Thyroid hormone signaling specifies cone subtypes in human retinal organoids

INTRODUCTION: Cone photoreceptors in the human retina enable daytime, color, and high-acuity vision. The three subtypes of human cones are defined by the visual pigment that they express: blue-opsin (short wavelength; S), green-opsin (medium wavelength; M), or red-opsin (long wavelength; L). Mutations that affect opsin expression or function cause various forms of color blindness and retinal degeneration. RATIONALE: Our current understanding of the vertebrate eye has been derived primarily from the study of model organisms. We studied the human retina to understand the developmental mechanisms that generate the mosaic of mutually exclusive cone subtypes. Specification of human cones occurs in a two-step process. First, a decision occurs between S versus L/M cone fates. If the L/M fate is chosen, a subsequent choice is made between expression of L- or M-opsin. To determine the mechanism that controls the first decision between S and L/M cone fates, we studied human retinal organoids derived from stem cells. RESULTS: We found that human organoids and retinas have similar distributions, gene expression profiles, and morphologies of cone subtypes. During development, S cones are specified first, followed by L/M cones. This temporal switch from specification of S cones to generation of L/M cones is controlled by thyroid hormone (TH) signaling. In retinal organoids that lacked thyroid hormone receptor β (Thrβ), all cones developed into the S subtype. Thrβ binds with high affinity to triiodothyronine (T3), the more active form of TH, to regulate gene expression. We observed that addition of T3 early during development induced L/M fate in nearly all cones. Thus, TH signaling through Thrβ is necessary and sufficient to induce L/M cone fate and suppress S fate. TH exists largely in two states: thyroxine (T4), the most abundant circulating form of TH, and T3, which binds TH receptors with high affinity. We hypothesized that the retina itself could modulate TH levels to control subtype fates. We found that deiodinase 3 (DIO3), an enzyme that degrades both T3 and T4, was expressed early in organoid and retina development. Conversely, deiodinase 2 (DIO2), an enzyme that converts T4 to active T3, as well as TH carriers and transporters, were expressed later in development. Temporally dynamic expression of TH-degrading and -activating proteins supports a model in which the retina itself controls TH levels, ensuring low TH signaling early to specify S cones and high TH signaling later in development to produce L/M cones. CONCLUSION: Studies of model organisms and human epidemiology often generate hypotheses about human biology that cannot be studied in humans. Organoids provide a system to determine the mechanisms of human development, enabling direct testing of hypotheses in developing human tissue. Our studies identify temporal regulation of TH signaling as a mechanism that controls cone subtype specification in humans. Consistent with our findings, preterm human infants with low T3 and T4 have an increased incidence of color vision defects. Moreover, our identification of a mechanism that generates one cone subtype while suppressing the other, coupled with successful transplantation and incorporation of stem cell-derived photoreceptors in mice, suggests that the promise of therapies to treat human diseases such as color blindness, retinitis pigmentosa, and macular degeneration will be achieved in the near future. ■

Pluripotent stem cells were well-maintained, and only cultures with minimal to no spontaneous differentiation were used for aggregation. To aggregate, cells were passaged in Accutase at 37°C for 13 min to ensure complete dissociation. Cells were seeded in 50 ul of mTeSR1 at 3,000 cells/well into 96-well ultra-low adhesion round bottom Lipidure coated plates (51011610, NOF). Cells were placed in hypoxic conditions (10% CO2 and 5% O2) for 24 hours to enhance survival. Cells naturally aggregated by gravity over 24 hours.
On day 10, aggregates were transferred to 15 mL tubes, rinsed 3X in DMEM (11885084, Gibco), and resuspended in BE6.2 with 100 nM SAG in untreated 10 cm polystyrene petri dishes. From this point on, media was changed every other day. Aggregates were monitored and manually separated if stuck together or to the bottom of the plate.
On days 13-16, LTR media with 100 nM SAG was added. Between days 11 and 16, retinal vesicles were manually dissected using sharpened tungsten needles. After dissection, cells were transferred into 15 mL tubes and washed 2X with 5 mLs of DMEM.
On days 16-20, cells were maintained in LTR and washed 2X with 5 mLs of DMEM, before being transferred to new plates to wash off dead cells.

CRISPR mutations
Cell line: All mutations were generated in H7 ESCs. Cells were modified to express an inducible Cas9 element. First, the puro-Cas9 donor plasmid was modified. The Puromycin N-acetyl transferase gene (puromycin-resistance gene) was replaced with Blasticidin S deaminase gene (blasticidin-resistance gene) using Xba I-Xho I restriction enzyme sites in the plasmid puro-Cas9 donor (58409, Addgene). This plasmid is referred to as the blast-Cas9 donor plasmid.
The integration of the targeting vectors into a previously genetically modified H7 human ESC line (59) was performed as follows: 0.25 million H7 Brn3B::tdTomato ES cells at 50% confluence were transduced using a DNA-In Stem kit (MTI-Globalstem, USA) with three plasmids (1ug each): Blast-Cas9 donor, M2rtTA donor (AAVS1-neo-M2rtTA: 60843, Addgene), and pSpCas9 (BB) plasmid (px459 v2.0: 62988, Addgene). gRNA sequences are listed below in the gRNA primer table. Cells were treated with Blasticidin (5ug/ml) and Geneticin (200ug/ml) for 5 days. Individual clones with both Blasticidin and Neomycin resistance survived, and were picked using sterile pipette tips and transferred to 96-well plates for clone identification. Positive clones carrying the correct insertion in both alleles were confirmed by PCR. Genotyping primers are listed below. Doxycycline induction was confirmed by qPCR and the verified clone iCas9 H7 Brn3B::tdTomato -24 was used for further experiments.

Primers for Cas9
Primer Name Primer Sequence Cloning gRNA plasmids: Plasmids for gRNA transfection were generated using pSpCas9(BB)-P2A-Puro plasmid modified from the pX459_V2.0 plasmid (62988, Addgene) by replacing T2A with a P2A sequence. gRNAs were cloned into the vector following the Zhang Lab protocol: https://media.addgene.org/cms/filer_public/e6/5a/e65a9ef8-c8ac-4f88-98da-3b7d7960394c/zhang-lab-general-cloning-protocol.pdf gRNA Primer Name Primer Sequence Transfection and mutation identification iCas9 stem cells were passaged in Accutase at 37°C for 13 min to ensure complete dissociation. Cells were seeded at 4 x 10 4 in 24 well plates for 24 hours in mTeSR with 5 μM Bleb. After 24 hours, media was removed and mTeSR was added. Cells were transfected with 2.5 ul DNA-In Stem (GST-2130, Life Technologies), 250 ng gRNA plasmid PX459v2 containing the gRNA and Cas9-p2a-puromicin-resistance genes in 50 ul of Opti-MEM (31985062, Gibco). Cells were incubated for 24 hours, then media was removed and mTeSR and 1 ug Doxycycline (D9891, MilliporeSigma) were added. After 24 hours, media was removed and mTeSR, 1 ug Doxycycline, and 0.3-1 ug of puromycin were added. After 24 hours, media was removed, and cells were washed 1X with mTeSR, and mTeSR was then added to the well. Surviving cells were passaged at single cell density, individual colonies were isolated, and mutations were confirmed by PCR sequencing. Gene diagrams of deletions are displayed in Fig. S2A.

Primer Name Primer Sequence
ThrB2_St_295_F GTGCTTGGAAATCTTGATGTTCAC

Immunohistochemistry
Retinal organoids: Retinal organoids were fixed in fresh 4% formaldehyde and 5% sucrose in PBS for 1 hour. Tissue was rinsed 3X in 5% sucrose in PBS, then incubated at 4°C in 6.75% sucrose in PBS for 30 min, 12.5% sucrose in PBS for 30 min, and 25% sucrose for 2 hours-overnight. Organoids were incubated for 2 hours in blocking solution (0.2-0.3% Trition X-100, 2-4% donkey serum in PBS). Organoids were incubated with primary antibodies in blocking solution for 16-36 hours at 4°C. Organoids were washed 3X for 30 min in PBS, and then incubated with secondary antibodies in blocking solution for 2 hours at room temperature. Organoids were incubated in 300 nM DAPI in blocking solution for 10 min and washed 3X for 15 min in PBS. At the end of staining, organoids were mounted for imaging in slow fade (S36940, Thermo Fisher Scientific).
Retinas: Human retinas were obtained from the National Disease Research Interchange (NDRI). Human retinal tissue was fixed by the NDRI in 10% formalin within 12 hours post-mortem and stored at 4°C until dissection. Retinas were dissected and whole-mounted, then rinsed 3X in PBS for 20 min, and blocked for 48 hours at 4°C in 0.3% Triton X-100 and 4% donkey serum. Retinas were stained with the same protocol as detailed above for organoids.
WERI-Rb1 cells: WERI-Rb1 cells were adhered to 0.01% w/v Poly-Llysine slides for 1-2 hours at 37°C and 5% CO2 and then washed 1X in PBS. WERI-Rb1 cells were fixed in fresh 4% formaldehyde for 20 min. Slides were washed with PBS 3X, and then incubated for 2 hours in blocking solution. Primary antibodies were added at 4°C overnight. Slides were washed 3X in PBS and incubated in secondary antibodies for 2 hours at room temperature in blocking solution.

Microscopy and image processing
Bright field images were acquired with a Nikon TE2000 or EVOS XL Core microscope. Fluorescent images were acquired with a Zeiss LSM710, LSM780, or LSM800 laser scanning confocal microscope. Confocal microscopy was performed with similar settings for laser power, photomultiplier gain and offset, and pinhole diameter. Maximum intensity projections of z-stacks (5-80 optical sections, 1.10 μm step size) were rendered to display all cones captured in a single organoid.
Opsin expression in different conditions: iCas9 H7 ESC-derived organoids for Thrβ2 KOs and controls were analyzed at day 200. Organoids for Thrβ KO, control, and wild-type + T3 were analyzed at two time points: 2 organoids were taken at day 199 for each group, and one was taken at day 277 for each group. T3-treated organoids were taken at time points between day 195 and day 200 for different differentiations. For each treatment group and genotype, organoids were compared to control organoids grown in parallel.

WERI-Rb1 siRNA and qPCR
Varying concentrations of WERI-Rb1cells were seeded onto 24-well plates with 500 uL of RPMI+Supplement. After ~24 hours, WERI-Rb1 cells were washed once with sterile DPBS (Gibco) and suspended in media with or without 100 nM T3. Negative control siRNA 1 (Thermo Fisher Scientific) or THRB ID:s14119 siRNA (Thermo Fisher Scientific) was incubated with lipofectamine RNAiMAX (Thermo Fisher Scientific) and Opti-Mem I Reduced Serum Medium (Thermo Fisher Scientific) according to manufacturer's instructions.
After 72 hours of incubation with RNAi, RNA was extracted from WERI-Rb1 cells using the Zymo Direct-zol RNA Microprep Kit (Zymo Research) according to manufacturer's instructions. RNA concentration was determined on a Nanodrop One (Thermo Fisher Scientific) and equal concentrations of RNA for each sample were used to generate cDNA using the RETROscript Reverse Transcription Kit (Thermo Fisher Scientific) according to manufacturer's instructions.
Quantitative Real-Time PCR was performed using TaqMan Gene Expression MasterMix (Thermo Fisher Scientific) on a StepOnePlus Real-Time PCR System (Applied Biosystems). Relative gene expression levels were determined for the genes THRβ (Hs00230861_m1 TaqMan probe from Thermo Fisher Scientific), and OPN1LW & OPN1MW (could not discriminate) (Hs01912094_s1 TaqMan probe from Thermo Fisher Scientific) and normalized to GAPDH (Hs02758991_g1 TaqMan probe from Thermo Fisher Scientific) using the delta-delta ct approach. For each condition, three biological replicates were performed and three technical replicates were run on the same plate for each primer set. Fold change was calculated relative to an siRNA negative control sample lacking T3.

Measurements and Quantification
Measurements of retinal area and cell morphology were done using imageJ software. Quantifications and statistics (except for RNA-seq data) were done in GraphPad Prism, with a significance cutoff of 0.01. Statistical tests are listed in figure legends. All error bars represent the SEM.

RNA-Seq time course analysis
Expression levels were quantified using Kallisto (version 0.34.1) with the following parameters: "-b 100 -l 200 -s 10 -t 20 --single". The Gencode release 28 comprehensive annotation was used as the reference transcriptome (60). Transcripts per million (TPM) values (Table S1) were then used to generate graphs in Prism and heatmaps in R using ggplot2. The distributions of transcripts were plotted to identify the best low TPM cutoff (Fig. S5A). The threshold was determined to be 0.7 Log(TPM+1), i.e. 5 TPM, and this value was used as an inflection point for heatmap. Heatmaps for Fig. S3A-C were made similarly, using CPM values from Hoshino, et. al (Fig. S5B)