Dysfunction of Heterotrimeric Kinesin-2 in Rod Photoreceptor Cells and the Role of Opsin Mislocalization in Rapid Cell Death

Loss of kinesin-2 function causes rapid death of rod photoreceptors. The cell death is dependent on the expression of opsin, which first accumulates along the route to the outer segment, but not on signaling by opsin-arrestin complexes or by light activation; the key element appears to be the accumulation of excessive protein in the wrong place.

Abstract

Due to extensive elaboration of the photoreceptor cilium to form the outer segment, axonemal transport (IFT) in photoreceptors is extraordinarily busy, and retinal degeneration is a component of many ciliopathies. Functional loss of heterotrimeric kinesin-2, a major anterograde IFT motor, causes mislocalized opsin, followed by rapid cell death. Here, we have analyzed the nature of protein mislocalization and the requirements for the death of kinesin-2-mutant rod photoreceptors. Quantitative immuno EM showed that opsin accumulates initially within the inner segment, and then in the plasma membrane. The light-activated movement of arrestin to the outer segment is also impaired, but this defect likely results secondarily from binding to mislocalized opsin. Unlike some other retinal degenerations, neither opsin–arrestin complexes nor photoactivation were necessary for cell loss. In contrast, reduced rod opsin expression provided enhanced rod and cone photoreceptor survival and function, as measured by photoreceptor cell counts, apoptosis assays, and ERG analysis. The cell death incurred by loss of kinesin-2 function was almost completely negated by Rho^−/−. Our results indicate that mislocalization of opsin is a major cause of photoreceptor cell death from kinesin-2 dysfunction and demonstrate the importance of accumulating mislocalized protein per se, rather than specific signaling properties of opsin, stemming from photoactivation or arrestin binding.

INTRODUCTION

A variety of genes encode proteins that are critical for the formation and maintenance of cilia. These genes underlie ciliopathies, which are manifest in different organs, such as the retina, kidney, and the cerebellum. In the vertebrate retina, the phototransductive compartment of the photoreceptor cells, the outer segment, is formed by extensive amplification of the plasma membrane of a primary cilium. The resulting phototransductive disk membranes undergo continuous renewal (Young, 1967), requiring the delivery of large amounts of newly synthesized opsin-containing membrane along the ciliary axoneme. In addition to membrane proteins, some soluble signaling proteins, such as arrestin, transducin, and recoverin, relocate between the inner and outer segments in response to changes in light intensity (Strissel et al., 2006). Protein trafficking along the photoreceptor cilium is thus especially busy, and defects in ciliary trafficking appear to lead to rapid photoreceptor cell degeneration.

Anterograde transport in all primary cilia requires heterotrimeric kinesin-2 and associated intraflagellar transport (IFT) proteins (Rosenbaum and Witman, 2002; Scholey, 2008). Genetic studies have shown that mutations in Kif3a or Kif3b, the genes encoding motor subunits of heterotrimeric kinesin-2 (Marszalek et al., 2000; Lin-Jones et al., 2003; Jimeno et al., 2006; Avasthi et al., 2009), or mutant IFTs (Pazour et al., 2002; Krock and Perkins, 2008; Sukumaran and Perkins, 2009) cause an abnormal accumulation of opsin in the photoreceptor inner segment and nuclear layer, followed by photoreceptor cell death. Homodimeric kinesin-2, which consists of KIF17 subunits, is required for the formation of the disk membranes, and thus appears to function in a later stage of the delivery process (Insinna et al., 2008; Insinna et al., 2009). [In the remainder of this paper, kinesin-2 refers to heterotrimeric kinesin-2, which consists of the motor subunits, KIF3A and KIF3B or KIF3C, and a KAP3 subunit, according to standardized nomenclature (Lawrence et al., 2004.] In KIF3A-depleted photoreceptor cells, arrestin failed to localize in the outer segment in the light, suggesting that kinesin-2 might also transport soluble proteins to the outer segment (Marszalek et al., 2000).

The loss of photoreceptor cells due to disruption of IFT is particularly swift. The synchronous deletion of Kif3a from rod photoreceptor cells results in the death of nearly all cells within the ensuing 2 wk. The redistribution of opsin that precedes cell death has been suggested as a primary cause of the death (Marszalek et al., 2000). Supporting this notion, mutations in the rod opsin gene that affect its targeting also result in cell death (Sung et al., 1991; Sung et al., 1994; Li et al., 1996; Green et al., 2000; Chapple et al., 2001; Concepcion et al., 2002; Tam and Moritz, 2006). Moreover, some other mouse models of syndromic ciliopathies (e.g., Rpgrip, Bbs2, Cep290, Ahi1 mutants) have now identified the mislocalization of opsin as a mutant phenotype preceding photoreceptor cell loss (Zhao et al., 2003; Nishimura et al., 2004; Chang et al., 2006; Louie et al., 2010). Nevertheless, an abnormal accumulation of opsin outside of the outer segment is not only associated with cell death in cases where the mistargeting of opsin is a primary lesion. It is a characteristic of photoreceptor cells before their death in a variety of retinal degeneration models, such as the rd and rds mice, and the RCS rat (Nir and Papermaster, 1986; Jansen et al., 1987; Nir et al., 1987; Usukura and Bok, 1987; Nir et al., 1989). These examples prompt the question of whether an abnormal accumulation of opsin is part of a requisite pathway leading to the cell death or is simply a secondary response in a dying photoreceptor cell.

In some cases, specific signaling from opsin has been suggested as necessary for cell death. A number of photoreceptor degenerations are light-dependent, suggesting the requirement of light activation of rhodopsin; although they may or may not require activation of the transducin-mediated phototransductive cascade (Hao et al., 2002). In cell culture studies with dissociated salamander retinas, light activation of mislocalized opsin appeared to elicit rod cell death through G protein stimulation of adenylate cyclase (Alfinito and Townes-Anderson, 2002), although in vivo studies with Xenopus showed that activation of mislocalized C-terminal truncated opsin was not required for cell death (Tam and Moritz, 2006). A proposed mechanism that does not require activation of the phototransductive cascade stems from genetics studies with Drosophila (Alloway et al., 2000; Kiselev et al., 2000). In these studies, the requirement of arrestin for degeneration was reported, suggesting signaling from phosphorhodopsin-arrestin complexes. More recently, the constitutively-active K296E opsin was found to bind arrestin and accumulate ectopically in transgenic mouse photoreceptors. The ensuing degeneration also appeared to result from transducin-independent signaling by opsin-arrestin complexes (Chen et al., 2006).

In the present study, we first dissected the time course of the mislocalization of opsin due to kinesin-2 dysfunction in a quantitative manner and thus describe a characteristic that is likely unique to a trafficking defect. Next, we addressed the possibility of a role for kinesin-2 in the translocation of arrestin into the outer segment. Part of this study demonstrated the light-dependent association of arrestin with mislocalized opsin in Kif3a-depleted rod photoreceptor cells. Therefore, we next tested whether light-dependent signaling from an opsin-arrestin complex was required for photoreceptor cell degeneration. These studies led to the demonstration of a clear relationship between the amount of opsin and the rate of photoreceptor cell death and loss, indicating the importance of mislocalized opsin in promoting photoreceptor cell degeneration.

MATERIALS AND METHODS

Mice

All procedures conformed to institutional animal care and use authorizations and were in accordance with regulations established by the National Institutes of Health. Mice were kept on a 12-h light/12-h dark cycle under 10–50 lux of fluorescent lighting during the light cycle. The genetic mutants used were Kif3a^flox/flox (Marszalek et al., 2000), RHO-Cre (line 8) (Jimeno et al., 2006), arrestin (Sag) knockout (Xu et al., 1997), alpha-transducin (Gnat1) knockout (Calvert et al., 2000), and rod opsin (Rho) knockout (Lem et al., 1999) mice. All mice had been backcrossed on to the C57BL/6HNsd (Harlan C57BL/6) genetic background, which is considered “wild-type” for the purposes of this study. Breeding strategies and genotyping of the individual mutations followed that described previously (Xu et al., 1997; Lem et al., 1999; Marszalek et al., 2000; Jimeno et al., 2006). RHO-Cre⁺ mice always contained only one transgenic allele, and RHO-Cre⁻ mice contained none. Breeders were checked every day for the birth of a litter during the previous night; the day a litter was first observed was regarded as postnatal day 0 (P0). Dark-reared mice were placed in total darkness on P0; they were handled thereafter using infrared light and infrared converting binoculars. For dark to light experiments (concerning the localization of arrestin), mice were kept in a dark room overnight and subjected to 600 lux for 15 min.

Microscopy

Eyes were enucleated and posterior eyecups were fixed by immersion. For conventional light and electron microscopy, eyecups were fixed in 2% glutaraldehyde + 2% paraformaldehyde in 0.1 M cacodylate buffer, followed by secondary fixation in 1% OsO4 and processing for embedment in Epon. Photoreceptor cell density was determined from images of dorso-ventral semithin (0.7 μm) sections, stained with toluidine blue or Azure II. Regions that were spaced at 0.5-mm intervals from the optic nerve head were identified; the central retina corresponds approximately with the region that is 0.5 mm dorsal to the optic nerve head. In each region, at least three representative columns of photoreceptor cell nuclei were counted to determine the thickness of the nuclear layer in terms of the number of rows of nuclei.

For immunoelectron microscopy, eyecups were fixed in 0.25% glutaraldehyde + 4% formaldehyde in 0.1 M cacodylate buffer, pH 7.4, and processed for embedment in LR White. Ultrathin sections (70 nm) were etched with saturated sodium periodate for 15 min, blocked with 4% BSA in antibody buffer (TBS + 1% Tween-20) for 1 h at room temperature and incubated with rod opsin antibody (1D4) in buffer overnight at 4°C. Sections were then washed and incubated with goat anti-mouse IgG conjugated to 12 nm gold (Jackson Lab) and stained with uranyl acetate and lead citrate. Sections from rhodopsin knockout mice were used as negative control. Immunolabeling density was determined by counting gold particles over the inner segment plasma membrane (defined as a 60-nm band, centered on the center of the membrane), and over the area enclosed by this membrane (designated as within the inner segment).

For immunofluorescence microscopy, eyes were fixed in 4% paraformaldehyde in PBS. Eyecups were washed in PBS and subjected to dehydration in ethanol and xylene. Samples were embedded in paraffin. Sections (5 μm) were mounted on glass slides. Before staining with antibodies, samples were rehydrated and blocked in 4% BSA in PBS. Autofluorescence was quenched with 50 mM ammonium chloride in phosphate-buffered saline (PBS). Antibodies were diluted into Antibody Buffer (AB; 2% goat serum and 0.01% Triton X-100 in PBS). Sections were incubated with primary antibody solutions overnight at 4°C and secondary antibody for 1 h at room temperature in the dark. Sections were mounted using anti-fading mounting media containing DAPI (Fluorogel II, EMS, USA) and analyzed on an Olympus FluoView 1000 confocal microscope. The antibodies used were as follows: 1D4 (mouse monoclonal anti-opsin), R7 (rat polyclonal anti-arrestin), anti-caspase 3 active form (rabbit polyclonal, Millipore, Bedford, MA), and anti-phospho-c-jun (Ser63) II (rabbit polyclonal, Cell Signaling). The secondary antibodies used were Alexa 568 or Alexa 488 goat anti-rabbit, goat anti-rat, or goat anti-mouse IgG (Molecular Probes, Eugene, OR).

Western Blot Analysis

Mouse eyecups were homogenized in lysis buffer [50 mM Tris, pH 7.4, 100 mM NaCl, 1 mM EDTA, 1 mM MgCl₂, 1 mM DTT and complete protease inhibitor cocktail (Sigma, St. Louis, MO)]. Equivalent amounts of sample were fractionated on a 4–12% Bis-Tris gel (Invitrogen, Carlsbad, CA) and transferred to PVDF membrane (Millipore). Membranes were blocked in PBS/0.05% Tween-20 with 4% BSA (blocking solution). The membrane was then probed with anti-opsin, pAb01 (1:10,000), in blocking solution, washed four times in PBS/0.05% Tween, and incubated with horseradish peroxidase-conjugated anti-rabbit antibody (1:30,000, Sigma). Bound antibody was detected using the ECL Dura Western Blotting detection system (Amersham, Piscataway, NJ).

The chemilumiscence signal detected was used to perform densitometry analysis in ImageJ, where the intensity was correlated with relative protein levels. Samples were analyzed in triplicate.

TUNEL Labeling

Detection of apoptotic nuclei by TUNEL was performed using the TACS TdT Kit (R&D Systems, Minneapolis, MN) according to the manufacturer's protocol. Briefly, the paraffin was removed from paraffin-embedded sections, which were then hydrated, followed by permeabilization with proteinase K. Sections were then subjected to quenching of endogenous peroxidase with Quenching Solution, and labeled with TdT for 1 h at 37 degrees. The labeling reaction was terminated with Stop Buffer. Detection was performed by incubation with streptavidin-HRP followed by diaminobenzidine (DAB) solution. Samples were analyzed on an Olympus FluoView 1000 confocal microscope. Positive and negative controls were used for the assay. As a positive control, a section of a WT retina was treated like the other sections, with the exception that before the quenching step, it was subjected to treatment with TACS-nuclease, to generate DNA breakage in most cells. As a negative control, a section of a WT retina was not labeled with the TdT labeling reaction mix. Quantification of the nuclei was performed using Image J software. Three different complete dorso-ventral sections per genotype were used. Because the negative controls showed no staining, stained (dark) nuclei were regarded as TUNEL-positive.

Electroretinographic Analysis

After overnight dark-adaptation, mice were anesthetized with an intraperitoneal injection of normal saline solution containing ketamine (15 μg/g) and xylazine (3 μg/g body weight). ERGs were recorded from the corneal surface of the eye after pupil dilation (1% atropine sulfate) using a gold loop corneal electrode together with a mouth reference and tail ground electrode. A drop of methylcellulose (2.5%), placed on the corneal surface, ensured electrical contact and corneal integrity. Responses were amplified (Grass CP511 AC amplifier, ×10,000; 3 dB down at 2 and 10,000 Hz) and digitized using an I/O board (National Instruments, PCI-1200) in a personal computer. Signal processing was performed with custom software (National Instruments, LabWindows/CVI). Mouse body temperature was maintained at a constant 38°C using a heated water pad.

All stimuli were presented in a large Ganzfeld dome (LKC Technologies), the interior surface painted with a highly reflective white matte paint (Eastman Kodak Corporation, #6080). Flash sequence and presentation frequency were under precise computer control. Rod-photoreceptor mediated ERGs were recorded to blue (Kodak Wratten 47A) light flashes over the range −4.67 to −0.44 log cd-s/m². Cone-photoreceptor mediated ERGs were recorded to white flashes (−0.54 to 0.70 log cd-s/m²) superimposed on a rod-saturating background (32 cd/m²) following 10 min of light adaptation.

RESULTS

Opsin Mislocalization in KIF3A-Depleted Photoreceptor Cells

Opsin mislocalization has been demonstrated by immunocytochemistry of KIF3A- or IFT-mutant photoreceptor cells (Marszalek et al., 2000; Pazour et al., 2002; Jimeno et al., 2006; Krock and Perkins, 2008; Avasthi et al., 2009; Sukumaran and Perkins, 2009). It is especially evident in mutant cone photoreceptor cells (Avasthi et al., 2009). The most obvious site of mislocalization is in the plasma membrane of the inner segment (Figure 1D) and nuclear layer (Supplemental Figure 1). However, there is also evidence of opsin within the inner segment of KIF3A-depleted photoreceptor cells (Marszalek et al., 2000; Jimeno et al., 2006). Here we quantified opsin immunogold labeling as a measure of opsin associated with intracellular and plasma membrane regions of the inner segments of RHO-Cre⁺;Kif3a^flox/flox rod photoreceptors at postnatal day 7 (P7), when opsin mislocalization can first be detected, and, a week later, at P14. The use of this RHO-Cre transgene results in synchronous excision of Kif3a^flox across the entire neonatal retina (Jimeno et al., 2006).

Figure 1.

ImmunoEM of opsin in the inner segment at P7 and P14. Electron micrographs of opsin immunogold labeled sections from wild-type (A and C) and RHO-Cre⁺;Kif3a^flox/flox (B and D) mice at P7 (A and B) and P14 (C and D). IS, rod inner segment; BB, basal body; CC, connecting cilium. Scale bar = 200 nm; all panels same magnification. Bar graphs, illustrating the density of opsin immunogold labeling within the inner segment (E) and along the inner segment plasma membrane (F) of rod photoreceptors, in relation to the genotype and age. Error bars represent ± SEM, where n = 3 eyes. Asterisks indicate p < 0.005, by Student's one-tail, t test. (G) Western blot analysis of rod opsin in P7 WT (labeled as Cre⁻) and RHO-Cre⁺;Kif3a^flox/flox (labeled as Cre⁺) eyecups, illustrating the same amount of opsin of the same mobility, despite depletion of KIF3A.

In P7 rod photoreceptors, the concentration of opsin labeling of the inner segment plasma membrane is similar between mutant and WT, but the amount of opsin labeling within the inner segment is significantly greater in mutant retinas (Figure 1). The overall amount of opsin in mutant and WT retinas is similar (Figure 1G), indicating that the additional opsin within the inner segment of mutant cells must be countered by less outer segment opsin. In P14 rod photoreceptors, there is still more opsin labeling within the inner segment of mutant compared with WT cells, but the striking difference is in the amount of opsin labeling associated with the plasma membrane of the inner segments, with 16-fold more in mutant retinas (Figure 1). These observations suggest that, in KIF3A-depleted rod photoreceptors, opsin is first backlogged along the intracellular trafficking route to the outer segment, then it accumulates in the plasma membrane outside of the outer segment. It is important to note that each KIF3A-depleted rod photoreceptor possesses a partly-developed outer segment [Figure 2B; see also Figure 5H in Jimeno et al. (2006) for immuno EM].

Figure 2.

Arrestin translocation in KIF3A-depleted rod photoreceptor cells. Paraffin sections (5 μm) of retinas from wild-type (WT) and RHO-Cre⁺;Kif3a^flox/flox (labeled as Kif3a^−/−) mice at P14. (A and B) Retinas labeled with anti-opsin (red) and DAPI (blue). Loss of KIF3A results in the abnormal accumulation of opsin in the IS and ONL (B). (C–J) Sections from dark and light adapted animals labeled with arrestin antibodies (green). Panels D, F, H, and J also show labeling with opsin antibodies (red) and with DAPI (blue), although no opsin is evident in D and H, because it is not mislocalized in these WT retinas. Yellow represents overlap between opsin and arrestin labeling. (C–F) After exposure to light, following the dark adaptation. (G–J) After the dark adaptation, before light exposure. (F′, J′) Higher magnification of the ONL region of the mutant retinas (areas magnified are outlined in F and J). There is a greater overlap of arrestin and mislocalized opsin in the light (F) than in the dark (J). Scale bars = 10 μm. OS, outer segment layer; IS, inner segment layer; ONL, outer nuclear layer; OPL, outer plexiform layer.

Figure 5.

Loss of KIF3A-depleted rod photoreceptors in mice with reduced rod opsin expression. (A–J) LM sections of photoreceptor cells in the region 0.5 mm dorsal to the optic nerve head, illustrating the extent of the photoreceptor cell nuclear layer, from 21-d old mice (A–E) and 28-d old mice (F–J). (A and F) WT, (B and G) RHO-Cre⁺;Kif3a^flox/flox (i.e., lacking KIF3A), (C and H) RHO-Cre⁺;Kif3a^flox/flox;Rho^+/− (i.e., lacking KIF3A and with ∼30% reduced rod opsin), (D and I) RHO-Cre⁺;Kif3a^flox/flox;Rho^−/− (i.e., lacking both opsin and KIF3A), (E and J) RHO-Cre⁻;Kif3a^flox/flox;Rho^−/− (i.e., lacking opsin, but with normal KIF3A, due to lack of Cre). All LMs are of the same magnification; scale bar = 50 μm. OS, outer segment layer; IS, inner segment layer; ONL, photoreceptor nuclear layer; OPL, photoreceptor synaptic layer. (K and L) Quantification of photoreceptor cell nuclei at 0.5-mm intervals from the optic nerve head in dorso-ventral sections of retinas from 21-d old mice (K) and from 28-d old mice (L). The depth of the photoreceptor cell nuclear layer is shown in terms of number of rows of cell nuclei. ON, optic nerve head; V1, V2 etc., 0.5 mm, 1.0 mm etc., from the optic nerve head on the ventral side; D1, D2 etc., 0.5 mm, 1.0 mm etc., from the optic nerve head on the dorsal side. M, Graph illustrating the time course of photoreceptor cell loss due to loss of KIF3A, and the retardation of this loss when rod opsin is reduced. The number of rows of the photoreceptor cell nuclei, in the region 0.5 mm dorsal to the optic nerve head, is shown. In K and L, error bars indicate ± SEM, where n > 3 animals for each genotype or condition. In all panels, RHO-Cre⁺;Kif3a^flox/flox has been indicated by Kif3a^−/− and RHO-Cre⁻;Kif3a^flox/flox by Kif3a^+/+.

Arrestin Translocation in KIF3A-Depleted Photoreceptor Cells

In P14 RHO-Cre⁺;Kif3a^flox/flox retinas, opsin mislocalization is evident by immunofluorescence and rod photoreceptor cell loss has only just begun (cf. Figure 2B with A) (Jimeno et al., 2006). We therefore chose this time-point to study the effect of kinesin-2 function on the localization of arrestin in light- and dark-adapted retinas by immunofluorescence.

As described previously (reviewed in Strissel et al., 2006), arrestin normally migrates to the outer segment in the light, so that its concentration in the inner segment and nuclear layers of WT retinas is much greater in the dark than in the light (cf. Figure 2, G and H with 2, C and D). In KIF3A-depleted photoreceptors, arrestin remained in the inner segment and nuclear layers in both the light and dark (Figure 2, E and I). This observation is consistent with transport of arrestin by kinesin-2; however, on closer examination, it was clear that most of the arrestin becomes colocalized with mislocalized opsin in the light (cf. Figure 2, F and J, including higher magnification images shown in F′ and J′, which illustrate an increased overlap of the opsin and arrestin immunolabeling in the light). Hence, it is plausible that arrestin is simply prevented from migrating to the outer segment due to binding mislocalized photoactivated rhodopsin. In any case, it is important to note for subsequent parts of the current study, that light exposure appears to promote the formation of opsin–arrestin complexes in the inner segment and nuclear layers of KIF3A-depleted photoreceptors.

Test for Requirement of Arrestin and Light in Eliciting Photoreceptor Cell Loss

To test whether complexes of mislocalized rhodopsin and arrestin were required for the photoreceptor cell loss, we examined retinas from P21 RHO-Cre⁺;Kif3a^flox/flox mice that also lacked arrestin (Sag^−/−). Further, to eliminate photoactivation of rhodopsin and the phototransductive cascade, we reared animals in darkness. P21 was the most appropriate age to sample, because the steepest decline in RHO-Cre⁺;Kif3a^flox/flox photoreceptor cell number occurs during the third postnatal week (Jimeno et al., 2006).

RHO-Cre⁺;Kif3a^flox/flox mice showed severe photoreceptor cell loss by P21, with only 1–2 rows of photoreceptor cell nuclei evident in the central retina. Many of the remaining nuclei are from cones (containing multiple small clumps of heterochromatin) or rods that appear to be apoptotic (Figure 3). Dark rearing had no effect on this loss of cells, nor did lack of arrestin (Sag) stem this loss, because RHO-Cre⁺;Kif3a^flox/flox;Sag^−/− retinas and RHO-Cre⁺;Kif3a^flox/flox retinas from animals, reared in normal lighting or in darkness, all showed a similar degree of photoreceptor degeneration (Figure 3). Note that, although the RHO-Cre⁺ transgene has been reported to effect retinal degeneration by itself, presumably due to Cre toxicity, degeneration is not evident until P49, with P28 animals showing no histological or electrophysiological defects (Jimeno et al., 2006).

Figure 3.

Loss of KIF3A-depleted rod photoreceptors in mice maintained in the dark or lacking arrestin. LM sections of photoreceptor cells in the region 0.5 mm dorsal to the optic nerve head, illustrating the extent of the photoreceptor cell nuclear layer in 21-d old mice that were (A) WT, (B) RHO-Cre⁺;Kif3a^flox/flox (i.e., lacking KIF3A), (C) RHO-Cre⁺;Kif3a^flox/flox, (i.e., lacking KIF3A, reared in constant darkness), (D) RHO-Cre⁺;Kif3a^flox/flox;Sag^−/− (i.e., lacking both arrestin and KIF3A), or (E) RHO-Cre⁻;Kif3a^flox/flox;Sag^−/− (i.e., lacking arrestin, but with normal KIF3A, due to lack of Cre). On the images, RHO-Cre⁺;Kif3a^flox/flox has been indicated by Kif3a^−/− and RHO-Cre⁻;Kif3a^flox/flox by Kif3a^+/+. All images were taken near the center of the retina, 0.5 mm dorsal from the optic nerve head. Except for C, all animals were reared under the normal light/dark cycle. All LMs are of the same magnification; scale bar = 50 μm. OS, outer segment layer; IS, inner segment layer; ONL, photoreceptor nuclear layer; OPL, photoreceptor synaptic layer. (F) Bar graph showing the counts of photoreceptor cells in relation to the different genotypes and lighting conditions; a–e correspond to the genotypes shown in A–E, respectively. The depth of the photoreceptor cell nuclear layer, in the region 0.5 mm dorsal to the optic nerve head, is expressed in terms of the number of rows of nuclei that span the layer (indicated as “ONL” in the micrographs). Error bars indicate SEM, where n > 3 animals for each genotype or condition. Neither light nor arrestin is required for photoreceptor degeneration due to loss of KIF3A.

Activation of the phototransductive cascade was also prevented by generating mice that lacked alpha transducin (Gnat1). The extent of photoreceptor cell loss in retinas from P21 RHO-Cre⁺;Kif3a^flox/flox;Gnat1^−/− mice was similar to that in retinas from P21 RHO-Cre⁺;Kif3a^flox/flox mice (Supplemental Figure 2), supporting the result obtained from rearing the mice in the dark. Photoreceptor cell degeneration due to loss of kinesin-2 function therefore does not require signaling from an opsin-arrestin complex, activation of normal phototransduction, or any photoactivation of rhodopsin.

Test for Requirement of Rod Opsin in Eliciting Photoreceptor Cell Loss

In the next set of experiments, we tested whether decreasing the expression of rhodopsin would affect the rate of photoreceptor cell loss. For the first month of life, retinas from heterozygous Rho knockout (Rho^+/−) mice contain the same number of photoreceptor cells as retinas of control (Rho^+/+) mice, although the individual rod photoreceptors contain only ∼70% of the amount of rhodopsin present in control mice (Figure 4) (Lem et al., 1999). Despite the reduced opsin levels, mislocalized opsin is still evident after the depletion of KIF3A (Supplemental Figure 3). In comparing control, RHO-Cre⁺;Kif3a^flox/flox, and RHO-Cre⁺;Kif3a^flox/flox;Rho^+/− retinas from P21 mice, we found that the presence of a single Rho⁻ allele, and thus reduced rhodopsin content, resulted in significant, although incomplete rescue of photoreceptor cells (Figure 5, A–C). RHO-Cre⁺;Kif3a^flox/flox;Rho^+/− retinas had approximately three times as many photoreceptor cells across the retina as did RHO-Cre⁺;Kif3a^flox/flox retinas (Figure 5K).

Figure 4.

Rhodopsin levels in retinas of mice, with two (WT, +/+), one (heterozygote, +/−), or no (null, −/−) functional Rho genes. (A) Western blot analysis of rhodopsin levels in Rho^+/+, Rho^+/−, and Rho^−/− mice. An equivalent portion of eyecup lysate was loaded for each genotype. The position of a 50-kDa molecular mass marker is shown on the left. (B) Bar graph showing the relative amount of rhodopsin in Rho^+/+, Rho^+/−, and Rho^−/− eyes, measured from Western blots, such as that shown in A, by densitometry, and expressed as a percentage of that in Rho^+/+ eyes. Loss of one copy of the rhodopsin gene results in a 28% reduction in opsin levels. Error bars represent ± SEM, where n = 3 animals.

By P28, all rod photoreceptor cells have been lost from RHO-Cre⁺;Kif3a^flox/flox retinas; the only remaining photoreceptor cell nuclei are from cone photoreceptor cells, which form a single row of cell nuclei (Figure 5G) (cf. Jimeno et al., 2006). RHO-Cre⁺;Kif3a^flox/flox;Rho^+/− retinas from P28 mice were also severely degenerate, although they contained ∼3 rows of photoreceptor cell nuclei, indicating that a significant number of rod photoreceptor cells still survived, and the rescue effect of Rho^+/− was still evident (Figure 5H).

Rho^−/− mouse retinas do not contain detectable rod opsin (Figure 4) (Lem et al., 1999). We therefore tested whether the complete absence of rod opsin would inhibit photoreceptor cell loss further. This proved to be the case. At both P21 and P28, RHO-Cre⁺;Kif3a^flox/flox;Rho^−/− mice have significantly more photoreceptor cells remaining than RHO-Cre⁺;Kif3a^flox/flox;Rho^+/− mice (Figure 5, D and I). This additional rescue is despite some photoreceptor cell loss due to the lack of RHO alone (Lem et al., 1999) (Figure 5E and J). Indeed, the contribution of KIF3A depletion to photoreceptor cell loss in Rho^−/− mice was found to be only minor, and counts taken from some regions of P21, P28, and P35 mice indicated no significant effect of KIF3A depletion (Figure 5, K–M).

Figure 5M shows time course plots of photoreceptor cell death with different expression levels of opsin. The effect of reducing opsin by ∼30% (as in Rho^+/− mice) results in a delay in photoreceptor cell loss of approximately one week. Elimination of opsin results in a larger delay, although this time course becomes confounded by the degenerative effect of Rho^−/−. The presence of the RHO-Cre transgene also eventually causes photoreceptor cell loss, but none is detectable at P28 (Jimeno et al., 2006).

Test for Requirement of Rod Opsin in Eliciting Photoreceptor Cell Apoptosis

Previously, we showed that aberrant opsin localization in IRBP-Cre⁺;Kif3a^flox/flox retinas was followed by the appearance of TUNEL-positive nuclei (Marszalek et al., 2000). Here, in addition to the TUNEL assay, which does not discriminate among autolytic, necrotic, and apoptotic cells (Grasl-Kraupp et al., 1995), we measured the presence of two more specific apoptotic indicators, immunofluorescence labeling with antibodies against c-Jun that is phosphorylated at Ser63 and with antibodies against activated caspase-3. We also counted condensed nuclei by electron microscopy. All four indicators were greatly increased in P14 RHO-Cre⁺;Kif3a^flox/flox retinas compared with P14 retinas from WT mice or mice lacking the floxed Kif3a gene or RHO-Cre.

Next, we tested whether the protective effect of reduced opsin expression, in terms of inhibiting the loss of KIF3A-depleted rods, was due to inhibition of apoptosis, as determined by these four indicators. We studied retinas from P14 mice of different genotypes. By the TUNEL assay, 30% of the photoreceptor nuclei in RHO-Cre⁺;Kif3a^flox/flox retinas were positive, approximately six times the number of positive nuclei in retinas that also lacked RHO (from RHO-Cre⁺;Kif3a^flox/flox;Rho^−/− mice) (Figure 6A). By electron microscopy, 20% of the photoreceptor nuclei appeared condensed in RHO-Cre⁺;Kif3a^flox/flox retinas, whereas no condensed nuclei were observed in retinas that also lacked RHO (from RHO-Cre⁺;Kif3a^flox/flox;Rho^−/− mice) or did not contain Kif3a^flox/flox (Figure 6B-D). Similarly, phospho-c-Jun (Figure 6E) and activated caspase-3 (Figure 6F) immunofluorescence was evident in RHO-Cre⁺;Kif3a^flox/flox retinas, but negligible in retinas that lacked RHO or were wild-type for Kif3a. These results indicate that the loss of KIF3A-depleted rod photoreceptor cells is due to caspase-dependent apoptotic cell death, and that this pathway is triggered by the mislocalization of opsin.

Figure 6.

Apoptosis in relation to rod opsin expression. (A) Bar graph illustrating counts of apoptotic photoreceptor nuclei detected by a positive reaction in the TUNEL assay. (B and C) Electron micrographs of ultrathin sections (70 nm) of retinas from WT (A) and RHO-Cre⁺;Kif3a^flox/flox (B) mice at P14. Scale bar = 0.5 μm. A typical condensed apoptotic nucleus is shown in C. Such nuclei were scored as apoptotic for the bar graph shown in D. (D) Bar graph illustrating counts of apoptotic nuclei detected by condensed nuclei in electron micrographs. Only RHO-Cre⁺;Kif3a^flox/flox animals (i.e., Kif3a^−/− and Rho^+/+) had apoptotic nuclei detectable by electron microscopy. Error bars indicate ± SEM, where n > 3 animals for each genotype or condition. (E and F) Paraffin sections (5 μm) of retinas labeled with antibodies against phosphorylated (Ser63) c-Jun (E) or against cleaved (activated) caspase 3 (F), shown in green, and with DAPI (blue). Scale bar = 15 μm. Mice contained rod photoreceptors that were either homozygous WT (+) or homozygous null (−) for Kif3a and Rho. Error bars indicate ± SEM, where n > 3 animals for each genotype or condition.

ERG Measurements of Photoreceptor Function

To test whether the protective effect of reduced opsin expression in KIF3A-depleted rods was evident in terms of photoreceptor function as well as survival, we measured ERG responses in mice of different genotypes at P21 and at P35. At P21, rod photoreceptor-mediated responses had higher amplitudes in RHO-Cre⁺;Kif3a^flox/flox;Rho^+/− mice in comparison with RHO-Cre⁺;Kif3a^flox/flox mice, indicating that the additional surviving rods, due to decreased RHO, are functional (Figure 7A). This increase in amplitude was evident despite reduced amplitudes in Rho^+/− mice compared with WT mice. Cone-mediated responses were also greater in the RHO-Cre⁺;Kif3a^flox/floxmice with reduced (Rho^+/−) or no RHO than in RHO-Cre⁺;Kif3a^flox/flox mice with Rho^+/+ (Figure 7C), indicating that the increased rod survival extends to increased cone function. At P35, cone responses in RHO-Cre⁺;Kif3a^flox/flox;Rho^−/− mice were clearly detectable, whereas we could not detect any cone responses in the RHO-Cre⁺;Kif3a^flox/flox mice with Rho^+/+ (Figure 7D). Nevertheless, the cone responses from retinas of RHO-Cre⁺;Kif3a^flox/flox;Rho^−/− mice were significantly smaller than those in the Rho^−/− mice. Together these ERG measurements show that the increased photoreceptor survival due to reduction of RHO in KIF3A-depleted rod photoreceptors also results in increased rod and cone responsiveness.

Figure 7.

Intensity-response functions for dark- (A and B) and light- (C and D) adapted retinas. ERG responses were recorded at P21 (A and C) and at P35 (B and D). In A–D, b-wave amplitude, defined conventionally as the amplitude from the trough of the a-wave to the peak of the b-wave, is plotted against increasing intensity. RHO-Cre⁺;Kif3a^flox/flox has been indicated by Kif3a^−/− and RHO-Cre⁻;Kif3a^flox/flox by Kif3a^+/+. For the dark-adapted ERGs (A and B) responses obtained to intensities below about −1.5 log cd-s/m² are assumed to be mediated by rods, whereas responses to higher intensities, while rod-dominated, also mask a small contribution from cones (Nusinowitz et al., 2007). Thus, the small dark-adapted response at the highest intensities in Rho^−/− mice likely reflects this residual cone signal. The light-adapted ERGs (C and D) are mediated by cones. Error bars indicate ± SEM.

DISCUSSION

Here, we have explored further the role of kinesin-2 in rod photoreceptor cells and in particular the basis of the very rapid photoreceptor cell loss that follows the loss the kinesin-2 function. We report that arrestin fails to translocate to the outer segment upon light exposure in KIF3A-depleted rods; it instead associates with mislocalized opsin. However, unlike in a number of other models of photoreceptor degeneration, signaling from opsin–arrestin complexes is not required for cell death. Indeed, light-activation of rhodopsin is not even required. Our results indicate that the critical event is simply mislocalization of opsin, which, as we show, accumulates in large amounts first within the inner segment and then in the plasma membrane of the inner segment and nuclear region. Reduction in the expression of opsin in KIF3A-depleted rods markedly inhibits apoptosis and photoreceptor cell loss, and preserves photoreceptor function.

Protein traffic along the photoreceptor cilium is extremely busy, thus representing an extraordinary example of IFT. Ten percent of the opsin-containing disk membranes of the outer segment in mammalian rods are renewed each day (Besharse and Hollyfield, 1979). The mislocalization of opsin that occurs in response to loss of function of kinesin-2 or IFT88 (Marszalek et al., 2000; Pazour et al., 2002; Jimeno et al., 2006) suggests that IFT powers the delivery of at least this protein. Opsin is the most abundant protein of the outer segment (Papermaster et al., 1975), at ∼7 × 10⁷ molecules per mouse rod (Lyubarsky et al., 2004). Nevertheless, arrestin is also very highly concentrated in rod photoreceptor cells, with about one molecule for every 1.3 molecules of opsin (Strissel et al., 2006). Given that the majority of arrestin (and other soluble signaling proteins, transducin and recoverin) translocates between the inner and outer segment at every light-dark transition (Broekhuyse et al., 1985; Strissel et al., 2004), the trafficking of soluble signaling proteins involves orders of magnitude more ciliary traffic than opsin and other disk membrane proteins. So, if IFT were responsible for the movement of these soluble signaling proteins, this would be, by far, its major task. Calvert et al. argued that diffusion could account for the delivery of arrestin and transducin to the outer segment (Calvert et al., 2006). Our results are consistent with their argument. Although arrestin is mislocalized in light-adapted KIF3A-depleted rods, our results suggest that this defect could be a result of binding to mislocalized rhodopsin, and so do not necessarily indicate that kinesin-2 transports arrestin.

The mislocalization of opsin observed in the first studies of genetic disruption of kinesin-2/IFT was taken as evidence that opsin is transported by kinesin-2/IFT along the photoreceptor cilium (Marszalek et al., 2000; Pazour et al., 2002). More recently, this suggestion has been questioned for rods, because, in comparison to cones, the mislocalization of opsin (and other outer segment proteins) seemed relatively minor (Avasthi et al., 2009). However, the present results document quantitatively that a backlog of opsin occurs very early in KIF3A-depleted rods (at P7, long before any cell death) and that rod opsin expression is key to the death of these cells. Both these observations are consistent with a direct link between kinesin-2 function and normal delivery of opsin to the outer segment, as suggested in earlier reports.

The high concentration of rhodopsin in a rod photoreceptor cell provides exquisite light sensitivity, but this level is close to the maximum that the cell can tolerate. An increase in opsin levels by just ∼23% is sufficient to induce photoreceptor cell degeneration (Tan et al., 2001). Our results indicate that aberrant accumulation of this highly abundant protein outside of the outer segment initiates cell death, and that the essential factor for apoptosis is the abundance of the protein, rather than its signaling properties as a G-protein coupled receptor. It is possible that the abnormal accumulation of other outer segment proteins may also promote cell death. However, the nearly complete negation of loss of KIF3A-depleted rods by Rho^−/− indicates that the accumulation of opsin is a major initiator of cell death.

A variety of models of inherited photoreceptor degeneration, such as rd1 and rds mice and the RCS rat, also demonstrate mislocalized opsin before cell death (Nir and Papermaster, 1986; Jansen et al., 1987; Nir et al., 1987; Usukura and Bok, 1987; Nir et al., 1989). However, the opsin in these examples appears to be restricted to the plasma membrane of the cell and is thus distinct from the excessive amount of opsin that accumulates initially within the inner segment, as a result of kinesin-2 dysfunction. It may represent redistribution from the outer segment, as a secondary response in ailing cells that are dying by a different mechanism. Interestingly, cell death in at least the rd1 and rds mice may be largely caspase-independent (Doonan et al., 2005; Lohr et al., 2006). In contrast to the increased capase-3 activity we detected in P14 KIF3A-depleted photoreceptor cells, labeling of active capase-3 was not detected in photoreceptor nuclei of P7–13 rd1 retinas (Doonan et al., 2005).

Different mutations in rod opsin have been shown to affect its delivery to the outer segment. Some of these mutations perturb the correct folding of the protein (Sung et al., 1991; Chapple et al., 2001). In cell culture, the P23H mutant opsin accumulates in aggresomes and is targeted to the proteasome system, which it impairs (Illing et al., 2002). In transgenic mice, P23H mutant opsin accumulates in perinuclear regions, at least in the absence of wild-type rod opsin (Frederick et al., 2001). However, there is no suggestion that the nontrafficked opsin in KIF3A-depleted photoreceptor cells is misfolded. The primary cellular effect of dysfunctional kinesin-2 in photoreceptor cells appears to be loss of transport along the cilium, with the accumulation of opsin in the inner segment, resulting from a backlog. Concanavalin A labels the accumulated opsin, indicating that it has been processed normally through the endoplasmic reticulum and Golgi apparatus (Marszalek et al., 2000).

A second class of opsin mutants appears to have defective targeting to the outer segment, and are therefore more comparable to kinesin-2 dysfunction. An important factor in the targeting of opsin to the outer segment is the presence of a localization signal and protein binding sites in the C-terminal region of the protein (Deretic et al., 1998; Tam et al., 2000; Deretic et al., 2005). Mutations in the C-terminal region of rod opsin result in defects in the targeting of opsin to the outer segment, although, interestingly, despite numerous reports on such models (e.g., Sung et al., 1994; Li et al., 1996; Green et al., 2000; Concepcion et al., 2002; Tam and Moritz, 2006), invariably, mislocalized opsin has been shown throughout the plasma membrane of the cell, with no significant opsin accumulation inside the cell. We demonstrate here that, with the dysfunction of a molecular motor, there is first a backlog of opsin along the route to the outer segment within the cell (Figure 1). This backlog occurs despite the presence of a partly-developed outer segment (Figure 2B, Suppl. Figure 3) (Jimeno et al., 2006). Nevertheless, opsin C-terminal mutations appear to be similar to kinesin-2 dysfunction in that apoptosis does not appear to require normal opsin signaling, since retinal degeneration still proceeds under darkness (Green et al., 2000; Tam and Moritz, 2006).

A third type of opsin mutation that poses an interesting contrast to our observations with depleted KIF3A is the constitutively active K296E opsin. As shown recently, this opsin is phosphorylated not just in the light, but also in the dark, and is mislocalized throughout the photoreceptor cell only in the presence of arrestin (Chen et al., 2006). Chen et al. have argued that this mutant opsin forms a complex with arrestin to activate apoptosis, similar to that reported for some Drosophila retinal degenerations (Alloway et al., 2000; Kiselev et al., 2000).

In addition to kinesin-2 and IFT proteins, there has been recent interest in proteins that are associated with the connecting cilium, and whose function may be related to transport along this structure. Many of these proteins are associated with retinal degenerations, such as Bardet-Biedl syndrome (BBS proteins), Senior Loken syndrome (NPHP proteins), Leber's congenital amaurosis (CEP290/NPHP6), and X-linked retinitis pigmentosa (RPGR, RPGRIP). In mouse models of some of these diseases, opsin has been reported to be present throughout the photoreceptor cell membrane before retinal degeneration (Zhao et al., 2003; Nishimura et al., 2004; Chang et al., 2006). In a closer examination of opsin distribution in Ahi1^−/− mice, another ciliopathic model with photoreceptor cell loss, an abnormal accumulation of opsin was found also within the inner segment before cell death (Louie et al., 2010). It seems likely that mislocalized opsin may be an important trigger for retinal degeneration in ciliopathies in general.

Because the ectopic opsin in KIF3A-depleted photoreceptor cells is not providing any necessary signal due to photoactivation or by forming a complex with arrestin, how does it activate apoptosis? Looking further afield, among neurodegenerations in general, there are clear examples of cellular catastrophe that ensues from blockage of transport. Huge accumulations of debris were demonstrated in axons of kinesin-mutant Drosophila (Hurd and Saxton, 1996). A variety of neurodegenerative diseases have since been linked to defects in axonal transport resulting from compromised molecular motor function (summarized in Perlson et al., 2010). Examples include Amyotrophic Lateral Sclerosis (ALS) and Alzheimer's disease (AD). A cause of familial ALS is mutant SOD1 (which encodes superoxide dismutase 1). Mutant SOD1 in mice has a number of consequences, but one includes possible effects on dynein and impairment of retrograde transport in axons of motor neurons (Chevalier-Larsen and Holzbaur, 2006). Mutations in APP (which encodes amyloid precursor protein) have been associated with familial AD. The axonal transport of APP is mediated by direct binding to a light chain subunit of kinesin-1 (KLC1) (Kamal et al., 2000). Lowering the expression level of KLC1 in APP-mutant mice promotes axonal swellings, containing accumulated cellular material, and increased amyloid deposition (Stokin et al., 2005). A common theme in these neurodegenerative disorders is that a relatively specific trafficking defect progresses to have a general effect on axonal transport. The loss of kinesin-2 function in photoreceptor cells has a similar consequence; first, opsin and perhaps some other cotransported proteins accumulate, then the trafficking of other cellular components, which use alternative pathways but still travel to the outer segment (e.g., the protein, peripherin/rds), is affected. The sheer mass of the resulting backlog appears to be the critical factor, and backlogs in different neurodegenerations might thus elicit a similar path to cell death.