Developmental Changes of Cone Opsin Expression but Not Retinal Morphology in the Hypothyroid Pax8 Knockout Mouse

Abstract

Purpose.: The effects of postnatal hypothyroidism on retinal development and spatial patterning of cone opsin expression were studied in Pax8-deficient mice. Pax8^−/− mice are incapable of synthesizing thyroxine and serve as a model for congenital hypothyroidism.

Methods.: Pax8^−/−, Pax8^+/−, and Pax8^+/+ littermates were studied. Serum thyroid hormone levels, body weight, and eye size were measured. Retinal cell-type–specific antibodies were used on frozen sections to examine the postnatal development of the major retinal cell classes and of retinal structure. The expression of short-wavelength–sensitive (S) and middle-to-long-wavelength–sensitive (M) cone opsins was assessed with opsin antibodies on retinal sections and whole retinas. The pattern of S opsin mRNA was assessed by in situ hybridization.

Results.: In Pax8^−/− mice, S opsin was upregulated in all cones, whereas M opsin was downregulated throughout the retina, the wild-type dorsoventral gradients of S and M opsin expression were absent. Otherwise, Pax8^−/− mice showed no overt mutant phenotype in eye size, gross retinal anatomy, and the time-course of structural differentiation of retinal photoreceptors, horizontal cells, bipolars, amacrines, ganglion cells, and Müller glia cells.

Conclusions.: Pax8^−/− mice show a pattern of cone opsin expression that differs substantially from the wild-type pattern, but exhibit no apparent alterations in general retinal development. The finding that a postnatal decrease in serum thyroid hormone yields changes in postnatal cone opsin expression is consistent with a ligand-dependent role of thyroid hormone receptor β2 in S opsin repression and M opsin activation.

Thyroid hormone (TH) is essential for many physiological and developmental processes (also in the nervous system) and acts on ligand-dependent nuclear TH receptors (TRs) that regulate gene expression. ¹ The role of TH in retinal development is well established. Its action is mediated by two TRs with different spatial and temporal expression profiles. Retinal TRα expression is ubiquitous and high throughout development. ^2–4 In contrast, TRβ2 is localized to cone photoreceptors only and reaches its highest expression during early cone development. ⁵  

Most mammalian retinas contain two spectral cone types characterized by the expression of either a middle- to long-wavelength–sensitive (M) cone opsin or a short-wavelength–sensitive (S) cone opsin (reviews in Refs. 6, 7). Wild-type mice show a special cone opsin expression pattern: M opsin expression is higher in the dorsal than in the ventral retina. In contrast, most ventral cones express S opsin whereas the dosal retina contains only a small proportion of S cones. ^8–10 Many S opsin–containing cones across the mouse retina coexpress M opsin. ^9,11 This cone opsin expression pattern develops during the first postnatal weeks. ¹²  

The requirement of TRβ2 for cone opsin regulation has been demonstrated in a TRβ2-knockout mouse. ¹³ This mouse shows a loss of M opsin and the expression of S opsin in all cones, resulting in an all S cone retina without the wild-type S opsin expression gradient. ¹³ The ligand dependence of TRβ2 in activating M opsin and repressing S opsin was established by receptor mutants with defective DNA or TH binding, showing the requirement of TH for cone maturation. ^14–16 In vivo studies addressing the role of TH during early retinal development often have used pharmacologic suppression of endogenous TH. ^17–20 Suppression was commonly started prenatally and continued postnatally, leaving a wide perinatal time window. Only very recently has cone development been studied in the Tshr^−/− mouse, an animal model of postnatal hypothyroidism. ²¹  

In the present study, we used the Pax8-knockout mouse ²² to investigate the effect of postnatal hypothyroidism on the development of various classes of retinal cells, as well as on cone photoreceptors and their opsin expression pattern. Pax8^−/− mice do not develop thyroid gland follicles and thus are incapable of producing thyroxine. A deficiency in serum TH occurs only after birth, when placental TH transfer from the euthyroid dam ceases. ²³ This deficiency makes Pax8^−/− mice a model of human congenital hypothyroidism, a disorder of thyroid function affecting approximately 1 in 3000 to 4000 newborn infants. ²⁴  

Materials and Methods

Animals

All procedures complied with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research, the NIH Principles of Laboratory Animal Care, and the corresponding German laws. Procedures were performed as approved by the local committee on animal experimentation. 

We studied the Pax8^−/− (Pax8^tm1Pgr) mouse, bred and characterized by Mansouri et al. ²² Pax8^+/− parents were crossed to produce +/+, +/−, and −/− Pax 8 littermates (genotyped by PCR analysis of genomic DNA), which were compared at postnatal days (P)7, P14, and P21 and up to postnatal week (W)22. At 10 AM, the animals' body weights were determined, and they were overdosed with isoflurane and decapitated. Blood was collected, and serum concentrations of free and total triiodothyronine (T3) and thyroxine (T4) were measured in a competitive immunoassay with direct chemiluminescence technology (ACS:180 Immunoassay System, Bayer HealthCare, Leverkusen, Germany). Measurements were performed at the Endocrinology Department of Frankfurt University Medical School. 

Fixation and Tissue Preparation

After death, the dorsoventral eye axis was marked in situ, and the eyes were enucleated. The axial length and equatorial diameter of the isolated eye were measured with a sliding caliper. The cornea and lens were removed, and the eye cup with the attached retina was immersed in 4% paraformaldehyde in 0.1 M phosphate buffer (PB, pH 7.4). Retinas used for wholemount preparations were fixed for 1 hour at room temperature (RT), tissue used for cryosectioning was fixed for 2 hours at RT, and tissue used for in situ hybridization (ISH) was fixed for 4 hours at RT. Fixation was followed by several washes in PB. Retinas assigned for wholemount staining were separated from the eye cup and stored either in PB at 4°C or in 30% sucrose in PB at −80°C until further processing. For transverse sections, the eye cup was cryoprotected in graded sucrose solutions (10%, 20%, and 30% in PB) and embedded in tissue-freezing medium (Jung, Nussloch, Germany). Sections of 20-μm thickness were collected on glass slides and frozen at −20°C. Some retinal pieces were dehydrated and embedded in Epon, for transverse, semithin sections (1 μm) to be cut. 

Immunostaining

Isolated whole retinas (wholemounts) were shock-frozen and thawed three times. The wholemounts were immunoreacted free-floating, sections on slides. For single and double labeling, the tissue was preincubated with 10% normal donkey serum (NDS), 0.5% Triton X-100, and 1% bovine serum albumin (BSA) in PB for 1 hour at RT. Incubation with primary antibodies was performed overnight at RT in 3% NDS, 0.5% Triton X-100, 1% BSA, and 0.05% NaN₃ in PB. Retinal cell-type–specific antibodies ^25,26 and concentrations used on the sections are listed in Table 1. Wholemounts were double-labeled with sc-14363 and JH 492 for cone opsins. After washes in PB, the tissue was incubated with secondary antibodies coupled with Alexa488 (1:500; Molecular Probes, Eugene, OR) and Cy5 (1:250; Jackson ImmunoResearch Laboratories, West Grove, PA) in the same medium as for the primary antibodies for 1 hour. After they were washed, the wholemounts (flattened onto slides with the photoreceptor side up) and sections were coverslipped in aqueous mounting medium (Aqua-Poly/Mount; Polysciences, Warrington, PA) and analyzed with a fluorescence microscope (Axioplan 2; Carl Zeiss Meditec, Oberkochen, Germany). Micrographs were taken with a CCD camera and the microscope system software (Axiovision; Carl Zeiss Meditec). Images were adjusted for brightness and contrast (Photoshop CS, ver. 8.0.1; Adobe Systems, San Jose, CA; and Axiovision 4.2; Carl Zeiss Meditec). 

Table 1.

 

List of the Antibodies Used in the Present Study

Antigen	Antiserum/Antibody and Host	Working Dilution	Source and Reference
M cone opsin	rb JH 492	1:2000	Jeremy Nathans, Wilmer Eye Institute at Johns Hopkins, Baltimore, MD 27
S cone opsin	gt anti-blue-sensitive opsin sc-14363	1:500	Santa Cruz Biotechnology, Santa Cruz, CA
Rod opsin	ms rho4D2	1:50	Robert S. Molday, University of British Columbia, Vancouver, BC, Canada 28
Calbindin	rb anti-calbindin CB 38	1:2000	Swant
Calretinin	rb anti-calretinin CR 7699/3H	1:2000	Swant
Choline acetyltransferase (ChAT)	rb anti-ChAT AB 143	1:50	Chemicon, Temecula, CA
Disabled-1 (DAB1)	rb anti-DAB1	1:500	Brian W. Howell, Neurogenetics, National Institute of Neurological Disorders and Stroke, Bethesda, MD 29
Glutamine synthetase (GS)	ms anti-GS G 45020	1:500	Transduction Laboratories-BD Biosciences, Lexington, KY
Glycogen phosphorylase (Glypho)	gp anti-Glypho rb anti-Glypho	1:1000	Bernd Hamprecht, Institute of Organic Chemistry University of Tübingen, Germany 10,30
Neurokinin receptor 3 (NK3-R)	rb anti-NK3	1:1000	Arlene A. Hirano, Los Angeles, Department of Neurobiology, UCLA, Los Angeles, CA 31
Protein kinase Cα (PKCα)	rb anti-PKCα P 4334	1:10000	Sigma-Aldrich, St. Louis, MO
Tyrosine hydroxylase	ms anti-tyrosine-hydroxylase	1:1000	Hermann Rohrer, Max Planck Institute for Brain Research, Frankfurt am Main, Germany 32

 gp, guinea pig; gt, goat; ms, mouse; rb, rabbit

Quantification

The retinal area was measured on photos of complete retinal flatmounts with ImageJ (developed by Wayne Rasband, National Institutes of Health, Bethesda, MD; available at http://rsb.info.nih.gov/ij/index.html). Total retinal thickness was measured on representative sections of Epon-embedded semithin sections and cryostat sections. In both measurements, only similarly treated material was compared, and only relative differences were relevant; hence, tissue shrinkage was not an issue. Densities (cells per square millimeter) of cones expressing S or M cone opsin were assessed in micrographs (40× lens; sample field size, 100 × 100 μm) of immunolabeled wholemounts at four defined positions (see Fig. 9) in the dorsal and ventral retina. One-way ANOVAs or t-tests (SigmaStat; SPSS, Chicago, IL) were used to evaluate significant differences between the genotypes. Data are presented as the mean ± SD. The number of retinas used for the quantifications are given in the respective text and figure legends. 

S Opsin ISH

For S opsin ISH, retinas of P21 mice (Pax8^−/−: n = 6; Pax8^+/−: n = 5; Pax8^+/+: n = 6) were fixed in 4% paraformaldehyde/PBS (pH 7.4), washed in PBS (10 minutes), acetylated in 0.1 M triethanolamine/0.25% acetic anhydride (pH 8.0; 10 minutes), and equilibrated in 5× SSC (5 minutes). Retinal wholemounts were prehybridized for 2 hours at 58°C with 20 μL hybridization buffer (50% deionized formamide, 5× SSC, 100 μg/mL salmon testes DNA, 100 μg/mL yeast tRNA, 2.5× Denhardt's solution, 5 mM EDTA, and 0.05% CHAPS). Hybridization was performed for 16 hours at 58°C in 30 μL fresh hybridization buffer with the addition of denatured DIG-labeled riboprobes to antisense mouse Opn1sw (50 ng/mL). Riboprobes were generated by in vitro transcription of a T7 promoter-coupled PCR template (nt 630-973, NM_007538), with RNA polymerase and DIG-labeled rUTP (DIG RNA Labeling Kit; Roche, Mannheim, Germany). Hybridized tissue was washed at 58°C in 2× SSC buffer (three changes) and 0.1× SSC buffer (three changes) before equilibration in 100 mM Tris-HCl and 150 mM NaCl (pH 7.5). Incubation with alkaline phosphatase-coupled anti-digoxigenin Fab fragments (1:2500; Roche) was performed in the same buffer supplemented with 1% BSA and 0.5% blocking reagent (Roche; 2 hours, RT). After they were washed in 100 mM Tris-HCl/400 mM NaCl (pH 7.5), the retinas were equilibrated in 100 mM Tris-HCl, 100 mM NaCl, 50 mM MgCl₂ (pH 9.5), and antibody binding sites were visualized by incubation in NBT (5 μL/mL)/BCIP (3.75 μL/mL) in the same medium. The color reaction was performed at RT and stopped after 5 hours in 10 mM Tris-HCl and 1 mM EDTA (pH 8.0). The retinas were mounted in glycerol. 

TH Replacement

To determine the effects of replacing TH in the Pax8^−/− mice, the mutants received daily SC injections of 18 ng thyroxine (T4)/g body weight in saline from P1 to W8. The control Pax8^+/+ mice were either injected with saline lacking T4 or received no injections; no differences were found between these two types of control. At W8, the animals were killed to determine blood serum levels of TH and retinal cone opsin expression patterns. 

Results

General Features

Hypothyroidism in Pax8^−/− mice emerges only after birth, as the embryos are supplied with hormone by their heterozygous mothers. Serum levels of free and total triiodothyronine (fT3 and T3) and thyroxine (fT4 and T4), measured at P14 and P21 and at W22, were strongly reduced in Pax^−/− mice compared with Pax8^+/+ and Pax8^+/− littermates. Figure 1 gives the concentrations of total T4 and of the biologically active form fT3. It is noteworthy that in Pax8 heterozygotes, these hormone levels did not differ from those in the Pax8 wild-types. 

Figure 1.

 

Serum thyroid hormone levels in Pax8 mice at P14 and P21 and at postnatal week W22. The Pax8^−/− mutants showed much lower hormone levels than did the wild-type and heterozygous mice. fT3, free triiodothyronine; T4, thyroxine. The number of +/+, +/−, and −/− Pax8 mice at P14: 3, 2, and 2; at P21: 15, 17, and 14; and at W22: 5, 7, and 3 (W22 ^−/− T4; n.d., not determined). Histograms show the mean and standard deviation. T4 levels in the Pax8^−/− mice were below the detection threshold of the assay. *Significant at P < 0.01.

Body weight was reduced in Pax8^−/− mice. At P7, mean body weight was slightly lower in −/− than in +/+ and +/− Pax8 mice (3.5 g vs. 4.8 g; n = 6, 7, and 9 for −/−, +/+, and +/−, respectively). At P14 and P21 the difference was significant: P14 mean weights were 5.1 g in −/− (n = 8) vs. 8.9 g in +/+ (n = 7) and 8.7 g in +/− (n = 7); P21 mean weights were 6.0 g in −/− (n = 7) vs. 11.9 g in +/+ (n = 5) and 11.7 g in +/− (n = 6) Pax8 mice. All groups contained male and female animals, and there were no weight differences between the sexes at these young ages. 

Eye axial length and equatorial diameter did not differ significantly between the genotypes at the ages studied. For example, mean axial length at P7 was 2.24 ± 0.27 mm in −/− (4 eyes), 2.25 mm in +/+ (4 eyes) and 2.31 ± 0.08 mm in +/− (6 eyes) Pax 8 mice; at P14 it was 2.52 ± 0.08 mm in −/−, 2.67 ± 0.08 mm in +/+, and 2.57 ± 0.13 mm in +/− (6 eyes for each genotype); and at P21 it was 2.68 ± 0.11 mm in −/−, 2.72 ± 0.04 mm in +/+, and 2.69 ± 0.11 mm in +/− (10 eyes for each genotype). These results indicate that ocular growth and eye size are not affected in Pax8^−/− mice. Retinal area and thickness were determined at P21 and W22 and did not differ significantly between the genotypes. 

Retinal Development

Retinal sections of stages P7, P14, and P21 were stained with cell-type–specific antibodies (Table 1) to assess the effect of postnatal hypothyroidism on the general structure of the retina and the postnatal development of its cell types. Müller glia cell morphology showed no difference between genotypes at the stages examined (Fig. 2). Glutamine synthetase immunoreactivity increased from P7 to P21, and at each stage, staining intensity was similar in +/+, +/−, and −/− Pax8 mice. Likewise, horizontal, amacrine, and ganglion cells labeled for calbindin showed similar features across the genotypes (Fig. 3). Typical of the mouse, three bands of calbindin signal in the inner plexiform layer (IPL) had reached adult appearance by P21. The outer two bands represent the processes of cholinergic amacrines, and the middle band those of NOS-containing amacrines. ²⁵  

Figure 2.

 

Retinal transverse sections immunolabeled for glutamine synthetase to reveal Müller glia cells in the +/+, +/−, and −/− Pax8 mice at P7, P14, and P21. OS/IS, outer and inner segments of photoreceptors; ONL, outer nuclear layer; OPL, outer plexiform layer; INL, inner nuclear layer; IPL, inner plexiform layer; GCL, ganglion cell layer. Scale bar, 100 μm.

Figure 3.

 

Retinal transverse sections labeled for calbindin to reveal horizontal, amacrine, and ganglion cells in the +/+, +/−, and −/− Pax8 mice at P7, P14, and P21. For abbreviations, see Figure 2. Scale bar, 100 μm.

Immunostaining of rod photoreceptors (anti-rod opsin) and rod bipolar cells (anti-PKC) showed no apparent differences between the Pax8 genotypes. The same was found for the neurokinin receptor 3 (NK3)-immunopositive cone bipolar cells (type 1 and 2 OFF-cone bipolar cells ³³ ). Likewise, dopaminergic (tyrosine hydroxylase-positive), cholinergic (cholinacetyltransferase-positive), and AII (disabled-1-positive) amacrine cells, as well as amacrine and ganglion cells, labeled for calretinin were similar in appearance in +/+, +/− and −/− Pax8 mice. These labeling patterns are not illustrated here. 

Cone photoreceptors were labeled with anti-glycogen phosphorylase, an antiserum that has been used as a general cone marker in the mouse, ¹⁰ to evaluate the morphology of the entire cone (Fig. 4). At the light microscopic level, no obvious morphologic differences were observed between the genotypes. Cone somata were distributed throughout the outer nuclear layer (ONL) at P7 and had migrated to the adult position near the ONL/innersegment (IS) border by P21 in all genotypes. 

Figure 4.

 

Retinal transverse sections labeled for glycogen phosphorylase to reveal all cones in the +/+, +/−, and −/− Pax8 mice at P7, P14, and P21. Only the outer retina is shown. For abbreviations, see Figure 2. Scale bar, 100 μm.

Taken together, these data show that the time course of retinal development and the general morphology of the retina were not affected in Pax8^−/− mice. The only difference we observed was the expression pattern of cone opsins. 

Development of Cone Opsin Patterns

Postnatal development of cone opsin expression was assessed in retinas labeled for S and M cone opsin. At P7, retinal sections of the three genotypes showed identical S opsin staining of cones. The S cones were labeled from their outer segments to their pedicles, and S cone densities were similarly high in the dorsal and ventral retina (Fig. 5). M opsin label was absent at P7 (not shown). By P14, the pattern of cone opsin expression changed (Fig. 6). M opsin was detectable for the first time, and many cones of the +/+ and +/− Pax8 mice coexpressed both opsins. In contrast, M opsin signal was low or absent in the Pax8^−/− mutants (Fig. 6). 

Figure 5.

 

S cone opsin expression at age P7. Retinal transverse sections labeled for S opsin, showing dorsal (D) and ventral (V) retinal regions in +/+, +/−, and −/− Pax8 mice at P7. For abbreviations, see Figure 2. Scale bar, 25 μm.

Figure 6.

 

S and M cone opsin expression at age P14. Matching fields in retinal transverse sections double-labeled for S opsin (top) and M opsin (bottom), showing dorsal regions in +/+, +/−, and −/− Pax8 retinas at P14. Comparison of the top and bottom labels showed that some of the cones express both opsins in Pax8^+/+ and Pax8^+/− mice. Pax8^−/− mice showed no M opsin immunoreactivity. For abbreviations see Figure 2. Scale bar, 25 μm.

Pax8^−/− mutants typically die approximately 3 weeks after birth. Therefore, we used P21 animals as the reference stage in the present study. Whole retinas were immunoreacted to assess S and M opsin expression in more detail (Fig. 7). The pattern of S opsin at P21 conformed to the adult wild-type pattern in +/+ and +/− Pax8 mice, with a high number of S opsin cones in the ventral retina and a sparse population of S cones in the dorsal retina. In contrast, Pax8^−/− mutants showed homogenous and high expression of S opsin over the whole retina. M opsin expression in +/+ and +/− Pax8 mice qualitatively conformed to the adult wild-type pattern. M opsin signal was strong in the dorsal retina, representing a dense cone population expressing M opsin at a high level, whereas cones in the ventral retina expressed M opsin at lower levels. In Pax8^−/− animals, M opsin was reduced over most of the retina, but patches with higher M opsin expression occurred in different parts of the retina, in deviation from the wild-type dorsoventral gradient of M opsin (Fig. 7). Between retinas, these patches varied in size, position, and number. They do not show up in the quantification (see Fig. 9) because the counts included all M opsin–containing cones, irrespective of their labeling intensity (opsin expression level), and because in many cases the patches were outside the sampled areas. 

Figure 7.

 

Staining of M and S opsin in flattened whole retinas of +/+, +/−, and −/− Pax8 mice at age P21. Each retinal image is assembled from several negatives of immunofluorescence micrographs; hence, the darker regions indicate more intense staining. S opsin staining was strongest in ventral retina in Pax8^+/+ and Pax8^+/−, but was intense across the entire retina in Pax8^−/−. M opsin staining was strongest in the dorsal retina in Pax8^+/+ and Pax8^+/− but was patchy in Pax8^−/−. D, dorsal; V, ventral. Scale bar, 1 mm.

Figure 8 shows the pattern of S cones at higher magnification for representative retinal locations. Figure 9 gives the results of the corresponding density counts. In the dorsal peripheral retina (75% eccentricity), S cone densities were higher in −/− (10,500/mm²) than in +/+ (7,000/mm²) and +/− Pax8 mice (7,500/mm²). In contrast, in the ventral retina, S cone densities were significantly lower (P < 0.01) in −/− (9,500–10,000/mm²) than in +/+ (13,000–15,000/mm²) and in +/− Pax8 mice (12,500–15,500/mm²). The density of M opsin–expressing cones was lower over the whole retina in −/− compared with +/+ and +/− Pax8 mice. In the central retina (positions 50% dorsal and 50% ventral), the differences in M cone density between −/− (8,000–10,000/mm²) and +/+ (11,500–15,500/mm²) Pax8 mice were significant (P < 0.01). 

Figure 8.

 

S opsin labeling in flattened retinas of +/+, +/−, and −/− Pax8 mice at age P21. Higher magnification views of representative positions in more central (50%) and more peripheral (75%) parts of the dorsal (D) and ventral (V) retina (see scheme in Fig. 9A). Pax8^+/+ and Pax8^+/− mice showed the wild-type gradient in S cone density, decreasing from the ventral to dorsal retina, whereas Pax8^−/− mutants showed a homogeneous distribution of S cones across the retina. Scale bar, 50 μm.

Figure 9.

 

Densities of S opsin– and M opsin–expressing cones at age P21. (A) The location of the retinal sampling field positions in a schematic flatmounted retina. Dorsal, top; ventral, bottom. Cone densities were assessed dorsally and ventrally at the more central positions (50% of the distance between the optic nerve head (○) and retinal margin) and at the more peripheral positions (75% of that distance). Densities of (B) S opsin– and (C) M opsin–expressing cones are represented as the mean density and standard deviation for +/+ (□; n = 10), +/− (▩; n = 11), and −/− (■; n = 9) Pax8 mice at the positions indicated in (A). *Significant at P < 0.01.

Some Pax8^−/− animals survived to older ages and were analyzed. Their serum TH levels remained at much lower levels than in their wild-type littermates (Fig. 1, see data for postnatal week W22). At the adult age of W22, the differences in S cone patterns between the genotypes were even more pronounced than at P21. In the dorsal retina of +/+ and +/− Pax8 mice, the density of cones expressing S opsin further decreased from P21 to W22, whereas in −/− animals, the pattern of S opsin remained unchanged, with similarly high S cone densities across the retina (Figs. 10, 11). In contrast, the topography of M opsin–expressing cones in Pax8^−/− mice was closer to the wild-type at W22 than at P21 (compare Figs. 11 and 9). This finding argues for a delayed maturation of the M opsin expression pattern (see the Discussion section). 

Figure 10.

 

S opsin labeling in flattened retinas of +/+, +/−, and −/− Pax8 mice at W22. Higher magnification views of representative positions in more central (50%) and more peripheral (75%) parts of dorsal retina (see scheme in Fig. 9A). In Pax8^+/+ and Pax8^+/− mice there was a wild-type–like dorsal decrease in S cone density, whereas in the Pax8^−/− mutants dorsal S cone density remained high. Scale bar, 50 μm.

Figure 11.

 

Densities of (A) S opsin– and (B) M opsin–expressing cones at W22. Data are expressed as the mean density and standard deviation for +/+ (□; n = 7), +/− (▩; n = 7), and −/− Pax8 mice (■; n = 3) at more central and more peripheral positions of dorsal and ventral retina (see scheme in Fig. 9A). *Significant at P < 0.01.

Double-labeling of transverse cryostat sections with general cone markers (peanut agglutinin PNA, anti-glycogen phosphorylase; see Fig. 4) and cone opsin antibodies showed that every cone expressed opsin in all three Pax8 genotypes (not illustrated). This allowed us to assess total cone densities from wholemount retinas double-labeled for S and M opsin. Total cone densities in W18 mice showed no consistent differences between the genotypes (Fig. 12). 

Figure 12.

 

Total cone densities at W18 for +/+ (□; n = 4), +/− (▩; n = 7), and −/− Pax8 mice (■; n = 2) at the more central and peripheral positions of the dorsal and ventral retina (see scheme in Fig. 9A).

ISH analysis of whole retinas at P21 indicated that changes in S opsin expression occurred at the mRNA level (Fig. 13). In the dorsal retina, more cones expressed S opsin in −/− than in +/+ and +/− Pax8 mice; in the ventral retina, the density of cones with S opsin mRNA was high in all three genotypes. This result is in line with those of S opsin immunolabeling (Fig. 8). Qualitative assessment suggested slightly lower ventral densities of these cells in Pax8^−/− mice than in the other genotypes (as expected from the data in Fig. 9B), but no quantitative evaluation of the in situ hybridized retinas was attempted. 

Figure 13.

 

Labeling of S opsin mRNA by ISH of whole retinas of +/+, +/−, and −/− Pax8 mice at age P21. The images were obtained from the same retinal positions as in Figure 8. Scale bar, 100 μm.

TH Replacement

To determine whether the observed opsin expression changes were due only to the reduced TH levels or also to some other peculiarities of the Pax8^−/− mutation, we injected Pax8^−/− mice with appropriate daily doses of T4 from P1 to W8, when the animals were killed for analysis. Thyroxine treatment of the mutants resulted in wild-type serum levels of fT3 and T4: fT3 levels were 4.3 ± 0.9 pM in treated −/− (n = 3) and 5.5 ± 0.7 pM in +/+ (n = 5) Pax8 mice; and T4 levels were 40.5 nM in treated −/− (n = 2) and 48 ± 13 nM in +/+ (n = 5). 

The treated Pax8^−/− mice showed a wild-type S and M opsin expression pattern across the retina. Figure 14 illustrates the S opsin expression in dorsal peripheral retina, where the differences between Pax8^+/+ mice and untreated mutants were most prominent. Thyroxine-supplemented mutants showed the same low density of S cones as the wild-type controls, which strongly contrasts with the high dorsal S cone densities in untreated mutants before and after W8 (Figs. 8, 10).The conclusion is that changes in cone opsin expression in Pax8^−/− mutants depend on the low postnatal TH levels and not on some other, hormone-independent mutant property. 

Figure 14.

 

S opsin labeling in flattened retinas of Pax8^+/+ and T4-treated Pax8^−/− mice at W8. Shown are corresponding regions in the dorsal periphery with similarly low, wild-type-like S cone densities. Scale bar, 50 μm.

Discussion

In our analysis of the Pax8^−/− mouse, a postnatal decrease in TH disturbed normal development of the spectral cone pattern by shifting cone opsin expression, whereas the structural development of retinal cell types was not affected. As will be discussed, these results differ from those obtained in mouse models in which hypothyroidism was induced prenatally, but confirm and extend recent findings in a different model of postnatal hypothyroidism, the Tshr^−/− mouse. 

The declining serum TH levels confirm that Pax8^−/− mutants are hypothyroid postnatally (Fig. 1). Prenatally, Pax8^−/− mice are supplied with TH by their heterozygous mothers, and hypothyroidism becomes effective only after birth. The low postnatal levels of serum fT3 may correspond to residual maternal TH ^34–36 that has been accumulated prenatally or may have been taken up with breast milk (which contains TH in other mammals and humans ^37,38 ). Our finding that postnatal TH replacement in Pax8^−/− mice restores the wild-type cone opsin expression pattern shows that lack of TH, not other consequences of the Pax8 mutation, produces the opsin phenotype. 

The role of TH in retinal development is well established, and its action is mediated by two TRs. TRβ2 is localized to cone photoreceptors only, and its expression is the highest during early cone development, ⁵ whereas retinal TRα expression is ubiquitous and remains high throughout development. ^2–4  

The present study for the first time addresses changes in the (structural) development of the different classes of retinal neurons under conditions of a postnatal onset of serum TH depletion. Our results show no apparent effects on eye size, retinal thickness, and area; on the overall development and cytology of cell types, including the cones; or on the expression of the molecular markers studied. Notably, this observation includes the largely postnatally generated rods, bipolar cells, and Müller cells, ³⁹ on which the most severe effects of postnatal hypothyroidism would have been expected. Our findings contrast with studies demonstrating clear changes in the structural development of the rat retina after prenatally initiated and postnatally continued pharmacologic TH suppression. These include reduced eye size, delayed development of retinal layers, and a reduced number of cells in all layers. ^17–20 The differences probably reflect normal prenatal retinal TH levels in Pax8^−/− mice; hence, our study narrows down the time window in which hypothyroidism has wide-ranging effects on retinal development to the prenatal period. It suggests that the postnatal development of retinal cell morphologies, whether TRα-dependent or not, is largely TH independent. Alternatively, it is possible that low residual TH levels suffice to influence TRα activity (but not TRβ activity). Of course, our anatomic study captured only light microscopic features. Potential ultrastructural alterations (e.g., at the synaptic level) may deserve further study. 

The requirement of TRβ2 for cone opsin regulation has been demonstrated in the TRβ2-knockout mouse. ¹³ The ligand dependence of this nuclear transcription factor in activating M opsin and repressing S opsin was established by receptor mutants with defective DNA or TH binding, showing the requirement of TH for cone maturation. ^14–16 Consistent with this TRβ2-mediated role of TH, we observed a clear difference in the spatiotemporal development of S and M cone opsins in the hypothyroid Pax8^−/− animals. Whereas Pax8^+/+ and Pax8^+/− mice had reached the adult pattern of cone opsins at P21, with the reverse dorsoventral gradients of M and S opsin expression, Pax8^−/− mutants retained high levels of S opsin expression and decreased M opsin expression over the whole retina. Differences in the patterning of S opsin between Pax8^+/+ and Pax8^−/− mice were present until W22, the oldest age studied, suggesting that repression of S opsin in dorsal Pax8^−/− retina is not delayed, but is in fact never achieved. 

The effects on M opsin expression were milder. In contrast to the TRβ2^−/− mouse, where M opsin expression is absent, ¹³ the Pax8^−/− mutant expressed reduced levels of M opsin. Also, the onset of immunocytochemically detectable M opsin levels occurred later than in the wild-type and heterozygous controls (Figs. 6, 7). However, by W22, M opsin was expressed in a larger proportion of cones than at P21, such that at W22, there was no longer a significant difference between Pax8^−/− and Pax8^+/+ mice in the number of M opsin–expressing cones. Nevertheless, the level of M opsin expression remained lower also in these adult Pax8^−/− mice (not illustrated). Hence M opsin expression is not exclusively dependent on normal TH levels. 

Total cone densities did not appear to differ between the Pax8 genotypes (Fig. 12). The low number of Pax8^−/− retinas (n = 2) available for total cone counts precluded a statistical evaluation, but the data are compatible with the interpretation that in Pax8^−/− mice, no cones are postnatally generated or deleted and that the reported differences represent opsin expression shifts in established cones. 

In conclusion, our findings confirm a requirement of TH for normal development of the spectral cone pattern that emerges by differential regulation of S and M opsin in a cone photoreceptor. ^19,15 The main effect of postnatal hypothyroidism is an upregulation of S opsin and a downregulation of M opsin, abolishing the dorsoventral asymmetry of the cone pattern present in wild-type mice. In contrast to TRβ2 deletion mutants ^13,15 or mutants deficient in T3-binding capabilities of the receptor, ^15,16,40 the postnatal TH deficiency in Pax8^−/− mutants reduces M opsin expression but does not abolish it. A similar effect has recently been reported in the thyrotropin receptor-deficient Tshr^−/− mouse, a different model of postnatal hypothyroidism. ²¹  

It remains unknown what sets the opposing dorsoventral gradients in S and M opsin expression in the normal mouse retina. TRβ2 expression is even from the dorsal to ventral retina, ¹⁵ whereas TH levels are higher in the dorsal retina ⁴⁰ where M opsin is high and S opsin is low, implying TH dose-dependent modulation of TRβ2-mediated S opsin repression and M opsin activation in the simplest model. In the wild-type mouse, local differences in TH availability may be generated by the differential expression of deiodinases, enzymes which activate (Dio2) or deactivate (Dio3) TH. ^41–43 Clearly, the cellular pathways of TH-mediated opsin regulation warrant further investigation before we fully understand its role in generating cone patterns. 

Footnotes

 Supported by Deutsche Forschungsgemeinschaft (DFG) Grant PE 38/16-1.

Footnotes

 Disclosure: A. Glaschke, None; M. Glösmann, None; L. Peichl, None

The authors thank Stefanie Heynck and Driss Benzaid for expert technical assistance and Heike Heuer and Karl Bauer (Leibniz Institute for Age Research, Jena, Germany) for providing Pax8^tm1Pgr founder animals.