**Background:** Heat shock protein 90 (HSP90) is an abundant molecular chaperone that stabilizes a select set of client proteins. Its two cytosolic paralogs, HSP90α (stress-inducible) and HSP90β (constitutive), share high sequence similarity, making it difficult to assign unique functions. Pan-HSP90 inhibitors used in cancer clinical trials cause night blindness as a major side effect, suggesting a critical role for HSP90 in photoreceptor cells. Previous work indicated that HSP90 interacts with the photoreceptor enzyme phosphodiesterase 6 (PDE6) and its cochaperone AIPL1, but whether HSP90α specifically is required for rod or cone survival was unknown. This study aimed to determine the role of HSP90α in the retina using a novel knockout mouse model.
**Methods:** The authors generated a global HSP90α knockout mouse (Hsp90α-/-) by CRISPR-Cas9-mediated deletion of exon 4 of the Hsp90aa1 gene, leading to a truncated, non-functional protein. Retinal function was assessed by electroretinography (ERG) at multiple ages (P30, P160, P250). Photoreceptor morphology and survival were evaluated by hematoxylin and eosin (H&E) staining, TUNEL assay, and transmission electron microscopy (TEM). Protein expression was quantified by immunoblotting and tandem mass tag (TMT) proteomics on P15 retinas (before degeneration). mRNA levels were measured by qRT-PCR. Immunoprecipitation was used to assess PDE6 complex assembly. An all-cone Nrl-/- mouse model was crossed with Hsp90α-/- to test cone-specific requirements.
**Key Results:** HSP90α was highly expressed in photoreceptor inner segments, while HSP90β was enriched in cone inner segments and inner retinal neurons. In Hsp90α-/- mice, rod photoreceptor function (scotopic ERG a-wave) was normal at P30 but progressively declined, with complete loss by P250. Cone function (photopic ERG) was preserved until P160 but lost by P250, coinciding with total rod degeneration. In the all-cone Nrl-/- background, cone function remained normal even at P250, confirming that cone loss in the full knockout was a bystander effect of rod death. Histology showed normal retinal development at P45, followed by progressive rod degeneration: ONL nuclei reduced from ~11 layers (control) to ~5 layers at P120, and to 1 layer by P250. TUNEL-positive cells appeared in the ONL at P45. TEM at P30 showed normal outer segment discs and cilia; at P60, whorl-like structures and vesicle accumulation were observed, but Golgi morphology remained normal. Immunoblotting at P15 revealed a 50% reduction in all rod PDE6 subunits (α, β, γ) and AIPL1, while cone PDE6α' was unchanged. HSP90β was upregulated ~2-fold. GRK1, rhodopsin, transducin, and CNGA1 levels were unaffected. mRNA levels of Pde6 subunits and Hsp90ab1 were unchanged, indicating post-transcriptional regulation. TMT proteomics quantified 7587 proteins; only 67 (<1%) were differentially expressed in Hsp90α-/- retinas. Downregulated proteins included rod PDE6 subunits, AIPL1, WDR17, WDR19, WDR35, and several cochaperones. Upregulated proteins included HSP90β, HSP105, and L-gulonolactone oxidase. Immunoprecipitation showed that residual PDE6 in knockout retinas still assembled into a complex. In AIPL1 heterozygous mice (50% AIPL1), PDE6 levels were normal, suggesting that the PDE6 reduction in Hsp90α-/- is not solely due to AIPL1 loss.
**Clinical Implications:** This study demonstrates that HSP90α is specifically required for rod photoreceptor maintenance by regulating rod PDE6 and AIPL1 protein levels. The selective vulnerability of rods explains the night blindness observed in patients treated with pan-HSP90 inhibitors. The preservation of cone function in the absence of HSP90α, likely due to compensatory HSP90β expression, suggests that isoform-specific HSP90 inhibitors might avoid retinal toxicity. The identification of WDR17 and other WD-repeat proteins as potential HSP90α clients opens new avenues for understanding photoreceptor degeneration. These findings underscore the need for careful design of HSP90-targeted therapies to minimize visual side effects.