Cone photoreceptor differentiation regulated by thyroid hormone transporter MCT8 in the retinal pigment epithelium

Significance

The differentiation of cone photoreceptors, which mediate daylight vision and color vision, depends critically upon thyroid hormone. This dependence upon endocrine signaling raises a key question for understanding cone function, or dysfunction in disease, concerning the route that conveys hormone to cones which are shielded behind the blood–retina barrier. By genetic manipulation in mouse models, we find that cone differentiation requires a plasma membrane transporter for thyroid hormone in the retinal pigment epithelium (RPE), the cellular interface that forms the outer blood–retina barrier. These findings suggest that in addition to provision of essential solutes and nutrients for photoreceptor homeostasis, the RPE mediates paracrine-like control of thyroid hormone signaling to promote cone-mediated vision.

Abstract

The key role of a thyroid hormone receptor in determining the maturation and diversity of cone photoreceptors reflects a profound influence of endocrine signaling on the cells that mediate color vision. However, the route by which hormone reaches cones remains enigmatic as cones reside in the retinal photoreceptor layer, shielded by the blood–retina barrier. Using genetic approaches, we report that cone differentiation is regulated by a membrane transporter for thyroid hormone, MCT8 (SLC16A2), in the retinal pigment epithelium (RPE), which forms the outer blood–retina barrier. Mct8-deficient mice display hypothyroid-like cone gene expression and compromised electroretinogram responses. Mammalian color vision is typically facilitated by cone types that detect medium-long (M) and short (S) wavelengths of light but Mct8-deficient mice have a partial shift of M to S cone identity, resembling the phenotype of thyroid hormone receptor deficiency. RPE-specific ablation of Mct8 results in similar shifts in cone identity and hypothyroid-like gene expression whereas reexpression of MCT8 in the RPE in Mct8-deficient mice partly restores M cone identity, consistent with paracrine-like control of thyroid hormone signaling by the RPE. Our findings suggest that in addition to transport of essential solutes and homeostatic support for photoreceptors, the RPE regulates the thyroid hormone signal that promotes cone-mediated vision.

Cone photoreceptors function in bright conditions and mediate color vision. Color vision is facilitated by cone populations with opsin photopigments for response to different regions of the light spectrum, namely medium-long (M, “green”) and short (S, “blue”) wavelength regions in most mammals (1, 2). Despite the fundamental contribution of cones to vision and the variety of visual disorders that involve cone defects (3–5), cone differentiation is incompletely understood. A reliance upon endocrine signaling was indicated by the finding that thyroid hormone receptor TRβ2 (encoded by Thrb) controls cone diversity (6, 7). TRβ2-deficient mice lack M opsin and instead express S opsin in all cones, suggesting that TRβ2 generates diversity by shifting precursors from a default S identity to an M identity (6). TRβ2 is a ligand-regulated transcription factor and controls expression of cone opsin and maturation genes (8). The Thrb gene and thyroid hormone promote opsin expression and sensitivity to medium-long wavelengths in diverse vertebrate species (9–14). Human THRB mutations have been associated with impaired responses to long-wavelength stimuli (15, 16), monochromacy (17), cone dystrophy (18), and in retinal organoid cultures, altered opsin expression (14).

The provision of thyroid hormone for cones requires careful regulation. Insufficient thyroid hormone in the circulation impairs cone opsin diversity and maturation (10, 19–21) whereas elevated thyroid hormone results in cone loss in rodent models (22–24) and has been associated with age-related macular degeneration (AMD) in human populations (25, 26). The thyroid gland biosynthesizes triiodothyronine (T3) the primary active form of thyroid hormone, and thyroxine (T4), a more abundant precursor form of the hormone, which are released into the circulation. Within target tissues, transporters mediate trans–plasma membrane uptake and passage of T4 and T3 (27, 28). Within some tissues, including the retina, the level of T3 can be amplified by conversion of T4 to T3 by type 2 deiodinase whereas both T4 and T3 may be depleted by type 3 deiodinase (29). Little is known of the cellular route that transports thyroid hormone to cones which are shielded behind the blood–retina barrier, a cellular interface that transports solutes to the retina and supports photoreceptor homeostasis (30). An outer barrier is formed by the retinal pigment epithelium (RPE), a tight junctional barrier between the choroidal capillaries and the photoreceptor layer of the retina (31). An inner barrier is formed by a tight junctional barrier of endothelial cells of the inner retinal capillaries.

The necessity for regulated uptake of thyroid hormone by tissues was indicated by finding mutations in human MCT8 (monocarboxylate transporter 8, SLC16A2), a membrane transporter for thyroid hormone in a syndrome of psychomotor retardation (32–34). However, within the cellular complexity of a tissue, transport routes to specific cells are unclear. Numerous candidate transporters belonging, for example, to organic anion and amino acid transporter families, have been identified with differing affinities for T4, T3, and other substrates (28, 35). In the brain and cochlea, additional controls have been proposed at the blood–brain (27, 36, 37) and blood–labyrinth (38) barriers, respectively.

The importance of thyroid hormone for cone function and its relevance to retinal disorders led us to investigate a cellular transport route of thyroid hormone to cones. We report that Mct8-knockout (KO) mice display cone opsin, transcriptome, and functional defects indicative of a hypothyroid retinal state. Single cone analyses reveal partial loss of M cone identity and a shift to an S identity indicating that MCT8, like TRβ2, controls cone diversity. The role of the RPE was established by RPE-specific deletion of MCT8 and reciprocal experiments to reexpress MCT8 in the RPE in Mct8-KO mice. These findings suggest that MCT8 in the RPE controls thyroid hormone signaling to promote cone diversity and maturation.

Results

Developmental Expression of Mct8 in the RPE and Eye Tissues.

We determined the expression pattern of Mct8 in reporter mice that express a lacZ knockin from the Mct8 locus (39). As pigmentation in the RPE would obscure colorigenic signals generated by the β-galactosidase product of lacZ, the Mct8 reporter allele was crossed onto a nonpigmented, albino C57BL/6J strain. The most prominent lacZ signals were detected in the RPE, with expression beginning as early as embryonic day 16.5 (E16.5) and persisting throughout development into adulthood (Fig. 1A). Expression appeared uniform over the RPE. In the early neonatal period, lacZ was also expressed transiently in a diffuse pattern in the immature retinal layers. During maturation after ~postnatal day 7 (P7), signals in the retina decreased and became restricted to Müller glia and minor cell populations in the inner retina (Fig. 1A and SI Appendix, Fig. S1 A–C). Mct8 was also detected in the embryonic ciliary margin zone and the nonpigmented epithelium of the adult ciliary body, a part of the blood–aqueous barrier that controls secretion of aqueous humor (Fig. 1B). Mct8 is X-linked and similar expression patterns were detected in males or females carrying the lacZ reporter.

Fig. 1.

Expression of Mct8 in the RPE and other eye tissues. (A) Mct8 expression in the RPE throughout development detected by histochemistry (blue) for β-galactosidase encoded by lacZ in the Mct8 gene. Transient expression occurs in precursor cells in the neonatal retina. Cryosections; ≥3 eyes. (B) Mct8 expression in the embryonic ciliary margin zone, adult ciliary body, and Müller glia in the INL. (C) Expression of candidate thyroid hormone transporters in purified RPE cells and a different set of transporters in the retina, detected by RNA-seq analysis. Transporter RNA levels did not differ between superior and inferior areas. Means ± SD, P > 0.11; unpaired t test; groups, 5 RPE samples, each representing 3 mice (6 eyes); 4 retinal samples from 4 mice, 2- to 3-mo-old males. In the retina, countergradients of cone M and S opsin RNA are detected over the superior-inferior plane (**P < 0.01). (D) Immunostaining for MCT8 (magenta) and Ezrin (apical RPE marker, cyan) detected MCT8 at the apical membrane of RPE cells. The dashed white line outlines the RPE basal surface. Specific signal was absent in Mct8-KO. Six eyes per genotype. (E) Diagram of a posterior eye section with cone opsin gradients indicated. Boxes, superior and inferior areas of RPE or retina taken for analysis. Abbreviations: CMZ, ciliary margin zone; GCL, ganglion cell layer; INL, inner nuclear layer; MG, Müller glia; NBL, neuroblast layer; ONL, outer nuclear layer; OS/IS, outer and inner segments; RPE, retinal pigment epithelium.

The expression of Mct8 in the RPE was corroborated by RNA-sequencing (RNA-seq) of purified RPE cells which revealed prominent expression of several candidate thyroid hormone transporter genes including Mct8, Slco1a4 (Oatp1a4),Slc7a5 (Lat1), Slc7a8 (Lat2), and Slco1c1 (Oatp1c1) (Fig. 1C). In contrast, RNA-seq of retinal tissue (40) detected expression of a distinct set of candidate transporters with Slco4a1 prominent but relatively low Mct8 RNA levels. Although cone photoreceptors are a primary target of thyroid hormone, maturing cones lacked Mct8 expression when analyzed by lacZ histochemistry and had low RNA levels in single cone datasets (40), suggesting that MCT8 is not a major transporter in cones (Fig. 1 A and C). In mice, cone diversity is reflected in the countergradients of opsins observed over the vertical plane of the retina with M opsin predominant in cones in superior areas and S opsin in inferior areas (2, 41, 42). Therefore, we analyzed RPE cells from superior and inferior locations to determine whether Mct8 or other transporters exist in gradients that correlate with opsin gradients in cones. However, no obvious gradients were detected in RPE cells for Mct8 or other transporter genes examined (Fig. 1C).

As the RPE is a polarized epithelial monolayer, transporters may segregate to basolateral or apical membranes of the cell where they could control transport between the choroidal vascular bed and the RPE or between the RPE and the retina, respectively. Immunostaining detected MCT8 signals primarily at the apical membrane of RPE cells which extend microvilli around the photoreceptor segments. The apical location of MCT8 immunofluorescence in the RPE was confirmed by colocalization with Ezrin, a marker of the apical membrane (43) (Fig. 1D) but not with MCT3 (SLC16A8), a marker of the basolateral membrane (44) (SI Appendix, Fig. S1D). Specific signals were absent in Mct8-deficient (Mct8-KO) mice. MCT8 signal in the RPE has also been reported using another antibody although without negative (Mct8-KO) control data (45). We also detected a segregated location of MCT8 at the basolateral membrane of the nonpigmented epithelium of the ciliary body (SI Appendix, Fig. S1). The apical location in the RPE suggests that MCT8 mediates transport between the RPE and the retina.

Hypothyroid-Like Opsin and Retinal Gene Expression in MCT8 Deficiency.

To investigate a requirement for thyroid hormone transporters for cone differentiation, we analyzed opsin expression in Mct8-KO mice. In mice, M and S opsin levels normally rise and acquire countergradients over the retina at juvenile stages spanning the period of eye opening (~P13) during postnatal weeks 2 to 3 (8). Compared to wild-type (WT) mice at P14, Mct8-KO mice had reduced M opsin immunofluorescence in cones in all retinal areas and elevated S opsin in superior areas (Fig. 2 A and B). These abnormalities persisted at adult ages when M opsin rose somewhat but remained below normal. The opsin changes were supported by qPCR analysis, which showed decreased Opn1mw (M) RNA and increased Opn1sw (S) RNA, mainly in superior areas (Fig. 2C). These results suggest a partial shift from an M to an S opsin identity, particularly in superior areas, representing a moderate version of the extreme M to S opsin shift in TRβ2-deficient mice (6). The similar phenotype of MCT8 or TRβ2 deficiency was supported by qPCR analysis of other reported TRβ2-dependent cone genes (8) (SI Appendix, Fig. S2). For example, Arr3, Kcne2, Pde6c, and Fzd10 all showed decreased expression in Mct8-KO mice as has been reported for TRβ2-deficient mice (8).

Fig. 2.

Opsin changes and hypothyroid-like retina in Mct8-KO mice. (A) Cone opsin immunostaining on eye cryosections. Mct8-KO mice have decreased M and increased S opsin, particularly in superior areas, diminishing the countergradients over the retina. (B) Quantification of immunofluorescence. M opsin differences persist but are milder at P60 than P14. Means ± SD, *P < 0.05; **P < 0.01, unpaired t test; groups, P14, 3 mice; 2 mo, 6 mice; 2 sections per retina area. (C) Expression of opsin genes and Hr and Dio3 indicator genes of T3-signaling status determined by qPCR of retinal RNA. Means ± SD, *P < 0.05, **P < 0.01; unpaired t test; n = 6 eyes from 6 mice, 2 to 3 mo of age. (D) Partial recovery of cone gene RNA expression in Mct8-KO mice by treatment with T3 in drinking water; qPCR analysis. Means ± SD, one-way ANOVA with Tukey's post hoc test; *P < 0.05, **P < 0.01; n = 5 or 6 male mice. (E) Heatmaps showing that retinal transcriptome changes in Mct8-KO mice resemble changes in hypothyroid Tshr-KO mice. RNA-seq, groups, 3 or 4 mice (each pooled from designated areas of 2 retinas). Means, P < 0.05, unpaired t test; fold change > 1.5; cutoff 2 cpm. (F) Hypothyroid-like expression changes in both Mct8-KO and Tshr-KO genotypes for example cone genes (Opn1mw, Opn1sw, Pde6c, Rdh12) and Hr and Dio3 indicator genes of T3-signaling status. RNA-seq, means ± SD, one-way ANOVA, with Tukey's post hoc test; *P < 0.05, **P < 0.01.

To support the proposal that the cone phenotype in Mct8-KO mice reflects impaired thyroid hormone signaling in the retina, we first analyzed Hr and Dio3, which can serve as marker genes of the thyroid hormone response status of a tissue (46). In the retina, Dio3 (type 3 deiodinase) is expressed in precursor cells and then decreases to minimal levels at mature stages (22). Hr (hairless, lysine demethylase and nuclear receptor corepressor) is expressed more widely at low levels in the retina (8). In various tissues, expression of both genes typically increases in response to elevated T3 signaling. Although Mct8-KO mice have moderately elevated circulating levels of T3 during postnatal development (38, 39), expression of Hr and Dio3 did not increase (Fig. 2C) but instead tended to decrease suggestive of a relatively hypothyroid response status of the retina. To support further the basis of the phenotype as a result of inefficient T3 signaling, we administered exogenous T3 to Mct8-KO mice from P5 to P60 (Fig. 2D). We predicted a rescue of cone gene expression based on the assumption that elevated T3 levels would bypass MCT8-mediated transport routes (38). T3 treatment partly recovered Opn1mw and suppressed Opn1sw RNA levels in the superior retina compared to untreated Mct8-KO mice, as shown by qPCR analysis (Fig. 2D). T3 treatment partly corrected expression of other Mct8-regulated genes including Arr3, Kcne2, and Hr, consistent with the Mct8-KO interfering with transport of T3 rather than other hypothetical substrates.

A hypothyroid-like tissue status in the Mct8-KO was also supported by finding that many changes in the retinal transcriptome recapitulated those of systemically hypothyroid Tshr-KO mice (Fig. 2E). Tshr-KO mice have very low T4 and T3 in the circulation due to a defective thyroid gland which results in decreased M opsin and increased S opsin levels (19). In Mct8-KO mice, RNA-seq detected 143 and 228 changed genes in the superior and inferior retina, respectively (P < 0.05, fold change > 1.5). These genes represented ontology categories including visual perception and membrane functions, consistent with defects in maturation of the retina (SI Appendix, Fig. S2). Many genes showed parallel expression changes in Mct8-KO and Tshr-KO mice consistent with a hypothyroid retinal status in both genotypes. Changes in expression of cone genes (e.g., Opn1mw, Opn1sw, Arr3, Kcne2, Pde6c, Rdh12) (Fig. 2F) and T3 signaling indicator genes (Hr, Dio3) suggested a hypothyroid-like response status of the Mct8-KO retina.

Impaired Cone Diversity in Mct8-KO Mice.

To support the proposal that MCT8, like TRβ2, regulates cone diversity, we performed high-resolution single cone analyses since bulk retinal analysis gives limited insights into the small cone population (which constitutes only ~2% of retinal cells). Cones were labeled using a Thrb^b2cre driver and then isolated individually from superior and inferior retinal areas, representing M- and S-dominant cone populations, respectively. To allow detailed analysis of the subtle distinctions underlying cone diversity, we generated RNA-seq libraries with substantial average depth (~9 million reads) (8). For control mice, a cell distribution (tSNE) plot yielded dispersed clusters representing M- and S-dominant cones (Fig. 3A), as expected (8). However, for Mct8-KO mice, these clusters were intermingled indicating diminished distinctions between cone types. Analysis of opsin genes, which are major contributors to diversity, revealed that Opn1mw was reduced whereas Opn1sw was elevated, particularly in the superior retina (Fig. 3B), reflecting a partial form of the extreme M to S opsin shift that occurs in TRβ2-deficient mice (6). The more emphatic change evident in M-dominant cones (in superior areas) is consistent with the view that an M opsin identity is acquired under control by T3, whereas S-dominant cones (in inferior areas) are closer to a default S opsin identity with less dependence on T3 (6, 8).

Fig. 3.

Diminished cone diversity in Mct8-KO mice revealed by single cone analyses. (A) Cell distribution (tSNE) plots for control mice reveal distinct clusters of M- and S-dominant cones from superior and inferior areas, respectively. In Mct8-KO mice, clusters intermingle, indicating loss of diversity. RNA-seq of 172 cones, groups of 42 to 44 cells (from 6 or 7 male mice; 2 to 3 mo of age). (B) Disrupted expression in the Mct8-KO of cone gradient (Left) and nongradient (Right) gene examples; a nonchange gene (Tulp1) is included. Cones ordered by increasing expression level in each group; crosses, medians. **P < 0.01, unpaired t test. (C) Heatmaps showing similar patterns of disrupted gradient gene expression in Mct8-KO and TRβ2-deficient mice based on 31 gradient gene examples identified in control cones in the present study (criteria: superior versus inferior, P < 0.01; fold difference > 1.2, cpm > 20). (D) Summary box plots of expression changes for all gradient (182) and nongradient response (263) genes in Mct8-KO mice analyzed in this study. Gradients are diminished in both superior and inferior bias categories. Box plots show Z score averages (normalized); crosses, means.

Apart from opsins, altered expression of other gradient genes such as Arr3 (superior bias) and Ccdc136 (inferior bias) contributed to the net outcome of poorly distinguished cone types in Mct8-KO mice (Fig. 3B and SI Appendix, Fig. S3). A comparison of 31 representative gradient genes revealed generally similar although not identical trends of disrupted expression in Mct8-KO and TRβ2-deficient mice (Fig. 3C), supporting the proposal that MCT8 regulates the TRβ2-mediated pathway for diversity. The net diminution of gradients in Mct8-KO mice was further supported by analysis of combined averages of all gradient genes identified in this study (182 genes; P < 0.01, fold difference > 1.2, threshold > 20 cpm) (Fig. 3D).

As in TRβ2 deficiency (8), the Mct8-KO also disrupted expression of nongradient cone genes, indicative of wider functions in cone maturation (Fig. 3D). Representative nongradient genes (e.g., Rd3L, Rdh12, Rtbdn) (Fig. 3B) were similarly altered in Mct8-KO and TRβ2-deficient mice (SI Appendix, Fig. S3), indicating that MCT8, like TRβ2, regulates a network of genes for cone maturation as well as cone diversity. These genes represented categories including visual perception, membrane, and synapse functions, consistent with cone maturation.

Impaired Cone Migration and Electroretinogram (ERG) Response.

During development, cone nuclei migrate to their mature location at the outer zone of the outer nuclear layer (ONL) by ~P12 (47). In mice lacking TRβ2 or type 3 deiodinase (Dio3), cone nuclei can be misplaced in the inner retina (6, 22). Using staining for Arr3 to label cone cells (Fig. 4 A and B), we observed that in WT mice at P14, 96% of cone nuclei were located in the outer one-third zone of the ONL. In contrast, in Mct8-KO mice, almost half of the cone nuclei remained in inner zones, indicating impaired migration. In adult Mct8-KO mice, cone nuclei numbers rose in the outer zone, but 14% still remained in inner zones. Overall cone numbers were not changed in Mct8-KO mice.

Fig. 4.

Cone migration and functional impairment in Mct8-KO mice. (A) Mislocated cone nuclei in Mct8-KO mice revealed by immunostaining for Arr3 at P14 and P90. Impaired migration is marked at P14 when most cones normally reach the outer zone of the outer nuclear layer (mislocated nuclei examples, blue arrows). (B) Counts in ONL zones showing decreased proportions of cone nuclei in the outer zone in Mct8-KO compared to WT. Means ± SD; **P < 0.001; unpaired t test. Groups, 4 retinas from 4 male mice (5 views per retina). Cone nuclei and total cone numbers counted in Arr3-stained, 12 µm thick cryosections. Total cone numbers, P > 0.51; unpaired t test, 4 mice at P90 (2 views per area). (C) Cone b-wave responses in photopic ERG analyses with stimuli at 515 nm and 367 nm, optimal wavelength regions for mouse M and S opsins, respectively (Top row). Mct8-KO mice have reduced responses to both stimuli. Rod b- and a-wave responses in scotopic analyses after dark adaptation (Bottom row). Means ± SEM; *P < 0.05, **P < 0.01, unpaired t test; 8 WT, 9 KO (rods), 9 WT, 12 KO (cones), males, 10 to 12 wk of age.

We assessed cone function in photopic ERG analyses. Mct8-KO mice had reduced responses to medium wavelength (515 nm) stimuli, consistent with the decreased expression of M opsin, indicating that MCT8, like TRβ2, promotes sensitivity to green regions of the light spectrum (Fig. 4C). Mct8-KO mice also had slightly reduced responses to short wavelength (367 nm) stimuli. This finding differs from the normal or modestly elevated responses to short-wavelength stimuli reported for TRβ2-deficient mice (22) and may reflect not only S opsin levels but also a wider influence of T3 in the retina in modifying the photopic responses. The scotopic ERG of rod photoreceptors was also impaired in Mct8-KO mice, suggesting that MCT8 regulates rods and potentially wider functions in the retina.

Cone Opsin Expression Shifts Following RPE-Specific Depletion of MCT8.

To test the proposal that MCT8 in the RPE controls thyroid hormone signaling to cones in the retina, we depleted Mct8 selectively in the RPE using a Cre/loxP conditional approach with a floxed (Mct8^f) allele and then assessed consequences on cone gene expression. To delete the Mct8^f allele, we used an RPE-specific Best1-Cre transgene that begins to express Cre at ~P6 (48), when M opsin expression begins in cones (8). To improve deletion efficiencies in RPE cells, Best1-Cre was backcrossed onto a 129S1 background (48), resulting in Cre-dependent activation of a Rosa26^Ai6 reporter in ~93% of RPE cells. Best1-Cre specifically deleted the Mct8^f allele in the RPE but not retina as demonstrated by genomic DNA analysis (Fig. 5A). As Mct8 is an X-linked gene, we maximized depletion in female progeny by introducing Best1-Cre into mice with one floxed and one constitutive null Mct8 allele (i.e., Mct8^f/null;Cre, abbreviated as f/null;Cre in Fig. 5). Depletion in males was achieved by introducing Best1-Cre into mice with an Mct8^f allele on the single X chromosome (i.e., Mct8^f/y;Cre, abbreviated as f/y;Cre in Fig. 5). In the RPE, ~90% and ~73% of Mct8 coding transcripts were inactivated in Mct8^f/null;Cre females and Mct8^f/y;Cre males, respectively (Fig. 5B).

Fig. 5.

Cone diversity disrupted by RPE-specific deletion of Mct8. (A) Genomic DNA analysis of the floxed region (f allele) of Mct8 demonstrating Best1-Cre-mediated deletion (del) in the RPE but not retina in Mct8^f/y mice. (B) Reduced expression of Mct8 RNA in RPE cells resulting from RPE-specific deletion by Best1-Cre. Analysis of RNA by qPCR of the floxed coding exon 3 of Mct8. (C) Curve plots showing dynamic response of M (decreased) and S (increased) opsin RNA levels associated with reduced Mct8 RNA levels in the RPE. The opsin expression shifts are consistent with partly diminished cone diversity resulting from depletion of MCT8 in the RPE. Dose–response best fit curves plotted using Prism v10. (D) RNA-seq analysis of retina of one matched, representative litter at P14 showing opsin genes, Hr indicator gene of a hypothyroid-like status, gradient gene Kcne2, and a nonchanged control gene, Tulp1. Means ± SD, **P < 0.01, unpaired t test; n = 3 f/null and 3 f/null;cre mice (1 or 2 retinas per mouse). (E) Reduced M opsin staining in mice with RPE-specific depletion of Mct8. Representative eye cryosections at P14. Arrows, Cre-positive nuclei in RPE cells; PNA, cone segments; Hoechst dye, nuclei. (F) Quantification of opsin immunofluorescence. Means ± SD; **P < 0.01; unpaired t test, n = 4 mice (4 retinas, 3 views per retina).

Best1-Cre-mediated deletion of the Mct8 gene depleted Mct8 RNA in the RPE and remarkably resulted in decreased M and increased S opsin RNA levels in the retina with the most marked changes in superior areas, when assessed by qPCR at P14 (Fig. 5C). A curve plot of individual samples revealed a dynamic, nonlinear relationship of opsin gene expression relative to Mct8 expression in the RPE: The lower the level of Mct8 RNA in the RPE, the greater the shift of opsin gene expression in the retina with a particularly sharp drop of Opn1mw expression at the lowest Mct8 levels. These opsin gene expression changes resembled those in the total Mct8-KO with a partial shift of M to S opsin cone identity. Expression of the Hr indicator gene decreased in all retinal areas of the conditional KO suggestive of a hypothyroid-like response status in the retina due to loss of MCT8 in the RPE. RNA-seq of the retina further confirmed that RPE-specific depletion of Mct8 shifted the expression of opsin genes from M to S, decreased expression of Hr indicating a hypothyroid-like response status, and altered several other genes, as shown in a representative analysis of matched individuals in a single litter at P14 (Fig. 5D) and in grouped data of several litters (SI Appendix, Fig. S4). Most gene expression changes mirrored those in the total Mct8-KO, although the degree of change was subtler, presumably reflecting the lack of depletion in ~10% of RPE cells. These findings highlight a dynamic sensitivity of opsin gene expression to Mct8 levels in the RPE, which was further supported by immunostaining of eye sections (Fig. 5 E and F). These results indicate that expression of opsins in the retina depends on MCT8 in the RPE.

Recovery of M Opsin Induction by Reexpression of MCT8 in the RPE.

In a gain-of-function experiment, we demonstrated that reexpression of MCT8 in the RPE in Mct8-KO mice is sufficient to recover M opsin induction (Fig. 6 A and B). An expression vector for MCT8 with a V5 epitope tag for visualization was introduced specifically into RPE cells in neonatal Mct8-KO mice by directed electroporation using the electrode polarity shown in Fig. 6B. A coelectroporated ZsGreen marker allowed independent identification of electroporated cells, which confirmed targeting into the RPE rather than the retina. In a parallel experiment, we electroporated a vector encoding G221R-MCT8, a mutant protein with disrupted transport activity but not subcellular localization, as identified originally in affected humans (49). The ectopically expressed WT and mutant MCT8 proteins localized appropriately at the apical membrane of RPE cells as shown by staining for the V5 tag.

Fig. 6.

Sufficiency of MCT8 in the RPE for induction of M opsin. (A) Ectopic expression of WT or mutant (G221R) Mct8 in the RPE following electroporation of expression vectors in neonatal mice with analysis at P14. A V5 tag fused on MCT8 allows visualization. Images show WT and mutant MCT8 proteins with an apical location in a representative RPE cell whereas an independent, coelectroporated ZsGreen marker extends to the basal side of the RPE cell. (B) After subretinal injection, electroporation drives vector DNA into the RPE rather than the retina using the electrode polarity shown. (C) Ectopic expression of WT-MCT8 in the RPE induces M opsin in Mct8-KO mice at P14. G221R- MCT8, lacking transport activity, fails to induce M opsin. Mct8-KO control mice (with empty vector) express minimal M opsin at P14. Hoechst dye, nuclei; PNA, cone segments. (D) Quantification of M opsin immunofluorescence at the electroporated RPE region. Means ± SD; **P < 0.001; one-way ANOVA with Tukey's post hoc test; n = 5, 7, and 6 mice for control empty vector, WT-, and G221R-MCT8 groups, respectively; three sections per retina. (E) Simplified model suggested for control of cone differentiation by MCT8 in the RPE. Thyroid hormone (TH) exits fenestrated vessels in the choroid and is taken up by undefined basolateral transporters (gray) in the RPE. MCT8 mediates apical transport of thyroid hormone from the RPE to cones where it stimulates transcriptional activity of TRβ2 to promote cone differentiation. A cone transporter is undefined. For simplicity, thyroid hormones T4 (precursor) and T3 (active ligand) are not depicted individually but both may be transported by MCT8.

Reexpression of WT MCT8 in the RPE recovered M opsin expression in Mct8-KO mice compared to Mct8-KO control mice in which M opsin expression remained notably reduced at P14 (Fig. 6 C and D). M opsin was induced in cones in proximity to the electroporated zone of the RPE (indicated by ZsGreen marker). Mutant G221R-MCT8 did not recover M opsin, indicating the necessity for MCT8 with transport activity in the RPE for M opsin expression. Reexpression of MCT8 did not consistently suppress S opsin for reasons that include technical limitations as the electroporation zone is typically near the central retina whereas the most sensitive area for suppression of S opsin is in superior regions (Figs. 2 and 3). These results establish the sufficiency of MCT8 in the RPE for promoting M opsin induction during cone maturation. We also investigated in mosaic RPE areas whether MCT8 in an RPE cell influences M opsin expression in cones distal to that RPE cell since transported thyroid hormone could potentially diffuse in the subretinal space to nearby cones. SI Appendix, Fig. S5 shows that at the boundary of an RPE area electroporated with WT-MCT8, M opsin was detected in cones adjacent to a given electroporated RPE cell and at limited, lateral distances (<60 µm) from that RPE cell, consistent with MCT8 transporting a diffusible signal. This suggestion was also supported by finding widespread M opsin expression in Mct8^+/− mice (SI Appendix, Fig. S5). As a result of random X-inactivation of one of the two Mct8 alleles in heterozygous females, the RPE is a mosaic of MCT8[+] and MCT8[−] cells. M opsin was detected relatively broadly under both MCT8[−] and MCT8[+] RPE cells in Mct8^+/− mice, again consistent with MCT8 transporting a diffusible signal.

Discussion

Cone Diversity and Maturation Regulated by MCT8 in the RPE.

Cone photoreceptor differentiation requires rising thyroid hormone levels in the circulation during development (9, 10, 19) but cones lack direct exposure to this general pool of hormone because of the blood–retina barrier. Our genetic analyses of the MCT8 thyroid hormone transporter suggest an intermediate step of paracrine-like control of thyroid hormone signaling by the RPE which forms an interface with the photoreceptors in the retina. The RPE constitutes the outer blood–retina barrier and is intricately involved in photoreceptor function through recycling of chromophore required for phototransduction, removal of disc tips during photoreceptor segment renewal, and transport of solutes. A role in thyroid hormone transport is consistent with the specialization of RPE cells for transport of electrolytes, oxygen, and nutrients between the circulation and photoreceptors (31, 50).

We propose that MCT8 contributes to a dynamic mechanism that balances the need for T3 for cone differentiation against premature or excessive exposure that causes cone death (22, 24). At embryonic stages, RPE cells lack strict control of permeability (31, 51) but the retina at this stage is protected against unregulated exposure to T3 by type 3 deiodinase, a thyroid hormone-inactivating enzyme that safeguards cone survival (22). Later, as tight junctions in the RPE mature, MCT8 in the RPE would become a critical mediator of thyroid hormone signaling for cones. A plausible model is that MCT8 in the RPE regulates a key step in transport of thyroid hormone from the fenestrated choroidal capillaries to TRβ2 in cones (Fig. 6E). MCT8 has high specificity for iodothyronines (i.e., T3, T4) (27, 52). Other candidate transporters in the RPE (Fig. 1C) have a broader spectrum of substrates (for example, amino acids), a lower preference for iodothyronines (53, 54) or may have a basolateral location where they could not substitute for MCT8 at the apical side. The RPE presumably requires basolateral transporters for uptake of T3 and T4 from the choroid. Candidates include SLCO1C1 (53) or SLC7A5 (Lat1) and SLC7A8 (Lat2) which may have a basolateral location in other epithelia (54) (Fig. 1). A functional transporter in cones is undefined but it may be expected to import T3 efficiently since cones express little or no type 2 deiodinase to convert T4 precursor into T3 ligand for the receptor (29, 40).

The evidence indicates that MCT8 in the RPE transports thyroid hormone although a minor involvement of other, undefined substrates cannot be excluded. The proposed role is consistent with the remarkable specificity of MCT8 for T4 and T3 in vitro (27) and with our experimental findings: i) the partial shift from an M to S opsin cone identity in the Mct8-KO resembles the specific, hallmark phenotype caused by deletion of TRβ2, a T3 receptor that controls cone diversity (Figs. 2 and 3); ii) administration of T3 partly restores opsin expression patterns in Mct8-KO mice, implicating deficient T3 signaling in the phenotype (Fig. 2D); iii) the retinal transcriptome in Mct8-KO mice is hypothyroid-like and resembles that caused by systemic hypothyroidism (Fig. 2 E and F). These and other observations also argue against a nonspecific structural role of MCT8 in the RPE that might lead to nonspecific photoreceptor defects. First, Mct8-KO mice do not show an obvious loss of RPE integrity, as the RPE retains appropriately located markers of the basolateral membrane, apical membrane, and tight junctions; also, the retina does not display overt signs of reactive gliosis or degeneration (SI Appendix, Fig. S6). Second, the retinal transcriptome in the Mct8-KO includes specific changes in cone gene expression similar to changes in hypothyroid or TRβ2-KO mice rather than general defects in all photoreceptors (for example, as caused by Crx mutation) (55) (SI Appendix, Fig. S6). Third, M opsin expression is only recovered in the Mct8-KO by reexpression in the RPE of WT MCT8 but not by an MCT8 protein with a normal subcellular location but defective transport activity, implying a need for a specific transport function (Fig. 6).

The Outer Blood–Retina Barrier, a Route for Hormonal Signaling to Photoreceptors.

Our study implicates the outer blood–retina barrier (i.e., the RPE) in thyroid hormone signaling for cones. The outer barrier route for transport of solutes to the retina is widely conserved: All vertebrates have a choroidal vasculature and RPE, whereas many species including some mammals (such as guinea pig) largely lack inner retinal vessels (56). As thyroid hormone regulates cone opsin expression in diverse vertebrate species, including species lacking inner retinal vessels such as the trout (57), the RPE may be the primary route of transport of thyroid hormone to cones. The RPE is also thought to be a primary route of glucose transport and metabolic control for photoreceptors (58).

The shift of M to S identity in cones in Mct8-KO mice is less extreme than that caused by TRβ2 deficiency (6) implying the existence of other limited or compensatory routes of transport of T3 to cones. This might involve Müller glia which extend projections spanning from the inner retinal capillaries to the photoreceptors. Müller glia express Mct8 (Fig. 1) and Dio2, encoding a T3-amplifying deiodinase (29), suggesting a possible role of these cells in T3 signaling in the retina. Interestingly, the RPE and Müller glia have been invoked in transport routes for vitamin A-derived chromophore for cone phototransduction (59). Other transporters perhaps in cell types that normally make little contribution to T3 signaling for cones might partly compensate for loss of Mct8. In this scenario, the outcome of RPE-specific deletion of MCT8 may underestimate the role of MCT8 in the RPE.

MCT8 expression is also detected in RNA-seq datasets of human RPE raising the possibility of related roles for MCT8 in human photoreceptor differentiation (SI Appendix, Fig. S1). Human MCT8 mutations result in neurological and motor disabilities (27, 32, 33) but retinal function has not been examined systematically. Our findings in mice suggest that ocular examination is merited in human cases with MCT8 mutations. Recent studies report that in zebrafish larvae with an mct8 mutation, gene expression changes included opn1mw2 (60), and in the chick embryo, knockdown of Mct8 in retinal progenitor cells subtly disturbed neuronal proliferation at the embryonic stages examined (61). These observations suggest possible functions for MCT8 in retinal cells at least in nonmammalian species. RPE function was not examined in these studies.

In the Mct8-KO, as in hypothyroid (Tshr-KO) mice, the range of retinal genes with disrupted expression is consistent with T3 regulating varied cell types in eye tissues (7, 40). The altered ERG responses of rods as well as cones suggest that rod photoreceptors, which mediate dim light vision, are also somewhat sensitive to loss of MCT8. Indeed, we detected altered expression of genes that are shared by rods and cones (e.g., Grk1, Rdh12, SI Appendix, Fig. S2). Rod impairment has been reported in mice lacking thyroid hormone-metabolizing deiodinases and in mice treated with T3 (24, 29).

An implication of our study is that damage to the integrity of the RPE such as occurs in AMD (30, 62), could deregulate access for thyroid hormone to the retina, thereby exacerbating impairment of cones and vision. Elevated T4 has been associated with AMD in human populations (25, 26) and could be a susceptibility factor for cone defects. Conversely, hypothyroid conditions might alleviate some consequences of a compromised RPE. At mature ages, opsin patterning (21) and cone survival (23, 24) remain sensitive to thyroid hormone changes, such that MCT8 in the RPE may also support photoreceptor maintenance during aging. Finally, retinal organoid cultures offer newer approaches to study photoreceptor differentiation (14). Our findings suggest that coculture of retinal organoids with RPE cells may be necessary to mimic the nuances of cellular control underlying the diversity and function of cones in vivo.

Materials and Methods

Mouse Genetic Strains.

The lacZ reporter allele of Slc16a2 (Mct8⁻) is also a KO and was used for phenotype studies. This allele was originally on a C57BL/6N x NMRI background (39) then at NIH was backcrossed for two generations onto a C57BL/6J background (Jackson Laboratory, RRID:IMSR-JAX: 000664). A retinal degeneration Crb1^rd8 mutation present in the C57BL/B6N background (63) was removed from all strains used during crossing. To detect lacZ expression in pigmented tissues, the Mct8⁻ (lacZ) allele was crossed onto a C57BL/6 albino (Tyr^−/−) background (Jackson Laboratory, RRID:IMSR-JAX: 000058). To label single cones for RNA-seq, the Mct8⁻ allele was combined with a Thrb^b2Cre cone-selective driver (8) and Rosa26^Ai6 reporter allele (RRID:IMSR-JAX:007906) (64). The Best1-Cre transgene strain C57BL/6-Tg(Best1-Cre)1Jdun/J (48) (RRID:IMSR-JAX:017557) was backcrossed >5 generations onto a 129S1/SvImJ background (Jackson laboratories, RRID:IMSR-JAX:002448) to enhance deletion in higher percentages of RPE cells (48). For RPE-specific deletion, an Mct8 floxed allele (Mct8^f, B6Brd;B6N-Tyrc-Brd-Slc16a2tm1c [KOMP]Wtsi/Wtsi) was obtained from Wellcome Trust Sanger Institute and Infrafrontier EMMA repository as cryopreserved sperm (EM: 12254; now listed as RRID:IMSR-EM:13413). For females, the Mct8^f allele was combined with a constitutive null Mct8 allele (Mct8^null, derived from the floxed allele by germline deletion). Experimental groups were generated by crossing Mct8^f/f females with Mct8^null/y males carrying Best1-Cre. Controls included analysis of Rosa26^Ai6 mice without the Cre transgene which lacked reporter expression in eye tissues. Genotyping was performed by PCR as reported in the references noted above except that the Mct8^f allele was genotyped with minor modification of the KOMP protocol (SI Appendix, Fig. S4). For thyroid hormone treatment, T3 (MilliporeSigma, cat# T6397) was provided to mutant pups (Mct8^−/− and Mct8^−/y) in the drinking water at 0.5 μg/mL from P5 to P60, as described (38). Animal studies followed protocols approved by the Animal Care and Use Committee of the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) at the NIH.

Histochemistry, Immunostaining, and Quantification.

Cryosections in the sagittal (vertical) plane at 12 µm thickness were prepared from eyes fixed in 2% paraformaldehyde (PFA) for 1 h at room temperature (RT) for X-gal staining and immunostaining for Mct8 or in 4% PFA for 2 h at RT for other immunostains. Colorigenic staining with X-gal substrate was performed using a β-Galactosidase Reporter kit (MilliporeSigma, cat# GALS-1KT) following the manufacturer’s protocol. Images were captured with a Nikon Ti2 microscope. Immunofluorescence staining followed established procedures (40) with images captured using a Nikon-SoRa spinning disk or Zeiss LSM 780 confocal microscope. For Mct8 staining, to reduce nonspecific signals, the antiserum was first preadsorbed by incubation on cryosections of the retina from Mct8-KO mice overnight at 4 °C.

For multiplex labeling of Ezrin, MCT8, and MCT3, after indirect immunofluorescence staining for MCT8 and Ezrin, sections were incubated with anti-MCT3 antiserum labeled with the Flexable CoraLite Plus 647 kit for Rabbit IgG (Proteintech, cat #KFA003) according to the manufacturer’s protocol. No antibody cross-reactivity was observed. Quantification of opsin fluorescence was performed with NIS Element AR software (v5.41.00, Nikon), measuring signal intensity across cone segments in 750 µm long retinal fields of view. Cone nuclei, identified by Arr3 staining, were counted in outer, middle, and inner zones evenly divided over the ONL depth, measured on 12 µm cryosections using ImageJ (RRID:SCR-003070). Groups included 4 mice, with an average of five counted fields (250 µm lengths of ONL) per mouse. Cone densities were determined as total Arr3+ cone numbers per 250 μm length of ONL. SI Appendix, Table S1 shows antibody information.

ERG.

ERG analysis using an Espion Electrophysiology System (Diagnosys LLC) followed described procedures (22, 40). Briefly, 2- to 3-mo-old mice were anesthetized by intraperitoneal injection of ketamine (80 mg/kg) and xylazine (10 mg/kg). Cone responses were tested in photopic conditions using LED-generated stimuli at 515 nm and 367 nm wavelengths, for M and S opsin responses, respectively (65). Rod responses were tested in scotopic conditions after overnight dark adaptation.

RNA Analysis by Real-Time qPCR.

Retinal pieces (~3 mm² areas) or RPE cells were isolated from superior and inferior regions around the vertical mid-line of the eye (8). RPE cells were collected as described by Wang et al. (66). Briefly, the dissected sclera/RPE was incubated in RNAprotect Cell reagent (Qiagen, cat#76526) to dissociate RPE cells from the sclera. RPE cells were pelleted (685 × g) and lysed for RNA extraction. RPE cells from 6 areas (from 6 retinas) were pooled to yield 1 superior or 1 inferior sample, respectively. First-strand cDNA was synthesized using Superscript IV Reverse Transcriptase (ThermoFisher, Cat #18091050), qPCR performed with FastStart Universal SYBR Green Mix (MilliporeSigma, Cat #4913914001) then analyzed using the 2−ΔΔCT method with Hprt as control (40). SI Appendix, Table S2 shows primer pairs for genes.

Single Cell RNA-seq and Tissue RNA-seq Analyses.

Single cell analyses.

Cones labeled with a Thrb^b2Cre driver were isolated from dissociated pieces of superior and inferior retina of Mct8-KO (Mct8^-/y; Thrb^b2Cre/+; Rosa26^Ai6/+) and control (Mct8^+/y; Thrb^b2Cre/+; Rosa26^Ai6/+) male mice at 2 mo of age as described (8). Cones were isolated individually by micromanipulator to minimize contamination by rods (or pieces of rods) which are >30-fold more abundant than cones and represent ~70% of retinal cells in mice. Isolated cells were examined visually for cone morphology and integrity before RNA preparation, library synthesis, and quality control analysis (8). For high-resolution analysis, single cell libraries were obtained with average depth of ~9 million reads.

Pooled RPE cells or bulk retinal tissue analyses.

~10 ng of total RNA was used to synthesize cDNA with a SMART-Seq v4 Ultra Low Input RNA kit (Takara, Cat #634890). The cDNA was sheared into 200 to 500 bp fragments with an ME220 Focused-ultrasonicator (Covaris), then ~1.5 ng used to make libraries with a ThruPLEX DNA-Seq Kit (Takara, Cat# R400676). Libraries were sequenced on Illumina HiSeq-3000 or NovaSeq-6000 instruments (at NIDDK or NHLBI Genomics Cores, respectively). For each library, ~20 million single-end 50 base reads were collected, then converted using bcl2fastq into fastq files, aligned on GRCm38/mm10 reference genome with STAR (v2.7.10b). Reads were counted by FeatureCounts (subread v2.0.6) and normalized as counts per million mapped reads (cpm). To compare transcript abundance, data were quantified using Kallisto (v0.46.0) as transcripts per million reads (TPM).

RNA-seq Data Processing.

RNA-seq datasets were analyzed using R (v4.2.1) and gene ontology using DAVID (Database for Annotation, Visualization, and Integrated Discovery) (https://david.ncifcrf.gov/home.jsp) with a whole-genome background reference for selected terms and P < 0.05. For single cone datasets, differentially expressed genes in Mct8-KO and control cones were analyzed as described (8) using Student’s t test (P < 0.01) (172 cells, including superior and inferior groups). Most genes analyzed had average expression >20 cpm. For tSNE plots, the top 3,000 expressed genes in single cones were analyzed (Rtsne: T-SNE, https://github.com/jkrijthe/Rtsne) with cluster analyses using Rtsne (v0.16) (8). Comparisons were made with RNA-seq datasets available at GEO for human adult RPE (GSE159435), human fetal primary RPE cells (GSE36695), retina for Crx mutant mice (GSE65506) and single cones of control and TRβ2-KO mice (GSE203481).

Vector Construction and In Vivo Electroporation.

Injection of expression vectors into the subretinal space and electroporation followed reported procedures (67, 68) except that electrode polarity was reversed to drive DNA into the RPE instead of the retina. Similar numbers of male and female mutants (Mct8^−/− and Mct8^−/y) at P2 were electroporated. Expression vectors were constructed in plasmid pUB incorporating a WT mouse Mct8 cDNA with a C-terminal V5 tag for detection of expressed protein. An inactivating mutation in mouse MCT8 (c.457G>A) equivalent to a known human mutation G221R (c.661G>A), was introduced by PCR-based, site-directed mutagenesis. WT and mutant vectors were sequenced in entirety. As a marker, pUb-Zsgreen was used to identify electroporated cells.

Statistical Analysis.

Statistical analyses included unpaired two-tailed Student’s t tests (significance P < 0.05) and one-way ANOVA with Tukey's post hoc test (significance P < 0.05). GraphPad Prism version 10 (GraphPad Software) was used for all analyses including fitting of linear and nonlinear regression curves.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

Genomic datasets generated in this work are available at Gene Expression Omnibus (GEO), https://www.ncbi.nlm.nih.gov/geo (accession no. GSE254669) (69). All other data are included in the manuscript and/or SI Appendix.