A thyroid hormone–mediated opsin switch initiates metamorphosis in a proto-vertebrate

Abstract

Thyroid hormones (THs) initiate metamorphosis in vertebrates, although the evolutionary origins of this process are uncertain. Here, we show that most TH processing genes are present in the proto-vertebrate model Ciona and play an instructive role in initiating metamorphosis, whereby swimming tadpoles with a chordate body plan are transformed into sessile filter feeders. Exogenous thyroxine (T4) accelerates metamorphosis, whereas TH inhibitors delay onset. Most notably, we present evidence that TH activates Opsin2 (Opsin1/2b) in a subset of photoreceptor cells in the simple brain of swimming tadpoles to initiate attachment, the first step in metamorphosis. Our findings suggest a deep evolutionary origin of TH-driven visual plasticity in vertebrates. We highlight the parallels between attachment of Ciona tadpoles and smoltification, whereby young salmonid fishes switch from UV to blue opsins for their transition from fresh to saltwater.

Thyroid hormone signaling activates Opsin2 in larval photoreceptor cells to initiate metamorphosis.

INTRODUCTION

Metamorphosis is the transformation of larvae into juveniles and a defining feature of many metazoan life cycles (1). In vertebrates, this process is coordinated by thyroid hormones (THs), which integrate external environmental cues with intrinsic developmental programs to specify stage-specific patterns of gene expression. Amphibians and flatfish provide classic models, whereby THs trigger extensive tissue remodeling alongside changes in behavior and habitat (2–5). THs also induce less notable postembryonic transitions, including changes in pigmentation (6), photoreceptor differentiation (7, 8), hatching (9, 10), and metabolic reprogramming (11). These effects are mediated by nuclear TH receptors, which transmit hormone availability into stage- and cell type–specific programs of gene expression.

Despite extensive studies in vertebrates, the core gene-regulatory architecture of TH signaling remains poorly defined. As the closest extant relatives of vertebrates (12), Ciona provide a simplified chordate system for understanding this process (13, 14). Larvae exhibit key chordate features, including a dorsal neural tube, notochord, and tail muscles, which are extensively remodeled or lost during metamorphosis (13). A compact genome, optical accessibility, small cell numbers, lineage maps, and extensive transcriptomic and proteomic atlases enable high-resolution dissection of gene regulatory networks underlying complex developmental processes (15–19).

Consistent with its evolutionary position as the sister group to the vertebrates, tunicates exhibit a variety of TH processes. For example, the endostyle is an iodine-concentrating pharyngeal organ believed to be an evolutionary precursor of the vertebrate thyroid (20–24). In juveniles and adults, different regions of the endostyle express genes implicated in TH synthesis, such as thyroid peroxidase and dual oxidases (25, 26). Tunicates also encode iodothyronine deiodinase homologs [e.g., hrDx in Halocynthia roretzi (27)] and a nuclear receptor that, unlike its vertebrate counterparts, does not exhibit high-affinity binding of thyroxine (T4) or T3 (28, 29). Nevertheless, T4 has been detected in larvae (30–32), and recent studies implicate T4 in TH-dependent apoptosis during tail regression (33).

Here, we explore the role of TH signaling in the development of sensory systems underlying the onset of metamorphosis. Of particular interest is the maturation of the visual system seen in the switch from larvae to juveniles in vertebrates such as salmon (34, 35) and coral reef fish (36). In Ciona, light- and gravity-guided swimming regulate dispersal and settlement, with the ocellus and otolith serving as principal sensory organs (37). The laser ablation of the ocellus disrupts larval photic swimming behaviors (38). We present evidence that TH receptor activity induces Opsin2 expression in photoreceptor cells of the sensory vesicle to promote attachment. Together, these findings suggest that the pleiotropic control of metamorphosis by TH predates the vertebrates.

RESULTS

Overview of TH signaling in Ciona

The Ciona genome has a conserved TH signaling toolkit for iodine uptake, hormone synthesis and transport, nuclear receptor binding, and tissue-specific regulation of gene expression (fig. S1A and table S1). Although Ciona lacks a canonical vertebrate hypothalamic-pituitary-thyroid axis [thyrotropin-releasing hormone, thyroid-stimulating hormone (TSH), and TSH receptor genes], TSH-like bioactivity has been reported in the endostyle and ovary (39). The lone TH receptor gene (Thr, aka CiNR1) (28) is first expressed in the endoderm during neurulation and peaks in swimming tadpoles (Fig. 1A). A 1.6-kb 5′ regulatory region drives reporter gene expression in the trunk endoderm and endodermal strand (Fig. 1B), which coincides with the expression of the endogenous Thr gene based on whole-mount in situ hybridization chain reaction (HCR) assays (fig. S1B).

Fig. 1. Characterization of the TH pathway.

(A) Schematic of the Ciona life cycle (top) and expression profiles of thyroid nuclear receptor (Thr, bottom). RNA [green, (16)] and protein [blue, (18)] levels peak during the swimming tadpole stage before attachment. RNA data are presented as means ± SEM (n = 2 to 3). Fert, fertilization; Hatch, hatching; Swi, swimming; Att, attachment; TR, tail regression; Juv, juveniles. (B) Larva expressing Thr membrane reporter (Thr>mNG::caax, green) in endodermal cells. en, endodermal trunk; est, endodermal strand. (C to E) Larvae expressing membrane reporters for different components of the pathway: synthesis-thyroid peroxidase (Tpo>mCherry::caax, mesenchymal cells, red), transporter-solute carrier family 5, member 5 (Slc5A5, encoding Nis) (Nis>mCherry::caax, mesenchymal cells, yellow), and activation/inactivation-iodothyronine deiodinase 1 (Dio1>fog::mNG::caax, muscle cells, cyan). Additional genes associated with the pathway in fig. S2. (F) Nkx2.1 membrane reporter (Nkx2.1>GFP::caax, orange) in endodermal cells of a hatching larva. (G) Schematic of the SplitGFP system (top) and larva coexpressing GFP reporters (Nkx2.1>GFP1-10 and Thr>GFP11, grey), with fluorescence reconstituted only in cells with both fragments (bottom). esp, endostyle primordium; post en, posterior endoderm. (H) Thr nuclear reporter (Thr>H2B::mCherry; top, magenta) and a version lacking the Nkx2.1 binding motif (ΔNkx2.1Thr>H2B::mCherry; bottom, magenta) in endodermal cells at mid tailbud stage. CRE: cis-regulatory element. (I) Quantification of nuclei expressing Thr nuclear reporter (left) and Nkx2.1 motif deleted reporter (right), related to (H). Violin plots are overlaid with boxplots showing median and upper and lower quartiles; individual points represent measurements from single tadpoles (n = 51 and 58, respectively). ***P < 0.001, one-way analysis of variance (ANOVA), followed by Tukey’s multiple comparisons test. White outlines mark embryo contours; lateral views with anterior to the left; yellow dotted outlines mark the ocellus; white dashed outlines mark the otolith. Scale bars, 100 μm [(B) to (G)] and 50 μm (H). n = 45 to 79 embryos, three to five biological replicates.

Conserved components of the TH signaling pathway exhibit localized expression in embryos and larvae (table S1). Tpo paralogs (Fig. 1C and fig. S2A) and associated oxidases Duox1 and Duox2 are expressed in mesenchymal cells (fig. S2, B to D). Similarly, sodium/iodide symporter Slc5a5 (Nis) (Fig. 1D) and accessory transporters such as Pendrin, monocarboxylate transporters (Mcts), and organic anion transporting polypeptides (Oatps) are also detected in mesenchyme and muscles (fig. S2, E to I). In contrast, deiodinases (Dio) and cytosolic carrier proteins (Pkm) exhibit localized expression in muscles (Fig. 1E and fig. S2J). Coregulators are also expressed in mesenchymal cells and muscles (fig. S2, K to M). For a subset of genes examined, HCR assays confirmed that the corresponding reporters recapitulate endogenous expression profiles (fig. S3). The localized expression patterns suggest broad communication across larval tissues for TH uptake, distribution, and activity (Fig. 1 and figs. S1 to S3). Ciona lacks vertebrate-specific thyroglobulin, although a recent study identified an alternative carrier in the related ascidian Styela (40).

Conserved thyroid gene regulatory network

In vertebrates, thyroid progenitors arising from the anterior endoderm coexpress a core set of transcription factors: Nkx2.1, Foxe1, Pax2/5/8, and Hhex (41–43). In Ciona, all of these genes except Hhex appear to play a role in the specification of the endostyle primordium (44). Nkx2.1 (aka Ttf1) expression begins at the 76-cell stage in endodermal blastomeres and persists through tailbud and larval stages (Fig. 1F and fig. S4), where it delineates the endostyle primordium in the anterior head endoderm (fig. S4C) (45, 46). Hhex is detected in mesenchymal cells (fig. S5A) (44). FoxE1 (aka Ttf2) overlaps with Nkx2.1 in endodermal cells from tailbud stages onward (fig. S5B) and is maintained via a dedicated shadow enhancer that reinforces late expression in the endostyle primordium in swimming tadpoles (fig. S5C) (22, 47, 48). Pax2/5/8a and Pax2/5/8b paralogs exhibit broader patterns of expression in neural precursors and mesenchyme but are also expressed in the endostyle primordium of hatching tadpoles (fig. S5, D and E) (47, 49).

We obtained evidence that Nkx2.1 serves as a direct activator of Thr expression in the endoderm. Split–green fluorescent protein (GFP) (50) assays confirmed spatial colocalization of Thr and Nkx2.1 (Fig. 1G), and CRISPR-mediated knockdown of Nkx2.1 in endodermal cells decreased Thr at mid-tailbud stages as measured by reverse transcription quantitative polymerase chain reaction (RT-qPCR) (fig. S5, F to H). Moreover, targeted mutations in a predicted Nkx2.1 binding site diminish the endodermal expression of an otherwise normal Thr reporter gene (Fig. 1, H and I). Together, these results support a vertebrate-like thyroid regulatory circuit used by the Ciona endoderm and endostyle primordium.

Thyroid receptor controls endodermal strand migration

The endoderm of Ciona tadpoles is composed of ~500 cells that are organized into trunk endoderm and an endodermal strand in the tail that lies beneath the notochord and has been likened to the vertebrate hypochord (51, 52). These cells migrate anteriorly into the trunk at the onset of metamorphosis before tail regression. Thr expression in the retracting endodermal strand of swimming tadpoles suggested a potential mechanism for the timing of tail regression. To test this, we manipulated Thr activity using an endoderm-wide Nkx2.1 5′ regulatory DNA (fig. S6) and an endodermal strand–specific Zip2 enhancer (fig. S7) (53, 54). The overexpression of full-length Thr accelerated the anterior migration of endodermal strand cells toward the trunk (Nkx2.1>thr), whereas a constitutive repressor form of the receptor (Nkx2.1>thr::WRPW) delayed or blocked migration (fig. S6A). Previous studies identified Hox10 as a key agent of endodermal strand cell migration (53). Consistent with this, the endoderm-wide overexpression of the wild-type Hox10 coding sequence (Nkx2.1>hox10) or constitutive Hox10 repressor (Nkx2.1>hox10::WRPW) phenocopied gain- and loss-of-function Thr activities, respectively (fig. S6B). Moreover, Hox10 overexpression rescued the migration defects induced by Thr::WRPW (fig. S6C). Similarly, the overexpression of normal or repressor forms of Thr via endodermal strand–specific manipulation either accelerated (Zip2>thr) or impaired cell migration (Zip2>thr::WRPW) (fig. S7). These epistasis experiments support a model in which Thr coordinates endoderm migration and the timing of tail regression through the regulation of Hox10.

T4 triggers and accelerates metamorphosis

To confirm TH responsiveness during metamorphosis, we treated embryos with exogenous L-thyroxine (T4) or triiodothyronine (T3) and monitored tail regression, a quantifiable process at the onset of ascidian metamorphosis (Fig. 2A) (55, 56). The continuous treatment of 70 nM T4 from fertilized eggs to swimming tadpoles led to precocious tail regression in more than 60% of larvae, with dose-dependent effects (Fig. 2B and fig. S8). RT-qPCR assays confirmed that these treatments led to transcriptional up-regulation of core pathway components (Thr, Tpo, and Dio; fig. S9). By contrast, T3 treatment had no effect on tail regression nor marked changes in gene expression (Fig. 2C and fig. S9). T4, or one its derivatives, therefore appears to be the primary TH signaling agent in Ciona, while T3 serves this role in vertebrates. Moreover, the pharmacological inhibition of TH synthesis or activation (methimazole, sodium perchlorate, and iopanoic acid) delayed or blocked metamorphosis in a dose-dependent manner (figs. S10 and S11), corroborating earlier thiourea-based results (30, 31, 33). Cotreatment with T4 partially rescued these phenotypes (fig. S10B), supporting a role for T4 in TH signaling in Ciona metamorphosis.

Fig. 2. T4 regulates larval attachment and tail regression.

(A) Representative bright-field images showing progressive tail regression at early (0%), mid (50%), and late (75%) stages. Scale bar, 100 μm. (B and C) Quantification of tail regression (%) at 6 hours posthatching after T4 treatment (B) and 8 hours posthatching after T3 treatment (C) at indicated concentrations. Boxplots show individual replicates (Rep 1 to 3). Additional experiments associated with the T4 tail regression in fig. S8. ***P < 0.001 and ****P < 0.0001; ns, not significant, ANOVA followed by Tukey’s multiple comparisons test. Sample sizes (n) are shown below each group. (D) Timeline of key developmental stages (hours postfertilization, hpf): swimming, attachment competence, and tail regression. T4 exposure windows for experimental groups (1 to 5) indicated alongside attachment and tail regression responses. (E) Schematic of attachment assay and pie charts showing percentage of larvae attached (red) versus swimming (gray) across experimental groups (control and groups 1, 2, 3, and 5) related to (D). n = 608 to 706, seven biological replicates. (F) Schematic of tail regression assay and violin plots with percentage tail regression for control and groups 1 to 5 related to (D). Boxplots (median and upper and lower quartiles) and individual data points from replicates (Rep 1 to 4) are overlaid. **P < 0.01, ***P < 0.001, and ****P < 0.0001, ANOVA followed by Tukey’s multiple comparisons test. hph, hours posthatching.

We next asked when T4 acts during metamorphosis and whether it controls the timing of larval attachment. Continuous T4 treatment from fertilization shortened the free-swimming phase and promoted earlier substrate attachment (Fig. 2, D and E). A T4 pulse during and after hatching (17.5 to 21 hours postfertilization) was sufficient to trigger precocious attachment (Fig. 2, D and E, condition 3), suggesting that tight regulation of T4 levels gates this transition. Precocious attachment was followed by accelerated tail regression (Fig. 2, D and F, condition 3). T4 treatment initiated after attachment was also found to accelerate tail regression (Fig. 2, D and F, condition 4). These data identify attachment as the primary T4-responsive developmental checkpoint, whereas tail regression represents a downstream developmental process. These hormone-timing results complement our previous analysis of the Thr-Hox10 control of endodermal strand migration.

A wave of T4 activity precedes tail regression

We first confirmed T4 staining in mesenchymal cells at hatching (fig. S12A), consistent with earlier observations (30, 31) and with the enrichment of TH signaling components in this cell type. A time course of histochemical staining revealed a dynamic, ordered sequence of T4 accumulation in swimming larvae, beginning shortly after hatching in photoreceptor cells of the sensory vesicle, identified by their proximity to the ocellus (fig. S12B), followed by the palps (fig. S12, C and D) and then posterior tail epidermis as previously reported (figs. S12, E to H, and S13) (33). At the stage when T4 first becomes detectable in the palps, a subset of larvae also shows T4 signal in mesenchymal cells, consistent with dynamic hormone production and distribution (fig. S12D). Tail regression proceeds as a polarized apoptotic wave that originates at the distal tip of the tail and propagates anteriorly toward the trunk (57, 58). T4 staining coincides with apoptotic cells and preceded the advancing front of cell death in the tail epidermis after hatching (figs. S12, E to H, and S13) (33). Thr expression expands from the endoderm and coincides with the evolving distribution of T4 during this process (fig. S14). In batch-matched embryos, both T4 and Thr signals progressed reproducibly from photoreceptors to the tail epidermis over time, consistent with coordinated pathway activation across different tissues. These studies do not rule out the possibility that T4 might act indirectly via local conversion to other bioactive ligands.

To determine the relationship between T4 signaling and apoptosis, we treated swimming larvae with T4 and monitored for caspases activation (33, 57). T4 exposure increased endogenous Caspase8 expression as measured using HCR assays (fig. S15A). It also induced earlier and higher TUNEL (terminal deoxynucleotidyl transferase–mediated deoxyuridine triphosphate nick end labeling) signals throughout the larva and at the tip of the tail as compared with control larvae (fig. S15B). Together, these observations are consistent with a T4-driven apoptotic trigger that parallels TH-associated tail regression in frog tadpoles (2).

The early T4 signal in photoreceptor cells is rapidly lost, indicating that this specialized TH input is transient and specific to late swimming stages. This may represent the initial step that signals the start of metamorphosis, a mechanism possibly conserved in fishes and frogs. Overall, the ordered sequence of T4 and Thr expression is consistent with a long-range, head-to-tail TH signaling program that coordinates both the timing of attachment and the directionality of tail regression.

An Opsin switch in photoreceptor cells

We next explored the possibility that transient expression of Thr in the sensory vesicle contributes to the timing of attachment (fig. S14). We therefore examined the temporal expression of visual pigments Opsin1 (aka Opsin1/2a) and Opsin2 (aka Opsin1/2b) in photoreceptor cells of the sensory vesicle. Opsin1 is broadly expressed in the developing sensory vesicle at tailbud stages and becomes progressively restricted to the photoreceptor cells of the ocellus (eye spot) after hatching (fig. S16A) (59, 60). Consistent with this pattern, Opsin1 was shown to be important for photic behaviors of swimming tadpoles following hatching (38, 59, 61–63). Opsin2 is only weakly expressed in a subset of photoreceptor cells during the late swimming stage when a behavioral shift from active swimming to attachment is observed (fig. S16, B and C). These Opsin2-positive photoreceptor cells arise from Opsin1-expressing cells (Fig. 3A). Coexpression analysis using HCR assays validated this developmental sequence, with Opsin1 expression preceding Opsin2 (Fig. 3B). Moreover, triple labeling assays identified a transient pulse of Thr expression in these photoreceptor cells at the time of the switch from Opsin1 to Opsin2 expression (Fig. 3B).

Fig. 3. Thyroid hormone receptor forms a transient circuit in photoreceptor cells.

(A) Swimming larvae expressing Opsin1 membrane-targeted reporter (Opsin1>GFP::caax, fuchsia) and Opsin2 nuclear reporter (Opsin2>H2B::mSc, orange), shown separately and merged. A subset of Opsin2-positive cells overlaps with the Opsin1 reporter. Boxed regions are higher magnification views of photoreceptor cells. (B) Sequential dynamics of Opsin1 (magenta), Thr (green), and Opsin2 (yellow) mRNA during early, mid, and late swimming phases detected by triple in situ HCR. (i) Early swimming: Opsin1 marks photoreceptors, whereas Thr expression is restricted to the endodermal trunk (maximum projection). cc, coronet cells; en, endoderm. (ii) Mid swimming: a transient wave of Opsin1-Thr coexpression appears in photoreceptors (arrows; optical section); Opsin2 is not detected. (iii) Late swimming: Opsin2 expression emerges in a subset of Opsin1-Thr-positive cells (arrowheads; maximum projection). Left panels show whole mounts; boxed regions are shown at higher magnification in the corresponding panels to the right. White outlines mark embryo contours; lateral views with anterior to the left; yellow dotted outlines mark the ocellus; white dashed outlines mark the otolith. Scale bars, 100 μm. (A) n = 38 to 41 embryos, two biological replicates. (B) n = 52 to 56 embryos, three biological replicates.

We next asked whether Thr directly activates Opsin2. We therefore expressed a constitutively active variant of Thr (thr::VP16) under the control of an Opsin1 enhancer (Opsin1>thr::VP16). This fusion gene triggered premature and expanded activation of an Opsin2 reporter (Opsin2>H2B::mNG) in the photoreceptor cells (Fig. 4A and fig. S17A). To test the specificity of activation, we mutated a predicted Thr-responsive element within the Opsin2 regulatory region (ΔThrOpsin2>H2B::mSc). Opsin1>thr::VP16 failed to activate the mutagenized Opsin2 reporter (Fig. 4B and fig. S17B), suggesting that Opsin2 is a direct transcriptional target of Thr.

Fig. 4. Thyroid hormone receptor in photoreceptors induces Opsin2 and promotes larval attachment.

(A) Higher magnification of the sensory vesicle expressing Opsin2 reporter (Opsin2>H2B::mNG, cyan) together with the control construct (Opsin1>lacz, left) or Thr active form (Opsin1>thr::VP16, right). Photoreceptor-restricted Thr::VP16 is sufficient to activate precocious Opsin2 reporter. Lower magnification of the larvae in fig. S17A. (B) Higher magnification views of the sensory vesicle: Thr activates wild-type Opsin2 reporter (Opsin2>H2B::mNG, cyan, left) but not the Thr binding site mutant reporter (ΔThrOpsin2>H2B::mScarlet, magenta, middle) in Opsin1-positive photoreceptors; right, merge. Lower magnification views in fig. S17B. (A and B) Yellow dotted and white dashed lines mark the ocellus and otolith, respectively. Scale bars, 20 μm. n = 75 to 86 embryos per condition, five biological replicates. (C) Larval attachment rates following photoreceptor-specific Opsin2 overexpression (Opsin1>opsin2) compared with control (Opsin1>lacz). Stacked bars show the fraction of attached (dark blue) versus swimming (grey) larvae; n values are shown above bars. Error bars indicate 95% Wilson score confidence intervals; **P < 0.01 and ***P < 0.001, binomial generalized linear-mixed effects model (GLMM). (D) Larval attachment rates following photoreceptor-specific CRISPR perturbations using the Opsin1 reporter. Stacked bars show the fraction of attached (red) versus swimming (grey) larvae for control single guide RNA (sgRNA), Thr sgRNA, and Thr sgRNA with rescue by Opsin2 mRNA. Sample sizes are indicated above bars. Error bars indicate 95% Wilson score confidence intervals; ****P < 0.0001, binomial GLMM.

To determine whether the Opsin1/Opsin2 switch is important for the timing of attachment, we examined the consequences of precocious and expanded expression of Opsin2. This was achieved by electroporating an Opsin1>opsin2 fusion gene whereby Opsin2 is broadly activated in the photoreceptor cells of the sensory vesicle using Opsin1 regulatory sequences. Swimming tadpoles expressing this transgene were found to undergo attachment more quickly than control larvae (Fig. 4C). Conversely, CRISPR-mediated Thr knockdown (Opsin1>cas9; Thr gRNA) resulted in delayed attachment. This delay was rescued by Opsin2 overexpression (Opsin1>cas9; Thr gRNA; Opsin1>opsin2) (Fig. 4D). Together, these findings suggest that the Opsin1/Opsin2 switch is critical for the timing of attachment and the onset of metamorphosis (see below).

DISCUSSION

We have presented evidence that tissue-restricted Thr activity coordinates different morphogenetic and behavioral processes in the Ciona tadpole, including directed migration of the endodermal strand, tail regression, and tissue remodeling during metamorphosis. Conserved transcription factors implicated in vertebrate thyroid lineage specification are also present in the Ciona endoderm, supporting the existence of a deeply conserved regulatory network linking the endostyle and thyroid to TH signaling (20, 26, 41, 43). In Ciona, Nkx2.1 and FoxE delineate the endostyle primordium, while Pax2/5/8 paralogs contribute to its subsequent development at hatching (47–49). TH signaling precedes the differentiation of a fully functional endostyle in Ciona, whereas in vertebrates canonical TH-driven remodeling is best characterized after the formation of a thyroid gland.

Localized Thr expression triggers anterior migration of the endodermal strand through Hox10 regulation, linking receptor action to a defined cellular rearrangement that precedes tail regression. In some ascidians, migration occurs before attachment (53), while in others it occurs afterward (64). In Ciona, tail regression is first anticipated by a localized patch of Thr expression in the posterior-most epidermis. This leads to a dynamic anterior-to-posterior wave of T4 accumulation and activation of caspases (33, 57, 58). Together, these observations support a model in which TH establishes temporal and tissue competence early, with later execution mediated by downstream effectors such as caspases, echoing TH-patterned remodeling in amphibian metamorphosis.

TH usage has diversified during chordate evolution. Early-branching chordates such as amphioxus rely on TRIAC (3,3′,5-triiodothyroacetic acid) (29); most vertebrates (fishes, frogs, and mammals) use T3 as the high-affinity ligand with T4 functioning largely as a prohormone (1–3, 7, 65), and lampreys are outliers marked by a sharp decline in TH levels at the onset of metamorphosis (66). In Ciona, we find that T4 is sufficient to trigger precocious attachment, tail regression, and onset of metamorphosis. The refinement of ligands and receptor affinity is well documented in other pathways such as retinoic acid signaling (67) and may have likewise sharpened TH signaling during the transition from a simple, iodine-concentrating endostyle to the metabolically demanding vertebrate thyroid.

Attachment is a critical event in the life cycle of ascidians because it irreversibly determines where and when motile larvae form sessile juveniles and adults in a new habitat (55). A key finding of our study is the link between TH signaling and attachment. The activation of Opsin2 by Thr provides a simple endocrine-to-sensory mechanism for controlling the timing of this process. Opsin2 activation occurs during the late phases of tadpole swimming. During this time, there is an increase in Thr activity in a subset of photoreceptor cells in the sensory vesicle. This triggers Opsin2 expression, and epistasis experiments place Opsin2 downstream of Thr in controlling the timing of attachment.

Ciona larvae exhibit multiple visually guided behaviors, including negative phototaxis and a dimming/looming-shadow response, mediated by partially overlapping groups of photoreceptors (68–70). Opsin1-expressing photoreceptors are required for larval photoresponses (59), and morpholino knockdown of Opsin1 abolishes these light-driven behaviors (61). Other studies showed that negative phototaxis emerges several hours after hatching and that dimming responses vary with light intensity (38, 70–73). Building on this framework, our data indicate that Opsin2 is induced within a subset of Opsin1 photoreceptors during late swimming stages and helps drive the behavioral transition toward settlement.

This regulatory architecture mirrors a broader principle in which activity of TH receptor shapes photoreceptor gene programs. The TH receptor–dependent regulation of opsins expression is broadly conserved in vertebrates, including mammals (e.g, retinal organoids) (8, 74). In fish, TH signaling is linked to spectral tuning during metamorphosis (3, 36, 75), as well as the switch from ultraviolet-to blue-sensitive opsins as salmonids prepare to enter seawater (34, 35, 76–78). In Ciona, the Opsin1-to-Opsin2 switch may tune the sensitivity of attachment to twilight, when ascidian tadpoles initiate metamorphosis in the wild (79, 80). Because opsins differ in wavelength sensitivity (34), we propose that TH-mediated Opsin switching provides a relay between environmental light cues and the decision to attach. We propose that this coupling of endocrine state to sensory cues is an ancient chordate strategy that predates the advent of the vertebrates.

MATERIALS AND METHODS

Ciona handling, electroporation, and fixation for reporter assays

Adult Ciona robusta (formerly Ciona intestinalis type A) were obtained from M-Rep (San Diego, CA) or Marinus Scientific (Long Beach, CA) and maintained in 33‰ artificial seawater at 18°C under constant light to prevent spontaneous spawning. Gametes were surgically collected, and embryos were dechorionated and fertilized following established protocols (81).

For reporter assays, 20 to 60 μg of plasmid DNA (table S2) was electroporated using a Bio-Rad GenePulser Xcell with a CE module. Electroporated embryos were cultured at 18°C in filtered artificial seawater (FASW) and fixed at appropriate developmental stages (Tunicanato; https://chordate.bpni.bio.keio.ac.jp/tunicanato/3.0/) for 15 to 30 min in minimum essential medium–paraformaldehyde (MEM-PFA) (4% paraformaldehyde, 0.1 M Mops (pH 7.4), 0.1 M NaCl, 1 mM EGTA, 2 mM MgSO₄, and 0.05% Tween 20). Fixed embryos were washed three to five times in PBT (phosphate-buffered saline with 0.1% Tween 20), each wash lasting ~5 min. Samples were stored in PBT at 4°C until further processing. Before imaging, embryos were mounted in FluorSave reagent (Sigma-Aldrich) and imaged using a Zeiss LSM 880 inverted confocal microscope (Carl Zeiss). Multiple replicate electroporations were performed for all experiments using independent batches of Ciona eggs. Each transgenic experiment was carried out at least three times in the same conditions using animals collected on different dates.

Molecular cloning

The KY21 identifiers for all genes mentioned in the manuscript, along with commonly used gene names, are listed in table S3. Enhancer sequences were PCR-amplified from genomic DNA using primers found in table S4. Amplified fragments were cloned into the pCESA vector or upstream of the basal Fog gene promoter using either restriction sites (Asc I, Not I, or Xba I) or NEBuilder HiFi assembly. The resulting reporter constructs included the fluorophores GFP::caax (caax is a palmitoylation motif targeting proteins to membranes), mNeonGreen::PH (mNG::PH, pleckstrin homology domain for membrane labeling), mNeonGreen::caax (mNG::caax), H2B::mNG, mCherry::caax, H2B::mCherry, and H2B::mScarletI3 (H2B::mSc). The regulatory regions for Nkx2.1 and Opsin1 have been described previously (45, 59, 60).

Point mutations in the putative Nkx2.1 binding sites were generated by plasmid PCR of the Thr>H2B::mNG construct using primers listed in table S4. Binding motifs were derived from published SELEX-seq datasets and the human NKX2.1 binding motif from the JASPAR database (82). Point mutations in the putative Thr binding sites were generated by plasmid PCR of the Opsin2>H2B::mNG or Opsin2>H2B:: mSc construct using primers listed in table S4. Binding motifs were derived from the human THR binding motif from the JASPAR database (82).

Coding sequences for Thr, Hox10, and Opsin2 were PCR-amplified from mixed-stage cDNA (mid-tail, late tailbud, hatching, and late swimming stages) with PrimeStar HS DNA polymerase (Takara). The lacz open reading frame was amplified from a previously described expression vector (17). These sequences were then cloned downstream of appropriate regulatory sequences of the pCESA vector using NEBuilder HiFi assembly (NEB), resulting in Nkx2.1>thr, Nkx2.1>hox10, Opsin1>thr, and Opsin1>opsin2. Additional constructs encoding the WRPW repressor motif or VP16 activation domain fused to Thr and Hox10 coding sequences were generated through PCR amplification and subcloning using the NEBuilder HiFi method from existing expression vectors, resulting in Nkx2.1>thr::WRPW, Nkx2.1>thr, Nkx2.1>hox10, and Nkx2.1>hox10, Zip2>thr::WRPW, Zip2>thr, and Opsin1>thr::VP16.

To drive robust and specific expression of Cas9 in the endoderm, the regulatory region of the Eef1a1>nls::Cas9::nls plasmid (83) was replaced with the Nkx2.1 enhancer using NEBuilder HiFi assembly. Subsequently, the T2A-mNG sequence was subcloned downstream of Cas9 using NEBuilder HiFi assembly, generating the Nkx2.1>nls::Cas9::nls::T2A:mNG construct. For photoreceptor-specific expression of Cas9, the Opsin1 enhancer was cloned upstream of the Cas9 coding region using NEBuilder HiFi assembly. Single guide RNAs (sgRNAs) targeting Nkx2.1 and Thr coding regions were designed using ChopChop software (https://chopchop.cbu.uib.no/) and are listed in table S5. A negative CRISPR control sgRNA (CTTTGCTACGATCTACATT) was included in every experimental replicate. The specificity and efficiency of sgRNAs were validated through electroporation experiments, followed by PCR amplification of targeted genomic regions and confirmation by Sanger sequencing.

T4 immunofluorescence and TUNEL staining

Embryos were fixed in MEM-PFA for 45 min at room temperature with gentle rocking, followed by three to five washes in PBST (PBS with 0.1% Triton X-100) for 5 min each. Samples were blocked in PBST containing 2% bovine serum albumin (BSA) for 1 hour at room temperature and then incubated overnight at 4°C with rabbit anti–l-thyroxine polyclonal antibody (1:1000; Sigma-Aldrich, T-2652) diluted in PBST with 2% BSA. After primary antibody incubation, embryos were washed three times in PBST (10 min each) and incubated with Alexa Fluor 488–conjugated donkey anti-rabbit immunoglobulin G (1:3000; Thermo Fisher Scientific, A21202) for 1 to 2 hours at room temperature or overnight at 4°C in PBST containing 2% BSA. Embryos were then washed three to five times in PBST (10 min each) and either processed immediately for TUNEL staining or stored at 4°C until imaging.

TUNEL staining was performed after T4 immunostaining using the In Situ Cell Death Detection Kit (TMR Red, Roche), following the manufacturer’s instructions and previously described methods (57, 58). Briefly, larvae were incubated in a reaction mixture (10% enzyme solution in labeling solution) for 1 hour at 37°C on a thermomixer (200 to 300 rpm), rinsed several times in PBST, mounted in FluorSave (Millipore), and imaged using a Zeiss LSM 880 confocal microscope. The embryos used in figs. S12 to S14 were from the same batch, split, and used in parallel for either T4 staining or in situ HCR.

Treatments with thyroxine, triiodothyronine, and dexamethasone

T4 (l-thyroxine, T2501, Sigma-Aldrich) and T3 (triiodothyronine, T6397, Sigma-Aldrich) were dissolved at 100 mM in dimethyl sulfoxide (DMSO) and stored at −20°C. Before experiments, THs were diluted in FASW to the required working concentrations.

To examine the effects of THs on tail retraction speed, 70 to 150 fertilized eggs were placed into each 12-well polystyrene cell culture nontreated plates, precoated with 0.05% gelatin. Embryos were exposed either to seawater alone, DMSO alone (vehicle control, Sigma-Aldrich; #D8418), or THs at final concentrations of 70, 140, or 600 nM. Three independent biological replicates were performed on separate days. At the designated developmental stages, embryos were fixed in MEM-PFA, washed in PBST, and the percentage of tail retraction was assessed using a microscope.

To define the temporal window of T4 activity, fertilized embryos with chorion were distributed into small petri dishes and incubated in 70 nM T4 at specific developmental time points. Control embryos were exposed to equivalent concentrations of DMSO. To assess larval settlement, a glass Pasteur pipette fitted with a rubber bulb was used to gently swirl the seawater; settled larvae were identified as those unable to move from their attachment sites. The number of larvae stably adhered to the petri dish was manually counted. The effect of T4 on metamorphosis was quantified by monitoring the fraction of larvae with fully resorbed tails under a dissecting microscope. When ~50% of control larvae (DMSO-treated) had metamorphosed, the proportion of metamorphosed larvae in T4 treatment was manually determined.

Differences among all treatment conditions (DMSO, T4, and T3) were assessed using the one-way analysis of variance (ANOVA), followed by Tukey’s multiple comparisons test. Statistical significance was defined as a P value < 0.05.

Thyroid hormone inhibitors

Sodium perchlorate (Sigma-Aldrich, 410241), an inhibitor of the sodium-iodide symporter, was dissolved directly in FASW at a final concentration of 2, 3, and 4 g/liter. Methimazole (Sigma-Aldrich, M8506), an inhibitor of thyroid peroxidase, was used at final concentrations of 1.5, 10, and 150 μM in DMSO (Sigma-Aldrich; Merck KGaA, #D8418). Iopanoic acid (Sigma-Aldrich, S14131), which targets deiodinases, was applied at final concentrations of 6 and 10 μM. Treatments were added immediately after fertilization to embryos with chorions intact. Embryos (100 to 150 per condition) were distributed into each wells of 12-well polystyrene nontreated cell culture plates. Both treated and untreated embryos were fixed in MEM-PFA and washed in PBST, and the progression of tail regression was quantified under a dissecting microscope. To evaluate the impact of THs pathway inhibitors, we compared the distribution of tail regression categories (0 to 30%, 30 to 50%, 50 to 80%, and 80 to 100%) between treated and control embryos using chi-square tests of independence. Pairwise comparisons between conditions were conducted using 2 × 4 chi-square tests with Holm correction for multiple testing. Statistical significance was set at P < 0.05. For rescue assays, embryos were exposed to sodium perchlorate with or without cotreatment with exogenous T4, then fixed, and scored for juvenile developmental categories; distributions were compared to controls using chi-square tests with Holm correction as above.

RNA isolation, cDNA synthesis, and RT-qPCR

To investigate the Nkx2.1-Thr gene regulatory network, embryos were electroporated with Nkx2.1>nls::Cas9::nls::T2A::mNG plasmids along with either control or Nkx2.1-specific guide RNAs designed with ChopChop as described above. One hundred fluorescent embryos were snap-frozen in TRIzol reagent (Invitrogen) at appropriate developmental stages. To evaluate the biological effects of T4 and T3 on gene expression, embryos were exposed to 70 μM T4 or 70 μM T3 immediately after dechorionation and fertilization. At the right developmental stage, 150 embryos were snap-frozen in TRIzol. Total RNA was isolated using the Direct-zol RNA MiniPrep Kit (Zymo), including a 10-min deoxyribonuclease I digestion step as per the manufacturer’s instructions. Approximately 500 ng of RNA was reverse-transcribed to cDNA using the iScript cDNA Synthesis Kit (Bio-Rad). RT-qPCR was performed using a ViiA 7 Real-Time PCR system (Applied Biosystems) with SYBR Green fluorescent dye (Thermo Fisher Scientific). Primer efficiencies and specificity were validated by standard curves from serial dilutions of cDNA and melting curve analysis for each primer set (table S4). Each biological replicate included technical replicates. Data were analyzed using the ΔCt method normalized to actin gene expression. Individual replicate values and means ± SEM are presented. Statistical analysis was conducted in R Script for the analysis is available on GitHub at https://github.com/andreamariossi/thyroid_hormone_Ciona.

Whole-mount in situ hybridization (HCR)

Electroporated eggs or chorionated embryos were fixed at the desired developmental stages in 100 mM Hepes, 500 mM NaCl, 2 mM MgSO₄, 2 mM ethylene glycol bis (succinimidyl succinate), and 1% formaldehyde as previously described (84, 85). Fixed samples were stored in 70% ethanol at −20°C until further processing. In situ HCR protocol was performed as previously described. Briefly, embryos were prehybridized in hybridization buffer for 2 hours at 37°C, followed by incubation overnight at 37°C with 1 pg of probes. Probes were either commercially obtained from Molecular Instruments or custom-designed with Easy_HCR (86). For the amplification step, embryos were incubated at least 4 hours or overnight at room temperature in the dark. Samples were then washed extensively (>4 times) in 5× SSCT (Sodium Chloride Sodium Citrate and 0.1% Tween-20) and stored at 4°C or mounted onto microscope slides using RapiClear 1.49 (Sunjin Lab) before imaging.

Endodermal strand measurement

Embryos were co-electroporated with Nkx2.1>GFP::caax or Nkx2.1>H2B::mCherry and one or a combination of the corresponding misexpression constructs (Nkx2.1>thr::WRPW, Nkx2.1>thr, Nkx2.1>hox10, or Nkx2.1>hox10:WRPW). To label the endodermal strand, we electroporated Zip2>H2B::mCherry together (53, 54) with the corresponding misexpression constructs.

To assess endodermal strand cell migration, subsets of embryos per electroporation were fixed at defined developmental stages (17.3, 22, 25, and 26 hours postfertilization) in MEM-PFA. Fixed embryos were washed in PBST, and a stereo microscope was used to classify the extent of cell migration at each time point. Statistical analysis was conducted using ANOVA, followed by Tukey’s multiple comparisons test.

Attachment measurement

Embryos were coelectroporated with Opsin1>nls::Cas9::nls::T2A::mNG together with either a control guide RNA or a Thr-targeting guide RNA (designed using ChopChop, as described above). For rescue experiments, embryos were additionally coelectroporated with the Opsin1>opsin2 reporter. To test the effect of Opsin2 alone, embryos were electroporated with Opsin1>lacz or Opsin1>opsin2 reporters. Embryos were distributed into small petri dishes. To check attachment, seawater was gently swirled using a glass Pasteur pipette fitted with a rubber bulb; settled larvae were scored as individuals that remained immobile and could not be displaced from their attachment sites. The number of larvae stably adhered to the dish was counted manually. Larval attachment was analyzed using binomial generalized linear-mixed effects models, with 95% Wilson score confidence intervals for proportions.

RNA-seq and protein mass spectrometry data analysis

The extended developmental time-series RNA sequencing (RNA-seq) data for Ciona development were sourced from (16) and processed using the RNA-seq nf-core pipeline. Briefly, RNA-seq data were downloaded from DNA Data Bank of Japan (accession number DRA003460) and mapped to KY21 reference genome using Kallisto with KY21 annotations and the Ensembl mitochondrial genome. Gene-level expression values were derived by aggregating transcript-level abundances [transcripts per million (TPM) values]. RNA-seq data from biologically replicates were pooled to calculate average gene expression levels at each stage. Proteomic data were sourced from (18).

Supplementary Materials

This PDF file includes:

Figs. S1 to S17

Tables S1 to S5