The guanine nucleotide exchange factor Arf-like protein 13b is essential for assembly of the mouse photoreceptor transition zone and outer segment

Arf-like protein 13b (ARL13b) is a small GTPase that functions as a guanosine nucleotide exchange factor (GEF) for ARL3-GDP. ARL13b is located exclusively in photoreceptor outer segments (OS) presumably anchored to discs by palmitoylation, whereas ARL3 is an inner segment cytoplasmic protein. Hypomorphic mutations affecting the ARL13b G-domain inactivate GEF activity and lead to Joubert syndrome (JS) in humans. However, the molecular mechanisms in ARL13b mutation–induced Joubert syndrome, particularly the function of primary cilia, are still incompletely understood. Because Arl13b germline knockouts in mouse are lethal, we generated retina-specific deletions of ARL13b in which ARL3-GTP formation is impaired. In mouse retArl13b−/− central retina at postnatal day 6 (P6) and older, outer segments were absent, thereby preventing trafficking of outer segment proteins to their destination. Ultrastructure of postnatal day 10 (P10) central retArl13b−/− photoreceptors revealed docking of basal bodies to cell membranes, but mature transition zones and disc structures were absent. Deletion of ARL13b in adult mice via tamoxifen-induced Cre/loxP recombination indicated that axonemes gradually shorten and outer segments progressively degenerate. IFT88, essential for anterograde intraflagellar transport (IFT), was significantly reduced at tamArl13b−/− basal bodies, suggesting impairment of intraflagellar transport. AAV2/8 vector-mediated ARL13b expression in the retArl13b−/− retina rescued ciliogenesis.

Primary cilia are hairlike protrusions involved in mechanotransduction (kidney epithelial cells), smell (olfaction), or vision (retinal photoreceptors). Ciliogenesis and maintenance of cilia depend on intraflagellar transport (IFT) 2 and many other factors, including Arf-like GTPases (ARL proteins) of the RAS superfamily (reviewed in Ref. 1). ARL13b was discovered in Caenorhabditis elegans and zebrafish cilia during a search for genes causative of Bardet-Biedl syndrome and cystic kidney disease (2,3). In mammals, ARL13b was expressed in cilia of all organs examined, including cerebellum, distal renal collecting ducts, olfactory epithelial cells, and photoreceptors (4 -7). A splice-acceptor site mutation in exon 2 of mouse Arl13b (hennin mutation, hnn) was associated with defects in neural tube patterning, limbs and eyes attributed to defects in the sonic hedgehog (shh) signaling pathway, and homozygous mutants did not survive beyond E14.5 (8). A kidney-specific deletion of Arl13b led to kidney fibrosis and decreased ciliogenesis (5,9). ARL13b mutations R79Q and R200C in the G-domain of the human ARL13b gene caused Joubert syndrome (JS), a syndromic ciliopathy affecting multiple tissues (10,11). Both mutant proteins disable ARL13b GTPase-activity and localize normally to cilia, but are unable to rescue ciliogenesis defects (12). ARL13b was shown to be the guanine nucleotide exchange factor (GEF) of ARL3 (13) and each, ARL13b-R79Q and ARL13b-R200C, impedes the GEF activity for ARL3. GEF activity was specific for ARL3; ARL2 was unaffected (13). The crystal structure derived from Chlamydomonas reinhardtii ARL13b revealed that the GEF activity of ARL13b is mediated by its G-domain and a coiled-coil C-terminal region (13) (PDB 5DI3).
We investigated the relationship among ARL3, its GEF, ARL13b (13), and its GAP, RP2 (14) in photoreceptor ciliogenesis as a means to understand disease etiology (Fig. 1). The photoreceptor outer segment is a modified primary cilium containing a stack of ϳ800 discs dedicated to reception of light and phototransduction (15). Unique among ciliated cells, the entire structure is replaced every 10 days by phagocytosis of the distal tip balanced by nascent disc assembly at the proximal outer segment, a process that requires massive protein synthesis in the inner segment and efficient trafficking pathways through the transition zone (TZ) (reviewed in Refs. [15][16][17]. ARL3 was shown to be required for ciliogenesis as a retina-specific deletion of Arl3 prevented formation of the photoreceptor transi-tion zone (18). Depletion of ARL3 in mature rods revealed that, additionally, ARL3 acts as a cargo displacement factor (CDF) disrupting the complex of lipidated cargo with the solubilization factors, PDE6D and UNC119 (18). RP2 null alleles in human and mouse, which prevent GTP hydrolysis and prolong the life-time of ARL3-GTP, are associated with X-linked retinitis pigmentosa (XLRP) (19,20). Abundance of ARL3-GTP in Rp2h Ϫ/Ϫ retina interfered with trafficking of lipidated proteins of the phototransduction cascade (PDE6 and GRK1) (21). Correspondingly, transgenic expression of dominantly active ARL3-Q71L leads to mistrafficking of lipidated proteins and retina degeneration (22).
Here, a conditional knockout of Arl13b ( ret Arl13b Ϫ/Ϫ ) was generated by expressing Cre recombinase in retina during embryonic development beginning at ϳembryonic day 9 (E9). The results show that ret Arl13b Ϫ/Ϫ photoreceptors degenerate rapidly after failing to form mature transition zones and outer segments. Depletion of ARL13b in the adult by tamoxifen-inducible Cre/loxP recombination ( tam Arl13b Ϫ/Ϫ ) destabilized axonemes and transition zones, leading to progressive photoreceptor degeneration. Significant reduction of IFT88 at tam Arl13b Ϫ/Ϫ basal bodies suggested impairment of IFT upon deletion of ARL13b. scARL13b-AAV8 vector, injected into the ret Arl13b Ϫ/Ϫ subretinal space at eye opening, partially compensated and rescued photoreceptor degeneration.

Generation of conditional Arl13b knock-out mice
ARL13b contains a G-domain characteristic of Arf-like proteins, a proline-rich domain, several coiled-coil domains for protein-protein interactions, and a VXPX-like ciliary targeting signal, RVEP ( Fig. 2A) (23,24). To generate a conditional knockout, we used Arl13b f/f mice in which exon 2 is flanked by loxP sites (Fig. 2B) (25). Cre/loxP-recombination was initiated at E9 by breeding Arl13b f/f mice with transgenic Six3-Cre (18,26) to yield Arl13b f/f ;Six3-Cre mice, abbreviated ret Arl13b Ϫ/Ϫ to indicate retina-specific knockout. LoxP-directed recombination is predicted to delete exon 2 resulting in a frameshift mutation and truncation of ARL13b at codon 2 of exon 3 (Fig. 2C). Genotyping of the floxed gene was performed using primers P1 and P2 flanking loxP in intron 2 with tail DNA (Fig. 2D), and primers P3 and P4 with retina DNA as template (Fig. 2E). Deletion of exon 2 in retina genomic DNA of P15 ret Arl13b Ϫ/Ϫ off-spring was confirmed using primers P1 and P4 flanking the loxP sites of introns 1 and 2 (Fig. 2F).
Polyclonal anti-ARL13b antibody (Proteintech Group Inc.) recognized two polypeptides (48 kDa and 60 kDa) by immunoblotting (Fig. 2G); the 48 kDa species corresponds to unmodified ARL13b (calculated M r 48,000), while the slower-moving polypeptide may be posttranslationally modified by N-terminal palmitoylation (8,27,28), or represent an unknown splice variant. Both polypeptides are detectable in P17 retina knock-out lysates, as far peripheral retina still elaborates outer segments (Fig. 2H, g-i). Immunohistochemistry using ret Arl13b ϩ/Ϫ frozen sections and monoclonal anti-ARL13b antibody (Neuro-Mab) showed that ARL13b is an outer segment protein expressed as early as P6 (Fig. 2H, a-c). At P6, ARL13b localized in nascent OS in close proximity to the basal body (Fig. 2Ha, enlargement in lower panel). At P10 and P15 maturing outer segments contained ARL13b (Fig. 2H, b and c, and lower panel enlargements). ARL13b was also detectable in the inner segment (Fig. 2H, b and c, arrows). By contrast, ARL13b was undetectable in ret Arl13b Ϫ/Ϫ central retina at P6, P10, and P15 ( Fig.  2H, d-f); enlargements show the presence of centrioles, identified by transgenic expression of CETN2 fused to EGFP (EGFP-CETN2), but absence of fully developed transition zones. While Six3-Cre-mediated recombination starts at E9 in the central retina (26), recombination is delayed in the peripheral retina because of delayed expression of Cre recombinase (18,29). Therefore, in the far peripheral retina, ARL13b was detectable and rudimentary outer segments still formed between P10 and P15 ( Fig. 2H, g-i).

ARL13b and the mouse photoreceptor transition zone
axoneme, was absent, consistent with failure to form a functional connecting cilium and outer segment (Fig. 3C, c and d). A whole retina immunoblot (P17) revealed significant traces of rhodopsin and GRK1 originating from far peripheral retina; rod transducin (T␣) and PDE6 expressions persisted more strongly (Fig. 3D). ARL3-GDP levels appear unaffected by ARL13b ARL13b and the mouse photoreceptor transition zone knockout. Scotopic a-wave amplitudes were severely attenuated even at low light intensities (Ϫ4.5 and -1.6 log cds⅐m Ϫ2 ). (Fig. 3E).

Absence of outer segments in ret Arl13b ؊/؊ cones
Cone pigment trafficking in heterozygous control and ret Arl13b Ϫ/Ϫ retinas expressing EGFP-CETN2 was also explored (Fig. 4A). Cone outer segments were identified by anti-ML-opsin (Fig. 4A, a and b) and anti-S-opsin (Fig. 4B, a and b) immunoreactivity in P10 and P15 control retinas. Control cone OS are shown connected to transition zones (yellow arrows) and basal bodies (white arrowheads) (Fig. 4, A and B, a and b, right panels). Although opsins distributed throughout mutant cones, intense accumulation was observed in bloated and deformed inner segments (Fig. 4, A and B, c and d, right panels). Peripheral proteins (GRK1, cone transducin-␥, and cone PDE6) were not detectable (not shown). Photopic b-waves, using flash intensities of Ϫ1.6 log cds⅐m Ϫ2 and higher, were highly suppressed at P15 (Fig. 4C). Optokinetic tracking (OKT) responses at 2 months of age were extinguished in retina-specific knock-out mice (p Ͻ 0.0001), indicating that the mice were unable to respond to light because of ret Arl13b Ϫ/Ϫ rod and cones being completely degenerated (Fig. 4D). By contrast, OKT responses of heterozygous knock-out mice were indistinguishable from wild-type (WT) littermates, suggesting haplosufficiency. ret Arl13b ؊/؊ rods degenerate faster than ret Arl3 ؊/؊ rods ARL3-GDP cannot be activated by GDP/GTP exchange in the ret Arl13b Ϫ/Ϫ retina, whereas in the ret Arl3 Ϫ/Ϫ retina, both ARL3-GDP and ARL3-GTP are absent. ARL3-GDP levels appear stable in the P17 ret Arl13b Ϫ/Ϫ retina (Fig. 3D). To explore differences in the degeneration rate, we compared the rate of ret Arl13b Ϫ/Ϫ retina degeneration with that of ret Arl3 Ϫ/Ϫ retina (Fig. 5). Plastic sections of the three genotypes (WT, ret Arl13b Ϫ/Ϫ , and ret Arl3 Ϫ/Ϫ ) ( Fig. 5A) showed severely reduced ret Arl13b Ϫ/Ϫ and ret Arl3 Ϫ/Ϫ OS/IS morphology indicating the ret Arl13b Ϫ/Ϫ and ret Arl3 Ϫ/Ϫ retinas degenerated as early as P15. ret Arl13b Ϫ/Ϫ ONL thickness decreased significantly faster at P15 and P30 than observed in ret Arl3 Ϫ/Ϫ (Fig.  5B). At 1 month of age only one nuclear row remained in ret Arl13b Ϫ/Ϫ retina, whereas ret Arl3 Ϫ/Ϫ photoreceptors retained three rows of nuclei (Fig. 5B); the different degeneration rates suggest that ARL13b may have an additional function. ret Arl13b ؊/؊ ultrastructure Ultrastructural analysis at postnatal days P10 and P15 revealed normal rod photoreceptor ciliogenesis in control ani-mals. In P10 WT retina, basal bodies dock to the photoreceptor cell membrane, transition zones extend, disc membranes assemble, and OS are elaborated (Fig. 6A, a and b). In P10 ret Arl13b Ϫ/Ϫ photoreceptors, basal bodies docked to the cell membrane (Fig. 6A, c and e), rudimentary transition zones formed (yellow arrow) but ended in membranous bags (white asterisk) without recognizable disc structure (Fig. 6A, d and f). At P15, WT photoreceptor structure appears advanced with well-aligned basal bodies, transition zones and disc membrane in the OS (Fig. 6B, a and b). By contrast, P15 ret Arl13b Ϫ/Ϫ basal bodies are either undocked to membrane or docked to membranes with distal and subdistal appendages (Fig. 6B, c-f). Docked basal bodies (Fig. 6Bd) developed stunted TZs connected to membranous bags. Stunted TZs, revealed by electron microscopy, were not identified by confocal microscopy despite presence of the EGFP-CETN2 transgene. This may perhaps be attributed to loss of CETN2 docking sites in the knockout, thereby eliminating EGFP fluorescence. The electron microscopy results indicate a requirement of ARL13b GEF activity and ARL3-GTP for normal photoreceptor TZ structure.

Arl13b depletion in the adult mouse
To explore TZ stability as a function of ARL3-GTP, we deleted ARL13b progressively at 1 month of age using a tamoxifen-inducible Cre/loxP recombination system, CAG-CreER (35). Nuclear translocation of CreER was induced by intraperitoneal injection of tamoxifen on five consecutive days, and the degeneration rate was assessed in retina cryosections of eyes harvested 2, 3, and 6 weeks later ( Fig. 7). At 2 weeks post tamoxifen induction (2WPI), ARL13b was present in treated WT outer segments, but undetectable in tam Arl13b Ϫ/Ϫ retina (Fig.  7A, compare a and b). Although tam Arl13b Ϫ/Ϫ outer segments appeared of normal length (Fig. 7A, d-f), rhodopsin, PDE6, and T␣ displayed retention in the tam Arl13b Ϫ/Ϫ inner segments ( Fig. 7A, d-f, lower panels). Tamoxifen treatment of tam Arl13b ϩ/ϩ mice had no effect on rhodopsin (Fig. 7Ac), PDE6, or T␣ trafficking (results not shown). At 3WPI, rhodopsin, PDE6, and T␣ ( Fig. 7B, a-d) were retained much more strongly in the tam Arl13b Ϫ/Ϫ inner segments suggesting outer segment shortening/degeneration. At 6WPI tam Arl13b Ϫ/Ϫ outer and inner segments were absent with only one row of nuclei surviving (Fig. 7, C, b-d, and G). Scotopic a-wave amplitudes were significantly reduced at 2WPI and 3WPI (Fig. 7, D and E) and extinguished at 6WPI (Fig. 7F).
Cone outer segments, identified by anti-ML-opsin antibody, appeared normal and were attached to their TZ (yellow arrows) , and flexible C terminus containing a proline-rich region (PRR). B, exon 2 is flanked by two loxP sites (orange arrowheads), both orientated in direction of translation. Arrows indicate primers for genotyping of genomic tail DNA. C, the deleted allele after Cre-induced recombination. ARL13b is truncated in out-of-frame exon 3, indicated by an asterisk. Arrows indicate primers for genotyping of DNA isolated from retinas. D, identification of the intron 2 loxP-site of Arl13b ϩ/ϩ ;Six3-Cre, Arl13b ϩ/f ;Six3-Cre, and Arl13b f/f ;Six3-Cre tail DNA with primers P1 and P2. E, identification of the intron 1 loxP-site of Arl13b ϩ/ϩ ;Six3-Cre, Arl13b ϩ/f ;Six3-Cre tail DNA with primers P3 and P4. F, verification of exon 2 deletion with primers P3/P4 in heterozygous and homozygous mutant retina DNA at P15. G, immunoblot of P17 and 2-month-old animals. ARL13b protein (molecular mass ϳ48 kDa) in the homozygous knockout is significantly reduced at P17 and absent at 2 months of age (48kDa and 60kDa); ␤-actin (42 kDa) is a loading control. H, retina cryosections of P6, P10, and P15 Egfp-Cetn2 ϩ ; ret Arl13b ϩ/Ϫ and Egfp-Cetn2 ϩ ; ret Arl13b Ϫ/Ϫ mice labeled with anti-ARL13b antibody (red). Egfp-Cetn2 ϩ ; ret Arl13b Ϫ/Ϫ sections were divided into central (middle column) and peripheral (right) regions. a-c, ARL13b localizes to the OS of heterozygous control mice. d-i, ARL13b is absent in central ret Arl13b Ϫ/Ϫ retina (d-f), whereas it localizes to shortened OS of the far periphery at P15 because ARL13b is incompletely deleted (g-i). Enlargements are shown as lower panels. Sections were contrasted with DAPI (blue) to reveal the outer nuclear layer (ONL). Scale bar, 20 m; 5 m in enlargements.

Impairment of intraflagellar transport in tam Arl13b ؊/؊ photoreceptors
Anterograde IFT along microtubules depends on association of IFT particles with kinesin-2 molecular motors and IFT particles (36). IFT88, an IFT-B particle required for anterograde IFT, is present at the 3WPI tam Arl13b ϩ/ϩ basal body where cargo is assembled (Fig. 8F, left panel, arrow heads), and in the proximal OS (Fig. 8F, arrows). In tam Arl13b Ϫ/Ϫ photoreceptors, IFT88 was significantly reduced or absent at the periciliary membrane (Fig. 8F, right panel, arrow heads). Quantitative evaluation (n ϭ 100) indicated 2-to 5-fold accumulation of IFT88 in the proximal OS (Fig. 8G) and 5-fold reduction at the basal body (Fig. 8H), suggesting that IFT is impaired in the absence of ARL13b.
The effect of ARL13b depletion on basal body/axoneme cytoskeleton was further studied by labeling with antibodies directed toward two well-characterized proteins, RP1 (Fig. 9A) and acetylated ␣-tubulin (AcTub) (Fig. 9B). In tam Arl13b ϩ/ϩ mice, the basal body and its TZ (green) are attached to the proximal axoneme (red), labeled with anti-RP1 (Fig. 9A, a and b) or anti-AcTub antibody (Fig. 9B, a and b). In ARL13b-depleted retina (Fig. 9, A, d and e and B, d and e), axonemes initially shorten (2WPI), become stunted (3WPI), and EGFP-CETN2 is unable to tightly associate with centrioles and TZs. Labeling with giantin (37), a 400-kDa protein that identifies the Golgi complex, reveals that the Golgi apparatus is present in the entire IS between outer limiting membrane and BB/TZ (Fig.  9C, a and b). In the tam Arl13b Ϫ/Ϫ inner segment (Fig. 9C, c  and d), EGFP-CETN2 aggregates at the Golgi complex near the outer limiting membrane. Axonemes are significantly reduced in length at 2WPI and 3WPI and absent at 6WPI (Fig. 9D).

Viral rescue of ret Arl13b ؊/؊ photoreceptors
A self-complementary AAV2/8 vector expressing ARL13b (scAAV-Arl13b) was generated to attempt the "rescue" of ARL13b and the mouse photoreceptor transition zone mutant photoreceptors. Virus was injected subretinally at P12-P15, and retinas were harvested 6 weeks later. Significant increase in ARL13b expression in WT mice injected with scAAV-Arl13b (Fig 10, upper), compared with uninjected controls (Fig. 10, lower) confirmed expression and correct localization. ONL thickness and OS protein trafficking in both central and peripheral retina signaled partial rescue (Fig 11, A and B) and specifically, expression of the microtubule-associated protein, RP1, in treated ret Arl13b Ϫ/Ϫ photoreceptors confirmed presence of nascent axonemes (Fig. 11A, bottom row).

Discussion
Mouse photoreceptor outer segments are modified primary cilia that form during early postnatal development (38 -42). Maturation of the mother centriole into the basal body, docking of the basal body to the cortex of the cell, generation of the transition zone and extension of the axoneme occur in several unsynchronized steps, employing numerous proteins, some of which harbor unknown function. At P0, mother (MC) and daughter centrioles (DC) are present in the rod cell cytoplasm, and the distal end of the MC is covered by a vesicle. Around P3-P4, the MC has matured and docks to the cell cortex, assuming its function as basal body and microtubule organization center (MTOC). The basal body is a highly conserved, barrel-shaped, microtubule-based structure that remains associ-ated with the daughter centriole through filamentous bundles. The MC is decorated with distal and subdistal appendages that enable microtubule nucleation and membrane anchoring; ultrastructural cross-sections reveal a symmetrical array of nine microtubules in a triplet arrangement (9(3)ϩ0). Tubulin subunits are incorporated into the emerging TZ, a doublet microtubule structure (9(2)ϩ0) anchored to the basal body. Around P6 -P7, a singlet microtubule (9(1)ϩ0) emerges forming the axoneme, and the first stacks of discs are synthesized (disc morphogenesis). Outer segment formation is complete at P21 (38).
In this communication, we explored the consequences of retina-specific deletion of ARL13b, the GEF for ARL3 (13) on basal body docking and TZ maturation. We monitored the formation of TZ by two independent methods, using immunohistochemistry with EGFP-CETN2 as a centriole and TZ marker (Figs. 2-4) and high-resolution electron microscopy (Fig. 6). We found that photoreceptor axonemes and outer segment discs did not form when ARL13b was absent before onset of ciliogenesis. ret Arl13b Ϫ/Ϫ basal bodies docked to the cell membrane and formed distal and subdistal appendages, but the TZ never formed correctly, and membranous bags attached to stunted TZs never contained OS disc structures (Fig. 6, A, c-f and B, c-f). These results align with our earlier result that, in the absence of ARL3 during early postnatal development, TZ for- Photoreceptor OS in ret Arl13b Ϫ/Ϫ and ret Arl3 Ϫ/Ϫ sections are absent at P10 and the ONL is reduced to ϳ6 -7 rows of nuclei in ret Arl13b Ϫ/Ϫ and ϳ7-8 rows of nuclei in ret Arl3 Ϫ/Ϫ (d-f). At 1 month of age, ret Arl13b Ϫ/Ϫ has only one nuclear row remaining in the ONL and ret Arl3 Ϫ/Ϫ has ϳ2-3 rows of nuclei left (g-i). Scale bar, 20 m. B, nuclei numbers at P10, P15, and 1 month of age in WT, ret Arl13b Ϫ/Ϫ , and ret Arl3 Ϫ/Ϫ retinas. Both knockouts are predegenerate at P10 and degenerate rapidly, whereas Arl13b retina degenerates faster than Arl3 retina (n ϭ 3; **, p Ͻ 0.05; ***, p Ͻ 0.001; ****, p Ͻ 0.0001). At 1 month of age, one ONL layer remains in ret Arl13b Ϫ/Ϫ and 2-3 layers in ret Arl3 Ϫ/Ϫ .

ARL13b and the mouse photoreceptor transition zone
mation is impaired and photoreceptor outer segments do not form, establishing ARL3-GTP, the downstream effector of ARL13b, as a factor in ciliogenesis and TZ formation (18). ret Arl3 Ϫ/Ϫ and ret Arl13b Ϫ/Ϫ morphological and physiological phenotypes are nearly identical except that ret Arl13b Ϫ/Ϫ photoreceptors degenerate faster, suggesting that ARL13b has additional function in the outer segment (Fig. 5).
The results in mouse photoreceptors are in contrast to C. elegans, zebrafish, and hTert-RPE1 cells where absence of ARL13b did not prevent formation of cilia. In C. elegans, studies in Arl-13 mutants (worm Arl-13 is the homolog of mammalian ARL13b) showed that truncated cilia were formed with various structural deformities (43,44). In zebrafish, knockdown of Arl-13 (sco mutant) led to multiple cilia-associated phenotypes,  a and b) and mutant (c-e) distal inner segments. The mother centriole, or basal body (BB), typically contacts a ciliary vesicle and extends nine microtubule doublets to form the proximal axoneme, which itself becomes surrounded by a sheath constituting the periciliary membrane. As the sheath fuses with the plasma membrane, the basal body docks to the photoreceptor cortex, stabilized by fibers, and extends a transition zone. Well-developed connecting cilia with outer segment discs (a and b) contrast with aborted or stunted transition zones of mutant photoreceptors (c-f). Because Six3Cre recombinase is driven in a central-to-peripheral sequence, central photoreceptors of the mutant are affected severely. BB docking occurs in P10 ret Arl13b Ϫ/Ϫ rods (c and e), but axoneme extension is aborted (d, yellow arrow) with elaboration of undefined membranous bags (asterisk) in slightly more peripheral retina (d and f). B, mature TZs (CC) and outer segments with stacks of disc membrane form in P15 WT photoreceptors (a and b). BB are shown docked to the plasma membrane (yellow arrows) of littermate mutant photoreceptors (c-f), decorated with distal and subdistal appendages, and forming unspecified membranous bags (f). BB with appendages are shown in tangential section (e and f).

ARL13b and the mouse photoreceptor transition zone
including left-right asymmetry, kidney pronephric cysts, and body curvature (45). In sco mutants, cilia are disorganized but their ultrastructure (9(2)ϩ1, nine microtubules doublets and one doublet in the center) appears to be intact. In a zebrafish arl13b null mutant, cilia and photoreceptor outer segments are shortened (46), and retina degeneration progressed very slowly over weeks. In hTert-RPE1 cell lines in which ARL13b was deleted, cilia were present but significantly shorter (24). In Arl13b null mice (hnn), cilia are formed but are short with spe-cific defects in the ciliary axoneme structure, accompanied by defects in sonic hedgehog (shh) signaling resulting in embryonic lethality (8,10,47). In kidney-specific deletions of mouse, ductal cilia were absent, a defect that led to rapid cyst formation and renal failure, and mutant mice died at P60 (5). Taken together, strong interspecies differences can be observed upon depletion of ARL13b. The phenotype of ARL13b deletion is strongest in mouse kidneys and photoreceptors where transition zones are stunted and axonemes are absent. Figure 7. Tamoxifen-induced ARL13b depletion in the adult affects rods. A, at 2 weeks post injection (2WPI), ARL13b was (a) present in tam Arl13b ϩ/ϩ outer segments, but (b) absent in tam Arl13b Ϫ/Ϫ outer segments. c, OS developed normally in tamoxifen-injected WT controls at 2 to 6WPI. In tam Arl13b Ϫ/Ϫ retina at 2WPI, rhodopsin (d), PDE6 (e), and rod transducin (f) accumulated in the inner segments. Areas of interest are enlarged (lower panels). B, in tam Arl13b ϩ/ϩ retina at 3WPI, rhodopsin localizes normally to the outer segments (a). In tam Arl13b Ϫ/Ϫ retina at 3WPI, rhodopsin (b), PDE6 (c), and transducin (d) mislocalized in the IS as OS degenerate. C, in tam Arl13b ϩ/ϩ retina at 6WPI, rhodopsin still localizes normally to the outer segments (a). In tam Arl13b Ϫ/Ϫ retina at 6WPI, rod photoreceptors are completely degenerated (b-d). D-F, scotopic a-wave amplitudes of tamoxifen-induced Arl13b knockouts at 2WPI (D), 3WPI (E), and 6WPI (F).

ARL13b and the mouse photoreceptor transition zone
In retina-specific deletion of ARL13b (this paper) and ARL3 (18), basal bodies appear to dock to the cell cortex, but mature TZ/OS are never formed. The consequence is accumulation of outer segment proteins in the inner segment, nuclear and synaptic regions, but the precise mechanism leading to abortion of ciliogenesis and rapid photoreceptor degeneration is unknown. Error bars indicate S.D., n ϭ 3. OKT threshold at 2WPI was measured with tam Arl13b ϩ/ϩ , tam Arl13b ϩ/Ϫ , and tam Arl13b Ϫ/Ϫ mice (C); the threshold was significantly reduced in tam Arl13b Ϫ/Ϫ mice only n ϭ 5; ***, p Ͻ 0.001). F, immunohistochemistry of 3WPI tam Arl13b ϩ/ϩ (left panels) and tam Arl13b ϩ/ϩ cryosections (right panels) with anti-IFT88 antibody. Enlargements show IFT88 accumulations at basal bodies (arrowheads) and proximal OS (arrows). G and H, quantitative evaluation of IFT88 fluorescence at the proximal OS (G) and basal body (H). n ϭ 100, p Ͻ 0.001.

ARL13b and the mouse photoreceptor transition zone
A possible trigger to cause degeneration may be the participation of ARL13b and ARL3-GTP in intraflagellar transport. IFT requires IFT-A and IFT-B particles, in addition to heterotrimeric kinesin-2, to transport cargo into cilia or flagella (48 -50). In C. elegans Arl-13 mutants, IFT-A and IFT-B particles required for IFT were disrupted suggesting that ARL-13 and, based on its GEF activity also ARL-3, regulate IFT particle integrity thus linking ARL-13 and ARL-3 to IFT (1,27,43). In hTert-RPE1 cells, ARL13b interacts with IFT-B particles, but its ciliary localization is independent of this interaction (24). In ARL13b knock-out RPE1 cells, which have shortened cilia, retrograde IFT is disabled, leading to accumulation of IFT-B and IFT-A particles at the ciliary tips (24).
IFT in photoreceptors is important for maintenance of the axoneme (29,51) that extends into the outer segment. Retinaspecific deletion of KIF3a, the obligatory subunit of heterotrimeric kinesin-2, prevented formation of photoreceptor TZ and axonemes, exactly as observed in retina-specific deletions of ARL3 and ARL13b. Tamoxifen-induced depletion of ARL13b showed increased solubilization of EGFP-CETN2 in rods and cones (Figs. 7 and 8). Apparently, EGFP-CETN2, normally tightly associated with TZ and centrioles, lost its ability to interact with centrioles/TZ, became increasingly soluble, and accumulated in the inner segment (Fig. 9, A and B). Furthermore, labeling axonemes with RP1 (Fig. 9A) and AcTub (Fig. 9B) in ARL13b knockouts indicated that axonemes shortened at 2WPI and became stunted at 3WPI, suggesting that failure to maintain the basal body-axoneme cytoskeleton may be a key event leading to photoreceptor degeneration. Gradual shortening of the axoneme at 2WPI and 3WPI was obtained when KIF3a was deleted in the adult mouse using tamoxifen induction (29), and even the degeneration rates with tamoxifen-induced deletions of KIF3a, ARL3, and ARL13b are nearly identical.  Fig. 7. Hatched boxes are enlarged as lower panels. RP1 is a MAP (microtubule-associated protein) and localizes to the proximal axoneme (red) which is attached to the TZ (green) (a and b). In tam Arl13b Ϫ/Ϫ retina at 2WPI (d) and 3WPI (e), the axoneme area interacting with RP1 is reduced and CETN2 interaction with centrioles/TZ is impaired (d and e, lower panels). B, anti-AcTub identifies acetylated ␣-tubulin which stabilizes the proximal microtubule axoneme (a-c). In tam Arl13b Ϫ/Ϫ retina at 2WPI (d) and 3WPI (e), AcTub labeling of the axoneme is reduced and EGFP-CETN2 solubilizes forming a long flare into the ONL (d and e, lower panels). At 6WPI, the photoreceptors have degenerated (Af and Bf) compared with controls (Ac and Bc). Scale bar ϭ 20 m; in enlargements, scale bar ϭ 20 m. Please note: Ac and Af and Bc and Bf are not on the Egfp-Cetn2 background. C, Egfp-Cetn2 ϩ ; tam Arl13b ϩ/ϩ (a and b) and Egfp-Cetn2 ϩ ; tam Arl13b Ϫ/Ϫ (c and d) sections were probed with anti-giantin antibody (red). CETN2 mislocalizes to the IS/Golgi of mutant retina (c and d). Scale bar, 5 m. D, reduced lengths (m) of RP1-responsive axonemes in tam Arl13b Ϫ/Ϫ photoreceptors at 2WPI, 3WPI, and 6WPI (****, p Ͻ 0.0001).

ARL13b and the mouse photoreceptor transition zone
We determined the localization of IFT88, an obligatory IFT-B particle essential for anterograde IFT in photoreceptors (36,52). The results show that IFT88 is nearly absent in ret Arl13b Ϫ/Ϫ basal bodies where cargo for IFT is assembled, thus strongly suggesting impairment of anterograde IFT upon deletion of ARL13b. Interference with transport of tubulin subunits and axoneme building blocks could explain fully the progressive OS shortening and eventual photoreceptor degeneration in tamoxifen-induced ARL13b deletions.

Animals
Procedures were approved by the University of Utah Institutional Animal Care and Use Committee and were conducted in compliance with the National Institutes of Health Guide for Care and Use of Laboratory Animals. Floxed Arl13b mice (Arl13b f/f ) were provided by Dr. Tamara Caspary (Emory) and maintained in a 12:12 h dark-light cycle. A transgenic mouse expressing EGFP-CETN2 fusion protein (The Jackson Laboratory, stock number 008234) was used to identify centrioles with fluorescence microscopy (32).

ARL13b and the mouse photoreceptor transition zone
ond) and contrast were kept constant as described (58). Visual acuities of mice were quantified by increasing the spatial frequency of the rotating grating (0.03-0.39 cycles/degree) until the maximum frequency threshold was tracked.

Transmission electron microscopy
Isolated mouse eyecups were fixed by immersion in 2% glutaraldehyde-1% paraformaldehyde in 0.1 M cacodylate buffer, pH 7.4, at 4°C overnight (32,46). The eyecups were 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 plastic sections were cut to face and orient photoreceptors near the optic nerve. Retina ultrathin (60 nm) sections were cut onto slot grids with carboncoated Formvar film (Electron Microscopy Sciences, Hatfield, PA) and poststained with uranyl acetate followed by lead citrate. Transmission electron microscopy was performed at 75 kV using a JOEL electron microscope.

Generation of the AAV2/8 shuttle vector and virus
Mouse Arl13b expression cassette under the control of the G protein-dependent receptor kinase (GRK1) promoter was placed in between the two inverted terminal repeats of an AAV2/8 shuttle plasmid (33). To generate a self-complementary vector, one inverted terminal repeats of the shuttle plasmid was mutated. The vector was packaged into AAV8 capsid. Production and purification of the vector were performed following a protocol described previously (59). Vector quantification was conducted by real time PCR using linearized plasmid standards and primers against the GRK promoter.

Statistics
SigmaPlot12 was used for statistical analysis using Student's t test and the level of statistical significance was set p ϭ 0.05.