Expression of the inactivating deiodinase, Deiodinase 3, in the pre-metamorphic tadpole retina

Thyroid hormone (TH) orchestrates amphibian metamorphosis. Thus, this developmental phase is often used to study TH-dependent responses in specific tissues. However, TH signaling appears early in development raising the question of the control of TH availability in specific cell types prior to metamorphosis. TH availability is under strict temporal and tissue-specific control by deiodinases. We examined the expression of the TH-inactivating enzyme, deiodinase type 3 (D3), during early retinal development. To this end we created a Xenopus laevis transgenic line expressing GFP from the Xenopus dio3 promoter region (pdio3) and followed pdio3–GFP expression in pre-metamorphic tadpoles. To validate retinal GFP expression in the transgenic line as a function of dio3 promoter activity, we used in situ hybridization to compare endogenous dio3 expression to reporter-driven GFP activity. Retinal expression of dio3 increased during pre-metamorphosis through stages NF41, 45 and 48. Both sets of results show dio3 to have cell-specific, dynamic expression in the pre-metamorphic retina. At stage NF48, dio3 expression co-localised with markers for photoreceptors, rods, Opsin-S cones and bipolar neurons. In contrast, in post-metamorphic juveniles dio3 expression was reduced and spatially confined to certain photoreceptors and amacrine cells. We compared dio3 expression at stages NF41 and NF48 with TH-dependent transcriptional responses using another transgenic reporter line: THbZIP-GFP and by analyzing the expression of T3-regulated genes in distinct TH availability contexts. At stage NF48, the majority of retinal cells expressing dio3 were negative for T3 signaling. Notably, most ganglion cells were virtually both dio3-free and T3-responsive. The results show that dio3 can reduce TH availability at the cellular scale. Further, a reduction in dio3 expression can trigger fine-tuned T3 action in cell-type specific maturation at the right time, as exemplified here in photoreceptor survival in the pre-metamorphic retina.


Introduction
To address the question of TH availability in specific retinal cell types, we used a number of complementary experimental approaches. First, we compared expression of dio3 using both in situ hybridization (ISH) and exploiting a pdio3-GFP construct, where GFP expression is controlled by part of the Xenopus dio3 promoter. Second, we compared dio3 ISH expression with immunohistochemistry for specific retinal cell types. Third, we determined which of these cells types responded to T 3 signaling by using a well-established TH/bZIP-GFP reporter system [22,23,24]. Taken together, the results show that dio3 expression could limit T 3 -dependent responses in specific retinal cells at specific developmental stages.

Transgenic and wild type animals
X. laevis tadpoles and juveniles were obtained by in-house breeding. To obtain eggs, adult frogs were injected with human chorionic gonadotropin (Chorulon) (400 U/female and 200 U/male). Some adult frogs were bought to the CRB (Xenope Biology Resources Centre, Centre de Ressource Biologie Xénope), France (University of Rennes1; http://xenopus.univ-rennes1.fr/).Animals were reared under a 12-hour light/12-hour dark cycle at 22-23˚C. Animals were staged according to Nieuwkoop and Faber (NF) (1956). Juveniles were used when they had developed beyond NF66 and weighed between 1 and 2 g.
To determine T 3 responsiveness in the retina, we used the X. laevis transgenic TH/bZIP reporter line [22,23,24]. The transgenic TH/bZIP reporter line contains a series of T 3 sensitive response elements upstream of the GFP coding sequence providing a direct T 3 readout. F1 founders were crossed together to produce F2 homozygous transgenic tadpoles.
All animal studies were carried out in accordance with the European Union regulations concerning the protection of experimental animals and approved by the Museum National d'Histoire Naturelle Animal Care and Use Committee, Paris, France. All procedures were approved by the institutional Ethics Committee (Animal Housing Agreement Number: C75-05-01-2, Committee Approval 68.031).

Generation of the Xenopus laevis deiodinase 3-GFP transgenic line
The X. laevis transgenic line deiodinase 3-GFP (pdio3-GFP) reporter was obtained by sperm nuclei-mediated REMI method transgenesis [25] using a dio3 promoter driven-GFP expressing construct.Injections were carried out so that the transgene was inserted into the genome prior to the first cleavage. The DNA plasmid construct used for this reporter line corresponds to a clone representing a part of the Xenopus tropicalis dio3 promoter locus and upstream regulatory elements. The DNA clone represents 1448 pb of the full Xt dio3 4915 pb promoter locus. The cDNA clone was from 454 pb of the 3'end Xt dio3 promoter locus. The PCR product was initially obtained with the ExTaq TaKaRa kit (Ozyme, ref: TAKRR013A) from X. tropicalis tadpole tail genomic DNA extract. PCR primers were designed against the promoter dio3 locus. Forward and reverse primers were 5'CGGGGAAGATATGTGAAGGA3' and 5'GGGCTCCCAGGATGATCTGA 3', respectively. The latest Xenopus tropicalis genome annotation, confirms the position of the clone in the promoter. This PCR product was sub-cloned into a pGlow-GFP vector (Invitrogen pGlowTOPO cloning kit ref: 45-0021) and verified by sequencing. F0 or F1 transgenic founders were crossed with wild type adult to produce F1 or F2 transgenic tadpoles that were used for the characterization experiments.

Sample collection and preparation
Tadpoles and juveniles of both transgenic reporter lines were deeply anesthetized by submersion in 0.1% tricaïne methanesulfonate anesthesia (MS-222, Sigma-Aldrich) and euthanized by decapitation. Whole heads of NF48 tadpoles and juveniles were fixed overnight at 4˚C in 4% paraformaldehyde (PFA, Sigma-Aldrich, Saint Quentin Fallavier, France) in phosphate buffered saline (PBS pH 7.4). Samples were briefly washed three times (15 min.) in PBS and placed overnight in 15% sucrose PBS. Samples were embedded in Tissue-Tek and stored at -80˚C. Coronal cryosections (16 micrometers) were performed on whole heads and analyzed on eyes. Sections were stored at -80˚C prior to immunohistochemistry or in situ hybridization.
The immunostained retinas were visualized with a LEICA DM 5500 B microscope equipped with a LEICA CTR 5500 lens and for epifluorescence with a PRIOR Lumen 200 system.Fluorescent image acquisitions were carried out using a ZEISS LSM 710 system confocal microscope by channels or spectral mode acquisitions and the Zen 2011 software acquisition at the J. Monod Institute (ImagoSeine platform, Jussieu University, Paris, France).

In situ Hybidization (ISH)
Probes for dio3 cDNA were isolated by RT-PCR using a pool of RNA extracted from embryos and tadpoles at several stages.The PCR fragment for pdio3 (296 bp, forward: TCGGTGCACAA TAGTCGGG and reverse:CTTCTGCCCGTGCCACAC) were cloned using the TOPO TA cloning kit dual promoter (Invitrogen, Carlsbad, CA, USA) and sequenced to check orientation. The dio3 mRNA probe was synthesized using T7 enzyme 25 (Roche, Basel, Switzerland) and labeled with digoxigenin (Roche, Basel, Switzerland). ISH was performed on cryosections. Slides were briefly rehydrated with PBS (pH 7.4) then post-fixed with 4% PFA during 10 min. Then slides were washed with PBS Tween 20 (0.1%). Tissues were permeated with proteinase k (5.10 −6 g/ml, Sigma, ref. P4850) for two minutes. Slides were then post-fixed with 4% PFA during 10 min follow by wash with PBS Tween 20 (0.1%). To prevent RNA interaction with protein, slides were incubated with 100 mM Triethanolamine/0.25% anhydride acid acetic during 10 min. Slides were washed with PBS Tween 20 (0.1%) and then pre-hydrated with hybridization buffer (formamide 50%, SSC 5X, ARNt grade VI à 1 mg/ml, heparin 100.10 −6 g/ml, Denhart's 1X, Tween 20 (0.1%), CHAPS 10 mM 0.1%, EDTA 10 mM) during 1 hour. After denaturation dio3 probe was used at 1 ng/μl concentration and incubated on slides overnight at 60˚C. Slides were washed several times with SSC buffer (saline sodium chloride citrate Sigma, ref.93017) at 65˚C and then blocked with 10% NGS conjugated to alkaline phosphatase (1/2000, Roche) in 5% NGS/PBS/ Tween 20 (0.1%). Slides were washed several times in maleic acid buffer. Signals were revealed using BM purple (Roche, ref.11442074001). Reactions were stopped by washes with PBS Tween 20 (0.1%). Slides were mounted with Prolong (Invitrogen, Carlsbad, CA, USA).ISH on retina sections was visualized using a LEICA DM 5500 B microscope equipped for visible light with LEICA CTR 5500. Image acquisition exploited a DFC 450 C Camera and the Leica Application suite LAS version 4.1.0 software acquisition.

RNA isolation and reverse transcription (RT)
RNA extraction was performed by RNA isolation from micro-scale, using RNAqueous Micro kit (Invitrogen AM 1931). For mRNA extraction from eyes, two steel balls were used for each sample in 100 microliters of lysis solution. Eyes were homogenized by using a tissue lyser at 30 Hz twice during 1 min. Ethanol 100% (50 microliters) was added and the lysate was placed on a micro-filter cartridge before centrifugation (16,400 x g, 20 sec.). The cartridge was rinsed three times with a wash solution before RNA elution. A warm (75˚C) RNA elution solution (18 microliters) was used before centrifugation (twice 16,400 x g, 30 sec. at room temperature). A DNAse was used (22 min. at 37˚C) and an inactivation reagent was added before centrifugation (16,400 x g, 2 min.). Concentrations of RNA were determined by using a NanoDrop (ThermoScientific, Rockford, IL). RNAs were stored in TRIS 10 mM /EDTA 0.1 mM (PH 7.4) at -80˚C.
Extracted total RNA (300 ng) was used for reverse transcription (RT) using a High Capacity cDNA RT kit (Applied Biosystems, Foster City, CA) with the addition of RNase inhibitor. A control for genomic DNA (RT-reaction: all reagents and RNA except reverse transcriptase) was performed for each group (Ctrl, T 3 , IOP, IOP+T 3 ).

Real-time PCR quantification
The amount of each RNA transcript was estimated by relative quantitative real-time PCR (qRT-PCR) using Power SYBR green master mix and a Quant Studio Flex 6 (Applied Biosystems). 5 to 8 biological replicates were performed for each group. A 1:10 dilution of each cDNA was run in triplicate on a 384-well plate for each primer pair (Table 1) (intra assay variability) by using thermal cycling parameters: 95˚C for 10 min, 95˚C for 15 sec and 60˚C for 1 min (40 cycles) and an additional step for dissociation curves was performed for all plates. Results were normalized with the expression of reference gene odc. DDCT method was used to estimate fold change of expression when compared to the untreated (CTL) group. PCR primer sequences are designed previously [26] for the housekeeping gene odc. Primers list is provided in Table 1.

Statistical analysis
For multiple comparison analysis, Non-parametric ANOVA was performed, followed by a Kruskall Wallis test (PRISM7). Heatmap was performed on Fold Changes (PRISM7), and correlation analysis was performed on XLSTAT.

Dynamic expression of retinal dio3 in the pre-metamorphic Xenopus laevis retina
To address the role of dio3 in pre-metamorphic retina, we cloned part of the X. tropicalis dio3 promoter and generated a transgenic pdio3-GFP reporter line. GFP immunostaining was used to follow pdio3-reporter GFP expression. To validate the dio3 transgenic GFP reporter signal in the retina, we compared it to the endogenous signal obtained with ISH using antisense probes against dio3 (Fig 1A-1F). Expression of dio3 (Fig 1A and 1B) and pdio3-reporter GFP Endogenous dio3 in situ hybridization on retina coronal sections from animals at NF45 (Fig 1A), NF48 (Fig 1B), NF66 (Fig 1C) stages. Fig 1D-1F: GFP immuno-labeling on retina coronal sections at NF45 (Fig 1D), NF48 (Fig 1E, 1H and 1I), NF66 (Fig 1F) stages. Comparison of GFP expression between tadpoles from the pdio3 transgenic line (Fig 1E) and from TH/bZIP transgenic T 3 reporter tadpoles at NF48 stage ( Fig 1G). Orientation is with lens on the left and retina on the right. Photoreceptor cell bodies (arrow head: Fig 1A-1F and 1I). Amacrine cells (arrow: Fig 1C, 1F, 1E and 1H). Bipolar neurons (magenta arrow: Fig 1E and 1I). Ganglion cells (asterisk: Fig 1E, 1H and 1G). Scale bars: 50 microns. https://doi.org/10.1371/journal.pone.0195374.g001 ( Fig 1D, 1E, 1H and 1I) were examined by ISH on head cryo-sections in tadpoles at stages NF45 and NF48 and in the froglet at stage NF66 (Fig 1C and 1F). At stage NF45, both the dio3 ISH signal and GFP expression were limited to a very few ganglion cells and in a subset of the photoreceptors of the outer nuclear layer, as judged by cell morphology and anatomical localization (Fig 1A and 1D). At stage NF48, both the dio3 ISH signal and GFP expression showed higher intensity and a wider distribution, being present in all retinal cell types: ganglion cells, inner nuclear layer cells (bipolar neurons, horizontal neurons and amacrine cells) and in the outer nuclear layer cells (basal part of the photoreceptors) (Fig 1B, 1E, 1H and 1I). Cell morphologies were determined according to Wong et al. [27]. At stage NF66 corresponding to a post-metamorphic juvenile stage, dio3 expression was more restricted and localized in the outer nuclear layer and in the inner nuclear layer. More specifically, both ISH and GFP immunocytochemistry showed dio3 expression in certain photoreceptor types and in Amacrine cells (Fig 1C and 1F).

D3 expression generally corresponds to lack of T 3 responsiveness
To determine which retinal cells express D3 and those that respond to T 3 , we compared the GFP expression in F2 tadpoles for two transgenic reporter lines, respectively pdio3-GFP and TH/bZIP-GFP. Retinal GFP immuno-labeling was examined on 16 micrometers coronal whole head cryo-sections at stage NF 41-42 (Fig 2) and NF48 (Fig 1E, 1H, 1I and 1G) (see Materials and Methods).
At stage NF41-42 D3 activity, as judged by GFP expression in pdio3-GFP tadpoles, is observed in numerous photoreceptors of the outer nuclear layer (yellow arrow heads, Fig 2A, 2C, 2E and 2F and Table 2) and in some bipolar neurons (blue arrow heads, Fig 2A, 2C, 2E and 2F and Table 2), whereas no T 3 -response is detected in photoreceptors nor in bipolar neurons of the inner nuclear layer (INL) (Fig 2G, 2I, 2J and 2L and Table 2). Furthermore, the photoreceptors that are GFP positive in pdio3-GFP tadpoles are in the dorso-central area (Fig 2A).
The horizontal neurons (magenta arrow heads) of the outer plexiform layer that are GABA positive cells (Fig 2B and 2H) are also GFP positive in pdio3-GFP (Fig 2C and 2F and Table 2) and in TH/bZIP reporter tadpoles (Fig 2I, Table 2). Similarly, numerous amacrine cells (white arrow heads) in the inner nuclear layer, that are GABA/PARV positive cells (Fig 2B, 2D, 2H and 2K) are GFP positive in pdio3-GFP (Fig 2C, 2E and 2F and Table 2) and in TH/bZIP-GFP lines (Fig 2I and 2L and Table 2). However, they are less numerous in the TH/bZIP reporters. A ganglion cells subset (white asterisk) is Parvalbumine PARV positive (Fig 2D) in the ganglion cell layer and some of them are also GFP positive, in pdio3-GFP tadpoles (Fig 2E and 2F and Table 2). In TH/bZIP-GFP tadpoles, GFP expression is seen in a large subset of ganglion cells (Fig 2K and 2L and Table 2).
At stage NF48, widespread GFP expression was observed in pdio3 tadpoles (Fig 1, 1E, 1H and 1I), a fact that was reflected by limited T 3 responses (lower TH/bZIP-driven GFP expression) notably in retinal ganglion cells, noted as " Ã " in Fig 1 and horizontal neurons (outer plexiform layer), noted as "+" (Fig 1G, Table 2). The pdio3-GFP expression displayed a strong dorso-ventral gradient (Fig 1E), which contrasted with the higher levels of TH/bZIP-driven GFP in the ventral area, mostly for the outer plexiform layer (Fig 1G). Thus, higher D3 levels appear to restrict T 3 -responsiveness in the dorsal retina at this stage.
Furthermore, at this stage, no TH/bZIP-driven GFP signal was seen in any cell type in the inner nuclear layer as in bipolar neurons or amacrine cells (Fig 1G, Table 2), which corresponded to the high expression of pdio3-driven GFP at the same stage ( (Fig 1E, 1H and 1I); note the magenta arrow indicating a bipolar neuron in Fig 1E and 1I and the white arrow indicating an amacrine cell in Fig 1E and 1H). Similarly, at stage NF48, the outer plexiform layer was GFP positive in both transgenic lines, reflecting a certain level of T 3 responsiveness despite discernable D3 activity. Indeed, some interneurons (horizontal cells, noted as "+") respond to T 3 (Fig 1G, Table 2) and yet display limited dio3 expression (Fig 1E and 1I and Table 2).
At stage NF48, most ganglion cells, (noted as " Ã ") did not express dio3 and therefore could respond to T 3 as confirmed by their positive TH/bZIP response (Fig 1G, Table 2). As expected, most photoreceptors express dio3 (noted as "white head arrow in Fig 1E and 1I) and do not respond to T 3 (Fig 1G, Table 2). Theses results suggest that dio3 expression limits T 3 responsiveness in specific cells too.

Opsin S-expressing cones, rods and bipolar neurons express pdio33-GFP at stage NF48
To determine more specifically in which retinal cell types pdio3-GFP expression was found we used double immunochemistry with established markers of each retinal cell type (see Materials and Methods). At stage NF48, pdio3-driven GFP expression was found to co-localize with markers of a large proportion of cone cells expressing Opsin Blue (Opsin S), noted by a magenta arrow in (Fig 3A, 3B and 3C), certain bipolar neurons (ChX10), noted by white head arrows (Fig 3D and 3H) and with all Rhodopsin positive rods, noted by white arrows (Fig 3E-3G).
Thus, as shown in Fig 3A, at stage NF48 GFP expression (green) was found in all retinal cell types in all layers, except in the ganglion cells layer where GFP expression is localized in few ganglion cells. In the outer nuclear layer, it only co-localizes with the body of the Opsin S cones (Fig 3A, in red). Co-localization of pdio3-GFP with several Opsin S cone body was observed and shown with magenta arrow in Fig 3B and 3C (zoom) (in red, with magenta arrow).
Another section revealed co-localization of GFP with Rhodopsin (in red), a specific marker of red rods (Fig 3E-3G for different magnifications). These results suggest that at stage NF48, dio3 expression limits T 3 availability, thereby controlling Opsin S cones number and rods survival.

dio3 expression prevents induction of T 3 -responsive genes
In order to determine whether dio3 expression contributes to local control of TH availability in the developing eye, we treated NF48 pdio3-reporter tadpoles with T 3 , IOP or IOP+T 3 (see Materials and Methods for details) and analyzed T 3 target genes expression (Table 1). A highly significant increase (p<0.001) in dio3 expression was seen between IOP and IOP+T 3 , but not between CTL and T 3 treated groups (Fig 4A). This difference in T 3 responsiveness can be interpreted in GFP/PARV co-labeling in amacrine cells (white arrow head) and ganglion cells (asterisks). Fig 2F: merge of all channels shown in Fig 2A- (Fig 4B). In the presence of T 3 , despite the presence of IOP, we observed an increased expression of the canonical T 3 -responsive genes klf9 and thibz. Of note, IOP treatment itself, induces no significant change (as compared to controls) in GFP, dio3 and T 3 target gene expression (Fig 4A). This result indicates that the local T 4 to T 3 conversion by D2 is insufficient to trigger T 3 target gene expression in the retina as a whole. In IOP+T 3 treated group, dio3, klf9, thibz and thrb expression increases when compared to control group, showing that the cells expressing these genes are T 3 -  (Fig 3A, 3B, 3D, 3E, 3F and 3H). Scale bars: 10 microns ( Fig  3C and 3G). https://doi.org/10.1371/journal.pone.0195374.g003 Deiodinase 3 in the tadpole retina responsive. When comparing dio3 and GFP expression (all observations pooled), a significant correlation (p = 0.007) is seen between dio3 and GFP transcript levels (r = 0,525) (Fig 4C). This observation strengthens the argument that GFP can be used as an indicator of dio3 expression levels.

Discussion
Several studies have shed light on the role of deiodinases in determining sensory organ development in vertebrates. This is especially the case in the eye, where TH-related development is  Fig 4A. Real-time q-PCR analysis of eGFP, dio3, klf9, thibz and thrb for their T 3 transcriptional response in NF48 eye from reporter transgenic line pdio3-GFP. Gene expression was normalized against odc. mRNA levels from vehicle controls (CTL) were used as reference values. Results pooled from two to three independent experiments are represented as scatter dot plots mean with SD. 14!n!6 per group. Non-parametric ANOVA, Kruskall Wallis with uncorrected Dunn's test (PRISM7) was used to assess statistical significance. Ã , p<0.05, ÃÃ , p<0.01; ÃÃÃ , p<0.001. Fig 4B. Working model for the local control of T 3 local availability. Fig 4C. Heatmap of mean expression for each group illustrating the correlation between endogenous dio3 expression and eGFP expression in pdio3-GFP tadpole retina. https://doi.org/10.1371/journal.pone.0195374.g004 Deiodinase 3 in the tadpole retina controlled by the timing of Dio3 expression in mouse retina [12,28] and dio3 in Zebrafish eye [5], as well as in Xenopus retina during metamorphosis [11]. In mouse retina, Dio3 is expressed in immature mouse retina. Ng et al. [12] showed that Dio3-/-mice lost 80% of cones through neonatal cell death. Their results suggest that Dio3 expression limits hormonal exposure of the cones ensuring cone survival and opsin (S and M) patterning, required for cone adaptive function during development. In zebrafish, knockdown of dio3 by morpholinos causes reduced eye size and a strong reduction in rods and all four cones types. This result suggests the importance of dio3 as a central player for zebrafish eye development. However, the roles of dio3 on the survival and the patterning of Opsin type photoreceptors have not yet been addressed in Xenopus retina development, notably during premetamorphic stages. What is more, in Xenopus, there were no data on TH availability nor on the role of deiodinase 3 expression in other retinal cell types.Our experiments were designed to address these questions: how TH availability is controlled, and more precisely, if the control occurs at the tissue or cellular level and in which cells.
In the mouse retina, TH availability appears to control survival and patterning of specific retinal cells. Notably, a TH gradient is observed that may play a role in establishing the gradient of M-opsin [29]. In contrast, in Xenopus retina, there is a spatial gradient of dio3 during metamorphosis, with a higher level in the dorsal retina [11]. These authors demonstrated that the dorsal ciliary margin zone (CMZ) cells are resistant to exogenous TH at stage 50-54, but they noted that an increase in proliferation of these cells was induced with a low concentration of T 3 when D3 activity was inhibited. In contrast, D3 overexpression inhibited TH-induced proliferation of the ventral CMZ cells [11]. This localized expression of dio3 in the dorsal CMZ leads to the asymmetric growth of the frog retina, but the question of the role of dio3 expression in the eyes of pre-metamorphic tadpoles was not addressed. Interestingly, in pre-metamorphic tadpoles, maximal dio3 mRNA levels are found in whole tadpoles at NF46-NF48 stages [22], crucial stages before the first pro-metamorphosis stage NF 53, where generalized competence to respond to T 3 is observed [30].
Our study shows a specific and dynamic spatio-temporal pattern for dio3 transcripts and dio3 promoter activity in the retina of pre-metamorphic Xenopus tadpoles and post-metamorphic juveniles. The results suggest that the timing of retinal maturation in pre-metamorphic tadpoles is mostly controlled by local, cell-specific D3 activity following modulations between stages NF 41 and NF48. Moreover, D3 activity follows a dorso-ventral gradient at NF48. And the differential-T 3 responsiveness of retinal cells in our TH sensor model strongly suggests that some photoreceptors and bipolar neurons are more specifically protected from TH-driven maturation at stages NF41 and NF48. Likewise, the same result is observed for amacrine cells at NF48. Our results suggest too that more ventral horizontal neurons and more ganglion cells needed an active TH signaling between NF41 and NF48 (Fig 5).
Therefore, we show that in the pre-metamorphic retina, TH availability is not controlled at the tissue level but at the cellular level by D3, in most of the cell types in each retinal layer. In particular, D3 is expressed in photoreceptors and bipolar neurons at NF41 and NF48 whereas at NF45, only a photoreceptors subset expressed D3. At stage NF48, D3 is expressed specifically in Opsin S cones, rods and bipolar neurons (Fig 5).
The more restricted expression of D3 at stage NF45 could allow the recruitment and the maturation of more cell types of the inner nuclear layer and more pre-existing photoreceptors types. In parallel, maternal stocks of TH decrease at this stage [22]. Then, as the tadpole thyroid gland is formed, T 4 levels increase at NF46. So, we can speculate that the small subset of photoreceptors expressing D3 at NF45 may correspond to a subtype protected from TH maturation, just before TH levels increase.
At stage NF48, the higher D3 expression observed in most photoreceptors and in cells of the inner nuclear cell layer, could be related to the fact that retina need to be protected from the step increase in TH levels that will initiate pro-metamorphosis at NF54 [31]. Pro-metamorphosis is a step where TH signaling is active and T 3 -dependent tissue remodeling starts [30], illustrated by the onset of green rod differentiation in retina [21].This result is in accordance with previous studies showing a retinal photoreceptors sensitivity to TH excess during retina development in several species. In vertebrates, TH and TRs are involved in opsin expression and cone development [10,28,32,33,34,35]. Furthermore, TH can modulate the M-opsin/S-opsin ratio in the developing mouse retina [29]. Moreover, in trout, excess of TH during smoltification results in a loss of UVS cones [34,36,37,38]. Likewise, another study shows that systemic TH may induce retinal remodelling in juvenile rainbow trout [39].
By analyzing dio3 and T 3 -target genes transcriptional regulation by T 3 in the presence or absence of a deiodinase inhibitor (IOP), we show that dio3 plays a pivotal role in controlling local T 3 availability in dio3 expressing cells. No effect of exogenous T 3 is observed when deiodinases are not inactivated by IOP. The fact that expression of T 3 -responding genes increases when exogenous T 3 is added in the absence of IOP reflects the fact that these genes must be expressed in cells devoid of dio3, such as in the ventral zone.
A final point is that the pdio3-GFP reporter tadpoles are a useful tool for following dio3 expressing cells and for displaying the regulation of TH availability by D3 at the cellular scale. Moreover, the expression level of dio3 and the reporter gene are reasonably well correlated, despite a certain variability due to the very dynamic D3 expression at specific stages, a fact that could be compounded by the difficulty of obtaining homogeneous batches of tadpoles. However, even if dio3 is transcriptionally responsive to T 3 , its range of T 3 -responsiveness is far below that of the TH/bZIP. As such the TH/bZIP reporter remains the best T 3 -sensor available to date in xenopus.

Conclusions
Our findings show that dio3 displays a dynamic and cell-specific expression in the pre-metamorphic retina between NF41 and NF48. At stage NF41 and NF48 most cells types express dio3, with a particularly high expression in rods and the S cones at NF48 whereas T 3 signaling is detectable in horizontal neurons and ganglion cells (Fig 5). After metamorphosis, dio3 expression was much more limited, only being found in amacrine cells and a sub-population of photoreceptors. The results show that dio3 plays a key role in determining TH availability during retinal development with a precise and cell-type specific timing. We suggest that D3 expression could be necessary to protect photoreceptors and bipolar neurons from out of phase TH signaling and hence that dio3 expression constrains effects of TH signaling in the retina prior to and during metamorphosis.