Evolutionary Conservation of Thyroid Hormone Receptor and Deiodinase Expression Dynamics in ovo in a Direct-Developing Frog, Eleutherodactylus coqui

Direct development is a reproductive mode in amphibians that has evolved independently from the ancestral biphasic life history in at least a dozen anuran lineages. Most direct-developing frogs, including the Puerto Rican coquí, Eleutherodactylus coqui, lack a free-living aquatic larva and instead hatch from terrestrial eggs as miniature adults. Their embryonic development includes the transient formation of many larval-specific features and the formation of adult-specific features that typically form postembryonically—during metamorphosis—in indirect-developing frogs. We found that pre-hatching developmental patterns of thyroid hormone receptors alpha (thra) and beta (thrb) and deiodinases type II (dio2) and type III (dio3) mRNAs in E. coqui limb and tail are conserved relative to those seen during metamorphosis in indirect-developing frogs. Additionally, thra, thrb, and dio2 mRNAs are expressed in the limb before formation of the embryonic thyroid gland. Liquid-chromatography mass-spectrometry revealed that maternally derived thyroid hormone is present throughout early embryogenesis, including stages of digit formation that occur prior to the increase in embryonically produced thyroid hormone. Eleutherodactylus coqui embryos take up much less 3,5,3′-triiodothyronine (T3) from the environment compared with X. tropicalis tadpoles. However, E. coqui tissue explants mount robust and direct gene expression responses to exogenous T3 similar to those seen in metamorphosing species. The presence of key components of the thyroid axis in the limb and the ability of limb tissue to respond to T3 suggest that thyroid hormone-mediated limb development may begin prior to thyroid gland formation. Thyroid hormone-dependent limb development and tail resorption characteristic of metamorphosis in indirect-developing anurans are evolutionarily conserved, but they occur instead in ovo in E. coqui.


INTRODUCTION
Direct development, a distinctive life-history mode in amphibians and other animals, has evolved in anurans multiple times from the ancestral biphasic life history; it characterizes many hundreds of living species (1). Even though directdeveloping frogs typically lack both a free-living aquatic larval stage and a discrete, post-hatching metamorphosis, many species display a cryptic metamorphosis before hatching: adult-specific features, such as limbs, form precociously in the egg, and numerous tadpole-specific features are present initially but then are lost [ Figure 1; (2,3)]. Because such changes in frogs with indirect development are mediated by thyroid hormone (TH), the primary regulator of metamorphosis (4), evolutionary change in thyroid axis function and timing may underlie the numerous heterochronies observed between direct-developing and indirect-developing species (5)(6)(7)(8)(9). Yet, there have been few attempts to precisely delineate the role of this or other pertinent physiological mechanisms.
Embryonic development of direct-developing frogs, as seen in the Puerto Rican coquí, Eleutherodactylus coqui, appears to comprise a mosaic of TH-independent and TH-dependent features. We use the term "embryonic" to describe all in ovo development in E. coqui, although this period encompasses both the initial formation of major organ systems as well as the patterning, morphogenesis and growth that follows. Many of the latter events correspond to metamorphic changes in biphasic anurans. It was once thought that embryonic development in direct-developing species was primarily TH-independent (5). However, subsequent studies with exogenous T 3 and with THsynthesis inhibitors suggested at least a partial role for TH in terminal stages of limb development as well as tail resorption (6,10). In E. coqui, for example, treatment with exogenous T 3 causes precocious tail resorption but has little to no effect on limb elongation (11). Similarly, treatment with methimazole, a THsynthesis inhibitor, inhibits only tail resorption and late stages of limb elongation but does not affect early limb differentiation or digit formation (8). The apparent TH-independence of early stages of limb development is correlated with the fact that limb bud, paddle and digit formation occur prior to formation of the embryonic thyroid gland [ Figure 1; (12,13)]. Thus, limb development in E. coqui comprises two periods: limb bud differentiation and paddle and digit morphogenesis, which precede formation of the thyroid gland and may be TH independent; and limb growth and elongation, which follow thyroid gland formation and are TH dependent. Experiments with TH-synthesis inhibitors, however, can only address the role of TH in the second period. The presumed TH independence of the first period remains to be verified experimentally.
All organs in the body are exposed to roughly the same concentration of circulating TH, primarily in the form of thyroxine (T 4 ) and lower concentrations of 3,5,3 ′triiodothyronine [T 3 ; (14,15)]. Hereafter, we use the term TH to refer to both T 4 and T 3 . However, tissue-specific differences in uptake, metabolism, and action provide for diverse effects of TH in different tissues. Thus, tissue-specific changes in TH metabolism and action likely contribute to the heterochrony of developmental events observed in direct-developing anurans relative to biphasic species. Alternatively, the principal locus of change in hormonal control may involve a shift in the source of THs and when they are present in the embryo. Maternally derived TH is present at early developmental stages of all vertebrates examined so far. In most vertebrates, maternal TH is in the yolk; in most mammals, maternal TH can pass from mother to fetus via the placenta or milk. Yet, the role of maternally derived TH in amphibian embryos is poorly understood (16)(17)(18). If maternally derived THs are present in early embryos of E. coqui, they could influence limb development prior to formation of the embryonic thyroid gland. Finally, three different deiodinase enzymes control cellular metabolism of T 4 in target tissues. In amphibians, two types of deiodinases play major roles during development. Deiodinase enzyme type II (Dio2) converts T 4 into T 3 , which has at least 10 times greater affinity for TH receptors (TRs) than T 4 . Deiodinase type III (dio3) converts T 4 to both T 2 and reverse triiodothyronine (rT 3 ), which are unable to bind TRs in most species. Thyroid hormones act by binding to two TR subtypes, designated alpha (α) and beta (β), to activate or repress transcription of TH target genes. Contrasting expression patterns of TRs and deiodinases may in part underlie the diverse, tissuespecific effects of TH in Xenopus species (19)(20)(21)(22)(23)(24)(25)(26), and it is likely that changes in the temporal or spatial expression of deiodinases or TRs influence TH competence and action in target tissues in E. coqui.
Here we tested the hypothesis that developmental changes in TR and deiodinase mRNAs in developing E. coqui limb and tail, and in whole body TH content are conserved relative to those seen during metamorphosis in indirect-developing frogs. We also investigated whether E. coqui tissues are capable of responding directly to T 3 action by mounting gene regulation responses similar to those seen in metamorphosing species. Taken together, our data support the hypothesis that limb development and tail resorption in E. coqui (8,12) are mediated by conserved components of TH signaling. Additionally, our results suggest that maternal TH could facilitate limb development prior to formation of the embryonic thyroid gland.

Animal Care
Live adult Eleutherodactylus coqui were field-collected from introduced populations in Hilo, Hawaii, with the permission of the U.S. Fish and Wildlife Service (permits EX-14-06, EX-16-07, and EX-17-11). They were brought to Harvard University and maintained as a breeding colony in the Hanken laboratory (IACUC protocol #99-09-03); embryos were obtained following spontaneous matings. Following removal of the overlying chorion with watchmaker forceps in 2% cysteine (pH 8.5) in 10% Holtfreter solution, embryos were reared in 10% Holtfreter solution in Petri dishes at 22.5 • C. Embryos were staged according to the normal table of Townsend & Stewart (TS;, which defines 15 stages from fertilization (1) to hatching (15). Following internal fertilization, the adult female deposits embryos at TS stage 1. (Middle) embryos removed from overlying egg membranes, and a TS stage 9 embryo (Right) stained for type II collagen, which reveals the cartilaginous skeleton and notochord. Scale bars, 1 mm.

Molecular Cloning and Sequence Validation
Partial cDNAs for dio2, dio3, thra, thrb, ribosomal protein L8 (rpL8), thyroid hormone induced bZip protein (thibz), and alpha-actinin 4 (actn4) (Genbank accession numbers MK784754, MK784753, MK784748, MK784749, MK784751, MK784750, MK784755) were isolated by PCR with exact primers ( Table 1) using cDNA generated from RNA isolated from whole TS stage 13 embryos, and the resultant DNA fragments were subcloned into the pCR II plasmid. Exact primers for dio2, dio3, thra, thrb, rpL8, and thibz were designed from predicted full-length cDNA sequences provided by L. Sachs, N. Buisine, and G. Kerdivel (personal communication), while actn4 primers were designed from genomic sequences provided by A. Mudd, R. Harland, and D. Roksahr (personal communication). We also subcloned a partial cDNA for krüppel-like factor 9 (klf9) by degenerate PCR (oligonucleotide primers designed using CODEhop) using the same cDNA described above (Genbank accession number MK784752). The sequences of the subcloned partial cDNA fragments were confirmed by direct DNA sequencing and by comparing them against the full-length cDNAs provided by the investigators listed above.

Whole Body Extraction and Quantification of Iodothyronines Using LC-MS/MS
The iodothyronines T 3 , rT 3 , T 4 , and T 2 were quantified from whole E. coqui embryos throughout development. Because embryos were not dissected from the yolk, all measurements include embryo and yolk TH content. Animals at different stages were anesthetized and snap frozen until extraction and LC-MS/MS analysis. Unfertilized oocytes were dissected from the ovaries of a newly sacrificed female and snap frozen. Between 15 and 20 embryos (∼ 600 mg) were pooled to make one biological replicate. Three or four biological replicates were used for each developmental stage. Tissues were extracted for thyroid hormone analysis as described by Denver (27,28) with the following modifications: stable isotope-labeled T 3 and T 4 ( 13 C 6 T 3 and T 4 , Sigma) were used as an internal standard to correct for differences in extraction efficiency, and solid phase extraction with a Supel-Select SCX cartridge (60 mg 3 mL, Sigma) was used to further purify the extracted tissue. After conditioning the cartridge with 3 mL methanol (HPLC Grade, Sigma) and equilibrating it with 5 mL of 2% formic acid in water (HPLC Grade, Sigma), the sample was loaded, rinsed first with 3 mL 2% formic acid in water and then with 3 mL methanol, and finally eluted with 2 mL of freshly prepared 5% ammonium hydroxide in methanol. It was then evaporated to dryness under nitrogen flow and resuspended in 100 µl of 0.1% formic acid in methanol. Samples were measured at the Harvard Small Molecule Mass Spectrometry facility by using gradient liquid-chromatography mass-spectrometry (LC-MS/MS). Ten microliters of samples were injected on a C18 column (Kinetex 2.6 µm, 100 Å pore size, 150 × 2.1 mm, Phenomenex) in an Agilent 1290 HPLC coupled with an Agilent 6460 Triple Quad Mass Spectrometer. See Supplementary Information for the LC and MS parameters (Supplementary Tables 1, 2). Calibration curves were made in 0.1% formic acid in methanol with pure standards and the same amount of internal standard as the samples. Quantification results with a signal-to-noise (S/N) ratio >10 were used for the statistical analysis. Results with a ratio between 3 and 10 (purple type; Supplementary Table 3) were included in the graph (Figure 2) but not used in the statistical analysis; those with a ratio below 3 were not used (red type; Supplementary Table 3).
We normalized iodothyronine content to the weight of the tissue extracted.

Quantitative PCR
Dechorionated embryos were anesthetized by immersion in 10% Holtfreter solution with drops of 2% neutral-buffered MS-222 added until the embryos no longer responded to toe pinches (between 30 and 60 s). Limbs and tails were dissected and homogenized in TriZol reagent (Invitrogen) and kept at −20 • C until RNA isolation. Total RNA was isolated following the manufacturer's protocol within 3 weeks of homogenization. Because qPCR primers did not span exon-exon boundaries, genomic DNA was removed with an Ambion DNA-free kit (cat. #AM1906). Controls with no reverse-transcriptase verified that removal of genomic DNA was complete. Total RNA was quantified with a Qubit Fluorometer 3.0 and checked for purity on a Nanodrop spectrophotometer. For SYBR Green RTqPCR assays, 200 ng of total RNA was used for input for each reaction.
For probe-based qPCR, 660 ng of total RNA for each sample was synthesized into cDNA with iScript Reverse Transcriptase Supermix for RT-qPCR (BioRad). Complementary DNA was kept at −20 • C until the qPCR assay was performed. mRNA levels were analyzed with either Ssoadvanced Universal Probes Supermix (BioRad) or an iTaq Universal SYBR Green One-Step kit (BioRad) on a CFX384 machine. See Supplementary Data for qPCR cycling conditions. Optimal qPCR conditions were determined with temperature gradient and cDNA dilutions for dynamic range of input. Standard curves showed high efficiency of reaction (90-105%), and R 2 was equal to or >0.98 for all primer sets. No template controls showed no amplification. All oligonucleotides are listed in Table 1. All SYBR and probebased qPCR experiments were done in simplex. The relative mRNA levels were determined as described by Schmittgen and Livak (29). For the developmental expression studies, target-gene expression was normalized to the reference gene rpL8, which did not show significant variation across development [rpL8 mRNA values are given in Supplementary Table 5; see also (8)].
In the in vivo and the tissue explant T 3 response experiments, E. coqui target gene mRNA levels were normalized to the reference genes rpL8 and actn4, which was unaffected by T 3 treatment. Small, statistically insignificant changes in reference gene mRNAs could have led to a small underestimation of the effect of T 3 in these experiments. For Xenopus tropicalis, qPCR primers and probes for thrb, klf9, thibz, dio3, elongation factor 1 alpha (eef1a1) and alpha-actinin 1 (actn1) were designed from publicly available sequences (Genbank accession numbers XM_012964865.2, NM_001113674.1, XM_018092557.1, NM_001113667.2, NM_001016692.2, and NM_001079198.1). For tissue explant experiments, X. tropicalis target gene expression was normalized to eef1a1 and actn1.

Treatment of E. coqui in vivo
Eleutherodactylus coqui embryos were dechorionated into 10% Holtfreter solution at least 24 h prior to immersion in T 3 . One mM stock T 3 in DMSO or 0.01 N NaOH was diluted to make 50 nM T 3 in 10% Holtfreter solution. We chose 50 nM T 3 because it has been shown to induce tail resorption in E. coqui (8), and a 46-h timepoint to allow enough time for induction of T 3 response genes. We chose TS stage 9 embryos because the last third of limb development is TH-dependent (8), but TS stage 9 is still prior to thyroid gland activation. T 3 treatment solutions were refreshed every 8-12 h. After 46 h (n = 12-14 TS-9 embryos), dechorionated embryos were anesthetized as described above and limbs and tails were dissected, from which total RNA was extracted using TriZol reagent.

Measurement of Environmental T 3 Uptake in X. tropicalis and E. coqui
To determine if E. coqui embryos are capable of taking up TH from their surrounding environment, we immersed dechorionated TS stage 9 E. coqui embryos or NF 51-55 X. tropicalis tadpoles in 30 mL (E. coqui) or at least 500 mL (X. tropicalis) 10% Holtfreter solution with either 1 nM (n = 4-6 biological replicates/treatment) or 50 nM (n = 3-4 biological replicates/treatment) stable isotope-labeled T 3 . We chose TS stage 9 E. coqui embryos to match the in vivo T 3 treatment experiments and selected X. tropicalis tadpoles with developing limbs with similar morphology to E. coqui TS stage 9. Approximately twenty E. coqui individuals (600 mg tissue) or two tadpoles were pooled were pooled to make one biological replicate. Tadpoles were either ordered from Xenopus1 (Ann Arbor, Michigan, U.S.A.) or derived from the Hanken lab colony. Stock 100 µg/mL stable isotope-labeled T 3 was diluted to either 1 or 50 nM T 3 . After either 8 or 24 h in 1 nM labeled T 3 solution or 46 h in 50 nM T 3 solution, X. tropicalis tadpoles and E. coqui embryos (with yolk removed) were anesthetized with neutralbuffered 2% MS-222, rinsed three times in PBS and snap frozen until extraction. On average, E. coqui embryos were more densely packed in T 3 solution (5.9 mg tissue per mL media) than X. tropicalis tadpoles (2.0 mg tissue per mL media); however, E. coqui embryos are routinely cultured in these conditions with no ill effects. Tissue was extracted as described above. Because we measured whole body content of stable isotope-labeled T 3 as a proxy for T 3 uptake, we used 25 ng of stable isotope-labeled rT 3 as an internal standard to correct for extraction efficiency.

Tissue Explant Culture and T 3 Treatments
To further investigate if thyroid axis components in the E. coqui limb and tail are functional, we cultured E. coqui and X. tropicalis limb and tail explants (30,31), treated them with T 3 , and assayed gene expression. We treated NF stage 52-54 (32) X. tropicalis tadpoles and TS stage 9 E. coqui embryos with 50 U/mL of penicillin-streptomycin added to aquarium or Petri dish solution for 24 h prior to dissection. Tadpoles and embryos were terminally anesthetized and dipped into 70% ethanol to sterilize the epidermis before dissection. Four X. tropicalis and two E. coqui individuals were pooled to make a single biological replicate of each species. Tissues were dissected into ice-cold 1:1.5-diluted Leibowitz-15 media (Gibco) containing 50 U/mL penicillinstreptomycin, 50 mg/mL gentamicin and 10 mM HEPES. Prior to T 3 treatment tissues were cultured overnight in media supplemented with insulin (500 ug/mL) on a laboratory bench at room temperature (21 • C) with gentle shaking (50 rpm). The next morning, stock T 3 was diluted in 0.01 N NaOH and added to the media to a final concentration of 50 nM. Media and T 3 were changed every 8-12 h. After treatment for 8 or 46 h, limb and tail explants were rinsed three times in phosphate-buffered saline (PBS) and homogenized in TriZol. RNA was isolated according to the manufacturer's protocol.

Statistical Analysis
Statistical analyses of qPCR data were done with RStudio version 1.0.136 and visualized with ggplot2 (https://ggplot2.tidyverse.org/). Developmental timeline qPCR and iodothyronine content data followed a non-normal distribution as determined by Q-Q plots and the Shapiro-Wilk test; Levene's test determined that TH content data additionally had unequal variance. Log 10 -transformed data were not normally distributed. Therefore, a Kruskal-Wallis test was used to determine if there were significant differences among groups, and a post-hoc Dunn's test with the Benjamini and Hochberg (BH) correction was used to identify stages that differ from each other while adjusting for multiple comparisons. We performed a least squares regression on T 4 , T 3 , and rT 3 data sets to investigate possible differences in iodothyronines kinetics during development. For the developmental timeline qPCR data, statistical tests were performed on data pooled from two independent experiments (see Supplementary Data for data from each experiment). For in vivo and in vitro T 3 treatment experiments, Student's t-test was used to identify significant differences between T 3 -treated groups and controls.

Predicted Proteins of Isolated E. coqui cDNAs Contain Conserved Domains
Most isolated cDNAs contained functional domains of orthologous proteins. The predicted E. coqui TRα and TRβ sequences cover amino acids 11-281 (65%), and amino acids 9-273 (69%) of the orthologous X. tropicalis proteins, respectively. Both predicted TR protein sequences contain the DNA-binding domain and most of the ligand-binding domain. Alignments show that the predicted protein sequence of the E. coqui TRα DNA-binding domain has 97% identity to the X. tropicalis DNA-binding domain, while the TRα ligand-binding domain shared between the predicted E. coqui and X. tropicalis sequences are 98% identical. The DNA-binding domain of the predicted E. coqui TRβ sequence is 100% identical to the DNA-binding domain in X. tropicalis TRβ, and the ligand-binding domain is 95% identical. The predicted partial E. coqui Dio2 sequence covers amino acids 2-254 (98%) of X. tropicalis Dio2 and the partial E. coqui Dio3 sequence covers amino acids 7-252 (90%) of X. tropicalis Dio3. Additionally, the predicted protein sequence of both dio2 and dio3 isolated cDNAs contain the selenocysteine site and the thioredoxin domain. Both thioredoxin domains share 86% identity with the orthologous X. tropicalis thioredoxin domain. The partial predicted amino acid sequence of E. coqui Klf9 covers amino acids 194-264 (25%) of X. tropicalis Klf9 and contains the three characteristic zinc-finger domains (100% identity) in the C-terminus of X. tropicalis Klf9. The isolated E. coqui thibz sequence covers amino acids 159-335 (53%) of X. tropicalis NFIL3-like (synonym for thbzip) and lacks the highly conserved basic leucine zipper domain. Even without the highly conserved basic leucine zipper domain, the predicted E. coqui protein sequence still clusters with other orthologous NFIL3-like proteins, rather than with other proteins with the basic leucine zipper domain (NFIL3 and CREB1) in maximum likelihood trees of these three orthologous vertebrate proteins (data not shown). Similarly, the other partial predicted E. coqui sequences cluster with other orthologous genes rather than with other closely related proteins containing similar domains (data not shown). We confirmed all isolated E. coqui cDNAs against the full-length transcript provided by investigators listed in the methods. Finally, we also performed BLASTx and BLASTn searches with the isolated E. coqui cDNA sequences. All cloned sequences have high similarity to predicted orthologous genes in frog species and other vertebrates (Supplementary Table 4).

Changes in Whole Body Iodothyronine Content During Embryonic E. coqui Development
Using LC-MS/MS, we detected the iodothyronines T 4 , T 3 , and rT 3 in unfertilized oocytes and at every stage of development (Figure 2). Thyroxine content (pg/mg body weight) was highest, followed by rT 3 and then T 3 . We detected T 2 only at TS stages 14 and 15, when hatching occurs, and at this point, T 2 content was less than all other iodothyronine content and ranged between 0.04 and 0.78 pg/mg body weight (Supplementary Table 3). The three quantifiable iodothyronines were low and relatively constant up to TS stage 8, after which stage they showed statistically significant increases [T 3 : Kruskal-Wallis rank sum test, X 2 = 43.2 (df = 15), p < 0.001; T 4 : Kruskal-Wallis rank sum test, X 2 = 43.7 (df = 15), p < 0.001; rT 3 : Kruskal-Wallis rank sum test, X 2 = 39.7 (df = 14), p < 0.001]. Whole body content of all three iodothyronines showed statistically significant increases between stages 8 and 13 (post-hoc Dunn's test; p = 0.048, 0.033, 0.035 for T 3 , T 4 , and rT 3 , respectively). The velocity of change was slower for rT 3  Tissue content of all three iodothyronines was highest at TS 15. Note also that oocytes and early embryos (TS 0-5) of E. coqui have large yolk deposits, which may increase the S/N ratio and cause an underestimation of iodothyronine content in the embryo and yolk at these stages.

Changes in Thyroid Hormone Receptor and Deiodinase mRNA Levels in the Embryonic Tail
Both thra and thrb mRNAs in the E. coqui tail showed statistically significant changes during development [ Figure 3A; thra: Kruskal-Wallis rank sum test, X 2 = 20.18 (df = 4), p < 0.001; thrb: Kruskal-Wallis rank sum test, X 2 = 26.78 (df = 4), p < 0.001]. Thyroid hormone receptor α and thrb mRNA in the tail bud are approximately equal at TS stage 5 ( Figure 3A). Thyroid hormone receptor α mRNA in the tail at hatching is between 2.1and 4-fold higher than the early tail (TS stages 5 and 7, post-hoc Dunn's test, p = 0.002 and 0.03, respectively). Thyroid hormone receptor β mRNA follows a similar pattern-it increased 4-fold between the onset of tail resorption (TS 13) and hatching (TS 15)-although thra increased only 1.8-fold over the same interval ( Figure 3A). Thyroid hormone receptor β mRNA at hatching (TS 15) is between 18-and 47-fold higher than in the early tail (TS stage 5 and TS stage 7, post-hoc Dunn's test, p < 0.001 and p = 0.002, respectively).

Changes in Thyroid Hormone Receptor and Deiodinase mRNA Levels in the Embryonic Hind Limb
Both thra and thrb mRNAs in the E. coqui hind limb showed statistically significant changes during development (Figure 4A; thra: Kruskal-Wallis rank sum test, X 2 = 20.66 (df = 4), p < 0.001; thrb: Kruskal-Wallis rank sum test, X 2 = 25.36 (df = 4), p < 0.001). The level of thra mRNA was greater than thrb mRNA in the limb bud until TS 10, when the thra mRNA level began to decrease and continued to decline through hatching ( Figure 4A). The peak thra mRNA level at TS 10 coincides with the appearance of thyroid follicles (13); thra mRNA in the hind limb at this stage was significantly higher than in the limb bud at TS 5 (post-hoc Dunn's test, p = 0.001), in the limb paddle at TS stage 7 (p = 0.009) and in the fully formed froglet limb at TS 15 (post-hoc Dunn's test, p = 0.002). At hatching, thra mRNA level was lower than thrb mRNA levels. Between paddle (TS 7) and toepad formation (TS 13), thrb mRNA rose ∼21-fold to a peak at TS 13. At TS 13, thrb expression was significantly higher than in the limb bud and paddle (Figure 4A; TS 5 and 7; posthoc Dunn's test, p < 0.001 and p = 0.001, respectively). Thyroid hormone receptor β mRNA drops almost 1.5-fold between TS 13 and hatching.

Exogenous T 3 Induced Gene Expression Responses in the TS 9 E. coqui Tail, but Not the Limb
To determine if E. coqui tissues are capable of mounting a gene regulation response to exogenous T 3 , we performed in vivo T 3 treatments (Figure 5A). Immersion of TS 9 E. coqui embryos in 50 nM T 3 for 8 h caused a significant induction of klf9 (Student's t-test, t = 5.61 (df = 21.74), p < 0.001) and thibz (Student's t-test, t = 6.20 (df = 12.42), p < 0.001) in the tail. Immersion in 50 nM T 3 for 46 h additionally significantly induced thrb mRNA (Supplementary Figure 4). In contrast, the identical treatment significantly increased only thibz expression ( Figure 5B; Student's t-test, t = 3.11 (df = 18.92), p = 0.006) in the limb.

E. coqui Embryos Took up Significantly Less T 3 From the Environment Than Did X. tropicalis Tadpoles
Because previous studies suggested that E. coqui limbs are insensitive to TH, and because we observed a weak TH response in our in vivo experiments, we wanted to confirm that immersion in T 3 increased tissue content of T 3 . We quantified stable isotopelabeled T 3 tissue content after immersing X. tropicalis tadpoles or E. coqui embryos in stable isotope-labeled T 3 solution under three conditions. We chose 50 nM T 3 and 46 h treatment to match the E. coqui in vivo T 3 response experiments. We also chose two conditions that represent relevant time points from previous studies of larval Xenopus species: (1) treatment with 1 nM T 3 for 8 h is sufficient for X. tropicalis' whole body T 3 content to surpass the T 3 concentration in the surrounding media (33), and (2) treatment with 1 nM T 3 for 24 h is sufficient to induce gene expression responses in X. tropicalis (31,34). After immersing E. coqui in 1 nM labeled T 3 for 8 and 24 h, we detected endogenous T 3 but not labeled T 3 . However, we detected labeled T 3 in X. tropicalis tissue at both 8 and 24 h ( Table 2). We detected stable isotope-labeled T 3 in both E. coqui and X. tropicalis tissue following 46-h treatment with 50 nM T 3 . Total content of labeled T 3 in X. tropicalis tissue was ∼63 times that found in E. coqui tissues [ Table 2, Student's t-test, t = −3.20 (df = 2.00), p = 0.085]. Additionally, X. tropicalis has ∼ 875 times more stable isotope-labeled T 3 than endogenous T 3 content. In contrast, stable isotope-labeled T 3 in E. coqui is approximately equal to endogenous T 3 content.  Each value represents the mean of 3-6 individuals ± standard error.  Figure 6D). In both species, the magnitude of increase for all genes was greater in the limb than in the tail ( Table 3). The same trends were observed after treating tissue explants with 50 nM T 3 for 46 h (Supplementary Figure 5). In tail explants, T 3 induced fold changes of a similar order of magnitude for thibz (between 42and 44-fold) and thrb (between 1.8-and 3.8-fold), but not for dio3 and klf9; in E. coqui, dio3, and klf9 mRNAs increased 2-and 9.9fold, respectively, while dio3 and klf9 mRNAs increased 11.9-and 12.5-fold in X. tropicalis. In limb explants, dio3 and thibz mRNA differed by an order of magnitude between species. Deiodinase type III mRNA increased 58.8-fold in X. tropicalis limb explants, while dio3 mRNA increased 3.7-fold in E. coqui limb tissue. Thyroid hormone induced bZip protein mRNA increased only 37fold in X. tropicalis limb explants, while dio3 mRNA increased 180-fold in E. coqui limb explants. Values represent the average fold increase above control (vehicle-treated) levels.

DISCUSSION
In this study we show that the core TH signaling components are evolutionarily conserved in Eleutherodactylus coqui limb and tail tissue. We also show that developmental patterns of thra, thrb, dio2, and dio3 mRNAs, and whole-body TH content in E. coqui closely match those reported during metamorphosis of Xenopus species. We also find maternal T 4 , T 3 , and rT 3 in unfertilized eggs and early embryos of E. coqui, which may mediate TR signaling prior to embryonic thyroid gland formation. This is the first published report of TH metabolites and maternally derived TH in a direct-developing frog. Additionally, we demonstrate that E. coqui tissues show robust gene expression responses to exogenous T 3 similar to those seen in metamorphosing species. Eleutherodactylus coqui embryos take up much less T 3 from the environment compared with X. tropicalis. This difference likely explains the relatively weak and variable gene expression responses seen in vivo in E. coqui, and was likely a significant confounding factor for previously published results.

Developmental Profiles of Whole Body Iodothyronine Content
Temporal dynamics of whole-body iodothyronine content in direct-developing E. coqui mirror those described for indirect-developing frogs, which retain the ancestral biphasic life history: Scaphiopus hammondii (28), Rana catesbeiana (35), Bufo marinus (36), Bufo japonicus (37), and Xenopus laevis (33). Anuran metamorphosis comprises three successive stages: premetamorphosis, when little to no TH is present; prometamorphosis, when TH concentrations slowly rise; and a rapid metamorphic climax characterized by a peak in TH concentrations. The temporal profile of TH content in embryonic E. coqui similarly defines three successive periods: (1) Low TH content characterizes the first half of development, prior to thyroid follicle formation (TS 1-8).
(2) After thyroid follicles appear, TH content gradually rises until tail resorption began (TS 9-12). (3) TH content dramatically increases, with a peak in TH at or just prior to hatching (TS [13][14][15]. In addition to amphibians, many other vertebrates experience peak concentrations of TH at life history transitions-at hatching in precocial birds (38), at the larval-to-juvenile transition in several fish species (39)(40)(41), at ∼14 days post-partum in rats and mice (42,43), and at birth in humans (44). Thyroid hormones are present throughout early embryogenesis and the subsequent period of pre-hatching development in E. coqui (TS 1-9), beginning up to eight days before thyroid follicles can be detected histologically (13). These hormones are almost certainly maternal in origin. Similarly, T 4 and T 3 have been detected in yolk and gastrulating embryos of four other anuran species-Bufo marinus (36), Rana catesbeiana (35), Bombina orientalis (45), and Xenopus laevis (16). Early Xenopus tropicalis embryos express key TH signaling components (46). Indeed, TH signaling is also functional in the Xenopus tadpole central nervous system (CNS) before thyroid gland formation (16,18). Maternally derived TH has a conserved role in vertebrate CNS development (47) and embryogenesis (17,48,49). Therefore, it seems likely that direct-developing frogs require maternal TH for normal neural development, as do most vertebrate species, although we do not evaluate that hypothesis here.
Maternal TH may regulate limb development occurring before the differentiation of the embryonic thyroid gland in directdeveloping frogs. In metamorphosing anurans, TH signaling is required for terminal limb differentiation (22), but the initial stages of limb development are TH-independent. For example, tadpoles immersed in methimazole, a TH-synthesis inhibitor, develop a long limb-bud-like structure (24), and thyroidectomized tadpoles develop calcification centers in the hind limb (50,51). In E. coqui, the limb bud proliferates and digits develop prior to the appearance of embryonic thyroid follicles (TS stages 9-10) [ Figure 3; (8,13)]. Two hypotheses could account for this observation: (1) E. coqui relies on maternal TH, rather than embryonically produced TH, to regulate early stages of digit patterning and growth (TS 6-9); or (2) paddle and digit formation in E. coqui proceed independently of TH. Our data show that requisite components of TH signaling are present at this time. Future investigation should evaluate the functional role of TH during this critical developmental period. A switch from embryonic to maternally synthesized TH for the regulation of early limb development, if it occurred, could explain the heterochronic shift in limb development and would represent an evolutionary novelty in direct-developing species.
Thyroid Hormone Receptor α, thrb, Dio2, and Dio3 mRNA Expression Patterns During Development and T3 Response in the Embryonic Tail Tail resorption in Xenopus tropicalis occurs late in metamorphosis and is mediated by TRβ (52). Because tail resorption in E. coqui occurs late in embryogenesis and requires T 3 (8), we expected that thra, thrb, dio2, and dio3 mRNA dynamics in the E. coqui tail would mirror those described in Xenopus. Our results support this hypothesis: in the E. coqui tail, a rise in thrb expression coincides with the rise in embryonic TH content, consistent with a role for thrb in mediating tail resorption.
Deiodinase type II and dio3 mRNA expression patterns in the developing E. coqui tail are also similar to those described in indirect-developing species in which these deiodinase enzymes are critical for coordinating metamorphosis (20). Elevated dio3 expression protects the tail from an early apoptotic response to T 3 until metamorphic climax in Xenopus (26); E. coqui tail resorption also begins at TS 13, when dio3 expression significantly decreases. Although they serve different functions, the tail serves a critical role in both species: the larval Xenopus tail is a critical locomotor organ, whereas the embryonic E. coqui tail functions in respiration. In both species, maintenance of the tail is accomplished in part by dio3 inactivation of T 4 and T 3 .
Given the conservation of mRNA dynamics in the E. coqui tail, we wanted to determine whether the tissue could respond to exogenous T 3 . In Xenopus species, treatment with exogenous T 3 induces transcription of direct T 3 response genes dio3, klf9, thibz, and thrb (19,(53)(54)(55)(56). Exogenous T 3 induces significant increases in the mRNA of three of these T 3 response genes, klf9, thibz, and thrb, supporting the hypothesis that TH signaling components are conserved and mediate tail resorption in E. coqui.
Thyroid Hormone Receptor α, thrb, Dio2, and Dio3 mRNA Expression Patterns During Development and T 3 Response in the Embryonic Hind Limb Thyroid hormone receptor α, thrb, dio2, and dio3 mRNA expression patterns parallel those described in Xenopus species in the period leading up to and during metamorphosis (33,57). In indirect-developing frogs, TRα has a critical role in controlling post-embryonic developmental timing (58)(59)(60) and in promoting proliferation in the hind limb during metamorphosis (61)(62)(63). Constitutive thra expression supports a proliferative and competence-establishing role for TRα in E. coqui. In the E. coqui limb, a rise in thrb expression coincides with the rise in embryonic TH content, consistent with TRβ autoinduction and tissue sensitization to TH described in Xenopus (64).
The tissue-specific patterns of dio2 and dio3 underlie the differential sensitivity of limb and tail tissue in metamorphosing frogs. Deiodinase type II expression is constitutive in the developing limb of Xenopus laevis, causing the limb to be sensitive to small amounts of T 3 produced during premetamorphosis (23). Similarly, elevated dio2 expression in E. coqui limbs throughout most of limb development, including several days prior to formation of the embryonic thyroid gland, supports a role for TH-mediated limb development and growth.
In indirect-developing species, including Xenopus and spadefoot toads (Scaphiopus), concentrations between 1 and 10 nM T 3 are sufficient to promote precocious metamorphosis, tail resorption, and gene expression responses in limbs and tail (31,65,66). However, previous studies report that the E. coqui limb has no morphological response to high doses of exogenous T 3 (11). Our study is the first to characterize mRNA expression changes in a direct-developing frog species in response to exogenous T 3 . Treatment of E. coqui embryos with exogenous T 3 prior to formation of the thyroid follicles increases expression of four direct T 3 response genes in the tail, consistent with studies in Xenopus species (19,(53)(54)(55). However, limbs of the same embryos do not respond to T 3 , despite the high dose administered (50 nM T 3 ). The lack of response previously observed in direct-developing species may be confounded by an inability of T 3 to reach the limb tissue. We observe a weak induction of T 3 response genes in TS stage 7 limbs, a full two days before E. coqui begins to produce TH (Supplementary Figure 3). It is possible that this response occurs because the adult epidermis is not yet fully formed and T 3 is better able to penetrate into the tissue, or because there is less endogenous T 3 present at TS 7 than at TS 9. In either case, the ability to respond to T 3 prior to thyroid gland formation is similar to biphasic species; tadpoles are also TH competent as soon as they hatch. Finally, the similar robust gene regulation response induced in E. coqui and X. tropicalis limb explants suggests that the limb tissue itself is similarly competent in both species. Overall, these data support the hypothesis that TH plays a role in E. coqui limb development and may do so prior to formation of the embryonic thyroid gland.
Here we support previous claims that later stages of limb development in E. coqui are TH-dependent but we additionally show that TH-signaling components are present during earlier stages, and that E. coqui limb tissue is sensitive to T 3 . Eleutherodactylus coqui eggs are provisioned with maternally derived TH, which may mediate organogenesis before differentiation and activity of the embryo's own thyroid gland. Altogether, our data suggest that the TH-mediated molecular module active during post-hatching metamorphosis in indirectdeveloping frogs has been shifted prior to hatching in directdeveloping species.

ETHICS STATEMENT
This study was carried out in accordance with the recommendations of the Harvard Faculty of Arts and Sciences Institutional Animal Care and Use Committee. The protocol was approved by the Harvard Faculty of Arts and Sciences Institutional Animal Care and Use Committee.

AUTHOR CONTRIBUTIONS
ML designed experiments and performed experiments, interpreted data, and wrote the manuscript. RD and JH contributed to experimental design, edited the manuscript, and discussed data interpretation.