**4. Discussion**

Studies involving genetic screening of families with autosomal recessive retinitis pigmentosa have implicated a founder missense mutation (K42E) in the gene encoding DHDDS [12,26]. DHDDS is required for N-glycosylation of proteins by adding multiple copies of isopentenyl pyrophosphate (IPP) to farnesyl pyrophosphate (FPP) to produce dehydrodolichyl diphosphate (Dedol-PP), a precursor of dolichol, which is utilized as a sugar carrier in protein glycosylation in the endoplasmic reticulum [6]. Even though the generation of isoprenoid chains is complex, involving multiple enzymes and enzyme complexes [27], only DHDDS and Nogo-B receptor are required for long-chain isoprenoid synthesis (C70-C120) [28]. Previous studies have reported that mutations in the opsin gene that abolish N-linked glycosylation cause retinal degeneration [28–30]. To understand the basis for the ocular pathology associated with DHDDS mutation, we utilized a Cre recombinase conditional knockout (*Dhddsflx*/*flx* CreRPE) driven under the control of the RPE-specific VMD2 promoter to achieve RPE-targeted excision of the loxP-modified *Dhdds* gene, which renders the enzyme non-functional.

The VMD2 construct was originally generated to be conditional on the presence of doxycycline; however, we found that the mice displayed a phenotype in the absence of doxycycline induction. While we cannot rule out the presence of some level of doxycycline in the standard chow mouse diet, it is more likely that Cre expression was due to "leaky" expression, which bypassed doxycycline control, as confirmed in Figure 2. Since the promoters used for cell-specific targeting of Cre expression turn on after development is complete, developmental changes that would otherwise preempt retina/RPE development were not observed.

The expression of Cre recombinase in the Cre lines was confirmed using a ZsGreen reporter strategy. The *Dhddsflx*/*flx* CreRPE mice exhibited about 90% coverage of Cre expression by PN 1 month (Figure 2) which persisted up to 3 months at least. We crossed the conditional *Dhddsflx*/*flx* lines with heterozygous VMD2 Cre lines to generate homozygous *Dhddsflx*/*flx* mice with RPE-specific knockout of DHDDS expression as the model for our study.

In vivo imaging suggested that ONL thickness and F.R.T. were comparable between age-matched WT and *Dhdds*+/*flx* CreRPE mice, but significantly reduced in *Dhddsflx*/*flx* CreRPE mice. While the specific pathology is observed in much greater detail in the histological (light microscopy; Figure 5) and ultrastructural (electron microscopy; Figure 6) images provided, it is clear that the pathology indicated in the SD-OCT tomograms (Figure 3D) is consistent with the histological observations. This altered structural integrity of *Dhddsflx*/*flx* CreRPE retinas also corresponded to a significant decrease

in a-wave and b-wave compared to age-matched WT at all flash intensities. Unexpectedly, the scotopic a- and b-waves also were reduced in *Dhdds*+/*flx* CreRPE mice, which may sugges<sup>t</sup> a functional change that occurs prior to any obvious retinal structural changes and suggests that 50% DHDDS activity is insu fficient in the RPE to maintain the required enzymatic activity level. Thus, carriers of *Dhdds* mutations also may develop visual defects, depending on the nature of the mutation, and other factors, such as genetic background and environment. This agrees with the rod-cone dystrophy reported in patients with autosomal recessive RP [16]. However, we cannot rule out the expression of a truncated protein that leads to a gain of function. This could explain the significant attenuation of the photoresponse in mice heterozygous for the floxed allele (Figures 7 and 8). It could also possibly explain the manifestation of the disease only in ocular tissues, if the gain of function is due to specific targeting of retina-specific protein complexes. Further experimentation will be required to sort this out.

We carried out further experiments to examine the fundus of older *Dhddsflx*/*flx* CreRPE mice at 6, 8 and 10 months PN. These fundus images obtained with fluorescein angiography showed the classical signs of RP including abnormal pigmentation in 8-month old *Dhddsflx*/*flx* CreRPE mice; vascular changes, including microaneurysms, increased vessel tortuosity and attenuation, were also observed by 6 months of age (Supplementary Materials, Figure S2).

Originally it was proposed that the cause of the pathology in DHDDS-related RP patients (RP59) is defective glycosylation of rod opsin because of reduced DHDDS enzyme activity [12,26]. While this assumption may explain the classic RP symptoms observed, it fails to explain the AMD-like macular involvement. Interestingly, Lam and colleagues [11] found no N-glycosylation deficiency in RP59 patients, based upon isoelectric focusing gel analysis of plasma transferrin (a systemic glycoprotein). In addition, in a related study (Rao et al., manuscript submitted for publication), utilizing a rod photoreceptor-specific knockout of *Dhdds* in mice, we found no evidence for a resulting lack of protein glycosylation in the retina; yet, there was a rapidly progressing photoreceptor degeneration, resulting in complete loss of photoreceptors by PN 6 weeks of age. Also, in a companion article in this Special Issue [31], we generated a mouse model harboring the global K42E homozygous *Dhdds* mutation associated with RP59 patients, but observed no retinal degeneration, even out to 9 postnatal months of age. Clearly, mutations that only partially diminish enzymatic activity would be far less severe than a complete ablation of the gene encoding the enzyme.
