**Background:** Trichromatic color vision in humans relies on three cone subtypes: S (blue), M (green), and L (red). The mechanisms specifying M vs. L cone fate have been unclear due to the high similarity of M- and L-opsin genes and proteins. Two models have been proposed: a stochastic model (random choice) and a temporal model (early-born cones become M, late-born become L). Retinoic acid (RA) signaling is known to regulate opsin gene arrays in zebrafish, but its role in human M/L specification was unknown.
**Methods:** The authors developed a colorimetric in situ hybridization method using 40- and 42-nucleotide probes targeting 8 differential nucleotides in M-opsin (OPN1MW) and L-opsin (OPN1LW) mRNA. Specificity was validated in HEK293 cells transfected with M- or L-opsin plasmids. The method was applied to human fetal retinas (day 122 and day 130 postconception), adult human retinas (3 donors, 5 mm punches from center, middle, and periphery), and human retinal organoids differentiated from H7 ESCs, H7 iCas9 ESCs, and EP1.1 iPSCs. Organoids were treated with 1.0 μm all-trans retinoic acid (ATRA) over different developmental windows: no RA after day 43 (control), RA days 43-60, RA days 43-130, or RA days 130-200. M and L cone ratios were quantified at day 200. Additionally, RNA-seq data from 12 prenatal human retinas (day 52/54 to day 136) and 3 adult retinas were analyzed for opsin and RA pathway gene expression. A genetic association study was performed in 738 males with normal color vision using flicker-photometric ERG to measure L cone ratio, combined with targeted sequencing of 21 gene regions including OPN1LW/MW, RA regulators (RARA, NR2F2/COUP-TFII, CYP26A1/C1), and other photoreceptor specification genes.
**Key Results:** In fetal retinas, M-opsin was expressed before L-opsin: at day 122, 100% of opsin-expressing cones were M (n > 500 cells); at day 130, >99% were M with sparse L cones. In adult retinas, the central and middle regions had significantly higher proportions of M cones compared to the periphery (center L vs. periphery L p < 0.01; middle L vs. periphery L p < 0.01; one-way ANOVA with Tukey's test). RNA-seq confirmed M-opsin expression from day 115 onward, while L-opsin was minimally detected during fetal development. RA-synthesizing enzymes ALDH1A1 and ALDH1A3 were highly expressed early (peak around day 59) and decreased over time in both central and peripheral retina. In organoids, addition of RA from days 43-130 yielded 98% M cones (vs. 13% M in no-RA controls; p < 0.005). RA from days 43-60 produced 74% M cones (p < 0.05 vs. control). Late RA (days 130-200) did not significantly alter M/L ratios (6% M, 93% L; p = 0.9635 vs. control). The genetic association study identified a significant association between SNP rs372754794 in the NR2F2-AS1 noncoding RNA upstream of NR2F2 (a nuclear receptor mediating RA signaling) and L cone ratio (β = −18.54, p = 1.67 × 10⁻⁵). The minor G allele, nearly exclusive to African American/African subjects (MAF = 0.0359), was associated with reduced L cone ratio. A second SNP (rs36102671) in RARA approached significance (β = −15.95, p = 6.54 × 10⁻⁵).
**Clinical Implications:** This study reveals that RA signaling acts early in human retinal development to promote M cone fate and suppress L cone fate, establishing a spatiotemporal gradient where early-born central retina is M-enriched and late-born periphery is L-enriched. This temporal model explains the natural variation in M/L cone ratios across the human retina and among individuals. The association with NR2F2 variants suggests that genetic variation in RA signaling components contributes to interindividual differences in cone ratios, which may influence color vision perception and susceptibility to cone-related diseases. Understanding the molecular mechanisms of cone specification could inform strategies for regenerative medicine, such as generating specific cone subtypes for retinal repair, and may provide insights into disorders like color blindness and cone dystrophies. The study also demonstrates the utility of human retinal organoids for investigating human-specific developmental processes.