Heterotrimeric Kinesin-2 (KIF3) Mediates Transition Zone and Axoneme Formation of Mouse Photoreceptors*

Background: Heterotrimeric kinesin-2 (KIF3) has been implicated in intraflagellar trafficking of photoreceptor membrane proteins by IFT. Results: KIF3 and IFT88 are required for transition zone and axoneme formation, but are dispensable for rhodopsin trafficking. Conclusion: Transmembrane proteins, including rhodopsin, traffic to the OS even when IFT is disabled. Significance: KIF3 builds and maintains the photoreceptor transition zone and axoneme that are essential for photoreceptor integrity and vision. Anterograde intraflagellar transport (IFT) employing kinesin-2 molecular motors has been implicated in trafficking of photoreceptor outer segment proteins. We generated embryonic retina-specific (prefix “emb”) and adult tamoxifen-induced (prefix “tam”) deletions of KIF3a and IFT88 in adult mice to study photoreceptor ciliogenesis and protein trafficking. In embKif3a−/− and in embIft88−/− mice, basal bodies failed to extend transition zones (connecting cilia) with outer segments, and visual pigments mistrafficked. In contrast, tamKif3a−/− and tamIft88−/− photoreceptor axonemes disintegrated slowly post-induction, starting distally, but rhodopsin and cone pigments trafficked normally for more than 2 weeks, a time interval during which the outer segment is completely renewed. The results demonstrate that visual pigments transport to the retinal outer segment despite removal of KIF3 and IFT88, and KIF3-mediated anterograde IFT is responsible for photoreceptor transition zone and axoneme formation.

Primary cilia are axoneme-stabilized protrusions that elongate from the basal body (mother centriole) of most vertebrate cell types (1)(2)(3). Cilia serve as a cell's "antenna" to sample a broad range of extracellular signals in olfaction, vision, taste, and mechanosensation (4,5). Primary cilia of kidney epithelia are mechanosensitive, detecting fluid flow through the tubule (6). Cilia of olfactory epithelia distinguish a large number of odorants (7). Photoreceptor sensory cilia detect light of different wavelengths and are responsible for dim light and color vision (8). The photoreceptor sensory cilium is composed of the outer segment (OS) 3 containing a stack of disc membranes, a basal body (microtubule-organizing center), and an axoneme. Each OS communicates with its inner segment (IS) through a connecting cilium, a structure equivalent to the transition zone of primary cilia (9). In the following, the term "connecting cilium" will be replaced by "photoreceptor transition zone" or PTZ.
Intraflagellar transport (IFT) is a bidirectional ciliary trafficking pathway conserved between invertebrates and vertebrates (5). Protein complexes and IFT particles (IFT-A and IFT-B) traffic along the flagellar/ciliary axoneme, a pathway integral for assembly, maintenance, and function of motile and primary cilia (10 -12). IFT defects cause human syndromic and nonsyndromic ciliopathies, a group of diseases affecting the function of sensory cilia and survival of entire organs (13). The IFT-B particle IFT88 (Tg737 or Polaris), located at basal bodies and axonemes, is required for anterograde trafficking and ciliary assembly (10,14). Chlamydomonas mutants lacking IFT88 are viable but do not form flagella (14), and zebrafish Ift88 mutants never develop photoreceptor outer segments (15). Germline deletions of IFT88 are embryonically lethal (14,16), and hypomorphic mutations (Tg737 orpk mice) cause complex ciliopathies affecting multiple tissues including photoreceptors, kidney, and olfactory neurons (17)(18)(19)(20). IFT88 was shown by pulldowns to be associated with rhodopsin, GC1, and a chaperone termed mammalian relative of Dna-J (MRJ) suggesting that outer segment transmembrane proteins are IFT cargo (21). However, the precise composition of IFT cargo is unknown.
This study addresses KIF3 and IFT88 in mouse photoreceptor ciliogenesis, IFT, and rhodopsin trafficking. We used a twotiered conditional approach to determine whether time of deletion affected the photoreceptor phenotype, i.e. deletion of each KIF3A and IFT88 in photoreceptor progenitors and depletion in adult photoreceptors by tamoxifen induction. Retina-specific deletion of either KIF3A or IFT88 during early development resulted in failure to form PTZs; depletion of either KIF3A or IFT88 by tamoxifen induction resulted in progressive, distal shortening of the OS axoneme, despite continued rhodopsin trafficking for at least 10 days. Our data indicate that the phenotype of KIF3 loss strongly depends on the time of Credriven recombination, and that KIF3-driven IFT functions primarily in photoreceptor ciliogenesis, PTZ formation, and axoneme stabilization.

EXPERIMENTAL PROCEDURES
Mice-Kif3a flox/flox and Ift88 flox/flox mice (29,36) were used to generate Kif3a and Ift88 conditional knock-out mice. Several Cre mouse strains, Six3Cre, CreER (The Jackson Laboratory stock number 004682), EIIaCre (The Jackson Laboratory stock number 003724), and iCre75, were used to disrupt conditionally the Kif3a and Ift88 genes (32,(37)(38)(39)(40). Rho Ϫ/Ϫ mice (41) and transgenic mice expressing S-opsin protein in rod photoreceptors (42) were obtained from Janis Lem (Tufts University) and Jason Chen (Baylor College of Medicine), respectively. A transgenic mouse expressing GFP-CETN2 fusion protein (The Jackson Laboratory stock number 008234) permits observation of centrioles and photoreceptor-connecting cilia with fluorescence microscopy (43). To induce Cre activity in CreER mouse, tamoxifen (150 mg/kg of body weight) was injected intraperitoneally at 1-2 months of age for 5 consecutive days. Tamoxifen stock solution (20 mg/ml) was prepared by dissolving tamoxifen powder in corn oil (each from Sigma-Aldrich). The Institutional Animal Care and Use Committee (IACUC) of the University of Utah, in compliance with statements for animal use of the Association for Research in Vision and Ophthalmology (ARVO), approved all experiments. Mice were maintained under 12-h cyclic dark/light conditions.
Western Blot-Mouse retinas were lysed by sonication in radioimmuneprecipitation assay buffer (150 mM NaCl, 1% NP-40, 0.5% sodium deoxycholate, 1% SDS, 50 mM Tris-HCl, pH 8.0). The supernatant of each lysate was separated by 10% SDS-PAGE (ϳ15 g of protein/well), and then transferred to a nitrocellulose membrane (Bio-Rad). The membrane was probed subsequently with primary antibodies, followed by HRP-conjugated secondary antibody. Phosphorescence (ECL system, PerkinElmer) was used to visualize the signal on x-ray film.
Immunohistochemistry-Mouse eyeballs were isolated and immediately immersion-fixed with 4% paraformaldehyde in 0.1 M phosphate buffer, pH 7.4, for 2 h on ice (45). After removal of the anterior segment, each eyecup was equilibrated sequentially with 15 and 30% sucrose in 0.1 M phosphate buffer for cryoprotection. The eyecups were then embedded in OCT and cut into 12-m-thick retina sections. The sections were incubated in each primary antibody at 4°C overnight after blocking with 10% donkey serum and washed three times with phosphate buffer. Sections were then incubated in fluorescent-conjugated secondary antibodies at room temperature for 1-2 h, washed three times, and mounted with VECTASHIELD medium (Vector Laboratories, Burlingame, CA). DAPI (1:50000, Invitrogen) was added to the secondary antibody solution to distinguish nuclei. Images were captured using an Olympus FluoView 1000 confocal microscope with either 60ϫ and/or 100ϫ objective lenses.
Electron Microscopy-Isolated mouse eyecups were fixed by immersion in fixative solution (2% glutaraldehyde, 1% paraformaldehyde in 0.1 M cacodylate buffer, pH 7.4) at 4°C overnight (32,46). The eyecups were then postfixed with 1% osmium tetroxide in 0.1 M cacodylate for 1 h, buffer-rinsed, stained en bloc with uranyl acetate, and subsequently dehydrated in an ascending series of methanol solutions. Eyecups were embedded in Epon resin (Ted Pella, Inc., Redding, CA) for sectioning. 1-m sections were cut to orient photoreceptors near the optic nerve. Retina ultrathin (60-nm) sections were cut onto slot grids with carbon-coated Formvar film (Electron Microscopy Sciences, Hatfield, PA) and post-stained with uranyl acetate followed by lead citrate. Transmission electron microscopy was performed at 75 kV using a JOEL electron microscope.
Electroretinography (ERG)-Full-field ERGs were performed on Kif3a mutant and control mice using a UTAS E-3000 system (LKC Technologies, Inc.; Gaithersburg, MD) as described (45). Briefly, dark-adapted mice were anesthetized by intraperitoneal injection with 10 mg/ml ketamine, 1 mg/ml xylazine (10 l/g of body weight), and pupils were dilated by applying a drop of 2.5% phenylephrine (Akorn, Inc., Decatur, IL). For scotopic ERG, mice were stimulated by flashes with different intensities ranging from Ϫ40 db (Ϫ3.4 log cd m Ϫ2 ) to 25 db (2.9 log cd m Ϫ2 ) without initial rod saturation, and their retinal responses were recorded. To record photopic ERG responses, a rod-saturating background light of 10 db (1.48 log cd m Ϫ2 ) was applied for 20 min before and during recording. Single-flash responses were usually recorded at stimulus intensities of Ϫ10 db (Ϫ0.6 log cd m Ϫ2 ) to 25 db (2.9 log cd m Ϫ2 ). Five or fewer flashes were averaged per intensity level with longer flash intervals with increasing intensity.
3-5 mice were tested for each group.
Data Analysis-Retina degeneration was evaluated by measurement of the outer nuclear layer (ONL) thickness in m (n ϭ 3) with standard deviation. About 100 -150 photoreceptors were examined for PTZ formation of each type, and the results were presented as percentages of cells with and without PTZ. Photoreceptor axoneme length was measured in ImageJ (50 photoreceptors each type) and analyzed with t test. ERG data (n ϭ 3) was analyzed by two-way analysis of variance.

KIF3 and Mouse Photoreceptor IFT
The absence of GFP-CETN2-labeled PTZs in P12 emb Kif3a Ϫ/Ϫ rods prompted us to examine the ultrastructure of emb Kif3a Ϫ/Ϫ PTZs and centrioles. In P6 emb Kif3a ϩ/Ϫ controls, we observed PTZ "buds" protruding from one basal body anchored at the photoreceptor cortex with subdistal and distal appendages (Fig. 3D, arrow and arrowhead, respectively). By P10, nascent OSs containing stacks of membrane discs appear at the PTZ distal portion (Fig. 3F). However, in emb Kif3a Ϫ/Ϫ photoreceptors, basal bodies docked at the apical IS with normal centriolar appendages (Fig. 3E, arrow and arrowhead), but failed to extend axonemes toward PTZ development at P6 (Fig.  3E) and toward OS at P10 (Fig. 3, G and H). Both photoreceptor ultrastructure and GFP-CETN2 labeling indicate aborted ciliogenesis in emb Kif3a Ϫ/Ϫ photoreceptors, thereby linking KIF3 to transition zone formation. The transition zone has been proposed as a gate for ciliary protein trafficking in primary cilia (50).
IFT Drives PTZ and Outer Segment Formation-A recent study of mouse embryonic fibroblasts revealed that KIF3 during initial stages of ciliogenesis assumes a novel role in centriolar organization that is distinct from its IFT motor function (51). In the absence of KIF3A, subdistal appendages, microtubule anchoring, and basal foot formation were abrogated. It is therefore important to discriminate the role of KIF3 in transition zone formation via IFT or its non-IFT organization of centriolar subdistal appendages organization. We disrupted photoreceptor IFT by deletion of IFT88, a critical subunit of the IFT-B complex that is essential for proper IFT in C. elegans (52,53), Chlamydomonas (14), Xenopus (54), mouse (17), and humans (55). By crossing an Ift88 floxed mouse line (36) with Six3Cre mice, we generated Ift88 f/f ;Six3Cre ϩ ( emb Ift88 Ϫ/Ϫ ) mice in which anterograde IFT is disabled in photoreceptor progenitors. We verified the absence of IFT88 in P10 emb Ift88 Ϫ/Ϫ photoreceptors by immunohistochemistry using an anti-IFT88 antibody (Fig. 3I), and assayed for PTZ formation in emb Ift88 Ϫ/Ϫ ; GFP-Cetn2 ϩ photoreceptors (Fig. 3J). Similar to the Kif3a mutant (Fig. 3B), PTZs failed to form at the apical ISs of P10 emb Ift88 Ϫ/Ϫ photoreceptors (Fig. 3J, right, and 3K) confirming that photoreceptor PTZ formation and subsequent OS development require KIF3-driven IFT.

KIF3A Depletion in the Adult Mouse Causes Outer Segment Degeneration-A rod-specific
Kif3a knock-out showed that rhodopsin trafficked to the OS even as the OS degenerated (32). However, KIF3 and IFT88 were associated with rhodopsin trafficking in cultured hTert-RPE1 cells (56), and rhodopsin was pulled down by IFT88 together with GC1 as IFT cargo (21), suggesting the involvement of KIF3 in rhodopsin trafficking. To solve this apparent discrepancy, we deleted KIF3A in the adult mouse using a tamoxifen-inducible Cre/loxP recombination system, CAG-CreER (38) (Fig. 4A). We induced nuclear translocation of CreER with tamoxifen on 5 consecutive days and followed the rate of degeneration in cryosections for 8 weeks (Fig. 4, C and D). Kif3a f/Ϫ ;CreER ϩ ( tam Kif3a Ϫ/Ϫ ) photoreceptors degenerated progressively, but the rate of degeneration was slower and less severe than that occurring in emb Kif3a Ϫ/Ϫ mice (Fig. 1, B and C). No morphological defects were obvious in the tam Kif3a Ϫ/Ϫ mouse retinas at 2 and 3 weeks post-tamoxifen induction (PTI), whereas the OS shortened at 4 weeks PTI and degeneration progressed with Ͼ90% reduced photoreceptor OS and ONL thicknesses by 7 and 8 weeks PTI (Fig. 4, C and D).
Rhodopsin and OS Proteins Traffic to tam Kif3a Ϫ/Ϫ Photoreceptor Outer Segments in the Absence of KIF3-If rhodopsin trafficking to outer segments requires KIF3, depletion of KIF3A should prevent rhodopsin trafficking to the ROS. After five consecutive daily inductions of Kif3a f/Ϫ ;CreER ϩ mice with tamoxifen, KIF3A was undetectable by immunohistochemistry in the tam Kif3a Ϫ/Ϫ photoreceptors as early as 1 week PTI (Fig.  4E, right) and undetectable by immunoblot at 2 weeks PTI (Fig.  4B). KIF3A levels were unchanged in tam Kif3a Ϫ/ϩ controls (Fig. 4, E-G, left). Rhodopsin localized normally in tam Kif3a Ϫ/Ϫ ROS at 1 and 2 weeks PTI (Fig. 4, H and I, right), a time interval during which ROS disc membranes should have been replaced completely (57). By 3 weeks PTI, when the tam Kif3a Ϫ/Ϫ OS layer was shortened more than 50% (Fig. 4J, right) relative to controls (Fig. 4J, left), rhodopsin was still localized correctly to the ROS, although minor mislocalization of rhodopsin became obvious (Fig. 4J, arrows). ML-and S-opsins were targeted correctly to the tam Kif3a Ϫ/Ϫ COSs up to 3 weeks PTI (Fig. 4, K-P, right), similar to the tam Kif3a Ϫ/ϩ controls. However, cone opsins started to be retained in inner segments and mistrafficked to synaptic termini at 2 weeks PTI (Fig. 4, L and O, right,  arrows), earlier than seen for rhodopsin. Cone pigment mislocalization only became severe after 3 weeks PTI (Fig. 4, M and P, right).
Rhodopsin Trafficking Is Independent of IFT88 -Excluding a possibility that IFT transports rhodopsin via motor(s) other than kinesin-2 proteins, we disabled anterograde IFT by depletion of IFT88, a core component of IFT-B protein complexes that moves cargo from the ciliary base to the axoneme tip. We generated a tam Ift88 Ϫ/Ϫ (Ift88 f/Ϫ ;CreER ϩ ) mouse line, conditionally disrupting IFT88 function in mature photoreceptors by tamoxifen-induced gene targeting. IFT88 localizes to the inner segment myoid region of heterozygous tam Ift88 ϩ/Ϫ photoreceptors (Fig. 8A). In tam Ift88 Ϫ/Ϫ photoreceptors, IFT88 was undetectable 3-5 weeks PTI (Fig. 8B). Rhodopsin trafficking, however, appeared unaffected by the absence of IFT88. Rhodopsin was present in tam Ift88 Ϫ/Ϫ OSs at 3 and 4 weeks PTI and only mislocalized in inner segments at 5 weeks PTI, a time when ROS structure has disintegrated and lost 90% of length (Fig. 8C) relative to the heterozygous control. This result suggests that rhodopsin transport occurs independently of IFT driven by plus-oriented molecular motors.
By ultrastructure at 2 weeks PTI, tam Kif3a Ϫ/Ϫ mice show normal rod and cone elaboration of the PTZ and OSs (Fig. 10, C and D) as the tam Kif3 ϩ/Ϫ control (Fig. 10, A and B). However, tam Kif3a Ϫ/Ϫ cone photoreceptors exhibited cytoplasmic condensation indicative of imminent cell death (Fig. 10D). tam Kif3a Ϫ/Ϫ rods, in contrast, showed normal structure and no cytoplasm condensation consistent with electroretinogram and histology results. When compared with emb Kif3a Ϫ/Ϫ retina, tam Kif3a Ϫ/Ϫ cone-rod degeneration was less severe. These results suggest that KIF3A is essential for axoneme maintenance and photoreceptor survival.

DISCUSSION
Under conditions of normal development, the basal body docks to the inner segment cortex during late ciliogenesis, with subdistal appendages forming the basal foot for microtubule anchoring and IFT beginning to deliver ciliary proteins. Extension of ciliary stalks and OS formation usually initiate around P3 and P7, respectively (60). Our results show that deletion of KIF3A or IFT88 in retinal progenitors prevented photoreceptor PTZ, axoneme, and OS formation, thus aborting ciliogenesis after the basal body docked to cell membrane. Expression of basal body/PTZ marker GFP-CETN2 in emb Kif3a Ϫ/Ϫ mice (Fig.  3, A and B) and ultrastructure of emb Kif3a Ϫ/Ϫ photoreceptors (Fig. 3, D-H) demonstrated the absence of PTZs and axonemes, although basal bodies docked to the cell membrane correctly (Fig. 3, E, G, and H). These experiments demonstrate that KIF3dependent IFT is necessary for axoneme and PTZ formation. The emb Ift88 Ϫ/Ϫ phenotype is distinct from that of the Oak Ridge Polycystic Kidney (ORPK) mouse carrying a hypomorphic mutation in the Ift88 gene. ORPK mice form normal PTZs but elaborate short disorganized OS and retinal degeneration. In the emb Ift88 Ϫ/Ϫ mouse, PTZs and axonemes never formed. In the emb Ift88 Ϫ/Ϫ and emb Kif3a Ϫ/Ϫ photoreceptors, membrane proteins destined to traffic to outer segments backed up in the inner segment, similarly as seen in KIF3A deletion based on early Cre expression driven by the interstitial retinoid binding protein (IRBP) promoter (29). Photoreceptor degeneration started at P12, when OSs rapidly developed in WT photoreceptors, and the degeneration was complete 2 weeks later (Fig. 1, B  and C). Consequently, IFT mediated by KIF3 and IFT88 is required for PTZ formation and axoneme extension.
In a second experimental series, we depleted KIF3A or IFT88 in fully developed photoreceptors by tamoxifen-induced gene targeting (Fig. 4). We focused on primary defects of either outer segment morphology or membrane protein trafficking. KIF3A was undetectable by immunohistochemistry in the ONL and inner segments 1 week post-tamoxifen induction (Fig. 4, E-G,  right). Trafficking of rhodopsin and other OS components proceeded normally in the absence of Kif3A for about 3 weeks post-tamoxifen induction (Fig. 4, H-J, and Fig. 6). In the adult mouse, rod and cone outer segments are completely replaced every 10 days, requiring resynthesis of 80 OS disks every days (57); therefore 3 weeks are sufficient to replace outer segments twice. Outer segment proteins continued to travel even after onset of OS degeneration (Figs. 4 and 6), a finding consistent with our previous study in which transgenic expression of iCre75 caused rod and ROS degeneration, but did not interfere with rhodopsin trafficking (32).
Surprisingly, cone pigment trafficking in tam Kif3a Ϫ/Ϫ cones progressed normally as well (Fig. 4, K-P, right), although cone pigment mislocalization and degeneration was observed earlier than in rods, about 2 weeks after the first injection of tamoxifen (Fig. 4, L-O, right). COS renewal proceeded normally in the absence of KIF3A. This contrasts with our previous study (32) in which transgenic Cre expression in cones under control of the human red/green promoter resulted in the early mislocalization of cone pigments and absence of OS components in tam Kif3a Ϫ/Ϫ COS. The discrepancy may be explained by differences in Cre drivers (transgenic versus tamoxifen-induced depletion), onset of Cre expression (early expression during cone development versus adult), and varying Cre expression levels. Tamoxifen-induced deletion of IFT88 did not affect OS protein trafficking as long as an OS and axoneme existed (Fig.  8C).
We were intrigued by the different rates of outer segment degeneration in rods versus cones and asked whether this may be caused by differences in outer segment structure, disc membrane organization, or intrinsic properties of visual pigment. Ectopic expression of S-opsin in Rho Ϫ/Ϫ rods restored outer segments and their ability to respond to light (42). S-opsin expressed in tam Kif3 Ϫ/Ϫ rods under rhodopsin knock-out or heterozygous background trafficked normally to the OSs at 2 and 3 weeks PTI, nearly indistinguishable from controls (Fig. 7,  D and F, right). In contrast, native ML-opsin in cones still mistrafficked in the S-opsin expressed in tam Kif3 Ϫ/Ϫ mice at 2 week PTI, and COS fell apart at 3 weeks PTI (Fig. 7, C-E, right). These experiments demonstrate that S-opsin expressed in rods follows the mistrafficking and degeneration pattern observed with rhodopsin in tam Kif3a rod OS.
The main effect of KIF3A deletion in the adult mouse retina was gradual shortening of the axoneme, as documented by labeling with antibodies directed against RP1 and Ac-tubulin (Fig. 9, D-F, I, and J). An axoneme extends from the basal body, through the PTZ into the OS to about half-length in the rod and to the cone tip in Xenopus (61). The precise axoneme extension in much smaller mouse rods and cones is unknown. Ciliary axonemes require microtubule plus-end tracking and continuous co-polymerization of ␣and ␤-tubulin subunits at their distal tips, and this process is highly dependent on a kinesin-2mediated IFT pathway (62,63). Our results are consistent with a role of KIF3 in axonemal maintenance, in which KIF3 powers IFT trains carrying axonemal tubulin subunits or other ciliary membrane building blocks. Cargo transports to the axoneme distal tip where axoneme assembly and turnover occur. Interruption of KIF3-based IFT is sufficient to deteriorate the axoneme distally, with progressive shortening of the OS and eventual photoreceptor degeneration.
If rhodopsin trafficking through the PTZ is independent of IFT, what trafficking routes are available for rhodopsin? Because of continuous OS turnover, rod trafficking requires a flow of ϳ70 rhodopsin molecules per second to be advanced through the PTZ. In Xenopus, an animal with very large photoreceptors, the flow increases to several hundred rhodopsins/ second (64). Alternative trafficking pathways have been proposed in which rhodopsin-containing vesicles stream toward the OS, without motor activity, inside the axonemal shaft to eventually expand into large disk membranes at the proximal OS (33,65). Clearly, water-insoluble rhodopsin requires membrane support for trafficking, either within freely floating vesicles or in membrane-embedded cargo. By immuno-EM, rhodopsin has been observed at the PTZ plasma membrane, supporting lateral membrane diffusion. As the nature of cargo (i.e. carrier vesicle constituents) is undefined, our knowledge is limited and incomplete.
In summary, KIF3a conditional knock-out in mouse retina progenitors prevents rod/cone axoneme formation and OS morphogenesis. Tamoxifen-induced KIF3A deletion in the adult mouse impedes photoreceptor axoneme maintenance, leading to progressive OS shortening from distal-to-proximal with subsequent photoreceptor degeneration. These experiments demonstrate collectively that IFT powered by heterotri- meric kinesin-2 is critical for axoneme initiation/maintenance, and consequently, photoreceptor OS structure. Further, kinesin-2 molecular motors and IFT88 are nonessential for ciliary trafficking of visual pigment and accessory OS proteins.