool for dissecting the cell type repertoire of a wide range of metazoans. These technologies allow for the comparison of molecular signatures and the investigation of evolutionary origins (1, 2). Furthermore, developmental single-cell transcriptomes have enabled the identification of signaling cascades and gene regulatory networks (GRNs) active during embryogenesis, morphogenesis, and organogenesis, revealing previously unknown cell-specific mechanisms (3, 4). This revolutionary approach has uncovered dynamic changes in the molecular signatures of cells constituting various cell types, even challenging traditional cell and molecular biology definitions of what a cell type is (5). Recently, substantial efforts have been made to identify and characterize cell types by combining omic technologies with morphology-based approaches (6).

Sea urchins are marine deuterostomes featuring a biphasic life cycle with free-swimming bilaterally symmetric planktotrophic or lecithotrophic larvae, followed by a radical metamorphosis leading to animals characterized by a pentaradial adult body plan. Sea urchins have been extensively studied in relation to biological phenomena such as oogenesis, fertilization, embryogenesis, morphogenesis, and organogenesis, at both molecular and GRN levels (7, 8). Recently, scRNA-seq studies have focused on understanding the identity and variety of echinoderm cell types in echinoid (9–12) and asteroid (13, 14) embryos and larvae. However, these studies were restricted to premetamorphic developmental stages, which feature less complex tissue compositions compared to postmetamorphic stages. All postmetamorphic echinoderms share synapomorphies that diversify them from other deuterostomes (15, 16). Such examples include their pentaradial symmetrical body plan and their decentralized nervous system organized in a central nerve ring and radial nerve cords along with peripheral nerves. Another one is in regard to their water vascular system (WVS) that is operating through the podia of echinoderms and is involved in many physiological processes including gas exchange, locomotion, and feeding (17).

The GRNs responsible for developing various sea urchin tissues, such as the skeleton, muscles, gut, and nervous system, have been thoroughly investigated in premetamorphic stages (18–21), and it has been shown that they produce corresponding cell types in juvenile and adult animals. Molecular studies on the premetamorphic sea urchin nervous system have shown that its composition is far from primitive, expressing many genes present in the vertebrate central nervous system (CNS), even during embryogenesis (10, 22–24). However, considerable differences in gene expression profiles and potentially in cell type signatures are present in postmetamorphic animals, in line with the vastly different environments that sea urchin larvae and adults inhabit. For instance, it has been demonstrated that diverse cell types and systems support complex behaviors, such as locomotion and feeding, across sea urchin development. In the planktonic sea urchin larva stage, movement and feeding rely on the coordinated beating of densely packed ciliated cells that constitute the ciliary band (25). The directed movement of the cilia enables swimming and directs food particles toward the larval mouth. In contrast, sea urchin juveniles have a benthic lifestyle, and the locomotion and feeding rely on the WVS. The WVS is a complex hydraulic network, composed of specialized cell types and structures (madreporite, canals, podia), that facilitates movement and supports feeding (25). Apart from cell type differences, also changes in gene expression have been reported by comparative transcriptomic analyses that have identified differential expression of distinct sets of regulatory and effector genes during the sea urchin larva to juvenile transition (26).

An interesting example of the cell type divergence between pre- and postmetamorphic sea urchins is in regard to photoreceptor cells (PRCs) (22, 27–30). In premetamorphic sea urchins, we and others have previously identified the expression of both a Go (opsin3.2) and an echinopsin (opsin2) (30–32). These are localized in neurons flanking the larval apical organ and in mesodermally derived cells placed at the tips of the larval arms, respectively, and are implicated in regulating the sphincter contraction within the digestive tract (30–32). Postmetamorphic sea urchins like most echinoderms lack specialized eye-like structures with the notable exception of postmetamorphic sea stars, which have compound eyespots at the tips of their arms made of lensless ommatidia. These eyespots are capable of true image formation, albeit at a low spatial resolution (33). In contrast, light perception in postmetamorphic sea urchins depends on PRCs distributed across their body. We and others have previously mapped the spatial expression profile of a ciliary opsin (opsin1) across the epidermis and podia, the echinoderm specific opsin2 in the podia region and the rhabdomeric opsin4 in clusters of neurons at the base of the podia and within the podia disc (27, 34, 35). In addition, we have identified the postmetamorphic expression of at least three other opsins at a transcriptional level, although their cell type domains remain undefined (35).

In this study, we investigated the cell type diversity in postmetamorphic juveniles of the sea urchin Paracentrotus lividus, with an emphasis on neuronal and PRC types. We used snRNA-seq to generate a cell type atlas, through which we explored cell type identities, examined the conservation of their genetic wiring, and compared their signatures to those found in larval stages. Spatial gene expression and ultrastructural analyses were used to characterize cell type identity at the (ultra)structural level. Our findings contribute to understanding the evolution and gene regulatory mechanisms controlling development and cell type function in echinoderms and across metazoan phyla.

RESULTS

Cell type identity and diversity in P. lividus juveniles

Sea urchin embryos and larvae have been extensively used as research subjects to decipher key developmental processes and to identify GRNs that drive the formation of distinct cell types, many of which are either evolutionary conserved or provide missing links essential for understanding the evolution of complex organs. Recent advances in single-cell transcriptomics have enabled the generation of several scRNA-seq and snRNA-seq atlases for premetamorphic stages of several echinoderm species (9, 10, 12, 13). However, very little is known regarding the cell types of the postmetamorphic sea urchin juvenile, and one reason for this is the difficulty to extract intact single cells primarily due to the presence of the extensive biomineralized endoskeleton.

To bridge this gap, we devised a single-nucleus extraction protocol (detailed protocol in Materials and Methods) and performed snRNA-seq on 2 weeks postmetamorphosis (wpm) P. lividus juveniles (Fig. 1, A to C). The choice of that developmental stage was based on the fact that 2 wpm allows enough time for ensuring the absence of any larval tissue remnants and the opening of the mouth and functionality of the adult digestive tract, thereby increasing the probability of highly diversified cell types to be present. Together, we constructed three snRNA-seq libraries, each containing transcripts corresponding to nuclei originating from six juveniles per library, that emerged from two genetically diverse animal cultures. Once the libraries were sequenced and mapped, computational analysis was carried out in RStudio using the Seurat pipeline. Initially, the three different snRNA-seq libraries were merged without performing any integration methods. While the overall distribution of nuclei per cluster was uniform we detected two clusters (20 and 32) that consisted of nuclei originating exclusively from libraries 1 and 2 and 3 and 4, respectively, an indication of batch effect (fig. S1A). To overcome this and to ensure that the clusters identified correspond to distinct cell type families and not batch-related artifacts, anchor-based integration was performed and resulted in the generation of the 2 wpm P. lividus juvenile cell type atlas displayed in Fig. 1. Our juvenile integrated atlas is composed of 25,000 nuclei (5000 nuclei from library 1, 10,000 from library 2, and 10,000 from library 3) distributed to 48 clusters, with nuclei from all three libraries contributing to the formation of the integrated cluster. Each one of the reconstructed clusters corresponds to either distinct cell types or to related cell type groups (Fig. 1 and fig. S1, B and C).

Fig. 1. Cell type diversity of the P. lividus 2 wpm juvenile.

(A) Two wpm P. lividus juvenile seen from the oral side. (B) Two wpm P. lividus juvenile seen from the aboral side. (C) Schematic representation of our snRNA-seq pipeline: from intact juveniles to single-nuclei, sequencing, and computational analysis. (D) Integrated UMAP of the three snRNA-seq libraries. Cells are color-coded in respect to the cell type groups recognized. (E) Dotplot showing the percentage of nuclei and the average expression of genes used to assign cluster identities. Color code is the same as in (D). (F) IHC for myosin heavy chain (MHC), labeling muscles. (G) IHC for synaptotagmin-B (SYT1), labeling neurons. (H) IHC for MSP130, labeling skeleton. (I) Overlay of MHC, SYT1, and MSP130. Juveniles are in oral view. bs, base of the spine; m, mouth; onr, oral nerve ring; p, podia; pn, podia neurons; js, juvenile spine; ps, primary spine; rnc, radial nerve cord. Scale bars, 50 μm.

To evaluate the overall quality of the clustering analysis, we plotted for the average expression and the percentage of nuclei expressing the top five marker genes of the 48 clusters that confirmed the diversification of their molecular fingerprint (fig. S2). Next, to find the identity of the generated clusters, we took advantage of (i) the extensive echinoderm literature providing us with a great number of genes known to label distinct embryonic and larval sea urchin cell types (8, 10), (ii) a (limited) number of known gene markers marking sea urchin juvenile cell types (27, 35, 36), and (iii) a gene marker toolkit originating from a recent publication describing tissue-specific gene expression patterns in postmetamorphic juveniles of the sea star Patiria miniata (37). In detail, plotting for the average expression of neural genes (syt1, syt4, syt6-1, syt6-2, syt7, syt7L, and syt9) (10, 13, 24), genes labeling the podia epithelium (pax6, six3, otx, barhl, and irxA) (35, 37), epidermal genes (pax2/5/8, hox1, and gbx) (37), pancreatic acinar-like genes (ptf1a and cpa2L) (38), posterior gut genes (pdx1, cdx, and hox11/13b) (39), WVS-enriched genes (ptch and six1/2) (37), skeletal genes (msp130 and sm50) (36), muscle genes (mhc and trop1) (40, 41), and immune system genes (fcolI/II/IIIf, spTrf, pks1, and gcm) (10, 42), we were able to group the clusters into eight distinct cell type groups corresponding to neurons, epidermis, podia epidermis, digestive tract, muscles, coelomocytes, WVS, and skeleton (Fig. 1, D and E). Noteworthy, looking at the distribution of nuclei per cluster, we detected that the most cell-enriched clusters correspond to epidermal and neuronal cells, while the nervous system itself is represented by 29 clusters of the 48 total ones (fig. S1D).

To perform unbiased reconstruction of the transcriptional relationships of the different clusters, we further performed cluster tree reconstruction analysis that enables the grouping of cell types based on how closely related they are (fig. S3) (10, 43). According to this analysis, a gene was defined as expressed if it showed an average expression greater than three transcripts per million in at least one of the 48 clusters. This analysis showed that most of the clusters corresponding to the nervous system form a well-defined group that contains the coelomocyte cluster (48) as an outlier. Moreover, the skeletal, WVS, and muscle clusters along with the coelomocyte cluster 47 are grouped together, reflecting potentially common mesodermal origins. Last, the three digestive tract clusters form a well-supported group, as well as clusters corresponding to general and podia epidermis, except for the podia epidermis cluster 31.

To confirm the snRNA-seq analysis predictions and our cluster annotation, we next performed immunohistochemistry (IHC) (Fig. 1, F to I). IHC on the sea urchin juveniles using sea urchin antibodies against Syt1 (formerly named SynB for synaptotagmin-B) labeled similarly as previously reported, the entire nervous system, including the oral nerve ring (ONR), the radial nerve cords (RNCs), and the podia neurons, thus confirming the numerous neurons present in the juvenile as well as highlighting the spatial restriction of the nervous system in the oral side of the animal (Fig. 1, F and I). Likewise, the IHC for myosin heavy chain (MHC) highlighted the muscles in P. lividus 2 wpm juveniles both at the level of the spines (Fig. 1, G and I) and within the body cavity, as previously reported (36, 44), and the same applies for MSP130 outlining the skeleton (Fig. 1, H and I) (36).

Once the identity of the cell types was established, we set out to explore how similar is the molecular signature of some of those cell types when compared to the larval ones. To do so, we calculated and plotted the z-score of genes that are either essential components of known sea urchin embryonic or larval GRNs or patterning genes giving rise to diverse cell types as heatmaps (Fig. 2).

Fig. 2. Regulatory signatures of the P. lividus 2 wpm juvenile cell types.

(A) UMAP highlighting the skeletal cluster in the P. lividus snRNA-seq data and heatmap showing the z-score, in the P. lividus snRNA-seq atlas, of genes constituting the S. purpuratus larval skeletogenic GRN. (B) UMAP highlighting the muscle cluster in the P. lividus juvenile snRNA-seq data and heatmap showing the z-score, in the P. lividus juvenile snRNA-seq atlas, of gene markers that are part of the S. purpuratus larval myogenesis GRN. (C) UMAP highlighting the digestive tract clusters in the P. lividus juvenile snRNA-seq data and heatmap showing the z-score, in the P. lividus snRNA-seq atlas, of genes patterning different domains of the S. purpuratus larval gut. (D) UMAP highlighting the neuronal clusters in the P. lividus juvenile snRNA-seq data and heatmap showing the z-score, in the P. lividus snRNA-seq atlas, of genes expressed in the sea urchin embryonic ANE and embryonic and larval peripheral neurons or necessary for embryonic neurogenesis.

First, we interrogated our snRNA-seq data for the expression of the genes constituting the skeletogenic GRN, one of the best characterized echinoderm GRNs to date (Fig. 2A). It has previously been postulated that elements of the skeletogenic GRN are reused in the postmetamorphic adult program (45). Plotting for the skeletogenic transcription factors and the differentiation gene battery, we found that, with the exception of the two transcription factors tgif and hex, and the terminal differentiation spicule matrix/C-lectin domain family member msp130r3, all skeletal embryonic and larval genes (alx1, tbr, ets1, erg, dri, mitf, jun, p16, p16rel1, p58, cara7LA, msp130, msp130r1, msp130r2, mtmr12L, clectin, sm20, sm30E, sm37, sm49, sm50, kirrelL, fgf9/16/20, fgfr2, vegfr, lasp, and frp) are operating in juvenile skeletogenic cells, indicating that the juvenile skeletogenic program is highly similar to the embryonic and larval one (Fig. 2A).

Second, investigating the expression profile of genes involved in the GRN controlling the specification and differentiation of sea urchin embryonic myoblasts (20), we found that core regulatory genes, including foxY, fgfR, sfrp1/5, ptc, myocardin, foxC, soxE, scratchX, twist, foxL1, foxF, and myod2 are expressed in the cluster corresponding to the juvenile muscles, as well as the differentiation genes corresponding to mhc, mlckb, tropomyosin, and troponin1 (Fig. 2B). We also found that many of these genes (foxY, fgfR, sfrp1/5, ptc, myocardin, foxC, soxE, twist, foxL1, foxF, and myod2) further acquired a potential novel role, as we found them to be expressed in cell types outside of the muscle lineage, such as in the nervous system, the epidermis, and the WVS (Fig. 2B).

In the case of the GRNs controlling the diversification of different larval gut cell types and domains, we found their components distributed across the three digestive tract juvenile clusters (39 to 41) (Fig. 2C). In detail, cluster 39 corresponds to the pancreatic acinar-like cells (38), as the typical larval gene markers of that cell type (38), including the transcription factors ptf1a, gataE, hnf4, and blimp1, and the terminal differentiation genes cpa2L, pnlip2/5, amy3, and trypsin were found coexpressed there. We also report the expression of the pancreatic transcription factor pdx1 in the acinar-like cells, a domain of expression that has not been reported so far in the sea urchin larval gut. Cluster 40 matched the transcriptomic profile of the larval midgut domain expressing genes such as foxA, gataE, blimp1, ptf1a, tgif, osr, cdx, pdx1, hnf4, runt1, and manrC1A (10, 39). Last, cluster 41 resembles the embryonic and larval posterior gut, including the pyloric sphincter, the intestine, and the anus, by expressing genes such as foxA, gataE, blimp1, foxD, pdx1, cdx, osr, hnf4, hb9, hox11/13b, and bra (10, 39).

Last, we explored the expression profile of transcription factors and signaling molecules that are essential for embryonic and larval sea urchin neurogenesis (10, 21, 46). Moreover, we looked at the expression of genes necessary for the specification of the anterior neuroectoderm (ANE) and for the differentiation of the serotonergic neurons that arise from that neurogenic domains as well as for genes expressed in the larval peripheral nervous system constituted by the ciliary band and esophageal neurons. Notably, we found most of the genes tested to be expressed by nuclei belonging to neuronal clusters, suggesting that juvenile neurogenesis is reusing the transcription factor repertoire operating during embryonic and larval development (Fig. 2D and fig. S4). Moreover, scoring for the average expression of the neuronal genes grouped into three distinct groups corresponding to embryonic/larval ANE, peripheral neurons and neuronal precursors (neurogenic) allowed us to assess whether patterns of combinatorial expression are present. We opted for the scoring of their expression approach versus plotting for their individual expression to account for a possible signal dilution due to either low expression levels or small percentages of nuclei containing transcripts for those genes. Outside the nervous system, we found the genes corresponding to the three gene groups present in different clusters corresponding to podia epidermis, while the groups composed of genes involved in the embryonic specification of ANE and neurons were also found in the WVS and epidermis cell type groups, respectively (Fig. 2D and fig. S4). Unexpectedly, we found very little overlap between the three different groups scored, suggesting a diversification of neuronal fate, similar to the embryonic and larval one. The ANE-associated genes were found to be differentially expressed in 9 of the total 29 neuronal clusters, with the highest score found in cluster 26, and the most widespread distribution in cluster 29, which, according to our snRNA-seq atlas, contains the highest number of nuclei. Of the ANE genes tested, six3, fezf, rx, frz5/8, and nkx2.1, which are members of an evolutionary conserved neurogenic anterior GRN in deuterostomes (47), were differentially expressed in the juvenile nervous system while absent from the nonneuronal ANE-positive clusters (Fig. 2D).

This observation prompted us to further investigate the expression and distribution of genes that in bilaterians are involved in the establishment of the anterior-posterior body axis. Since these genes are typically involved in the establishment of the bilateral body plan, we emphasized on the expression domain of these genes in clusters constituting the main body of the animal such as epidermis and nervous system. To this end, z-scores for genes involved in the establishment of the head, head/trunk boundary, and trunk regions were calculated and plotted as a heatmap (Fig. 3A). This analysis revealed that the anterior head (frz5/8, nkx2.1, rx, otp, sfrp1/5, six3, and hh) and posterior head (irxA, dmbx1, tlx, otx, barhl, and pax6) region genes were all widely expressed in the clusters corresponding to juvenile nervous system and epidermis, including the podia one. To test the predicted expression patterns, we performed hybridization chain reaction (HCR) for the anterior head markers fzd5/8 and nkx2.1 as well as for the posterior head markers irxA, otx, and barhl. Our HCRs showed that fzd5/8 and nkx2.1 are expressed in cells within the ONR and RNC regions (Fig. 3, B to E), while irxA, otx, and barhl are expressed in cells corresponding to the podia nervous system. Moreover, irxA and barhl were also found expressed in the RNC regions while otx in the ONR (Fig. 3, F to K). Regarding the juvenile expression of the genes involved in the establishment of the head/trunk boundary (engrailed, gbx, pax2/5/8, wnt3, and hox1), our snRNA-seq atlas predicted that their predominant expression is confined in clusters corresponding to epidermis and podia epidermis. HCR for pax2/5/8 confirmed the single-nucleus predictions revealing expression in epidermal cells related to the main body, the spines, and the podia (Fig. 3, L and M). Last, in line with previous observations in echinoderms, our analysis showed that the genes involved in the specification of trunk (hox3, hox5, hox6, hox8_1, hox8_2, hox11/13a, hox11/13b, and hox11/13c) are enriched in clusters that correspond to nonneuronal and nonepidermal cell type groups such as the digestive tract and the WVS (Fig. 3A). The only evidence of cumulative coexpression of the trunk genes within epidermal or neuronal domains is the coexpression of hox3, hox8_1, hox8_2, hox11/13a, hox11/13b, and hox11/13c in the neuronal cluster 22, which expresses also anterior head, posterior head, and head/trunk boundary markers.

Fig. 3. Localization and balance of the head versus trunk genes in the P. lividus 2 wpm juvenile.

(A) Heatmap showing the z-score, in the P. lividus snRNA-seq atlas, of genes involved in the establishment of the anterior-posterior body plan axis. (B and C) HCR for fzd5/8 with (B) and without (C) nuclei labeling [4′,6-diamidino-2-phenylindole (DAPI)]. (D and E) HCR for nkx2.1 with (D) and without (E) nuclei labeling (DAPI). HCR for irxA with (F) and without (G) nuclei labeling (DAPI). HCR for otx with (H) and without (I) nuclei labeling (DAPI). HCR for barhl with (J) and without (K) nuclei labeling (DAPI). HCR for pax2/5/8 with (L) and without (M) nuclei labeling (DAPI). pd, podia disk. Scale bars, 50 μm. (N) Schematic representation of the expression patterns of the head and trunk genes in P. lividus juveniles. Asterisks indicate previously published expression patterns of related genes in P. lividus juveniles (35). (O) Cartoon summarizing the spatial distribution of anterior head, posterior head, and head/trunk boundary genes in sea urchin.

Following our gene candidate–based analysis, we also used the SAMap pipeline to compare our P. lividus snRNA-seq atlas to the already available scRNA-seq atlas of the Strongylocentrotus purpuratus 3 days postfertilization (dpf) larva (Fig. 4A). This analysis allowed us to explore the similarity of the transcriptional programs operating between the two developmental stages of the two distinct species. We observed that juvenile muscle, skeletal, and digestive tract clusters (respectively, clusters 46, 45, and 39 to 40) align with the respective ones of the S. purpuratus larva, further supporting our previous analysis (Fig. 4, A to C). The larval esophagus, intestine, and anus do not align with any of the juvenile ones. Moreover, we found that juvenile coelomocyte clusters 47 and 48 align with the larval blastocoelar and immune cells, respectively. The larval blastocoelar cells also aligned with two juvenile WVS clusters (42 and 44). We also found a population of juvenile neurons (cluster 10) aligning with the larval cluster corresponding to the aboral ectoderm. The aboral ectoderm larval cluster also aligned with three juvenile general epidermis clusters (35 to 37) and two juvenile podia epidermis clusters (32 and 33). The three general epidermis clusters (35 to 37) also aligned with the larval ciliary band and apical plate clusters, although with lower alignment scores. Likewise, the two juvenile podia epidermis clusters (32 and 33) aligned with the larval upper and lower oral ectoderm clusters. Moreover, we found that the cardiac sphincter cluster of the larva, representing endodermal muscles, exhibits low alignment with the muscle (46) and the digestive tract clusters (40) of the juvenile atlas. The larval anus and intestine cell type families weakly align with the digestive tract clusters 40 and 41 of the juvenile atlas. The juvenile digestive tract cluster 40 also exhibited low alignment with the acinar-like and cardiac sphincter clusters of the larva. Among the clusters with low alignment score, we also found three juvenile clusters corresponding to epidermis (32 and 33) and podia epidermis (36) matching the larval esophagus cluster. Last, we report the low and moderate alignment of most of the juvenile clusters with a larval one (undefined) that has been previously hypothesized to represent undifferentiated/differentiating cells. Although we previously found expression of the larval neurogenic patterning genes in the juvenile nervous system (Fig. 4D), the SAMap analysis failed to find any cluster alignment between the juvenile and larval neuronal clusters. To further evaluate the validity of this comparison and to determine whether the lack of any similarity between the juvenile and the larval neuronal clusters is linked to the finer resolution of the P. lividus atlas, which contains 29 neuronal clusters compared to the single one of the S. purpuratus larval atlas, we repeated our analysis after computationally merging the 29 P. lividus neuronal clusters into a single one (fig. S5). We observed that by collapsing the P. lividus nervous system into a single cluster allows SAMap to successfully align the two single neuronal clusters confirming the transcriptional similarities of the two diverse neuronal programs.

Fig. 4. SAMap cell clusters alignment scores between P. lividus juvenile and S. purpuratus 3 dpf larva.

(A) Mapping of the cell clusters between the snRNA-seq P. lividus juvenile dataset and the scRNA-seq S. purpuratus larva dataset. Black boxes indicate high alignment scores. Color code for the P. lividus juvenile clusters is the same as in Fig. 1E. (B) S. purpuratus 3 dpf UMAP atlas projecting the high-scored P. lividus juvenile cell types. (C) P. lividus 2 wpm UMAP atlas highlighting the high-scored S. purpuratus larval cell types.

Noteworthy, both our analyses failed to detect similarities between any of the juvenile clusters and the larval coelomic pouches, a domain where a germ/stem cell–like molecular signature has been previously reported (Fig. 4 and fig. S5). However, genes typical of germ cells and stem cells, including vasa/ddx4, piwi, seawi, nanos2, tudor boule, msl3, msy, and pum, were found expressed throughout the snRNA-seq atlas in various clusters (fig. S6, A and B). Furthermore, subsetting our P. lividus juvenile atlas by selecting the nuclei that are expressing either of the aforementioned genes generated an atlas consisting of 8797 nuclei equally distributed across all the clusters and cell type groups (fig. S6C). The lack of restriction of these genes in a specific cluster suggests that, at 2 wpm, a committed germline lineage is absent. Plotting of the average expression of genes involved in cell proliferation such as pcna, cyclins (A, B, B3, D, E, and S) blb3h, cdt1, dna2, mcm6, and pst1 highlights the proliferative potential of those nuclei (fig. S6D). Together, we propose that these genes, at this stage, mark stem cell–like populations, similar to what has been initially proposed for their embryonic and larval counterparts (48). On the basis of their widespread distribution across all tissues, we hypothesize that they are involved in building new juvenile structures, regenerating damaged ones and/or are incorporated into somatic regulatory programs.

Characterization of the P. lividus juvenile nervous system

Our results showing that the early larval and juvenile nervous systems are highly diverse in cell type content prompted us to further characterize the juvenile neuronal clusters. As mentioned previously, the juvenile nervous system is composed of millions of interconnected neurons and ganglia (fig. S7A) (35, 44), and this diversity is reflected by the high number of clusters found to correspond to neurons in our snRNA-seq atlas (fig. S7B). To characterize the identity of those neurons, we interrogated our data for the presence of genes involved in the biosynthesis of neurotransmitters and neuropeptides (Fig. 5, A and B). Overall, we found that 19, of the 29 neuronal clusters, contain transcripts for at least one of the seven major neurotransmitter enzyme/transporter families investigated (Fig. 5B and fig. S7C). These genes correspond to tryptophan hydroxylase (tph), tyrosine hydroxylase (th), dopamine β-hydroxylase (dbh) choline acetyltransferase (chat), glutamate decarboxylase (gad), glutamate transporter (vglut), and histidine decarboxylase (hdc), which are respectively involved in the biosynthesis of serotonin, dopamine, norepinephrine, acetylcholine, γ-aminobutyric acid (GABA), glutamate, and histamine.

Fig. 5. Molecular characterization of the P. lividus 2 wpm juvenile nervous system.

(A) UMAP highlighting the neuronal clusters and their identities in the P. lividus juvenile snRNA-seq data. (B) Heatmap showing the z-score of neuropeptides and neurotransmitters across the P. lividus juvenile neuronal clusters. HCR for syt1 (C and D), th (E and F), hdc (G and H), and gad (I and J) with (C, E, G, and I) or without (D, F, H, and J) nuclei staining (DAPI). (K and L) Double HCRs for th (magenta) and gad (yellow) and with (K) or without (L) nuclei staining (DAPI). (M and N) Double HCRs for th (magenta) and hdc (yellow) and with (M) or without (N) nuclei staining (DAPI). Juveniles are whole-mount and in oral view. Scale bars, 50 μm.

On the basis of our snRNA-seq data, we defined a diverse array of neuronal types in the postmetamorphic juvenile nervous system. Those include dopaminergic (cluster 1), serotonergic (clusters 10 and 24), cholinergic (clusters 2, 4, 5, 14, 20, 22, and 23), GABAergic (cluster 11), glutamatergic (cluster 27), and histaminergic (clusters 13 and 21) neurons (Fig. 5B and fig. S7C). In addition, we identified several clusters with a dual neuronal identity corresponding to serotonergic/histaminergic (clusters 3 and 25), serotonergic/cholinergic (cluster 15), GABAergic/cholinergic (cluster 16), and dopaminergic/noradrenergic (cluster 26), indicative of a high degree of neuronal diversity and specialization. Notably, the neuropeptidergic signature varied substantially among all the neuronal types identified, indicating a high degree of postmetamorphic functional diversification. Within the cholinergic neuronal clusters, cluster 2 expressed transcripts for ngfffap, luqin, and kisspeptin; cluster 4 for somatostatin1, glyph3, ppln1, glyph1/2, and ilp1; cluster 5 for somatostatin1, adam/tsL6, luqin, and ilp2; cluster 14 for ilp1, np20, kisspeptin, and ilp2; cluster 20 for secV, np20, and bursiconb; cluster 22 for ngfffap, glyph1/2, secV, np20, and somatostatin2; and cluster 23 for ngfffap, somatostatin1, secV, np20, and somatostatin2. In the serotonergic neuronal clusters, cluster 10 expressed adam/tsL6 and secV, while cluster 24 expressed ngfffap, adam/tsL6, secV, np20, and nesf. The serotonergic/histaminergic clusters showed expression of ppln1, glyph1/2, adam/tsL6, kisspeptin, and bursiconb in cluster 3 and glyph1/2, ppln1, ilp1, adam/tsL6, np20, kisspeptin, nesf, somatostatin2, and an in cluster 25. Regarding the histaminergic neuronal clusters, cluster 13 expressed ngfffap, glyph3, and luqin, while cluster 21 expressed ppln1, secV bursiconb, and an. The serotonergic/cholinergic cluster (15) expressed somatostatin1, ppln1, np20, and bursiconb, whereas the GABAergic/cholinergic cluster 16 expressed bursiconb and ilp2. In the GABAergic cluster (11), we detected transcripts for ilp1, salmfap, np20, and somatostatin2, while the glutamatergic cluster (27) expressed ngfffap, glyph3, glyph1/2, ilp1, secV, np20, and ilp2. The dopaminergic cluster (1) expressed nesf and an, and the dopaminergic/noradrenergic cluster (26) expressed secV and ilp2. Together, these results reveal a remarkable degree of molecular and functional diversification in the juvenile nervous system, shaped by highly combinatorial use of neurotransmitter and neuropeptidergic signaling.

The neuronal identity of 10 clusters (6, 7, 8, 9, 12, 17, 18, 19, 28, and 29) could not be further defined in terms of their neurotransmitters due to the absence of enzymes/transporters involved in the biosynthesis/transport of serotonin, dopamine, norepinephrine, glutamate, acetylcholine, GABA, or histamine. For those clusters, we analyzed the neuropeptidergic molecular signature (Fig. 5B and fig. S7C). We found transcripts encoding for glyph1/2, adam/tsL6, secV, np20, and nesf in cluster 6; ilp1, salmfap, nesf, somatostatin2, secV, and ilp2 in cluster 7; np20 and nesf in cluster 8; gnrh, glyph3, adam/tsL6, and secV in cluster 9; ppln1, ilp1, secV, salmfap, np20, nesf, somatostatin2, and an in cluster 12; np20 and nesf in cluster 17; gnrh, glyph3, glyph1/2, adam/tsL6, secV, np20, an, and ilp2 in cluster 18; ngfffap, kisspeptin, and somatostatin2 in cluster 19; ilp1, secV, np20, kisspeptin, and an in cluster 28; ngfffap, somatostatin1, ppln1, glyph1/2, ilp1, secV, np20, kisspeptin, and ilp2 in cluster 29.

Spatial expression analysis, using HCR for syt1, an echinoderm pan-neuronal marker, and several other neuromodulators, corroborated the predictions from our snRNA-seq data. While syt1 labeled the ONR, RNC, and podia neurons (Fig. 5, C and D), th HCR labeled exclusively the RNC (Fig. 5, E and F), and transcripts for hdc and gad were detected both in the RNC and podia neurons (Fig. 5, G to J). To test whether there was any colocalization of th and gad as well as th and hdc in the RNC, we performed double HCRs for these genes. Both experiments revealed no coexpression but complementary expression patterns in RNC neurons, thus validating the single-nuclei predictions (Fig. 5, K to N). Overall, our findings provide insight into the neuronal types present in the postmetamorphic P. lividus juvenile, highlighting a high degree of nervous system diversity.

PRC signatures in the P. lividus juvenile nervous system

Upon the identification of the neuronal types present in the 2 wpm P. lividus juvenile, we set out to identify the putative PRC types as well as the PRC cell type candidate responsible for sea urchin vision. To do so, we computationally subsetted the fraction of the neuronal cell type families that contain transcripts of at least one of the seven opsin genes encoded in the P. lividus genome and performed further clustering analysis. This resulted in the identification of 15 neuronal PRC clusters (Fig. 6A).

Fig. 6. P. lividus 2 wpm juvenile neuronal photoreceptor repertoire.

(A) UMAP showing the subclustered and reanalyzed opsin-positive neuronal clusters. (B) Dotplot showing the percentage of nuclei and the average expression of the opsin genes encoded in the P. lividus genome and of transcription factor orthologs involved in the establishment of the PRC fate in other animals. (C and D) IHC using a sea urchin specific opsin1 antibody. (E and F) Fluorescent in situ hybridization using an antisense RNA probe against opsin2. Nuclei are stained with DAPI (cyan). HCR using a specific probe against opsin3.2 (G and H) and opsin4 (I and J). Nuclei are stained with DAPI (cyan). pb, podia base; pd, podia disc. Scale bars, 25 μm.

To characterize the molecular signatures of these clusters, we subsequently used genes identified from other animal taxa as essential for the specification of the PRCs and/or as important elements of the PRC identity. Such gene markers included opsins, retinal transcription factors, and phototransduction molecular cascade genes. Each of them had a distinct molecular signature as found by plotting the top 10 differentially expressed marker genes (fig. S8) and are expressing orthologs of genes that are known structural components of cilia (e.g., tubulins and dyneins) (fig. S9), as well as the phototransduction molecular cascade in other taxa (fig. S10).

To understand the opsin distribution across the generated clusters, we plotted for the average expression of the seven sea urchin opsin genes within the clusters (fig. S11), together with transcription factors whose orthologs are either essential for the specification of the PRCs and/or important elements of the PRC identity (Fig. 6B).

Our analysis showed that opsin2 was the most widely expressed opsin, i.e., being present in seven (2, 4, 6, 7, 12, 14, and 15) of the 15 reconstructed PRC clusters, followed by opsin1, opsin3.1, and peropsin expressed in five of the 15 PRC clusters (1, 3, 4, 10, and 11; 1, 3, 6, 10, and 14; and 1, 3, 5, 8, and 11, respectively). opsin3.2 transcripts were detected in only four clusters (1, 3, 9, and 10), and opsin4 and opsin5 were found solely in clusters 9 and 13, respectively.

Moreover, our analysis revealed that many of the PRC clusters are defined by more than one opsin gene. We report the coexpression of opsins 1, 2, 3.1, 3.2, and peropsin in PRCs (1); opsins 1, 3.1, 3.2, and peropsin in PRCs (3); opsins 1 and 2 in PRCs (4); opsins 2 and 3.1 in PRCs (6); opsins 3.2 and 4 in PRCs (9); opsins 3.1 and 3.2 in PRCs (10); opsin1 and peropsin in PRCs (11); and opsins2 and 3.1 in PRCs (14). Nonetheless, there were also some PRC clusters expressing a single opsin, such as clusters 2, 7, 12, and 15 expressing only opsin2, cluster 13 that contains transcripts only for opsin5, and clusters 5 and 8 where only peropsin transcripts were detected.

Transcription factors known to control the establishment and maintenance of the PRC identity in other animals were further detected in all PRC clusters. Except for the PRC cluster 12, all PRC clusters contained transcripts for more than one of these transcription factors, suggesting that PRC specification and differentiation genes are reused during animal evolution and that different PRC types use different transcription factor toolkits. For instance, we found pax6, hlf, otx, six1/2, six3, and isl1 coexpressed in PRCs (1); prox1, tbx2/3, rx, six3, acsc, and irxA in PRCs (2); tbx2/3, neuroD, barhl, hlf, irxA, and pax6 in PRCs (3); rx and isl1 in PRCs (4); prox1, rx, ato8, isl1, acsc, otx, hlf, and pax6 in PRCs (5); brn3, otx, hlf, irxA, and pax6 in PRCs (6); isl1 and tbx2/3 in PRCs (7); ato8, six1/2, barhl, irxA and pax6 in PRCs (8); irxA, hlf, and barhl in PRCs (9); acsc, six1/2, six3, and ato8 in PRCs (10); pax6, brn3, neuroD, rx, and tbx2/3 in PRCs (11); hlf and ato8 in PRCs (13); six1/2, six3 and neuroD in PRCs (14); and six3, otx, and irxA in PRCs (15).

To check which opsin-expressing cell type cluster might correspond to a potential visual PRC candidate, we analyzed spatially the expression domains of some of the opsin genes expressed by the P. lividus 2 wpm juveniles using IHC for opsin1, fluorescent in situ hybridization for opsin2, and HCR for opsin3.2 and opsin4. We found opsin1 expression in the podia (Fig. 6, C and D) as well as in various ectodermal tissues, consistent with previous observations (35). By comparison, opsin2, opsin3.2, and opsin4 were confined to the podia domain (Fig. 6, E to J). opsin3.2 and opsin4 transcripts were found in overlapping domains at the base and in the disc of the juvenile podia, respectively, with additional opsin3.2-positive cells being present in the surrounding epidermis (Fig. 6, G to J). Those expression patterns corroborate the snRNA-seq predictions showing opsin3.2 being coexpressed with opsin4 in the PRC cluster 9 (Fig. 6B).

Reconstructing the PRC map of the sea urchin juvenile nervous system, along with the spatial expression patterns of opsin genes, enabled us to refine our search for cell type candidates that could mediate the sea urchin visual behaviors. While PRCs occur widely throughout large portions of the animal’s epidermis and within other tissues, including those expressing opsin1, opsin2, and opsin3.2, these cells likely mediate a variety of physiological functions requiring photoreception (49). However, sea urchin vision has been hypothesized to depend on specific PRC “units,” as suggested by a neuromathematical model of sea urchin vision (50). From our analysis, only two PRC clusters matched these criteria: the PRC cluster (13) positive only for opsin5 and the PRC cluster (9) exclusively positive for opsin4 (Fig. 6B). opsin5 codes for an echinoderm-specific opsin type, named echinopsin, which has now no known function in sea urchins and is phylogenetically distinct from any of the opsin types typically involved in metazoan vision (51). In contrast, melanopsin-expressing PRCs have previously been hypothesized to facilitate image-forming vision in sea urchins (27, 50). Our analysis revealed that a single PRC subcluster expresses melanopsin (opsin4) in the P. lividus juvenile (Fig. 6B and fig. S10). opsin4 mRNA and protein were detected exclusively in two morphologically distinct cell aggregations: one at the base (Fig. 7A) and another within the disc of the juvenile’s podia (Fig. 7, C and D). To characterize the identity of the opsin4 positive cell type, we plotted the average expression of the top 50 differentially expressed marker genes of PRC (9) (Fig. 7B). Among these genes, we detected transcripts for opsin4, opsin3.2, beta-arrestin, cryptochrome1 (vcry), and nitric oxide synthase, all of which are known components of phototransduction and ultraviolet avoidance pathways.

Fig. 7. Molecular and morphological reconstruction of the P. lividus 2 wpm juvenile opsin4 (melanopsin) cell type.

(A) HCR for opsin4 (magenta) paired with IHC using a sea urchin specific OPSIN4 antibody (yellow). Nuclei are stained with DAPI (cyan). (B) Dotplot showing the percentage of nuclei and the average expression of the top 50 marker genes of the PRC (9) cell cluster. cGMP, cyclic guanosine 3′,5′-monophosphate. Double HCR using a specific probe against opsin4 paired with opsin3.2 (C and D), arrestin (E and F), vcry (G and H), and barhl (I and J). Nuclei are stained with DAPI (cyan). (K) Overview of one of the embedded 2 wpm P. lividus specimens before being processed for SBF-SEM. (L) SBF-SEM isolated section showing the opsin4-positive PRCs at the base of the podia. (M) TEM isolated slice showing the opsin4-positive PRCs at the podia disc. (N) 3D reconstruction of the opsin4-positive PRCs at the base of the podium. Podium 3D segmented membranes depicted in gray. (O) Close-up of the 3D reconstructed basal PRCs. (P) 3D reconstruction of the podium disc PRCs. Disc 3D segmented membranes are in pale green. (Q) Close-up of a 3D reconstructed disc PRC. Ciliary rootlet is labeled in pale blue. n, nucleus; v, vesicle. Scale bars, 25 μm (A and C to J), 4 μm (N), and 1 μm (L, M, and O to Q).

To validate the spatial coexpression of genes predicted by our snRNA-seq analysis (Figs. 6B and 7B), we performed double HCR on whole-mount P. lividus 2 wpm juveniles. We confirmed coexpression with opsin4 for opsin3.2 (Fig. 7, C and D), beta-arrestin (Fig. 7, E and F), vcry (Fig. 7, G and H), and barhl (Fig. 7, I and J). In all of our in situ validations, coexpression was detected in both opsin4 expression domains, the base and in the disc of the juvenile’s podia. Given that our snRNA-seq analysis consistently identified only a single cluster containing opsin4 transcripts (PRC cluster 9) (Fig. 6B), this suggests that the opsin4-positive cells at the base and in the disc of the podia belong to the same cell type family.

To further investigate whether juvenile P. lividus has only one melanopsin-positive cell type or whether there is a convergence of molecular PRC signatures, we examined the morphological characteristics of these two cell populations using serial block face scanning electron microscopy (SBF-SEM) and high-resolution transmission electron microscopy (TEM). Serial sectioning of resin-embedded P. lividus juveniles (Fig. 7K) revealed enlarged microvilli-bearing PRCs in both the basal podia region and the podia disc. Densely packed, large membrane vesicles inside the PRC cytoplasm, which have been shown to carry melanopsin in S. purpuratus (27), were clearly distinguishable in both the basal podia (SBF-SEM; Fig. 7L) and the podia disc regions (TEM; Fig. 7M).

In P. lividus 2 wpm juveniles, we further identified at the base of the podium two bilaterally symmetrical PRC aggregations, each containing three to four PRCs (Fig. 7, N and O), and consistent with previous immunostainings against sea urchin melanopsin in the developing rudiment of P. lividus (35), we here identified in 2 wpm juveniles two bilaterally symmetrical PRC aggregations flanking the podium base, each containing three to four PRCs (Fig. 7, L and N). A three-dimensional (3D) reconstruction of their morphology revealed flask-shaped cells extending a thin dendrite and numerous long microvilli projecting from the apical cell surface (Fig. 7O). Some PRCs bore an unmodified cilium (Fig. 7O), corroborating the snRNA-seq–predicted expression of tubulins, dyneins, rootletin, and cfap91 in this cluster and thus meeting the structural requirement for a Go-Opsin (opsin3.2) potentially functioning in those PRCs. In cells lacking a cilium, a ciliary rootlet was observed, suggesting that the cilium might have been lost during chemical fixation. At this developmental stage of P. lividus, we observed only two to three PRCs in each podium disc (Fig. 7, P and Q) forming epidermal patches with other ciliated cells with presumed receptor function. Except for the differing spatial PRC distribution (single PRCs in the podium disk versus PRC aggregations at the podium base), the cells feature an identical morphology and ultrastructure.

DISCUSSION

Signature, conservation, and divergence of pre- and postmetamorphic cell types

Analyzing the cell type atlas of postmetamorphic juvenile stages of the sea urchin P. lividus enabled us to assess questions such as: What are the cell types constituting a postmetamorphic sea urchin juvenile? How similar are those cell types to the larval ones? What components constitute the juvenile nervous and photoreceptive systems?

Previous studies using molecular markers have already reported on the distribution of juvenile muscle, skeletal, and neuronal cell types in postmetamorphic developmental stages (35, 36, 44, 45). Our analysis confirmed these findings at the single-nucleus level and provided additional knowledge on these cell types as well as disclosed the molecular signatures for five additional major cell types. Together, our snRNA-seq survey highlighted the molecular signatures of eight major P. lividus cell type groups: muscles, neurons, epidermis, podia epidermis, skeleton, coelomocytes, WVS, and digestive tract, all of which are distributed across 48 distinct cell type clusters.

At the regulatory signature level, the only available data regarding postmetamorphic juveniles pertain to the skeleton. Previous studies have shown that a few of the genes constituting the embryonic and larval skeletogenic GRN (vegfr, alx1, sm37, sm50, and mps130) are expressed in the juvenile skeletal cells, suggesting a partially conserved skeletogenic regulatory program between embryos, larvae, and juveniles (36, 45). Our findings here corroborate this statement. Our study reveals that, with the exception of three genes, almost the entire embryonic/larval reconstructed skeletogenic GRN, composed of 27 genes, including the regulatory genes and the differentiation gene batteries, operates in postmetamorphic skeletal cells, exemplifying how a cell type can be formed de novo using conserved genomic and regulatory information.

Overall, our analysis demonstrates that many juvenile cell types show molecular similarity to their larval counterparts. For instance, juvenile muscle cells exhibit high molecular similarity to larval ones, using most of the regulators involved in the embryonic myogenesis program. Furthermore, we identified distinct clusters of juvenile coelomocytes that align with either larval immune cells, a cell type family that contains globular and pigment cells (10), or larval blastocoelar cells, all of which are components of the sea urchin premetamorphic immune system (52). Notably, the alignment of larval blastocoelar cells with juvenile WVS clusters further suggests that the larval blastocoelar cell network may function akin to a vascular system.

Of the three juvenile digestive tract clusters identified in our analysis, only one corresponds to the larval mid/posterior gut, one to the larval acinar-like cells, and one appears to be specific to the juvenile stage. Notably, we found no significant alignment between the larval clusters corresponding to the anus, intestine, and foregut with any of the juvenile digestive tract clusters, suggesting that while similar transcription factors are expressed in those cells, the cell type outcome is different. This is not so unexpected given that these larval tissues are known to be reabsorbed during the metamorphic process (53). Another explanation for the lack of similarity between those cell types could be the different feeding strategies of sea urchin larvae and juveniles feeding on microorganisms and algae, respectively. Such differences must dictate the presence of diversified gut cell types involved in proper food digestion and nutrients uptake. For acinar-like cells, our previous research demonstrated their molecular and morphological homology to the exocrine acinar cells of mammals (38). Their presence in juveniles represents evidence of a pancreatic-like cell type in postmetamorphic echinoderms. While our findings indicate here that juvenile digestive tract cell types such as esophagus, intestine, and anus are formed putatively de novo based on the low transcriptional similarities revealed by SAMap, the larval stomach and acinar-like cells may be retained and used as a scaffold for completing juvenile gut organogenesis, as suggested by previous morphological studies (53). Thus, the digestive tract represents a particular example of cell types mostly inherited from the larva, rather than newly formed in juveniles.

The echinoderm germ line

In our snRNA-seq atlas, no cluster corresponding to the germ line could be identified. In sea urchins, gene markers known to label both germ and stem cells in other taxa (54–56) were identified in numerous embryonic and larval cells with a substantial enrichment in the embryonic small micromere descendants, which, during embryogenesis, populate the foregut region while, in the larval stages, are located within the coelomic pouches (10, 57). On the basis of on those gene expression data, it was initially hypothesized that these cells could correspond to multipotent stem cells that can differentiate to a variety of somatic tissues (48). Later on, it has been proposed that these cells could represent the germ line and that sea urchins may use an inherited mode of germline specification (54–56, 58). However, the functional role of these genes remains unclear. Both our gene candidate approach and SAMap analysis failed to identify a juvenile cluster matching the coelomic pouches cluster of the larva, where cells with germ/stem molecular signatures have been reported (10, 57). Moreover, germ/stem cell molecular signatures were found scattered in a broad spectrum of clusters corresponding to somatic cell types. Notably, cells expressing either of the germ/stem cell markers were found to populate all the clusters of every cell type group. Several scenarios could explain this observation. A first possibility is that the germline cells may be present in juveniles but simply few in number or embedded within deeper tissue layers that our dissociation protocol failed to extract. In disagreement with this hypothesis is, however, the fact that we retrieved in our snRNA-seq atlas rare cell types, such as opsin4-positive cells, as well as digestive tract–related cell types, which are located in the deepest layers of the juvenile body cavity. Another possibility is that the embryonic and larval germ/stem cell populations represent stem cells rather than germ cells, supporting the initial identity assessment of these cells (48). This hypothesis is well-supported by their localization in the larval coelomic pouches, a domain where the juvenile rudiment forms. These cells could thereby contribute to juvenile formation rather than directly to the germ line, and consequently, the germ line could be specified later, during juvenile or adult development through inductive mechanisms.

Evidence of an “all-brain” echinoderm state

The most abundant and diverse group of cell types in terms of cluster representation, as revealed by our analysis, corresponds to neurons. While we identified the expression of most embryonic and larval neurogenic patterning genes in the juvenile nervous system clusters, we found low similarity between larval and juvenile neurons even when reducing the 29 P. lividus neuronal clusters into a single one. This finding suggests that although the same genetic toolkit is used to generate neurons, the outcomes of the neurogenic program differ substantially between the two analyzed life stages. A notable exception is a single juvenile neuronal cluster (cluster 10) that exhibited similarity to an ectodermal larval cluster corresponding to the aboral ectoderm. Previous studies have shown that the aboral ectoderm cells of sea urchin larvae express several genes also involved in the specification of the larval apical plate (10) and that neurogenesis occurs in the aboral part of the animal plate (59). Together, this could either mean that the larval aboral ectoderm and the neuronal cluster 10 share expression of key patterning genes or that, similar to the digestive tract, some larval neurons present in the aboral ectoderm have been inherited by the juveniles, during metamorphosis.

Traditionally, the echinoderm nervous system has been characterized as “simple” due to the absence of a centralized brain. It was previously postulated that the expression of key CNS orthologs in the adult sea urchin pentaradial nervous system is indicative of a fivefold interconnected CNS duplication (60). The large number of neuronal clusters we identified, comprising more than half of the juvenile cell atlas clusters, together with the high diversity of their molecular signatures, including dopaminergic, serotonergic, cholinergic, GABAergic, glutamatergic, and histaminergic ones, neurons with dual neuronal identity (serotonergic/histaminergic, serotonergic/cholinergic, GABAergic/cholinergic, and dopaminergic/noradrenergic), as well as neurons whose function depends solely on neuropeptidergic signaling, suggest a far more sophisticated nervous system than a mere network of interconnected neurons and ganglia. Moreover, we found that these neurons, even of the same type as defined on the basis of the neurotransmitter they produce, use diverse neuropeptides, suggesting the lack of neuronal type–specific neuropeptidergic modules and high functional specialization.

Our analysis of the anterior-posterior patterning genes strongly supports the hypothesis that the sea urchin postmetamorphic body plan exhibits a head-like organization similar to what has been recently proposed for the sea star and brittle star postmetamorphic body plan (37, 61). This interpretation is elaborated by the expression profile and spatial distribution of genes involved in the definition of anterior head, posterior head, and head/trunk boundary across the juvenile nervous system and epidermal regions and the restriction of the trunk-related genes in internal structures and organs such as the digestive tract and the WVS of the juvenile. Notably, our snRNA-seq atlas shows overlapping expression of the trunk, anterior head, posterior head, and head/trunk boundary genes in the cholinergic neurons cluster 22, suggesting a convergence of molecular signature in nervous system patterning rather than the presence of a dedicated cell type corresponding to a sea urchin trunk.

In line with this hypothesis, we show that the anterior head genes frz5/8 and nkx2.1 are expressed in both ONR and RNC, which together comprise the largest part of the sea urchin adult nervous system. The spatial expression profile of both genes corroborates previous findings in sea stars and brittle stars (37, 61). In addition, we show that the posterior head genes irxa and barhl are expressed in the podia neurons and the RNC, while otx is expressed in the ONR and podia neurons. Notably, many of the genes involved in the definition of the anterior head domain in bilaterians (fzd5/8, nkx2.1, rx, sfrp1/5, and six3) are also involved in the specification of the ANE, which, in marine invertebrate larvae, is required for the formation of neurons with an analogous role to the vertebrate CNS. The evolutionary conservation of these two GRN programs are highlighted by a recent study in chordates that demonstrated the presence of an orthologous subset of the sea urchin ANE genes, including nkx2.1 and frz5/8 constituting the anterior GRN that controls the forebrain identity (47). Together, the data of our study suggest that most of the anterior and posterior head genes are expressed within the sea urchin postmetamorphic nervous system. Therefore, our data, which demonstrate the expression of several vertebrate CNS homologs in tissues throughout the sea urchin juvenile nervous system, not only support the head-like hypothesis for echinoderm axial patterning but also suggest that the nervous system of these allegedly “brainless” animals features an “all-brain” organization. Last, the conservation of these molecular fingerprints in other echinoderms (37, 61) suggests that the brain-like organization hypothesis may apply to the entire echinoderm clade.

Melanopsin and Go-opsin are coexpressed in a deuterostome photoreceptor

Adding to the molecular complexity of the sea urchin nervous system is the unexpected high diversity in neuronal PRC types. This suggests that large portions of the juvenile and adult sea urchin nervous system are under light-dependent control. Fifteen clearly distinguishable PRC types, each expressing a specific combination of evolutionarily conserved retinal transcription factors, highlight the remarkable complexity of the sea urchin photoreceptor system.

Despite the numerous opsin-expressing neurons and in contrast to the extensive gene duplications observed in melanopsin-expressing PRCs of brittle stars (62, 63) and crinoids (64), our snRNA-seq data revealed the expression of only seven specific opsin genes in the juvenile P. lividus, with no evidence of opsin gene duplication. Apart from opsin2-expressing PRCs, which have recently been shown to modulate larval swimming behavior in the sea urchin Hemicentrotus pulcherrimus (32, 65), the specific functions of most echinoderm opsins remain poorly understood.

PRCs involved in animal vision predominantly express either a ciliary opsin (localized inside often-modified PRC cilia) or a melanopsin/rhabdomeric opsin (localized inside often-modified PRC microvilli) (66). Our spatial expression analysis of the sea urchin ciliary opsin in juvenile P. lividus corroborates previous findings in other sea urchins where a ciliary opsin is expressed across the entire juvenile body (34, 35), rendering it an unlikely candidate for sea urchin vision. In contrast, P. lividus melanopsin (opsin4) expressing PRCs exhibit a similar localization and ultrastructure to those described in S. purpuratus (27). PRCs at the podial base and disc were morphologically indistinguishable from one another, a finding in accordance with our molecular analysis showing only one single melanopsin-positive PRC subcluster. We thus propose only one melanopsin-expressing PRC type in P. lividus and potentially in other sea urchins, which have shown comparable spatial melanopsin expression patterns (67).

P. lividus melanopsin–expressing PRCs coexpress a Go-opsin (opsin3.2), a type of opsin previously reported in only a small number of marine invertebrates (51, 68–70). In sea urchins, the Go-opsin has so far only been documented in larval cells. Valero-Gracia and colleagues (31) first identified it in two cells flanking the larval apical organ. Valencia and colleagues (71) thereafter reported some transcription factors typically associated with ciliary PRCs, while other typical ones were absent. Cocurullo and colleagues (30) also identified Go-opsin within modified cilia of neurons. Last, Yaguchi and Yaguchi (72) demonstrated a role for Go-opsin3.2 in light-dependent pyloric opening in sea urchin larvae.

The presence of a PRC coexpressing melanopsin and Go-opsin represents a novelty within deuterostomes. Go-opsin–expressing PRCs seem to be absent in vertebrates and have so far only been found in the chordate Amphioxus (73, 74), although no spatial expression or functional data are available to date in these animals. Phylogenetic analyses place sea urchin Go-opsins within the same clade as those found in the marine polychaete worm Platynereis dumerilii (49), for which Gühmann and colleagues (68) reported coexpression of a Go-opsin with melanopsin in larval eye photoreceptors. Here, the Go-opsin mediates spectral tuning of the melanopsin photopigment, thereby enhancing PRC efficiency. Given the dimly lit marine environments inhabited by sea urchins such as P. lividus, similar spectral tuning could certainly benefit their visual capabilities.

While the presence of Go-opsin–expressing PRCs at the base of Bilateria is undisputed (49), our findings provide compelling evidence of such an opsin being coexpressed with a canonical melanopsin in deuterostomes. This also underscores that although the same opsin genes, in this case opsin3.2, are expressed in both pre- and postmetamorphic animals, their functional roles may differ substantially.

In summary, by comparing cell type gene expression profiles between sea urchin larvae and juveniles at a single-cell/nucleus level, we here demonstrate the developmentally conserved and diverse building blocks of the postmetamorphic body plan. It is important to note that, although the SAMap analysis presented here is sufficient to identify developmentally conserved cell types and thus provides a valuable framework for cross-species comparisons, it nonetheless infers gene expression similarity rather than true cell type conservation. Future investigations will be required to elucidate the precise developmental program dynamics of these similar cell types in the pre- and postmetamorphic stages. Last, the complexity of the sea urchin nervous system, as characterized by the diversity of postmetamorphic neuronal cell type signatures and their integration of diverse PRC systems, leads us to propose that the sea urchin nervous system in its entirety comprises an “all-brain” rather than a “no-brain” state. Different PRC systems, including the proposed visual PRCs coexpressing a melanopsin and a Go-opsin feed into this complex “all-brain” network, play an indispensable role in sea urchin photobiology.

MATERIALS AND METHODS

Animal husbandry

Adult P. lividus individuals were collected from both the Gulf of Naples (Italy) and the Bay of Villefranche-sur-Mer (France) and were housed, respectively, at Stazione Zoologica Anton Dohrn (SZN) and Institut de la Mer de Villefranche (IMEV), in circulating seawater aquaria. All animal work described in this study was carried out in accordance with relevant institutional and national guidelines and was approved by the appropriate ethics committee. Sea urchins being noncephalopod invertebrates are not subject to regulation under European Union legislation, governing the use of animals for scientific purposes (Directive 2010/63/EU). Nevertheless, all individuals were treated with the utmost respect, appreciation, and compassion throughout the course of this study. At SZN, spawning of gravid animals was induced by vigorous shaking. At IMEV, gametes were collected from dissected gonads. For fertilization, at SZN and IMEV, sperm was diluted 1:1000 and added to a beaker containing unfertilized oocytes in ~100 ml of natural filtered sea water (FSW). Fertilization was monitored through microscopic observation and confirmed by the presence of the fertilization envelope. Zygotes were rinsed twice with FSW, to remove the excess of sperm. At SZN, embryos and larvae were reared at a density of five embryos/larvae per milliliter in FSW at 18°C and fed three times per week with Dunaliella tertiolecta. Twice per week, half of the culture seawater was exchanged with fresh FSW. Upon successful metamorphosis, the juveniles were transferred to fresh containers and kept at 18°C, and water was changed twice per week. Once the mouth was formed, juveniles were fed with Ulva lactuca algae until they reached the desired stage which was 2 wpm. At IMEV, embryos and larvae were reared at a concentration of, respectively, 70 embryos and one larva per milliliter at 18°C under constant mechanical stirring and fed five times per week with a mixture of Dunaliella salina and Rhodomonas salina. Three times per week, two-third of the culture sea water was exchanged with fresh FSW. Upon induced metamorphosis using Dibromomethane (75), the juveniles were transferred to fresh containers and kept at 16°C under constant sea water flow. The sea water was pumped at a depth of 5 m, decanted, and cooled down to 16° to 18°C. Once the mouth was formed, the juveniles were fed with Tetraselmis suecica algae until they reached the 2 wpm stage. Juveniles reared at SZN were used for IHC, fluorescent in situ hybridization, HCR, and electron microscopy experiments. Juveniles reared at IMEV were used for the snRNA-seq and HCR experiments.

Isolation of intact nuclei and snRNA-seq

P. lividus juvenile nuclei were isolated using a previously described and well-established method that allows for the generation of high-quality single-nuclei suspensions (76), with minor modifications. A total of 18 juveniles were used, issued from two distinct cultures and collected at 2 wpm. Three independent snRNA-seq libraries were prepared, each with six juvenile individuals (all from culture 1, all from culture 2, or a 1:1 mix from cultures 1 and 2). All the following steps described here were performed on ice. Juveniles were placed in the lid of a 5-ml Protein LoBind tube (Eppendorf, #0030108302) containing 560 μl of prechilled Homogenization Buffer. Homogenization Buffer is a hypotonic buffer composed of 250 mM sucrose, 25 mM KCl, 5 mM MgCl₂, 10 mM tris buffer (pH 8.0), 1 mM dithiothreitol, 1× protease inhibitor (Sigma-Aldrich, #11873580001), ribonuclease inhibitor (0.4 U/μl; Invitrogen, #AM2682), SUPERase-In ribonuclease inhibitor (0.2 U/μl; Invitrogen, #AM2694), 1 mM spermine, and 0.3% Triton X-100. For each library, the six juveniles were cut in half using a microsurgical knife (Electron Microscopy Sciences, #72047-30), while the tissues from three of them were further gently scraped using a microsurgical blade and fine forceps (Fine Science Tools, #10316-14 and #11251-20, respectively) under a dissecting microscope. Juvenile large fragments were transferred, using a P1000 micropipette, into a clean 1.5-ml Protein LoBind tube. The rest of the supernatant, before its transfer into the same tube, was filtered through a 40-μm cell strainer (PluriSelect Mini strainer), to remove debris. Nuclei were extracted through application of mechanical force via pipette aspiration for 10 times using a P1000 micropipette and 15 times using a P200 micropipette. The procedure was performed gently to avoid the formation of bubbles that can lead to nuclei loss or damage. Specimens were left until the large juvenile fragments settled, and the supernatant containing the isolated nuclei was transferred into a fresh clean 1.5-ml Protein LoBind tube. Isolated nuclei were spun down for 10 min at 500g and 4°C using a 5424 R Eppendorf microcentrifuge. The presence of pellets was visually confirmed (brown colored pellets). All the supernatant was discarded except from 50 μl (surrounding the pellet). Six hundred microliters of Wash Buffer [1× Dulbecco’s phosphate-buffered saline (DPBS), 2% bovine serum albumin (BSA), SUPERase-In ribonuclease inhibitor (0.2 U/μl), 1 mM spermine, and 5 mM MgCl₂] was carefully added (drop by drop) using a P200 micropipette, in order not to disrupt the pellet. Next, the nuclei pellets were resuspended through pipette aspiration for five times, ensuring that the resuspended nuclei stay close to the bottom of the tube. Isolated nuclei were once more spun down for 10 min at 500g and 4°C using a 5424 R Eppendorf microcentrifuge. Most of the supernatant was discarded apart from 25 μl (covering the pellet). Last, 25 μl of Resuspension Buffer [3× DPBS, 2% BSA, ribonuclease inhibitor (0.4 U/μl), SUPERase-In ribonuclease inhibitor (0.2 U/μl), and 1 mM spermine] were added to the pellet. Pellet was resuspended through gentle pipette aspiration for 20 to 30 times using a P200 micropipette. Nuclei suspensions were filtered through a 10-μm cell strainer (PluriSelect Mini strainer) to avoid aggregates. The successful generation of single-nucleus suspensions and the lack of debris were estimated by microscopic observation of the preparations stained with trypan blue to enhance the contrast as well as of nonstained specimens. Similarly, trypan blue was also used to estimate the nuclear integrity and quality, while the numbers were estimated by using a Malassez hemocytometer. Isolated nuclei (25,000 per library) were loaded on the 10x Genomics Chromium Controller according to the manufacturer’s instructions. cDNA libraries were prepared using the Chromium Single Cell 3’ Reagent Kit (v3.1 Chemistry Dual Index). Libraries were sequenced by the Institut de Génomique Fonctionnelle de Lyon (IGFL) sequencing platform (Plateforme de Séquençage de l’IGFL) using the Illumina NextSeq 500, and 146 to 171 million reads were obtained per sample.

Mapping of snRNA-seq reads

Before mapping, the P. lividus genome (77) version 1.0 annotation gene models were extended in the 3′ direction by 5 kb, to ensure no overlap with downstream genes as described in (76) and improve mapping rates. Then, the reads were mapped to this genome and modified annotation using the Cell Ranger Software Suite v7.1.0 (78) with --include-introns = true, --no-bam, --nosecondary, and --force-cells = N flags, where N is the number of cells to force for each sample. This number was determined by mapping the data without forcing a set number of nuclei and visually assessing the barcode rank plot from the web summaries generated by Cell Ranger. This was done to ensure that only real nuclei end up in the final analysis. Cell Ranger output matrices, features, and barcodes were used for further analysis using the R package Seurat v 5.1.0 (79).

Analysis of snRNA-seq data

Seurat objects were created by excluding from the analysis genes that are transcribed in less than three cells and cells that have less than a minimum range of 350 to 500 (depending on the library) and a maximum of 5000 transcribed genes. The generated datasets were normalized, and variable genes were found using the variance stabilizing transfer method with a maximum of 2000 variable features. First, the three different objects were merged, followed by scaling and principal components analysis (PCA). Jackstraw was used to evaluate PCA significance. A sharing nearest-neighbor (SNN) graph was computed with 50 dimensions (resolution, 1.0). Uniform Manifold Approximate and Projection (UMAP) was used to perform clustering dimensionality reduction. Upon analysis of the unintegrated dataset, only two clusters demonstrated unequal distribution among the three specimens, a potential evidence of batch effect. Therefore, integration of the three different objects was performed, and the integrated object was used for the subsequent analyses. The three different objects corresponding to the three independent libraries were integrated through identification of gene anchors (FindIntegrationAnchors) between the different datasets. After integration of the objects, scaling and PCA were performed, and the PCA significance was evaluated by Jackstraw. The SNN graph was computed with 50 dimensions (resolution, 1.0), and UMAP was used to perform clustering dimensionality reduction. The final object consisted of 25,000 nuclei that successfully passed the quality controls. The FindAllMarkers command was used to identify differentially expressed marker genes. To further confirm P. lividus gene IDs, nucleotide BLAST v2.6.0+ (80) searches were performed against S. purpuratus using the -max_target_seqs 1 and -max_hsps 1 options to select a single top hit per P. lividus gene sequence and the output setting -outfmt “6 qseqid sseqid” to get both query and subject sequence IDs. Cell type tree reconstruction based on pairwise expression distances between clusters, using all expressed genes, was performed using 10,000 bootstraps as previously described (10). Subclustering analysis of the PRCs was performed using the subset function available in Seurat, by using all the clusters corresponding to the nervous system (clusters 1 to 29) and only the nuclei that express at least one of the seven opsin genes encoded by the P. lividus genome. The subsetted objects were further analyzed by means of calculation of variable genes, scaling, and PCA analysis. Jackstraw was used to determine the number PCAs to be used for clustering. Nineteen PCAs were selected, and clustering dimensionality reduction was performed. To compare the S. purpuratus 3 dpf larva single-cell atlas and the P. lividus juvenile single-nucleus atlas, SAMap v1.02 (81) was used as previously described (82). First, mapping between the P. lividus and S. purpuratus protein sequences was performed using the custom script map_genes (81). The two Seurat objects were converted into .h5ad files and used as an input. Pairwise alignment was performed and resulted in the construction of a stitched SAM object. The resulting mapping table (n_top = 0) was stored as a DataFrame and used for downstream plotting in R. The average score of ANE, peripheral, and neurogenic markers was estimated using the AddModuleScore function incorporated in the Seurat R package, which computes the average expression of a given gene set for each cell, subtracted by the average expression of randomly selected matched control gene sets, that accounts for expression biases. The data were binned on the basis of averaged expression and stored as a new metadata column in the Seurat object. The gene IDs used to generate the heatmaps, dotplots, and featureplots can be found in the Supplementary Materials.

Immunohistochemistry

Postmetamorphic juveniles were fixed in 4% paraformaldehyde (PFA) in FSW for 20 min at room temperature (RT). Next, specimens were washed once in 100% ice-cold methanol (MeOH) for 1 min at RT, followed by multiple rinses with 1× PBS and 0.1% Tween 20 (PBST). Specimens were incubated in a blocking solution containing BSA (1 mg/ml) and 4% sheep serum in PBST for 1 hour at RT. Primary antibodies were appropriately diluted in PBST and incubated for 1 hour and 30 min at 37°C [or overnight (ON) at 4°C in the case of OPSIN1]. The primary antibodies used were against as follows: MSP130 (gift from D. R. McClay) to label skeletogenic cells (1:20), SYT1 (gift from R. Burke) to mark the nervous system (1:20), MHC to label muscles (1:100) (custom-made, PRIMM srl), and OPSIN1 (1:50) (custom-made, PRIMM srl) and OPSIN4 (1:50) (custom-made, PRIMM srl) to mark different photoreceptor types. Following primary antibody incubation, specimens were washed multiple times with PBST and incubated for 1 hour with the appropriate secondary antibody (Alexa Fluor, Thermo Fisher Scientific, #A11029, #A21428, and #A21050), diluted 1:1000 in PBST. OPSIN4 IHC paired with Pl-Opsin4 HCR was performed upon the completion of the HCR procedure. The only differences in the IHC protocol in this case are that the fixation step was skipped and the specimens were kept in the dark throughout the procedure. Specimens were imaged using Zeiss LSM 700 and Leica TCS SP5 confocal microscopes.

Fluorescent in situ hybridization

Fluorescent in situ hybridization for opsin2 was performed as previously described (35, 83). Briefly, specimens were fixed in 4% PFA in Mops buffer [0.1 M Mops (3-morpholinopropane-1-sulfonic acid), pH 7, 0.5 M NaCl, and 0.1% Tween 20 in nuclease-free water] for 1 hour at RT, gradually dehydrated, and kept at −20°C in 70% ethanol. The signal was developed using the tyramide signal amplification technology (Akoya Biosciences), and specimens were imaged on a Zeiss LSM 700 confocal microscope.

Hybridization chain reaction

Probes for HCR were generated using the coding sequence of the genes of interest. The opsin3.2, opsin4, beta-arrestin, vcry, otx, and barhl probes were synthesized by Molecular Instruments. Probes for pax2/5/8, fzd5/8, nkx2.1, irxA, syt1, hdc, th, and gad were designed using the custom software developed by Kuehn et al. (84) and synthesized by Integrated DNA Technologies (IDT). The stock concentration of each probe set was 1 μM. Information on the probe/initiator pairs can be found in the Supplementary Materials. HCRs for opsin3.2, opsin4, beta-arrestin, vcry, otx, and barhl were performed at SZN and Museum für Naturkunde (MFN) using a modified protocol from (37), while for pax2/5/8, fzd5/8, nkx2.1, irxA, syt1, hdc, th, and gad HCRs were carried out at IMEV as described in (37, 85) with minor modifications. For the HCRs performed at SZN and MFN, the procedure was carried out as follows. Two wpm juveniles were collected and fixed in 4% PFA in a fixative buffer containing 0.1 M Mops, 0.5 M NaCl, 2 mM EGTA, 1 mM MgCl₂, and 1× PBS ON at 4°C. The day after, specimens were rinsed three times (5 min each) in fixative buffer at RT and stored in 100% MeOH at −20°C until use. On the day of the HCR experiment, juveniles were rehydrated in decreasing concentrations of MeOH (75, 50, and 25% in nuclease-free water), and each wash lasted 10 min. Juveniles were washed three times (10 min each) in PBSTr (1× PBS, 0.1% Triton X-100). Specimens were incubated in a decalcification buffer containing 5% EDTA in nuclease-free water for 4 hours at RT and, after this step, washed again in PBSTr for three times (5 min each). Juveniles were incubated in permeabilization solution composed of 1% Triton X-100 in 1× PBS at 4°C ON. The day after, the permeabilization buffer was removed, and samples were digested with proteinase K (4 μg/ml) for 6.5 min at 37°C. Samples were washed in PBSTr for two times (5 min each) and postfixed in 4% PFA for 1 hour at RT. Samples were washed again for five times (5 min each) with PBST and afterward incubated in the probe hybridization buffer (Molecular Instruments) at 37°C for 2 hours. Within this time interval, probe hybridization buffer was exchanged once. Probe solution was prepared by diluting each probe in the probe hybridization buffer to a final concentration of 0.04 μM. Specimens were incubated with the probe solution ON at 37°C. Following the probe hybridization, specimens were washed three times (5 min each) and two times (30 min each) with preheated probe wash buffer (Molecular Instruments). Specimens were washed several times with 5× SSCTr (5× SSC, 0.1% Triton X-100). A preamplification step was performed by incubating the sample for 30 min at RT in an equilibrated to RT amplification buffer (Molecular Instruments). In the meantime, 2 μl of each hairpin from Molecular Instruments (of 3 μM stock) was placed individually in tubes. The hairpins h1 and h2 were heated up separately by placing the tubes at 95°C for 90 s. After this step, the hairpins were stored on ice, in the dark, for 30 min. The hairpins were then added to the amplification buffer at a final concentration of 0.06 μM. The specimens were incubated in the hairpin solution at RT, in the dark, for at least 20 hours. The following day, the samples were washed several times in 1× SSCTr. The juveniles were transferred to a solution containing 4′,6-diamidino-2-phenylindole (DAPI; 1 μg/ml) in 1× SSCTr, then transferred to 50% glycerol [in 1× PBS (pH 7.4)], and mounted for imaging. Images were acquired using Zeiss LSM 700 confocal and Leica TCS SP5 microscopes.

At IMEV, HCRs were carried out as follows. Two wpm juveniles were fixed ON at 4°C in 4% PFA and 10 mM 4-(2-hydroxyethyl)piperazine-1-propanesulfonic acid (EPPS) diluted in FSW. Upon fixation, juveniles were washed once for 5 min at RT in 1× PBST and once for 5 min at RT in cold (−20°C) MeOH, before being stored at −20°C in MeOH. Fixed juveniles were rehydrated through progressive washes in 75, 50, and 25% MeOH in 1× PBST (5 min each) at RT. Juveniles were then washed twice (10 min each) at RT in 100% 1× PBST, followed by three washes (10 min each) at RT in ribonuclease (RNase)–free water. For tissue clearing purposes, juveniles were incubated ON at 4°C in 50% tetrahydrofuran (THF) (#186562, Sigma-Aldrich, Saint-Quentin-Fallavier, France) in RNase-free water (86). The THF was thereafter discarded by three washes (30 min each) at RT in RNase-free water and two washes (5 min each) at RT in 1× PBST. Permeabilization of the juveniles was conducted by incubation in detergent solution, containing 50 mM tris-HCl (pH 7.5), 1 mM EDTA, 150 mM NaCl, 1% SDS, and 0.5% Tween 20, for 1 hour at RT. Juveniles were washed four times (5 min each) at RT in 1× PBST, before they were transferred into 96-well plates and prehybridized for 3 hours at 37°C in 50 μl of hybridization buffer containing 30% formamide, 5× SSC, 9 mM citric acid (pH 6.0), heparin (50 μg/ml), 1× Denhardt’s solution, 10% dextran sulfate, and 0.1% Tween 20. During the prehybridization step, each probe pair set was diluted in 50 μl of hybridization buffer, to a final concentration between 0.008 nM and 0.04 μM depending on the targeted gene. Probe pair dilutions were then added to the specimens and placed in a humid chamber ON at 37°C to allow probe pair hybridization. Probe pairs were washed out by four washes (5 min each) at 37°C and then two washes (30 min each) at 37°C, all using preheated probe wash buffer [30% formamide, 5× SSC, 9 mM citric acid (pH 6.0), heparin (50 μg/ml), and 0.1% Tween 20]. Next, two additional washes were performed (5 min each) at RT in 5× SSCT (5× SSC, 0.1% Tween 20), followed by two washes (5 min each) at RT using 1× SSCT. Juveniles were subsequently incubated in 50 μl of amplification buffer (5× SSC, 10% dextran sulfate, and 0.1% Tween 20) for 30 min at RT. During the preamplification step, h1 and h2 hairpins (Molecular Instruments) were prepared as described above for HCRs in SZN and MFN, except that only 1 μl of each hairpin (3 μM stock) was used. Upon cooling down, h1 and h2 hairpins were transferred together in 50 μl of amplification buffer. Diluted hairpins were added onto the specimens and incubated ON in the dark at RT. Hairpin removal was carried out by several washes, all performed in the dark at RT. These washes included two washes (5 min each) in 1× SSCT, one wash (30 min) in 1× SSCT containing 1:1000 Hoechst 33342 for nuclei staining, two washes (30 min each) in 1× SSCT, and four washes (5 min each) in 1× PBST. Juveniles were mounted in an EZ clear mounting medium prepared as described in (86). Images were acquired on a Leica SP8 confocal microscope.

Transmission electron microscopy

Postmetamorphic juveniles were fixed in 1.25% glutaraldehyde in artificial seawater (ASW) + 1% osmium at 18°C for 1 hour at RT. Samples were subsequently washed in PBS (0.05 M phosphate/0.3 M NaCl) (pH 7.4) for 10, 30, and 60 min, then cooled to 4°C, and washed again after 4 hours and ON. The samples were postfixed in 1% osmium/1.5% ferrocyanide in 0.2 M cacodylate buffer for 1 hour at RT. Following two 15-min washes in distilled water, samples were incubated in 0.3% tetracaine hydrochloride (TCH) in 0.2 M cacodylate buffer (freshly prepared from a 1% TCH stock solution) for 15 min. Samples were then washed twice in distilled water for 15 min each, followed by incubation in 1% osmium in cacodylate buffer for 1 hour. After two additional 15-min washes in distilled water, samples were dehydrated in a graded ethanol series (30, 50, 70, 80, 90, and 95%) for 10 min each, followed by three 15-min washes in 100% ethanol. After this step, specimens were dehydrated in a graded alcohol series and embedded in plastic resin (Araldite M, Fluka; art. no. 10951) for 48 hours at 60°C. Serial sections were cut using a diamond knife and a Reichert ultramicrotome S and stained with uranyl acetate and lead citrate (Phoenix Ultrastain, Staining Technologies). Examination was carried out using a ZEISS EM 900 transmission electron microscope.

Serial block face scanning electron microscopy

Two wpm P. lividus juveniles were collected and fixed on ice for 1 hour in 1.25% glutaraldehyde and 1% osmium in ASW. After fixation, samples were kept on ice and washed with cold PBS [0.05 M phosphate buffer/0.3 M NaCl (pH 7.4)] for 10 min, 30 min, 4 hours, and ON at 4°C. All postfixation steps were performed at RT. Samples were incubated in 0.2 M cacodylate buffer with 1% osmium and 1.5% ferrocyanide for 1 hour. The specimens were then rinsed twice in PBS (15 min each) and incubated for 15 min in 0.3% TCH in 0.2 M cacodylate buffer, followed by another two PBS washes. Subsequent incubation in 1% osmium in 0.2 M cacodylate buffer for 1 hour was followed by two final PBS washes before samples were transferred to distilled water. For decalcification, samples were incubated in a 1:1 mixture of 2% ascorbic acid and 0.3 M NaCl at RT until the skeleton was completely dissolved. The samples were then washed with distilled water, dehydrated in a graded alcohol series, and embedded in plastic resin (Araldite M, Fluka; art. no. 10951) for 48 hours at 60°C. Acquisition was performed at the Electron Microscopy Core Facility at the European Molecular Biology Laboratory, Heidelberg, Germany. Targeting of the region of interest before SBF-SEM acquisition was accomplished using x-ray imaging (Bruker Skyscan, 1272). Target region trimming was performed using a Leica UC7 ultramicrotome. Imaging was carried out using a Zeiss GeminiSEM 450 equipped with a Gatan 3View system and an OnPoint backscattered electron detector, using an acceleration voltage of 1.5 kV and a current of 350 pA. The 3View microtome was set to remove 40 nm from the block surface at each cutting iteration. After each cut, a low-resolution overview image of the entire block face was acquired, along with a high-resolution tiled image of the region of interest. The latter was acquired with a 10-nm pixel size and a dwell time of 1.6 μs. To cover the entire region of interest, 6 to 10 tiles (4096 × 3072 pixels) were acquired per section, with an overlap of 350 pixels. The subsequent alignment of tiles resulted in a dataset of serial images of the region of interest with a voxel size of 10 nm by 10 nm by 40 nm. Segmentation and analysis were performed using the software package Amira (Thermo Fisher Scientific), versions 2021.2 and 2022.2. Nuclei were segmented semiautomatically using thresholding and Magic Wand tools in Amira. Plasma membranes and cell organelles were segmented manually. The segmented objects were rendered in three dimensions using the surface rendering module. A low-resolution reference volume (scaled down by a factor of 16) was used for initial nuclei segmentation, enabling reliable 3D orientation in regard to the juvenile anatomy. The alignment of the high-resolution EM subvolumes to the reference volume was performed manually using the transform editor module and computationally refined using the multiplanar channel module with rigid transformation. Areas of interest comprising the rhabdomeric photoreceptor candidates were segmented at a resolution of 10 nm by 10 nm by 40 nm (x, y, and z).

Supplementary Materials

The PDF file includes:

Figs. S1 to S11

Legends for files S1 to S4

Legends for tables S1 and S2

Other Supplementary Material for this manuscript includes the following:

Files S1 to S4

Tables S1 and S2