Ciliary neurotrophic factor (CNTF) protects retinal cone and rod photoreceptors by suppressing excessive formation of the visual pigments

The retinal pigment epithelium (RPE)-dependent visual cycle provides 11-cis-retinal to opsins in the photoreceptor outer segments to generate functional visual pigments that initiate phototransduction in response to light stimuli. Both RPE65 isomerase of the visual cycle and the rhodopsin visual pigment have recently been identified as critical players in mediating light-induced retinal degeneration. These findings suggest that the expression and function of RPE65 and rhodopsin need to be coordinately controlled to sustain normal vision and to protect the retina from photodamage. However, the mechanism controlling the development of the retinal visual system remains poorly understood. Here, we show that deficiency in ciliary neurotrophic factor (CNTF) up-regulates the levels of rod and cone opsins accompanied by an increase in the thickness of the outer nuclear layers and the lengths of cone and rod outer segments in the mouse retina. Moreover, retinoid isomerase activity, expression levels of RPE65 and lecithin:retinol acyltransferase (LRAT), which synthesizes the RPE65 substrate, were also significantly increased in the Cntf−/− RPE. Rod a-wave and cone b-wave amplitudes of electroretinograms were increased in Cntf−/− mice, but rod b-wave amplitudes were unchanged compared with those in WT mice. Up-regulated RPE65 and LRAT levels accelerated both the visual cycle rate and recovery rate of rod light sensitivity in Cntf−/− mice. Of note, rods and cones in Cntf−/− mice exhibited hypersusceptibility to light-induced degeneration. These results indicate that CNTF is a common extracellular factor that prevents excessive production of opsins, the photoreceptor outer segments, and 11-cis-retinal to protect rods and cones from photodamage.

been shown to inhibit rhodopsin expression in the rodent retina (21)(22)(23), but it induced expression of rod and cone opsins in the chick retina (24,25). In addition, CNTF increased RPE cell survival and mitotic activity (26).
A gene profiling study showed that intravitreal injection of CNTF induced numerous genes associated with inflammation and gliosis in the Muller cells (43). In addition, CNTF increased secretion of neurotrophin-3 and decreased secretion of vascular endothelial growth factor, interleukin-8, and transforming growth factor-␤2 in the RPE cells (14). These altered gene expression and protein secretion might cause many secondary effects in the CNTF-treated patients and animals, and therefore, increased difficulty in defining the primary role and mechanism of CNTF function in regulating development, function, and protection of the retinal photoreceptors and RPE. In this study, we analyzed neuroretinal and RPE phenotypes in Cntf Ϫ/Ϫ mice, and identified CNTF as a common extracellular signal that down-regulates expression of opsins in rod and cone photoreceptors, as well as RPE65 and LRAT in the RPE, to avoid excessive formation of both light-sensing organelles and lightsensitive visual pigments, which mediate phototransduction or photoreceptor degeneration, depending on the light intensity they have captured.

Rod excessive development is associated with up-regulation of rhodopsin in the Cntf ؊/؊ retina
CNTF added into retinal cultures displayed distinct effects on rod development depending on species: it promoted chick rod development and rod-opsin expression (24), but it inhibited rat rod differentiation and rod-opsin expression (21). To define the effect of CNTF deficiency on rod development and rodopsin expression, we first compared expression levels of rhodopsin in WT and Cntf Ϫ/Ϫ mice. As shown in Fig. 1A, expression levels of rhodopsin (Rho) in 3-and 5-week-old Cntf Ϫ/Ϫ mice were 45-50% higher than those in age-matched WT mice. Rho mRNA expression levels in 3-week-old Cntf Ϫ/Ϫ retina were increased by ϳ53% as compared with those in the same age WT retina (Fig. 1B).
To confirm these results, we analyzed morphology of WT and Cntf Ϫ/Ϫ retinas. Immunohistochemistry showed that length of the rod outer segments (ROS) in a 3-week-old Cntf Ϫ/Ϫ mouse was 15% longer than that of WT mouse (Fig. 1C). Differential interference contrast microscopy of retinal sections confirmed that both ROS length and thickness of the outer nuclear layer (ONL) in 3-week-old Cntf Ϫ/Ϫ retina are clearly greater than those in age-matched WT retina (Fig. 1D). To know if numbers of rods are increased in the Cntf Ϫ/Ϫ mouse, we counted nuclear numbers in a 400-m wide region of the superior ONL that is 600 m away from the optic nerve head. As

CNTF inhibits visual pigment generation to protect retina
shown in Fig. 1E, numbers of rods in the region of the Cntf Ϫ/Ϫ retina were 11% greater than those in WT retina. Numbers of TUNEL-positive cells in the ONL and the outer plexiform layer (OPL) of the Cntf Ϫ/Ϫ retina at postnatal day 8 were significantly smaller than those in age-matched WT retinal ONL-OPL (Fig. 1F).

Increase of cone-specific proteins in over-elongated OS of Cntf ؊/؊ cones
Exogenous CNTF has been shown to promote cone differentiation in the developing chick retina (24, 25), suggesting that CNTF deficiency may reduce expression of cone-specific proteins. Unexpectedly, quantitative immunoblot analysis showed that cone arrestin (CAR), which expresses in both M-and S-cones (44), was increased ϳ50% in 3-and 5-week-old Cntf Ϫ/Ϫ mouse retinas, as compared with age-matched WT mouse retinas ( Fig. 2A). In agreement with this result, immunohistochemistry showed that length and numbers of CARpositive outer segments in 3-week-old Cntf Ϫ/Ϫ retina were clearly greater than those in WT retina (Fig. 2, B and G).
To know whether M-and S-opsins are also increased in the Cntf Ϫ/Ϫ retina, we performed quantitative immunoblot analy-sis for cone opsins. As shown in Fig. 2, C and E, expression levels of M-and S-opsins in 3-and 5-week-old Cntf Ϫ/Ϫ retinas are significantly higher than those in age-matched WT retinas. This increase in expression of cone opsins is associated with an increase in M-and S-cone numbers as well as length of cone outer segments in the Cntf Ϫ/Ϫ retina (Fig. 2, D and F-H). Quantitative RT-PCR revealed that transcripts for M-and S-opsins were increased by 60ϳ70% in 3-week-old Cntf Ϫ/Ϫ retina, as compared with those in age-matched WT retina (Fig. 2I).

Up-regulation of visual cycle enzymes in Cntf ؊/؊ RPE
Increased rod and cone opsins may need an increase in 11cRAL synthesis to form functional visual pigments. We therefore tested whether the visual cycle enzymes are changed in the Cntf Ϫ/Ϫ RPE. Immunoblot analysis showed that protein levels of RPE65 in 3-and 5-week-old Cntf Ϫ/Ϫ RPE were 40 -60% higher than those in age-matched WT RPE (Fig. 3A). Similarly, expression levels of LRAT are also significantly increased in the Cntf Ϫ/Ϫ RPE, as compared with WT RPE (Fig.  3B). In agreement with these results, activities of the retinoid isomerase in the 3-and 5-week-old Cntf Ϫ/Ϫ RPE were at least 45% higher than those in age-matched WT RPE (Fig. 3C).

CNTF inhibits visual pigment generation to protect retina
To know if the number of RPE cells is increased in Cntf Ϫ/Ϫ mouse, we counted DAPI-positive RPE nuclear numbers in sections taken from the dorsal-ventral midline of the eye. We then calculated RPE cell density by dividing RPE nuclear numbers with arc length of retina in the sections. RPE cell densities in Cntf Ϫ/Ϫ sections (28.6 Ϯ 2.0 nuclei/mm, n ϭ 4) were similar to those in WT sections (27.8 Ϯ 1.8 nuclei/mm, n ϭ 4). RPE morphology in the Cntf Ϫ/Ϫ retinal section was similar to that in WT retinal section (Fig. 3D). Distribution patterns of RPE65 and LRAT in the Cntf Ϫ/Ϫ RPE were also similar to those in WT RPE (Fig. 3E). Consistent with these results, expression levels of the RPE microvilli protein Ezrin in the Cntf Ϫ/Ϫ RPE were similar to those in WT RPE (Fig. 3F).
To determine whether CNTF up-regulates RPE65 expression at the transcription level we performed quantitative RT-PCR. As shown in Fig. 3G, mRNA expression levels of RPE65 were increased ϳ30% in the Cntf Ϫ/Ϫ mouse, as compared with WT mouse. To confirm this result, we incubated Cntf Ϫ/Ϫ eyecup RPE with media of 293T-LC cells transfected with pRK5 or pRK-CNTF. Quantitative RT-PCR and immunoblot analysis showed that both mRNA and protein levels of RPE65 were reduced in the eyecup RPE incubated with media containing CNTF (Fig. 3, H and I).

Rod hyperpolarization, but not rod ON bipolar depolarization, was increased in Cntf ؊/؊ mouse
Because rod-opsin is up-regulated in the Cntf Ϫ/Ϫ retina, we tested whether the Cntf Ϫ/Ϫ rods display an increased phototransduction. We recorded scotopic ERG responses of darkadapted WT and Cntf Ϫ/Ϫ mice to a series of increasing light flashes. Representative ERG responses elicited with 0ϳ2 log candelas ϫ s/m 2 (cd ϫ s/m 2 ) flashes are shown in Fig. 4A. Amplitudes of a-waves elicited with 1ϳ2 log cd ϫ s/m 2 flashes were significantly greater in Cntf Ϫ/Ϫ mice, as compared with

CNTF inhibits visual pigment generation to protect retina
those in WT mice (Fig. 4, A and B). Unexpectedly, b-wave amplitudes in Cntf Ϫ/Ϫ mice were similar to those of WT mice (Fig. 4

, A and C).
To know if the synaptic function between rod and rod ON bipolar cell (ON-BC) was altered in the Cntf Ϫ/Ϫ mouse, we measured a-and b-wave implicit times (times from stimulus onset to a-or b-wave peak). As shown in Fig. 4D, a-wave implicit times of Cntf Ϫ/Ϫ rod ERG were similar to those of WT rod ERG. The b-wave implicit times of Cntf Ϫ/Ϫ rod ERG were also similar to those of WT rod ERG elicited with Ϫ4ϳ0 log cd ϫ s/m 2 flashes. When 0.7ϳ1.7 log cd ϫ s/m 2 flashes applied, the b-wave implicit times of Cntf Ϫ/Ϫ rod ERG were 10.5ϳ12.2% shorter than those of WT rods (p ϭ 0.056, n ϭ 6; Fig. 4D). Immunoblot analyses showed that expression levels of the ON-BC markers (mGluR6 and PKC␣) in the Cntf Ϫ/Ϫ retina were similar to those in WT retina (Fig. 4, E and F).

Faster recovery rate of rod light sensitivity is underpinned by accelerated visual cycle in Cntf ؊/؊ mouse
Because rod opsin and the visual cycle enzymes are increased in the Cntf Ϫ/Ϫ mouse, we tested whether the recovery rate of rod light sensitivity is accelerated in Cntf Ϫ/Ϫ mouse. We exposed dark-adapted WT and Cntf Ϫ/Ϫ mice to 800 lux light for 5 min, and then returned the mice to darkness. At different times, we recorded scotopic ERG responses to three different flash stimuli (10, 25, and 50 cd ϫ s/m 2 ). WT and Cntf Ϫ/Ϫ mice kept in darkness for 5 min displayed similar a-wave amplitudes (Fig. 5A). However, a-wave amplitudes of Cntf Ϫ/Ϫ mice kept in darkness for 10 -30 min were higher than those in WT mice under the same light conditions (Fig. 5A).
To know if this faster recovery of rod light sensitivity is supported by an accelerated visual cycle, we measured the regeneration rates of 11cRAL in WT and Cntf Ϫ/Ϫ mice kept in darkness for different times after photobleaching the visual pigments. Because both the ONL thickness and the OS length are greater in Cntf Ϫ/Ϫ retina than those of WT retina, we expressed the levels of 11cRAL in photobleached WT and Cntf Ϫ/Ϫ eyes as 100%. After keeping the mice in darkness for 15 or 30 min, the amounts of 11cRAL were increased 55 or 83%, respectively, in Cntf Ϫ/Ϫ mice, but only 15 or 30% in WT mice (Fig. 5B). These data indicate that the regeneration rates of the 11cRAL in the Cntf Ϫ/Ϫ eyes are faster than those in the WT eyes.

Hyper visual function of cones in Cntf ؊/؊ mouse
The up-regulation of cone opsins in the Cntf Ϫ/Ϫ retina ( Fig.  2) prompted us to test if cone visual function was enhanced in the Cntf Ϫ/Ϫ mouse. Under a rod-saturating background light, we recorded photopic ERG responses of WT and Cntf Ϫ/Ϫ mice to a series of increasing achromatic flashes. As shown in Fig. 6, A-C, amplitudes of a-and b-waves elicited with a given intensity of flash were significantly higher in Cntf Ϫ/Ϫ mice than WT mice. Implicit times to photopic b-wave peaks elicited with 0.5ϳ2 log cd ϫ s/m 2 flashes were markedly shorter in Cntf Ϫ/Ϫ mice versus WT mice (Fig. 6D), suggesting that synaptic function between cone and cone ON-BC is normal or enhanced in Cntf Ϫ/Ϫ mice. To confirm these results, we recorded photopic flicker ERG. Consistent with the above results, amplitudes of b-waves elicited with 10-and 20-Hz flashes at 10 cd ϫ s/m 2 intensity were significantly increased in Cntf Ϫ/Ϫ mice, as compared with those in WT mice (Fig. 6, E and F). These results indicate that CNTF deficiency causes a "super vision" mediated by cones. These results also suggest that CNTF may have different roles in the development and/or function of cone ON-BC versus rod ON-BC systems.

Rods and cones in Cntf ؊/؊ mouse exhibited hyper susceptibility to photodamage
Because the photoresponsiveness of rods and cones is increased in Cntf Ϫ/Ϫ mice, we investigated the effects of intense light on the retinal function and structure of Cntf Ϫ/Ϫ mice. We exposed mice to 12,000 lux light for different times (30ϳ75 min), and then kept them in darkness for 5 days. Retinoid analysis showed that the contents of 11cRAL and the total retinoids were significantly reduced in the Cntf Ϫ/Ϫ eyes exposed to the intense light for 75 min, as compared with those in WT mice under that same light conditions (Fig. 7A). Cntf Ϫ/Ϫ , but not WT, mice under that same light conditions displayed a significant reduction in rod visual function, as measured by scotopic ERG (Fig. 7B). Notably, both a-and b-wave amplitudes are markedly reduced in Cntf Ϫ/Ϫ mice (Fig. 7B). This reduction of rod visual function is associated with a dramatic decrease in rod OS length and contents of rhodopsin in the Cntf Ϫ/Ϫ retina (Fig.  7, C and D). Morphological analysis showed that 75 min exposure of 12,000 lux light caused a severe retinal degeneration in Cntf Ϫ/Ϫ , but not WT, mice (Fig. 8, B and C). The superior ONL in Cntf Ϫ/Ϫ retina was reduced in thickness to 2-8 nuclei versus 9 -10 nuclei in WT retina whereas the inferior ONL in the Cntf Ϫ/Ϫ retina was reduced to 4 -8 nuclei versus 10 -11 nuclei in WT retina (Fig. 8D).
Generally, cones are resistant to light-induced degeneration (45,46). To know if CNTF deficiency increases susceptibility of cones to photodamage, we did immunohistochemistry for cones in WT and Cntf Ϫ/Ϫ mice exposed to the intense light for 75 min. Lengths of both S-cone and M-cone OSs are significantly shortened in Cntf Ϫ/Ϫ mice (Fig. 9, B and C). Consistent with these observations, immunoblot analysis showed that expression levels of S-and M-opsins were reduced at least 40% in the Cntf Ϫ/Ϫ retinas, as compared with those in WT retinas (Fig. 9, D and E).

Discussion
Phototransduction and the visual cycle are the most important functions in sensing and converting light signal into biochemical and electrical signals in the retina. Light sensitivities of both rod and cone visual pigments rely on their 11cRAL chromophore provided by the visual cycle. In this study, we showed that CNTF deficiency resulted in up-regulation of opsins and the visual cycle enzymes, leading to an increase in thickness of ONL, length of rod and cone OSs, and the rates of the 11cRAL regeneration. These molecular and morphological changes led to 1) hyper phototransduction of rods and cones, 2) faster recovery of rod light sensitivity, and 3) increase in susceptibility of rods and cones to photodamage in Cntf Ϫ/Ϫ mouse.
Up-regulation of opsins and the visual cycle enzymes (Figs.  1-3) are the most important primary phenotypes in the

CNTF inhibits visual pigment generation to protect retina
Cntf Ϫ/Ϫ visual system. CNTF might down-regulate these proteins through the heterotrimeric receptor complex (17), which can activate at least three distinct downstream pathways: Jak-STAT3, Ras-MAPK, and PI3K-AKT pathways (22,47). The Jak-STAT3 pathway is the most studied pathway in CNTF/LIF signaling. Exogenous LIF activated STAT3 in mouse photoreceptors (48) and down-regulated rhodopsin (49), whereas CNTF did not activate STAT3 in any photoreceptors but down-regulated rhodopsin (23). CNTF might down-regulate rhodopsin by acting on Muller cells (23) or through the Ras-MAPK and/or PI3K-AKT pathways in rods.
Opsin up-regulation was accompanied by morphological changes in the Cntf Ϫ/Ϫ retina. The ONL thickness and lengths of rod and cone OSs were increased in Cntf Ϫ/Ϫ mice. The increase in the ONL thickness might be associated with the decrease in programed cell death in the ONL of postnatal Cntf Ϫ/Ϫ mice (Fig. 1). CNTF has been shown to promote programed death of postmitotic rod precursor cells; and blocking the CNTF/LIF pathway reduced cell death during mouse retinal development, resulting a thicker ONL (50).
CNTF deficiency also resulted in an increase in lengths of both rod and cone OSs. This observation is consistent with

CNTF inhibits visual pigment generation to protect retina
three previous studies: 1) CNTF induced a reversible rod OS shortening in WT rat (23), 2) CNTF reduced cone OS length in the rds mouse (51), and 3) transgenic LIF inhibited rod and cone OS maturation and/or elongation (52). In a rat line carrying a rhodopsin mutation, however, CNTF promoted regeneration of cone OS (32). The different effects of CNTF on cone OS morphology may depend on the presence or absence of strong inflammatory stimulation (53), which produce numerous signaling proteins in the retina (54).
11cRAL bound with opsins functions as a molecular switch for initiating phototransduction in response to light stimuli. Light-mediated isomerization of 11cRAL to all-trans-retinal induces opsin activation and phototransduction. To restore light sensitivity to opsins that have lost 11cRAL, 11cRAL must be regenerated and recombined with apo-opsins. RPE65 and LRAT are critical enzymes in the visual cycle that generates 11cRAL. We found that the synthesis rate of 11-cis-retinol in the Cntf Ϫ/Ϫ RPE was increased due to up-regulation of RPE65

CNTF inhibits visual pigment generation to protect retina
and LRAT. RPE65 and LRAT might be up-regulated in part through the gp130-STAT3 pathway in the RPE (49). We further found that recovery of rod light sensitivity was accelerated due to an increase in 11cRAL synthesis in the Cntf Ϫ/Ϫ RPE (Fig. 5).
Increase in opsin expression and 11cRAL synthesis suggest that the Cntf Ϫ/Ϫ rods and cones contain more light-sensitive visual pigments. This might cause the enhanced responses of Cntf Ϫ/Ϫ rods and cones to light stimuli (Figs. 4 -6). The hyper phototransduction led to greater hyperpolarization of rods and cones, reflected in elevation of a-wave amplitudes in both scotopic and photopic ERGs of Cntf Ϫ/Ϫ mice.
Although scotopic a-wave amplitudes were increased in Cntf Ϫ/Ϫ mice b-wave amplitudes, implicit times were similar to those in WT mice. In addition, expression levels of mGluR6 and PKC␣ in the Cntf Ϫ/Ϫ retina were also similar to those in WT retina. These results suggest that synaptic function between rod and rod ON-BC, which produce b-wave responses (55,56), is not significantly changed or may be slightly reduced in Cntf Ϫ/Ϫ mice. Overexpression of CNTF or LIF has been shown to cause BC disorganization (51,57). Exogenous CNTF has also been shown to promote BC differentiation (21, 51) through increasing Ath3 expression and pleiotrophin secretion (58,59). These studies and the similar b-wave amplitudes in Cntf Ϫ/Ϫ and WT mice suggest that CNTF plays an important role in the development, differentiation, and function of rod ON-BCs.
Interestingly, cone b-wave amplitudes and implicit times in photopic ERG elicited with certain flash intensities were significantly higher or shorter in Cntf Ϫ/Ϫ mice versus WT mice. These results are consistent with the previous studies that show significant reduction of photopic b-waves in WT animals overexpressing CNTF (41,42). Because b-waves of photopic and flicker ERGs are mainly from depolarizing cone ON-BCs (60, 61), our results suggest that the synaptic function between cone and cone ON-BC, as well as cone-induced depolarization of cone ON-BCs are enhanced in Cntf Ϫ/Ϫ mouse.
One of the pathological consequences of CNTF deficiency was retinal photodamage. Cntf Ϫ/Ϫ rods exhibited hyper susceptibilitytolight-induceddegeneration.Thisphenotypeisconsistent with the well-known phenomenon: the rodent superior retina has much higher susceptibility to photodamage, as compared with the inferior retina. Length of photoreceptor OS in the superior retina is ϳ30% longer than that in the inferior retina (62). The longer OSs in the Cntf Ϫ/Ϫ and the rodent superior retinas may contain more rhodopsin, the primary mediator of retinal degeneration induced by light (9). Activated rhodopsin may increase the DNA-binding activity of the transcription factor AP-1 to promote light-induced apoptosis of photoreceptors (9,63). In Drosophila, activated rhodopsin induced photoreceptor apoptosis by promoting clathrin-dependent endocytosis of rhodopsin-arrestin complexes (64).
It has been known that the higher expression level and activity of RPE65 cause an increase in susceptibility of retina to photodamage (10 -12). Conversely, decrease in the visual cycle rate or RPE65 activity reduce retinal photodamage (65)(66)(67). RPE65 promotes retinal photodamage by facilitating the visual cycle that provides 11cRAL to generate light-sensitive visual pigments. Up-regulated RPE65 and LRAT have accelerated the synthesis rate of 11cRAL in the Cntf Ϫ/Ϫ mouse. This acceler-ated visual cycle plus increased expression of opsins resulted in an increase in the formation of light-sensitive visual pigment, therefore promoted retinal photodamage in Cntf Ϫ/Ϫ mice.
In general, cones are resistant to photodamage (45,46). In this study, however, we observed that intense light caused a severe degeneration of both M-and S-cones in the Cntf Ϫ/Ϫ mouse. This result suggests that CNTF deficiency itself and increased formation of cone visual pigments contributed to cone photodamage in Cntf Ϫ/Ϫ mouse. In addition, increased visual cycle and phototransduction rates may result in elevation of oxidative stress in cones due to an increase in contents of 11cRAL and all-trans-retinal in the Cntf Ϫ/Ϫ cones. Retinaldehydes have been shown to play a critical role in retinal photooxidative damage (68 -70). Cones are known to have high susceptibility to oxidative stress (71)(72)(73)(74). Overexpression of NRF2, a master antioxidant transcription factor, effectively protected cones from degeneration in animal models of retinal degeneration (75). These studies and our results suggest that reduction of neuroprotective signals and increase in oxidative stress promoted light-induced cone degeneration in Cntf Ϫ/Ϫ mice. In summary, we identified CNTF as a critical signal that down-regulates rod and cone opsins, as well as RPE65 and LRAT, to suppress visual pigment over formation and retinal photodamage.

Animals
We crossed Cntf Ϫ/Ϫ mice (76) with WT 129S2/Sv mice (Charles River Laboratories), then intercrossed the heterozygous offspring to yield Cntf Ϫ/Ϫ mice homozygous for the Leu-450 allele of the Rpe65 gene. The homologous Cntf knockout mutation and the Leu-450 alleles were confirmed by PCR and DNA sequencing, as described previously (12,76). Except where noted, mice were maintained in 12 h cyclic light at ϳ30 lux. All animal experiments were performed on both sexes of WT and Cntf Ϫ/Ϫ mice at 3, 5, or 6 weeks of age in accordance with the Association for Research of Vision and Ophthalmology statement for the use of animals in ophthalmic and vision research and the protocols approved by the Institutional Animal Care and Use Committee for LSU Health Sciences Center.

Eyecup RPE, cell culture, and transfection
After removal of the anterior section and neural retina, the RPE in 3-week-old Cntf Ϫ/Ϫ mouse eyecups were maintained in DMEM/F-12 medium (Thermo Fisher Scientific Inc.) supplemented with 10% heat-inactivated fetal bovine serum (FBS) and antibiotics (77). The 293T-LC cells maintained in DMEM (Invitrogen) with 10% FBS (78) were transfected with pRK5 or pRK5-CNTF plasmid DNA using the PolyJet transfection reagent (SignaGen Laboratories). After transfection, the cells were maintained overnight in DMEM/F-12 medium containing 5% FBS. These media with or without CNTF were incubated with the eyecup RPE for 6 h. All cultures were maintained in an incubator containing 5% CO 2 .

Immunohistochemistry
Retinal cryosections were prepared from the dorsal-ventral midline of mouse eye as described previously (83). Briefly, enucleated mouse eyeballs were fixed overnight with 4% paraformaldehyde in 0.1 M phosphate buffer (PB). After removing cornea and lens, eyecups were immersed in 15% sucrose in 0.1 M PB for 2 h, in 30% sucrose in 0.1 M PB for 2 h, and then in a 1:1 mixture of 30% sucrose and optimal cutting temperature (OCT) medium (Sakura Finetechnical) overnight at 4°C. After embedding eyecups in OCT, 15-m thick sections were cut on a Shandon Cryotome SME cryostat (Thermo Scientific). The sections were immunostained with the primary antibodies listed in the method of immunoblot analysis and secondary antibodies, as described previously (82). Nuclei were counterstained with DAPI (Sigma). Images were captured with a Zeiss LSM710 Meta confocal microscope with a ϫ20 objective lens or ϫ40 oil-immersion lens. Lengths of rod and cone OS as well as numbers of cells were measured using ImageJ software (National Institutes of Health).

TUNEL assay
This assay was performed using the in situ cell death detection kit (Roche Applied Science) following the manufacturer's protocol. Briefly, retinal cryosections washed with 0.1% sodium citrate in 0.1% Triton X-100, PBS were incubated with terminal deoxynucleotidyl transferase (TdT) and fluorescein-dUTP for 1 h at 37°C. After rinsing three times in PBS containing 0.05% Tween 20, nuclei were counterstained with DAPI. Numbers of TUNEL-positive nuclear in the ONL-outer plexiform layer and in the inner nuclear layer-inner plexiform layer of whole retinal sections were counted separately using an Olympus BX61VS microscope equipped with a digital camera and VS-ASW FL software.

Quantitative RT-PCR
Total RNA was extracted from mouse retina or RPE using a PureLink RNA mini kit (Invitrogen), and was reverse-transcribed to cDNA using SuperScript III (Invitrogen). Quantitative PCR was performed on an iCylcer iQ TM Real-Time PCR Detection System (Bio-Rad) using a two-step qRT-PCR kit with SYBR Green (Invitrogen) and primer sets specific for mouse RPE65, opsin, and 18S rRNA. Three mice of each genotype were analyzed and all samples were run in duplicates. Starting templates were normalized after determining 18S rRNA C t val-ues for each sample (84). Relative mRNA levels of RPE65 were determined from the ⌬C t values.

Electroretinography (ERG)
Overnight dark-adapted 6-week-old mice were anesthetized with an intraperitoneal injection of 100 mg/kg of ketamine and 10 mg/kg of xylazine. The pupils were dilated with 1% tropicamide. ERG was recorded from the corneal surface using a silver-silver chloride wire electrode referenced to a subcutaneous electrode in the mouth. A needle electrode in the tail served as the ground. A drop of 2.5% methylcellulose was placed on the cornea to ensure good electrical contact and to prevent corneal desiccation during the entire procedure. Single-flash ERG recordings were performed in a Ganzfield dome (Espion e2, Diagnosys LLC) under dark-adapted (scotopic) and lightadapted (photopic) conditions. For scotopic ERG, single flash stimuli were presented with various intensities, reaching from Ϫ4 log cd ϫ s/m 2 to 2.5 log cd ϫ s/m 2 , in interstimulus intervals of 0.5ϳ2 min (depending on the stimulus intensity). Three to four responses were averaged for each step. For photopic ERG, animals were light adapted for 10 min by exposing to a white 32 cd/m 2 light, and ERG responses were obtained with white flashes (Ϫ0.3ϳ2.4 log cd ϫ s/m 2 ) on the rod-saturating background light (32 cd/m 2 ). Five responses to 10-s interval flashes were averaged for each step. Intensity-response amplitude data were displayed on log-linear coordinates using SigmaPlot 11 software. Flicker ERG responses were recorded in light-adapted mice with flicker flashes on the rod-saturating background. The flicker stimuli had an intensity of 10 cd ϫ s/m 2 with frequencies of 10 or 20 Hz.

Retinoid isomerase assay
RPE homogenates in 20 mM HEPES buffer (pH 7.4) containing protease inhibitor mixture were prepared from mouse eyecups without the neural retina. The homogenates were irradiated for 10 min on ice with 365-nm light from a Spectroline Model EN-140L UV light source to destroy endogenous retinoids. Each assay mixture contained 200 g of cell homogenate, 10 M all-trans-retinol, and 6% BSA. After incubating for 2 h in darkness at 37°C, retinoids were extracted with hexane and analyzed by HPLC, as described below.

Analysis of retinoids
Retinoids in mouse ocular tissues or in vitro enzyme assays were extracted with hexane and analyzed by normal-phase HPLC (84). In brief, retinoids in hexane extractions were evaporated, dissolved in 100 l of hexane, and separated on a silica column (Zorbax-Sil 5 m, 250 ϫ 4.6 mm, Agilent Technologies) by gradient (0.2-10% dioxane in hexane at 2.0 ml/min flow rate) or nongradient (10% dioxane in hexane at 1.0 ml/min flow rate) elution of mobile phase on an Agilent 1100 HPLC system equipped with a photodiode array detector (Agilent Technologies). Spectral data were acquired for all eluted peaks. Quantitation was performed by comparison of peak areas to calibration curves established with authentic retinoid standards.

Light microscopy
Mouse retinal sections were prepared as described previously (84). In brief, mice were fixed by intracardiac perfusion CNTF inhibits visual pigment generation to protect retina with a mixture of 2% paraformaldehyde and 2.5% glutaraldehyde in 0.1 M PB (pH 7.4). Light cautery was applied at the superior pole of the cornea to mark the orientation before enucleation of the eyeball. A window was cut in the cornea, and the eye was immersed in primary fixative and rotated at room temperature for 2 h. The anterior segment was removed, and the remaining eyecup was refrigerated overnight in primary fixative. The eyecup was trimmed into temporal and nasal hemispheres and immersed in 1% osmium tetroxide in 0.1 M PB (pH 7.2) for 1 h. Following dehydration in a graded series of alcohols, the hemispheres were embedded in an Epon/Araldite mixture (5:3, v/v). Care was taken to orient the eyecups so that sections were obtained through the vertical meridian to ensure sampling of both the superior, green-sensitive cones and the inferior, blue-sensitive cones. Sections were cut at 1 m and stained with 1% toluidine blue and 1% sodium borate, then photographed using a ϫ20 objective lens or a ϫ60 oil-immersion lens in the Olympus BX61VS microscope mentioned above. Except for whole retinal images, all images were obtained from retinal sections at a distance of 600 m superior or inferior to the optic nerve.

Light-induced retinal degeneration
WT and Cntf Ϫ/Ϫ mice were dark-adapted for 3 days. After dilation of the pupils under dim red light (Kodak Wratten 1A), mice were exposed to 12,000 lux of white fluorescent light for 30, 60, and 75 min and then kept in darkness for 5 days. Retinoid contents, visual function, and retinal structures of these mice were analyzed as described above.

Statistical analysis
SigmaPlot Version 11 (Systat Software, Inc.) was used for statistical analyses. Data were expressed as the mean Ϯ S.D. of three or more independent experiments indicated in the figure legends. Differences between WT and Cntf Ϫ/Ϫ mice were determined by single comparisons with an unpaired two-tailed Student's t test. The significance threshold was set at 0.05 for all statistical tests.