Tunicates are the closest living relatives of vertebrates. Recent phylogenies place the little-studied, free-swimming thaliaceans—including salps—within sessile ascidians, highlighting a remarkable ecological transition. Historical reports hinted at a parallel developmental shift. Salp embryogenesis diverges from that of ascidians and involves unique maternal cells called calymmocytes. Here, we provide foundational resources for two distantly related salp species, Salpa fusiformis and Thalia democratica. Using advanced microscopy, we generated developmental staging tables showing that while embryogenesis is stereotyped within species, it differs in cleavage patterns and blastomere positioning between them. We traced the origins of calymmocytes and confirmed their conserved role in separating blastomere clusters that form adult tissues. Apoptosis contributes to the progressive elimination of maternal calymmocytes. Finally, we show that calymmocytes express embryonic developmental regulators, suggesting co-option of an embryonic gene program. These findings provide an advanced framework for studying embryogenesis evolution in a previously underexplored chordate lineage.

Fig 1. Salps branch out of the paraphyletic ascidians, alternate sexual and asexual reproduction, and have a divergent embryogenesis.

(A) Lifecycle of Thalia democratica. (B) Consensus phylogenetic tree of tunicates and table showing the state of some characters in each order: benthic or pelagic lifestyle of the adult, presence/absence of a developmental stage with a “tadpole” morphology (at the larva, juvenile or adult stages), and presence/absence of a stereotypic ascidian-like embryogenesis (bilaterally symmetric embryos developing in an invariant manner).

https://doi.org/10.1371/journal.pbio.3003636.g001

Improved resolution of early maternal-zygotic interactions and extraembryonic structure formation

By applying confocal imaging approaches, we first compared the initial invasion of the embryo by calymmocytes, building on earlier studies that were largely limited to S. fusiformis [25,27,32–34]. Our analysis spans key early stages: from immature oocytes up to the 13-cell stage in T. democratica, and from the 6- to the 12-cell stages in S. fusiformis.

Before the 8-cell stage, blastomeres are readily distinguishable from follicle cells, polar bodies, and fertilization duct cells based on morphology. Blastomeres have larger cell and nuclear volumes and a lower nuclear-to-cytoplasmic ratio than polar bodies or follicle cells (Fig 2Aa and 2Ad). Polar bodies are located opposite to the fertilization duct remnant, are smaller than blastomeres, and often with irregular chromatin possibly indicative of apoptosis (Fig 2Aa and 2Ad). Follicle cells surround the blastomeres and polar bodies in an epithelium-like sheet. They are small, with condensed chromatin and a high nuclear-to-cytoplasmic ratio (Fig 2Aa and 2Ad). Fertilization duct cells are located between the follicle cells and the overlaying atrial epithelium and have similar morphological characteristics to follicle cells. At the 8-cell stage, we confirmed that in both species, maternal calymmocytes begin invading between blastomeres from the follicular sac and possibly from remnants of the fertilization duct (Fig 2Ab and 2Ae). Calymmocytes keep features of the follicle and fertilization duct cells: they can be differentiated from blastomeres by their smaller, less rounded shape, more condensed chromatin, and higher nuclear-to-cytoplasmic ratio—features maintained throughout embryogenesis (S1 Fig and S1 and S2 Data).

Concurrently, we investigated the formation of maternal periembryonic structures that surround the embryo: the incubation chamber and placenta. In S. fusiformis, previous studies described the presence of the incubation chamber, which protects the embryo from the atrial cavity fluid [25,27,32,33]. We examined how this structure develops and whether a counterpart exists in T. democratica. Prior to the 8-cell stage, embryos of both species are surrounded by a single layer of follicle cells and located within the blood sinus, overlayed by a specific region of the maternal atrial epithelium: the epithelial cone (Fig 2C and 2D). Around the 8-cell stage, the onset of calymmocyte invasion coincides with the translocation of the embryo into the atrial cavity, across the epithelial cone. The incubation chamber forms from the folding of the epithelial cone and surrounding atrial epithelium. In S. fusiformis, by the 18-cell stage, the embryo is enclosed by the two epithelial layers that form this chamber (Fig 2C). The improved resolution brought by our imaging approach and higher sampling of embryos enabled the first detection of a morphological structure similar to an incubation chamber in Thalia democratica (Fig 2D). In both species, the placental knob forms from the follicle cells left protruding in the blood sinus by the translocation of the embryo through the epithelial cone. This placental knob forms part of the future placenta and likely serves as an exchange zone between the mother and the embryo.

The timing of incubation chamber formation differs between species. In S. fusiformis, translocation precedes chamber formation, which completes around the 18-cell stage (Fig 2C). In T. democratica, by contrast, the epithelial cone folds between the 2- and 4-cell stages to create a transient empty cavity above the follicular sac (Figs 2D, S2A, and S2Ba–b), which the embryo enters at the 8-cell stage (Figs 2D and S2Bc–d). A slight heterochronic shift is also observed in placental knob formation. In S. fusiformis, this structure forms around the 10- to 12-cell stage. In T. democratica, a homologous structure arises from the same follicular epithelium but between the 12- and 24-cell stages.

These findings reveal that both salp species form similar incubation chambers and placental knobs, though with species-specific timing and morphology. Importantly, the identification of the incubation chamber in T. democratica contradicts earlier studies [25] and suggests greater conservation of maternal periembryonic structures among salps than previously recognized. Together, these results reveal that both salp species construct comparable, possibly homologous, maternal periembryonic structures but through slightly divergent developmental trajectories.

Developmental staging tables for Salpa fusiformis and Thalia democratica

A developmental staging table is a fundamental tool to federate a community of researchers working on a model organism, as it standardizes anatomical descriptions and provides a framework to describe deviations from the wild-type and compare species.

Previous work by Heider [32] in Salpa fusiformis led to a first table mostly relying on external morphology and macroscopic features of the embryo, such as the development of the incubation folds and the placenta, and complemented with drawings of embryo sections. No developmental table was available for Thalia democratica. Encouraged by the quality of confocal imaging of early embryos from both species, we set out to complement Heider’s Salpa fusiformis staging table (S1 Table and S3 Fig) and generate an equivalent Thalia democratica table (S2 Table and S4 Fig). In the case of Thalia democratica, the lack of high quality drawings to facilitate the rapid identification of the indicated stages in the lab led us to augment this table with brightfield microscopy pictures. To facilitate access to the developmental staging tables and associated figures for both Salpa fusiformis and Thalia democratica, we organized the data into an online web portal (https://octopus.obs-vlfr.fr/public/salps). This interface currently hosts searchable, stage-by-stage tables, representative images, and links to 3D reconstructions (when available). The portal is temporarily hosted on the LBDV’s Octopus server and is intended as a dynamic resource for the salp and broader tunicate research community.

In S. fusiformis, 3D confocal stacks allowed us to refine Heider’s table (1895) with additional substages based on internal features, such as the formation of internal cavities by calymmocytes. This staging table includes the 11 main Heider stages (A to L) that can be recognized with a stereoscope. Two stages (E and F) were separated in substages that can only be distinguished from the confocal images (labeled by a lowercase letter). The developmental table begins with early cleavage (stage A) and ends with muscle band formation (stage L) (S1 Table and S3 Fig). For each stage and substage, S1 Table provides a characteristic feature used to identify the stage with a simple stereoscope, an estimate of the size of the embryo, of the number, size and chromatin appearance of blastomeres when known, a detailed description of the embryo, of its cavities, and a pointer to representative drawings or pictures.

In T. democratica, a staging table was created de novo based on the observation and confocal imaging of 90 embryos. Seven stages (I to VII) were labeled with roman numbers and defined by morphological characteristics easily observable under a stereoscope. Seventeen sub-stages, labeled by a lowercase letter following the main stage number, were defined based mostly on the embryo’s internal morphology. This staging table covers development from the zygote (stage Ia) to stolon-producing embryo ready for release (stage VIIc) (S2 Table and S4 Fig). It provides a foundational framework for studying T. democratica development with greater precision and consistency.

To enable interspecies comparisons and highlight both conserved and divergent aspects of salp development, we synthesized the staging tables of Salpa fusiformis and Thalia democratica into a unified comparative matrix. Rather than enforcing a strict one-to-one correspondence between individual stages, which is complicated by differences in cleavage geometry, calymmocyte organization, and heterochronies in formation of periembryonic structures, we grouped stages into broader developmental periods anchored around shared morphological transitions (e.g., onset of calymmocyte invasion, blastomere segregation, organogenesis). This period-based approach allowed us to propose developmental equivalences while preserving species-specific dynamics. The criteria used for each period are outlined in the table and supported by descriptive data and imaging presented in the following sections. Together, this synthesis (Table 1) provides a practical framework for comparative developmental analysis across salp species and highlights both homology and divergence in their embryogenesis. Note that as all samples were collected from the field and processed immediately upon return from the sea, the duration of each stage is unknown.

These annotated staging tables constitute foundational resources for in-depth developmental analyses within each species, offering reproducible criteria to guide future investigations. However, comparative analyses between species remain challenging, particularly when attempting to identify stage homologies between S. fusiformis and T. democratica. Nonetheless, these tables establish a standardized framework, and the data presented in the following sections help address these challenges (see Discussion).

Intraspecific reproducibility and interspecific variability in early embryonic blastomere positions and geometries

We next used our three-dimensional imaging and cell and nuclear surface reconstructions to evaluate the degree of reproducibility of the position, size, and shape of early embryonic blastomeres within and between species, which techniques used in previous studies could not address [25,27,32].

In S. fusiformis, at the 10- to 12-cell stage (n = 4) the embryo exhibits bilateral symmetry, with two anterior pairs of blastomeres (three pairs at the 12-cell stage), two ventral pairs, and one posterior pair (Fig 3A and 3B). Some transient asymmetry is sometimes observed due to asynchrony in cell division between the two sides (Fig 3A, see embryo 2), yet the transient nature did not allow us to assess their consistency. Blastomere size, shape, and contacts are highly consistent. For instance, within the ventral group, blastomeres of the central pair are larger and have a relatively small contact area between them, while the lateral pair blastomeres are smaller and contact the larger blastomere above (Figs 3A, S5E, and S5F and S5 Data).

T. democratica embryos also exhibit stereotyped blastomere positions and shapes, although with a less strictly bilateral arrangement—some blastomeres align along the embryonic midline, contributing to a mixed symmetry pattern. At the 4-cell stage, blastomeres occupy consistent size and stereotypic positions (n = 4, Figs 3C, 3D, S5, and S5B). Two of the large blastomeres and the small blastomere are in direct contact with the remnant of the fertilization duct (Fig 3D). At the 12- to 13-cell stage (n = 3), the embryo shows bilateral symmetry, but in contrast to S. fusiformis, some unpaired blastomeres are positioned on the midline. The symmetry plane divides a pair of dorsal blastomeres, a pair of anterior blastomeres, a pair of ventral blastomeres and two central pairs, into right and left halves (Figs 3E, 3F, S5C, and S5D), whereas the ventro-central blastomeres (one at 12-cell stage, two at 13-cell stage) and a single, posterior blastomere are positioned along the midline. At this stage, all blastomeres are round in shape. Compared to the 4-cell stage (S5B Fig), the sizes of the blastomeres of each group appear more variable between analyzed samples, particularly in the medio-lateral (purple) and posterior (yellow) groups (S5D Fig). In summary, the position, volume, and shape of individual embryonic blastomeres are highly reproducible within each species but differ between S. fusiformis and T. democratica. While S. fusiformis maintains strict bilateral blastomere pairing throughout early cleavage, T. democratica embryos reproducibly produce a minority of unpaired blastomeres.

A comparative analysis of Thalia democratica and Salpa fusiformis embryonic maternal structures

We next analyzed the reproducibility of the calymmocyte structures and periembryonic maternal tissues, within and between the two species.

S. fusiformis, at stage E (see S1 Table) calymmocytes are organized in reproducible epithelium-like sheets and form stereotypic cavities (S3 Fig) as previously described [32]. In contrast, no clear monolayered sheet structures are observed in T. democratica. The only cavities observed in the embryo are organ primordia formed by blastomeres during organogenesis (S4 Fig).

The morphology of the incubation chamber differs between the two species. In S. fusiformis, from the 12-cell stage (between stage C and D), the two incubation folds grow on either side of the embryo to form a long, dorsal crest along the antero-posterior axis of the embryo. This delineates the incubation chamber which remains visible as a cavity throughout most of embryogenesis (Figs 2C and S3, see stage K). In contrast, in T. democratica, between the 2- and 4-cell stages, the epithelial cone invaginates in a ring shape and closes in a single point, forming the incubation chamber (stage Ib, Figs 2D, S2, and 4). This cavity, still visible at the 14-cell stage, (Figs 2D and S2), quickly reduces until no space is visible between the internal layer of the incubation chamber and the embryo (see stage IIa in S4 Fig).

These descriptions reveal strong reproducibility of calymmocyte positions, behaviors, and periembryonic tissue organization within each salp species, but also highlight marked differences between species. These interspecific differences emphasize the developmental variation of salps and contrast sharply with the extreme conservation typically observed in ascidian embryogenesis. This plasticity may reflect a relaxation of the stringent developmental constraints usually seen in other tunicates.

Clearing of maternal cells by apoptosis

To investigate whether calymmocytes could be eliminated by apoptosis, we conducted a TUNEL assay on whole embryos at various developmental stages in both species.

Positive control (treated with DNase I), and negative control (without TUNEL enzyme) confirmed the proper penetration of the TUNEL enzyme and the lack of natural autofluorescence, respectively (S6 Fig). In both species, some of the TUNEL-positive nuclei (TUNEL+) exhibit smaller or complex shapes, indicative of apoptotic blebbing (Fig 4Aa′ and 4Ca′,b′). Some TUNEL+/Hoechst+ nuclear fragments (1–2 µm in diameter) were observed, suggesting the presence of apoptotic bodies (S6B Fig).

In S. fusiformis, TUNEL+ calymmocytes were observed at stage E, but they were rare and scattered (Fig 4Aa–a′, n = 4). TUNEL+ nuclei were also detected in the periembryonic incubation chamber: from stage E, the inner layer contains a few TUNEL+ nuclei, and by stage F, the entire inner layer is TUNEL+ (Fig 4Ba–b, n = 5).

In T. democratica TUNEL+ calymmocytes were observed scattered throughout the embryo from stage IIIa to stage IVc (Fig 4Ca–a′ and S2 Table, n = 5), while at the onset of organogenesis (stage V, S2 Table), only few TUNEL+ nuclei were detected within the embryo (Fig 4Cb–b′, n = 5). At a later stage, TUNEL+ nuclei were also visible at the future aperture of the oral and atrial siphons (S6Aa Fig). TUNEL+ nuclei were also detected in periembryonic tissues: from stage IIIb to stage IVc, TUNEL+ nuclei were found in the outer epithelium of the incubation chamber (Fig 4Da–a′, n = 2). By stage V, the entire outer epithelium of the incubation chamber is TUNEL+ and begins to degenerate (Fig 5Db–b′, n = 5). By stage VI, the inner and outer epithelia of the incubation chamber have fully regressed, thus exposing the embryo in the atrial cavity. TUNEL+ nuclei are still detectable at the base of the placenta, likely corresponding to the remains of the incubation chamber epithelia (S6Ab Fig).

Although we could not analyze all the stages due to limited access to samples and tissue penetration issues at later stages, these findings support a role for apoptosis in the elimination of at least some calymmocytes and in the regression of periembryonic tissues. The higher frequency of TUNEL+ nuclei and fragments in T. democratica could be due to variations in development timing and duration. Assuming the cellular duration of apoptosis remains constant, the longer embryogenesis of S. fusiformis could dilute the occurrence of apoptotic calymmocytes over time.

Developmental regulatory genes Otx and Rar are expressed in calymmocytes

We next examined whether calymmocytes express key developmental regulators classically involved in the patterning of embryonic blastomeres in ascidians and other species. We adapted in situ hybridization protocols to the salp species and examined the expression patterns of two key developmental regulators of morphogenesis and tissue differentiation: the retinoic acid receptor Rar and the homeobox transcription factor Otx.

We identified in published transcriptomes two paralogs of Rar for T. democratica and S. fusiformis, a single ortholog of Otx in S. fusifomis, and two Otx in-paralogs in T. democratica (S7 and S8 Figs). For each species, we focused our analysis on one of these genes, selecting the salp paralog that most closely match a Ciona robusta (former intestinalis) gene [38], based on BLAST e-value. In the ascidian Ciona robusta, Rar is expressed at the 64-cell stage in the presumptive ectoderm and in neural precursors [38], while Otx is expressed in the presumptive neuroectoderm, endoderm, and mesoderm (including tail muscle and mesenchyme) at the same stage [38], comparable patterns of expression have been also documented in other ascidian species [39,40].

In Salpa fusiformis, at stage D (S1 Table) Rar.a transcripts were detected in two domains: one in a ring of calymmocytes at the embryo periphery and another in a group of calymmocytes on the dorsal side (n = 3, Fig 5Aa–a′, compare with controls in S9 Fig). At stage Ea (S1 Table), Rar.a transcripts were expressed in a monolayer of calymmocytes around the periphery of the embryo (n = 2, Fig 5Ab–b′) and in one pair of blastomeres in the dorso-lateral side of the embryo (Fig 5Ac–c′, arrowheads). By stage Fb (S1 Table), Rar.a expression persisted in the same layer of calymmocytes, now 1–3 cells thick, partially encompassing the embryo (n = 1, S10 Fig), suggesting that this expression pattern is maintained from stage E to stage F. In Thalia democratica, transcripts of Rar.b were found in two bilateral layers of calymmocytes (n = 4, Fig 5B, compared with controls in S9 Fig). No blastomeres were positive for the Rar.b probe at the examined stages (IIIa and IVa/IVb, S2 Table).

In Salpa fusiformis, Otx expression was first identified at stage Ea in the dorsal-most layer of calymmocytes between the closing incubation folds (Fig 5Ca–a′). At this stage, this layer directly overlies the dorsal blastomeres. At stages Eb (S1 Table), two distinct zones of expression were observed: one in a dorsal ring of calymmocytes surrounding the dorso-anterior blastomeres (Fig 5Cb–b′), and another ventral ring of calymmocytes (n = 2, Fig 5Cc–c′). At stage Fa, in addition to expression in calymmocytes a group of five dorsal blastomeres also expresses Otx (S10 Fig). In Thalia democratica, Otx.b is expressed in a few calymmocytes in the dorsal part of the embryo overlaying the dorsal blastomeres at stage IIb-early IIIa (n = 2, Fig 5D).

Intraspecific stereotypy and interspecific variability

The lack of comprehensive staging systems has hindered understanding of salp embryogenesis. Prior to this study, the only systematic staging for salp embryos were those of Heider (1895) [32] for Salpa fusiformis and the basic size-based one of Foxton (1966) [41] for Salpa thompsoni. A major contribution of this study is the establishment of two comprehensive developmental staging tables, one for each species (S. fusiformis and T. democratica), providing a much-needed standardized framework for staging salp embryos. The image-annotated tables included with this paper are designed to enable cross-species comparisons and facilitate reproducibility in future work. The resources here provided, which facilitate cross-species comparisons and future investigations on salp embryogenesis, were facilitated by made possible by the highly stereotypical nature of early embryonic stages within each species. This is evident in the consistent positions, sizes, and shapes of blastomeres, along with the previously documented stereotypy of calymmocytes [25,27,32]. The consistent topology of calymmocytes and their interactions with blastomeres support the existence of robust developmental mechanisms within each species. Currently, although the position of blastomeres during embryogenesis might hint at their possible fate, we were unable to follow their lineages. However, the reproducibility of early embryonic stages across individuals could make it possible to track cell fates with a sufficient number of embryos and a high temporal resolution.

The staging scheme and developmental descriptions we established for these two species, when placed within a broader phylogenetic framework, provide a valuable basis for reconstructing the evolutionary history of salp embryogenesis. While the developmental patterns of T. democratica and S. fusiformis are highly consistent within each species, notable interspecific differences arise between these two phylogenetically distant species [28], particularly in cleavage patterns, calymmocyte architecture and behavior, and the morphology of the incubation chamber. According to the most updated phylogeny from Damian-Serrano and colleagues [28] Thalia democratica belongs to the sister group of the majority of described salp species, including Ilhea punctata and all Salpa and Cyclosalpa species. The structural organization observed in S. fusiformis embryos, characterized by epithelial-like calymmocytes arranged in sheets, internal cavities, and incubation folds, has also been reported in other species such as Salpa maxima [25], Cyclosalpa pinnata [42,43] and Ihlea punctata (though the incubation folds were described to be incomplete in the latter) [25,33,44], and may thus be ancestral to this group. In contrast, T. democratica embryos are smaller, have a more compact structural organization, and develop faster [5]. This condensed development, might reduce the need for extensive signal diffusion, potentially explaining the role of calymmocytes-lined cavities in larger embryos (see below), where they may aid in signal dissemination.

Testing various evolutionary scenarios within salps will benefit from the study of additional, more basally branching species such as such as Pegea confoederata or Thetys vagina [28], as well as from a better description of the duration of embryogenesis across species.

Uncovering potential homologies

While establishing clear morphological benchmarks enables detailed developmental tables for both species—facilitating precise comparisons of developmental timing, cellular dynamics, and morphological traits—it remains challenging to define strict one-to-one homology between stages across the two species. Yet, a comparative analysis of embryogenesis in Salpa fusiformis and Thalia democratica reveals both conserved features and striking developmental divergence (Table 1). Early cleavage stages are broadly comparable: both species initiate development with a spherical oocyte surrounded by follicle cells and progress through stereotyped early divisions. The 8-cell stage marks a key point of similarity, as it coincides with the first invasion of maternal calymmocytes in both species. This stage also marks the beginning of blastomere compartmentalization by calymmocytes, a process that continues through subsequent stages and appears essential for establishing the embryonic architecture in both taxa, as in all described salp embryos [25,33,44]. The onset of organogenesis, which begins around stage late F/G in S. fusiformis and stage V in T. democratica, is another point of similarity, with similar formation of organ primordia by primary lumen formation. Apoptosis of calymmocytes is detected by TUNEL in both species before organogenesis onset, suggesting a shared mechanism for removing these maternal cells during tissue differentiation. Embryogenesis in both species culminates in the formation of a stolon-bearing oozooid [25], marking a homologous endpoint in body plan establishment.

Despite this conserved framework, significant divergence emerges at intermediate stages. One of the clearest differences lies in the architecture and behavior of calymmocytes. In S. fusiformis, calymmocytes form organized epithelial-like sheets that generate distinct internal cavities between blastomere clusters. This structural arrangement contributes to a well-defined bilateral symmetry and axial organization. In contrast, T. democratica exhibits compact, dense aggregates of blastomeres, with calymmocytes interspersed but lacking clear layering or cavity formation.

Additional divergences reinforce this pattern. The incubation chamber forms with a distinct morphology and at different time relative to embryo development, early via epithelial cone invagination in T. democratica, and later in S. fusiformis as the epithelial cones rises as two folds around the embryo. Differences also arise in blastomere topology: S. fusiformis embryos exhibit early and persistent bilateral symmetry pairing, whereas T. democratica shows a combination of bilateral and midline-aligned arrangements. Moreover, only S. fusiformis calymmocytes present epithelial-like sheet organization and cavity formation by calymmocytes, features entirely absent in T. democratica. These interspecific differences reflect underlying evolutionary plasticity in morphogenetic processes and likely arise from variations in embryo size, developmental timing, and ecological or reproductive constraints. Together, these findings emphasize that while S. fusiformis and T. democratica share core developmental milestones—including cleavage patterns, calymmocyte invasion, and juvenile morphogenesis—the intermediate steps by which they construct their embryos have diverged substantially. This highlights both the reproducibility of salp development within species and the diversity between them.

Maternal calymmocytes as regulators of embryonic architecture and signaling

Calymmocytes play active roles in shaping the embryonic microenvironment. By migrating between blastomeres, they impede direct signaling between them until contact is restored. They also contribute to the structural organization of the embryo, which is particularly visible in S. fusiformis, where they form cavities and epithelial-like sheets. Their presence may obscure or constrains blastomere movements, making it difficult to define classical gastrula or neurula stages.

Sustained physical interactions between calymmocytes and blastomeres are likely to play an important role in the coordination of the complex fate and behaviors of blastomeres. We have examined the expression of salp orthologs of two genes that, in ascidians, are associated with the specification of particular embryonic cell types [14,45,46]. Otx (a homeodomain transcription factor) and Rar (a nuclear receptor). In both species, the two genes are predominantly expressed in groups of calymmocytes. The expression of embryonic regulatory genes by calymmocytes suggests that these cells may perform functions usually associated with embryonic tissues, such as structural organization, cell signaling, and the establishment of embryonic compartments. For instance, Rar expression in calymmocytes surrounding the embryo suggests a role in patterning or signaling at the embryo-maternal interface. Otx is generally involved in the definition of cell identity during development and might have been co-opted to specify dorsal calymmocytes identity.

Conversely, blastomeres may also instruct calymmocytes fates and functions. The presence of the retinoic acid receptor Rar in calymmocytes is consistent with the possibility that these maternal cells are competent to respond to RA cues produced by embryonic blastomeres and/or other maternal tissues. Such RA responsiveness could contribute to calymmocytes regionalization, the organization of periembryonic compartments, or signaling dynamics at the embryo–maternal interface. Testing these possibilities will require functional perturbations (e.g., RA pathway manipulation and assessment of calymmocyte behavior/organization and downstream transcriptional responses).

Interactions between follicle cells and oocyte or embryonic cells have been rarely studied in ascidians. Some studies on follicle maturation have been conducted (reviewed in [47]), and there is solid evidence suggesting that test cells play a role in the later stages of embryogenesis particularly in tunic formation and metamorphosis [48,49]. An intriguing possibility is that the intimate follicle/embryo interactions observed in salps reflect a co-option of developmental programs originally deployed during asexual budding in colonial tunicates, where developing buds likewise engage in extensive, patterned interactions with surrounding adult tissues. Under this hypothesis, the cellular behaviors and molecular circuits that mediate adult–bud integration may have been redeployed during sexual embryogenesis in thaliaceans, thereby facilitating the emergence of viviparity. Testing this idea will require direct comparisons between embryogenesis and budding, e.g., assessing whether homologous cell types and conserved signaling/regulatory modules are shared between follicle/embryo interfaces and adult/bud interfaces across colonial ascidians and thaliaceans.

Elimination or integration? Tracking the fate of maternal cells across development

The fate of the calymmocytes during salp embryogenesis has been a topic of considerable debate. Early researchers, such as Salensky [34] (cited by Brien [25]), suggested that calymmocytes might contribute to the formation of future oozooid organs. However, there is general agreement that these cells degenerate [25,27,50]. This study suggests that at least some calymmocytes are eliminated through apoptosis, addressing a long-standing question regarding their fate. Using a TUNEL assay, we observed that apoptotic calymmocytes are present throughout various rather early developmental stages in both species.

However, the scattered distribution of apoptotic cells and the timing of their appearance suggest that the disappearance of calymmocytes may not be entirely accounted for by apoptosis during the early stages analyzed. Given that the onset of organogenesis—and thus the major phase of calymmocyte clearance—likely occurs after these stages, it remains possible that apoptosis plays a more prominent role later in development. Additional data covering a broader range of stages are needed to fully assess its contribution. During organogenesis, apoptosis of blastomeres may also play a morphogenetic role. This is supported by the presence of TUNEL+ cells at the sites of future oral siphon and cloaca opening in Thalia democratica. The role of apoptosis in siphon opening during ascidian juvenile development has not been documented, but morphogenetic roles of apoptosis have been documented in both solitary and colonial tunicates. In solitary tunicates apoptosis removes most larval tissues at metamorphosis [51] as well as extraembryonic test cells [52]. In the colonial ascidian Botryllus schlosseri apoptosis also play a prominent role during asexual development and organ homeostasis [53]. In addition to calymmocytes, we observed TUNEL+ nuclei in the epithelial layers of the incubation chamber of both species, beginning slightly before the organ primordia become visible. This process is likely critical for removing protective periembryonic layers around the embryo, facilitating its growth at later developmental stages.

Unlocking salp development: new tools for an emerging model

This study revisits and significantly expands upon previous, often fragmentary descriptions of salp embryogenesis, uncovering new dimensions of the dynamic interactions between maternal calymmocytes and zygotic blastomeres. By revealing structural and developmental features that had previously gone unnoticed, our findings highlight the complexity of salp development and challenge the prevailing notion of tunicate embryogenesis as uniformly conserved and stereotyped. Through a careful synthesis of novel observations with historical data, we provide robust and accessible anatomical resources that lay a solid foundation for the modern study of thaliacean developmental biology.

Importantly, recent advances now make it possible to breed salps in captivity and close their life cycle [54], a critical step that removes one of the main barriers to experimental research. Coupled with the availability of genomic (GCA_965202585.1) [55] and transcriptomic datasets [56], the field now has the essential tools to investigate this last major group of chordates that still lacks in-depth study. Approaches such as single-cell RNA sequencing (scRNA-seq) could yield unprecedented insights into the molecular identity and roles of calymmocytes, as well as their interactions with embryonic cells, shedding light on their evolutionary origins and developmental functions.

Together, our findings, in conjunction with these technological and methodological advances, underscore the urgency and promise of re-examining thaliacean biology with modern tools. The uniquely derived embryogenesis of salps—shaped by their transition to a pelagic lifestyle—offers an exceptional system to explore evolutionary plasticity of embryogenesis, maternal-zygotic interactions, and evolutionary innovation beyond the established ascidian models and the still-underexplored appendicularians. This reinvigorated focus should catalyze a new era of research into thaliaceans, enriching our understanding of chordate diversity and development.

Collection and fixation of the embryos

Blastozooids of Salpa fusiformis or Thalia democratica were collected during the spring blooms in the bay of Villefranche-sur-mer using a plankton net with a 150 µm mesh at a depth of 10–40 m. The larger embryos were exposed to the fixative by cutting the mother blastozooid or dissecting out the embryo before fixation. The smaller embryos were fixed inside the blastozooid. Animals reserved for morphological analysis were fixed in 4% formaldehyde or 4% paraformaldehyde (PFA) overnight at 4 °C before rinsing and conserved in phosphate-buffered saline (PBS). Animals reserved for WMISH and TUNEL were fixed in 4% PFA overnight at 4 °C and dehydrated in graded ethanol baths before storing at −20 °C as described in Christiaen and colleagues (2009) [57]. ProtectRNA RNase Inhibitor (Sigma R7397) 1× was added to all solutions for WMISH.

Staining, imaging,s and 3D-reconstruction whole-mount TUNEL assay

To study the anatomy of the embryos, we stained nuclear DNA with DAPI (Sigma 32670), Hoechst 34222 (Sigma 14533), or Draq5 (Thermo Fisher 62251) and either cellular actin with a fluorophore-conjugated phalloidin (Sigma P5282) or the membranes with FM4-64 (Fisher 10717864).

Embryos (about 90 for T. democratica and about 40 for S. fusiformis) were imaged by biphoton or confocal microscopy on a Zeiss confocal LSM 780, a Leica SP8 or a Leica Stellaris microscope. Images were analyzed with ImageJ and Imaris. The descriptions of Heider [32], Brien [25], and Sutton [27] were used as a framework for our analysis of the early development of Salpa fusiformis, and that of Brien [25] for Thalia democratica.

To quantify and compare the morphology of blastomeres and calymmocytes, one 13-cell stage embryo of each species was analyzed by manually reconstructing the cell and nuclear/chromatin surface of blastomeres and a random subset of calymmocytes from confocal images with Imaris (version 9.6.1). Several morphometric parameters were then calculated from these reconstructions to illustrate the differences observed visually.

To study the reproducibility of morphology and position of blastomeres, very similar stages were analyzed as described above with Imaris. Using the cortical actin/membrane/chromatin staining as a reference, we manually reconstructed four 4-cell stage and 12/13-cell stage (n = 3) T. democratica embryos, as well as 10- to 12-cell stage (n = 4) S. fusiformis embryos. Several morphometric parameters were then computed from these reconstructions to compare the embryos. The raw confocal images as well as the 3D reconstructions are available on the online web portal https://octopus.obs-vlfr.fr/public/salps/ and have been deposited in Zenodo (https://doi.org/10.5281/zenodo.18165317).

Whole-mount TUNEL assay

Dehydrated embryos were rehydrated with an incremental percentage of 0.1% Tween20 in PBS (PBT) mixed with ethanol, dissected out of the mother, and put in baskets in 24-well plates for further treatments. Samples were digested with a Proteinase K (Promega V3021) treatment at a concentration of 10 µg/ml for 30–45 min at 37 °C. The digestion was stopped with a quick incubation in ice-cold 2 mg/ml glycine in PBT. Samples were post-fixed in 4% PFA in PBT at room temperature and then washed in PBS. For positive and negative controls, samples were treated for 30 min at room temperature with 100 U/ml DNaseI (Fisher 10649890) to generate DNA breaks recognized by the TUNEL assay, then washed. The TUNEL labeling was done according to the kit (In Situ Cell Death Detection Kit, Sigma 11684795910) instructions, with 1h30 incubation at 37 °C. The embryos were then counterstained with Hoechst 33342 and mounted in PBS.

Phylogenetic analysis of Otx and Rar

Sequences of S. fusiformis were retrieved from the transcriptomes of Delsuc and colleagues 2018 [1], while T. democratica sequences were recovered from our in-house transcriptome database and have been deposited in GenBank under accession numbers PX826397, PX826398, PX833381, and PX833380 for Otx.a, Otx.b, Rar.a, and Rar.b, respectively. For Otx, blastp searches were performed using sp|P32242|OTX1_HUMAN as a query and 1e-70 as a threshold e-value. For Rar, blastp searches were performed using sp|P10276|RARA_HUMAN as a query and 1e−23 as a threshold e-value. Blast searches were performed against the non-redundant protein database (accessed through NCBI, Jan. 2025) for vertebrates, tunicates (with the exception of thaliaceans and styelidae), echinoderms, cephalochordates, hemichordates and a lophotrochozoan (the annelid Platynereis dumerilii). Additional blast searches were performed against the ascidians gene models in ANISEED database (v4.0) and against previously published Stolidobranch transcriptomes ([58], https://octopus.obs-vlfr.fr/public/botryllus/blast_botryllus.php). Sequences were aligned with MAFFT using G-INS-i strategy, then alignments were visually edited to remove ambiguously aligned regions using BioEdit. For Otx, we kept only the region corresponding to the homeodomain. A preliminary phylogenetic tree was constructed using PhyML, with the Q.insect+G+I for Otx and the Q.insect+R model for Rar (automatically selected best-fitting models). Appropriate outgroups were selected based on these trees. Then IQ-TREE 2 was used to build a Maximum Likelihood phylogeny with the best-fit model LG+G4 (for both Otx and Rar alignments) selected by ModelFinder following the Bayesian information criterion and with 10,000 ultrafast bootstrap replicates. The amino-acid alignments are provided in S6 and S7 Data.

Fluorescent WMISH

Primers (see S3 Table) were designed with the Primer3plus website (https://www.primer3plus.com/index.html), avoiding the conserved domains, if possible, then mapped by BLAST on the transcriptome to check for possible off-targets.

A mix of oozooids and blastozooids at various stages were collected, their gut and testes removed to limit contamination, then snap-frozen in liquid nitrogen. RNA was extracted from the frozen powdered sample with the kit Nucleospin RNA Mini (MACHEREY-NAGEL 740609.50). 0.5–1 µg of RNA was reverse transcribed with Superscript RT (Fisher 10328062). The probes were prepared by PCR with the designed primers and cloned into a pGEM-T Easy vector for selection and amplification. After linearization of the vector, Digoxygenin-labeled probes and sense probes were transcribed using T7 or SP6 RNA polymerases and purified with RNeasy kit (Qiagen 74104).

The WMISH protocol was adapted from Christiaen and colleagues (2009) [57] (performed as described until step 27 antibody washes). All solutions until antibody incubation contain ProtectRNA 1×. After washing with TNT as described, the embryos were then washed thoroughly with 0.5% BSA in PBT and PBS. They were then incubated with an in-house FITC-Tyramide mix for 3 h, counterstained with DAPI, and mounted in Vectashield (Cliniscience H-1000).