Arf-like Protein 2 (ARL2) Controls Microtubule Neogenesis during Early Postnatal Photoreceptor Development

Arf-like protein 2 (ARL2) is a ubiquitously expressed small GTPase with multiple functions. In a cell culture, ARL2 participates with tubulin cofactor D (TBCD) in the neogenesis of tubulin αβ-heterodimers, the building blocks of microtubules. To evaluate this function in the retina, we conditionally deleted ARL2 in mouse retina at two distinct stages, either during the embryonic development (retArl2−/−) or after ciliogenesis specifically in rods (rodArl2−/−). retArl2−/− retina sections displayed distorted nuclear layers and a disrupted microtubule cytoskeleton (MTC) as early as postnatal day 6 (P6). Rod and cone outer segments (OS) did not form. By contrast, the rod ARL2 knockouts were stable at postnatal day 35 and revealed normal ERG responses. Cytoplasmic dynein is reduced in retArl2−/− inner segments (IS), suggesting that dynein may be unstable in the absence of a normal MTC. We investigated the microtubular stability in the absence of either ARL2 (retARL2−/−) or DYNC1H1 (retDync1h1−/−), the dynein heavy chain, and found that both the retArl2−/− and retDync1h1−/− retinas exhibited reduced microtubules and nuclear layer distortion. The results suggest that ARL2 and dynein depend on each other to generate a functional MTC during the early photoreceptor development.

Within the mammalian retina, ARL2 localizes in the photoreceptor IS, with an enrichment in the basal body area/rootlet [21]. ARL2 has a relatively weak affinity to nucleotides and presumably exchanges GDP with GTP without the assistance of the Guanine nucleotide Exchange Factor (GEF) [22,23]. The GAPs for ARL2 include ELMOD1-3 (ELMO Domain

Generation of Retina-and Rod-Specific Knockout Mice
We acquired a cell line (UC-Davis KOMP repository, allele name Arl2tm1a(KOMP)Wtsi) with a gene trap flanked by loxP and FRT sites in intron 1 and a third loxP site in intron 3. Gene-trapped mice were generated at the University of Utah Transgenic Core Facility by a blastocyst injection. The germline transmission of the Arl2 GT allele was verified by PCR using primers upstream and downstream of the loxP site in intron 1. A floxed Arl2 allele (Arl2 F/+ ) was generated following the recombination of FRT-FLP with Flprecombinase. Mice carrying the Arl2 F allele were mated with C57BL/6J mice to remove the rd8 mutation inherent in KOMP mice [38]. Arl2 F/F mice were crossed with Six3Cre [36] or iCre75 transgenic mice [37] to generate retina-specific Arl2 F/+ , Six3Cre ( ret Arl2 +/− ) or rodspecific Arl2 F/+ , iCre75 ( rod Arl2 +/− ) heterozygous mice. The mice were then backcrossed to Arl2 F/F to generate experimental animals. For the identification of the centrioles and the connecting cilia, select mice were maintained on an Egfp-Cetn2 background.

Western Blot
The retinas were lysed in PHEM buffer (60 mM PIPES, 25 mM HEPES, 10 mM EGTA, and 4 mM MgSO4). A total of 30 ug of protein (Lowry assay) were separated by a 16% Bis-Tris gel using low-MW buffer (50 mM of MES (compound-2 N-morpholino) ethanesulfonic acid); 50 mM of Tris, 1 mM of EDTA, and 0.1% SDS). After the protein transfer to a 0.45 µm nitrocellulose membrane (ThermoFisher, Logan, UT, USA), the membrane was blocked using 5% milk in Tween Tris-buffered saline (TTBS) for 1 h and then incubated overnight at 4 • C in a polyclonal Arl2 antibody (diluted 1:1000, gift from Richard Kahn, Emory University) in the same buffer. The membrane was washed (10 min × 3) and incubated for 2 h in LI-COR secondary antibodies (mouse IR680 1:10 K and rabbit IR800 1:4 K). Following washes in 1X TTBS, the membrane was scanned using a LI-COR Odyssey imaging system (Lincoln NE).
The tubulin immunohistochemistry was performed on tissue fixed, as described above, with the addition of 'antigen retrieval' after sectioning. Briefly, the slides with sections were warmed for 30 min and were then washed (10 min × 2) in TBS followed by an antigen retrieval for 5 min in 0.1% SDS in TBS at room temperature. The slides were then rinsed quickly (5 min) in TBS. The following antibodies were diluted in 5% goat serum in TBS and incubated overnight at Basal body labelling required a light fixation of 4% paraformaldehyde for 10 min only. The slides were incubated at 37 • C for 30 min, rehydrated in 1X TBS by washing (10 min × 3), blocked in 5% normal goat serum in 1X TBS, and incubated for either 1 h at RT or overnight at 4 • C. The primary antibodies were diluted in a blocking buffer (5% serum in TBS) to cover the wells and were incubated for 2 h at RT or overnight at 4 • C. The following antibodies were used: the CEP250/CNAP1 antibody (14498-1-AP, Proteintech Group Inc., Rosemont, IL, USA; rabbit anti-CEP164 (EMD Millipore, St. Louis, MO, USA, ABE 2621) 1:250; or anti-CEP164 (1:350, Sigma-Aldrich, St. Louis, MO, USA). After washing the slides in 1X TBS (10 min × 3), the secondary antibodies were diluted in a blocking buffer, applied to the sections, and the sections were incubated 1 h at room temperature in the dark.
Images were acquired using a Zeiss LSM 800 confocal microscope with a 63× objective. Some figures were post-processed with Airyscan. All of the genotypes of a given age and antibody were imaged at a single z-plane using identical settings for the laser intensity and the master gain. The digital gain = 1 for all of the images. The pinhole size was set for 1AU on the red channel (39 µm for the 40× objective). The post-processing of the non-saturated images consisted of equal adjustments to the brightness/contrast of the control and knockout images using Adobe Photoshop but without affecting the conclusions which were made. A red channel separation was obtained by isolating the R-channel in the "blender options" of Adobe Photoshop.

Electroretinography
Scotopic and photopic ERG measurements were performed using P15-P18 for Six3Cre experiments and P35 for the iCre75 experiments. Prior to ERG, the mice were dark-adapted overnight and anesthetized with an intraperitoneal injection of 1% ketamine/0.1% xylazine at a 10 µL/g body weight. The mice were kept warm during ERG by using a temperaturecontrolled stage. The scotopic and photopic responses were recorded as described [39][40][41] using a UTAS BigShot Ganzfeld system (LKC Technologies, Gaithersburg, MD, USA). The scotopic single-flash responses were recorded at stimulus intensities of −4.5 log cd s·m −2 [log candela seconds per square meter] to 2.4 log cd s·m −2 ). The mice were light-adapted under a background light of 1.48 log cd s·m −2 for 5 min prior to measuring the photopic responses. The photopic single-flash responses of the control and knockout were recorded at stimulus intensities of −0.1 log cd s·m −2 to 1.9 log cd s·m −2 .

Statistical Analysis
We performed an unbalanced two-factor ANOVA to compare the experimental and control animals for their quantified A-and B-wave ERG measurements across multiple ages. A post hoc multiple comparison was performed using Tukey's honestly significant difference criterion. The statistical significance was determined using an alpha value of p < 0.05. ERG statistics were computed using MATLAB's statistical toolbox "anovan" and "multcompare" functions.

3.1.
Generation of the Retina-( ret Arl2 −/− ) and Rod-Specific ( rod Arl2 −/− ) ARL2 Knockouts Mouse ARL2 (184 residues) is a small GTPase featuring a G domain and small coiledcoil domains ( Figure 1A) encoded by a~6500 bp gene with 5 exons ( Figure 1B). To enable conditional knockouts, we acquired a cell line in which a gene trap flanked by loxP and FRT sites was placed in intron 1 and a third loxP site was placed in intron 3 ( Figure 1C). A floxed Arl2 allele (Arl2 F ) ( Figure 1D) is generated following the FRT-FLP recombination with Flp-recombinase. The deletion of ARL2 in the retina was achieved by mating with Six3Cre transgenic mice [36] expressing Cre recombinase at embryonic day 9 (E9) to yield Arl2 F/F ; Six3Cre knockouts (abbreviated as ret Arl2 −/− ) ( Figure 1E), or by mating with iCre75 transgenic mice expressing Cre recombinase under the control of the rhodopsin promoter to yield rod knockouts ( rod Arl2 −/− ). In iCre75 mice, a Cre expression occurs with Flp-recombinase. The deletion of ARL2 in the retina was achieved by m Six3Cre transgenic mice [36] expressing Cre recombinase at embryonic day 9 (E Arl2 F/F ; Six3Cre knockouts (abbreviated as ret Arl2 −/− ) ( Figure 1E), or by mating w transgenic mice expressing Cre recombinase under the control of the rhodopsin to yield rod knockouts ( rod Arl2 −/− ). In iCre75 mice, a Cre expression occurs duri ond postnatal week when the photoreceptors are postmitotic. The deletion of e 3 truncates ARL2 at residue 20 after exon 1 as exon 4 is out of frame. Genotypi 1F-H) confirmed the presence of loxP (F), the presence of Six3Cre (G, band of 45 the loss of exons 2 and 3 at P6 (H) in both of the knockouts (for details see Figur and Methods). Western blots with the P10 ret Arl2 −/− and P15 rod Arl2 −/− retina con absence of ARL2 ( Figure 1I,J).

ret ARL2 −/− −/− Outer Nuclear Layers Display Abnormal Histogenesis
We generated plastic sections at P10-P15 to study the phenotype of the ret nas relative to the controls. The control sections displayed normal photorecep OS as well as correctly laminated nuclear layers (Figure 2A-C, left panels). B

ret ARL2 −/− −/− Outer Nuclear Layers Display Abnormal Histogenesis
We generated plastic sections at P10-P15 to study the phenotype of the ret Arl2 −/− retinas relative to the controls. The control sections displayed normal photoreceptor IS and OS as well as correctly laminated nuclear layers (Figure 2A-C, left panels). By contrast, the nuclear layers, outer limiting membrane (OLM), and outer plexiform layers (OPL) of the ret Arl2 −/− retinas were severely distorted. The ONL and INL thicknesses varied and the OPL was malformed with the absence of the outer segments, the OLM interruptions, and the photoreceptor nuclear migration into the subretinal space nuclei (Figure 2A-C, right panels). We observed a similar photoreceptor phenotype of retina ONL/INL distortion in P6 ret Dync1h1 −/− mice in which cytoplasmic dynein heavy chain 1 was deleted [41].
Cells 2023, 12, x the nuclear layers, outer limiting membrane (OLM), and outer plexiform layers (O the ret Arl2 −/− retinas were severely distorted. The ONL and INL thicknesses varied an OPL was malformed with the absence of the outer segments, the OLM interruption the photoreceptor nuclear migration into the subretinal space nuclei (Figure 2A-C panels). We observed a similar photoreceptor phenotype of retina ONL/INL distort P6 ret Dync1h1 −/− mice in which cytoplasmic dynein heavy chain 1 was deleted [41].

Cytoplasmic Dynein Is Unstable in ret Arl2 −/− Photoreceptors
Based on the ret Arl2 −/− phenotype with an ONL distortion (Figure 2), we susp that the cytoplasmic dynein, a multi-subunit complex organized around the heavy DYNC1H1, may be affected by the distorted/damaged MTC. The dynein transports toward the minus-end of the microtubules at the basal body and are essential for the lamination, nuclear positioning, vesicular trafficking, and inner/outer segment ela tion [41]. The immunohistochemistry (IHC) of P6 and P10 control retina cryosection anti-DYNC1H1 showed the normal lamination of the nuclear layers ( Figure 3A

Cytoplasmic Dynein Is Unstable in ret Arl2 −/− Photoreceptors
Based on the ret Arl2 −/− phenotype with an ONL distortion (Figure 2), we suspected that the cytoplasmic dynein, a multi-subunit complex organized around the heavy chain DYNC1H1, may be affected by the distorted/damaged MTC. The dynein transports cargo toward the minus-end of the microtubules at the basal body and are essential for the retina lamination, nuclear positioning, vesicular trafficking, and inner/outer segment elaboration [41]. The immunohistochemistry (IHC) of P6 and P10 control retina cryosections with anti-DYNC1H1 showed the normal lamination of the nuclear layers ( Figure 3A-D, left panels). A low magnification image of the entire P6 retina reveals the presence of DYNC1H1 in IS, OPL, and IPL of the control retina ( Figure 3A, left panel), but distorted ONL/INL layers and the absence of DYNC1H1 (green) in the IS and OPL of the ret Arl2 −/− retina ( Figure 3A, right panel). At P6, the control OSs begins to form as guanylate cyclase 1 (GC1, gene nomenclature GUCY2E), a component of the phototransduction cascade, is detectable in budding OSs ( Figure 3B, left panel. inset). GUCY2E is undetectable in ret Arl2 −/− sections ( Figure 3B, right panel). In the P6 controls, the MTC labeled with anti-Ac-α-tubulin is well developed ( Figure 3C, left panel) but severely distorted in the absence of ARL2 ( Figure 3C, right panel). At P10, the C-terminal binding protein 2 (CTBP2 alias RIBEYE) is located in the synaptic region of the controls ( Figure 3D, left panel). The ret Arl2 −/− ( Figure 3D [41,42]. By contrast, the r erage ERG traces at P35 (n = 5) and scotopic a-and photopic b-wave amplitude tion of the light intensity ( Figure 4E-H) indicate the near-normal function of rod and cones. Scotopic a-waves and photopic b-waves were indistinguishable at from −4.5 to 2.4 log cd s m −2 , ruling out significant morphological changes in th outer segments. We conclude that the rod Arl2 −/− rod and cone OSs developed no  Figure 4C,D). Residual scotopic and photopic a-and b-waves are attributed to the formation of the short photoreceptor OS, commonly observed with Six3Cre knockouts in the retina periphery where the expression of Six3Ce is delayed [41,42]. By contrast, the rod Arl2 −/− average ERG traces at P35 (n = 5) and scotopic a-and photopic b-wave amplitudes as a function of the light intensity ( Figure 4E-H) indicate the near-normal function of rod Arl2 −/− rods and cones. Scotopic a-waves and photopic b-waves were indistinguishable at intensities from −4.5 to 2.4 log cd s m −2 , ruling out significant morphological changes in the inner and outer segments. We conclude that the rod Arl2 −/− rod and cone OSs developed normally despite the reduced αβ-tubulin heterodimer biosynthesis.

Effect of ARL2 Deletion on Pericentriolar Material
As ARL2 interacts with the basal body and the rootlet [14], we investigat of the deletion of ARL2 on the localization of the basal body (BB) markers CEP250, employing EGFP-CETN2 to serve as a centriole and connecting ci [39,41]. CEP164 is a distal appendage protein surrounding the BB distal e ( Figure 5A) and is required for BB docking during the development of the p [14,45]. BB docking and CC extension in ret Arl2 −/− photoreceptors is indistingu

Effect of ARL2 Deletion on Pericentriolar Material
As ARL2 interacts with the basal body and the rootlet [14], we investigated the effects of the deletion of ARL2 on the localization of the basal body (BB) markers CEP164 and CEP250, employing EGFP-CETN2 to serve as a centriole and connecting cilium marker [39,41]. CEP164 is a distal appendage protein surrounding the BB distal end [39,43,44] (Figure 5A) and is required for BB docking during the development of the photoreceptor [14,45]. BB docking and CC extension in ret Arl2 −/− photoreceptors is indistinguishable from the controls ( Figure 5A ,B ) but the connecting cilia do not extend axonemes and OS. Rather than being located at the inner segment cortex, ret Arl2 −/− BBs mislocalize within the ret Arl2 −/− ONL ( Figure 5B). Cells 2023, 12, x 9 of 15 the mother and daughter centrioles of both the control and ret Arl2 −/− photoreceptors ( Figure  5C',D'), suggesting that the absence of ARL2 does not affect centriole cohesion. In Arl2 knockout panels, the BB-CC structures are mislocalized into the ONL. CEP164 still enables docking of the basal body and extension of CC, and CEP250 still connects mother and daughter centrioles. Mice were kept on an EGFP-CETN2 transgenic background to mark centrioles and CC. Image was postprocessed with Airy Scan of the LSM800 confocal microscope.

Microtubules Are in Disarray in the ARL2 Knockout
The results shown in Figure 3C are consistent with an MTC defect, generated by the absence of ARL2. The photoreceptor MTC is controlled by the microtubule organizing center (MTOC) consisting of the basal body and daughter centriole. The microtubule minus ends are anchored in the pericentriolar matrix and the basal body [48]. To test the status and stability of the ret Arl2 −/− MTC in the retina ONL and INL, we used anti-acetylated α-tubulin (Ac-Tub) and anti-polyglutamylated tubulin (polyE-Tub) antibodies to probe the P6 and P10 retina cryosections ( Figure 6). The acetylation of lysine 40 of α-tubulin and the addition of glutamate and glycine to both α-and β-tubulin, referred to as polyglutamylation and polyglycylation, are known to stabilize the microtubules (reviewed in [49]. Ac-Tub (white) and polyE-Tub (red) are labeled strongly at P6 and traverse the ONL of the Arl2 F/F retina ( Figure 6A,C). At this age, Mueller glia, which penetrate the ONL at later ages, are not detectable in the ONL/INL [41] (see discussion). At P10, photoreceptor OS are emerging, but the labeling of the connecting cilia and axonemes by Ac-Tub and PolyE is relatively weak ( Figure 6C, right panel, inset), in contrast to a previous report where polyglutamylated microtubules are enriched at the photoreceptor connecting cilium [50]. At P6 (Figure 6B), the ret Arl2 −/− ONL is strongly distorted, acetylated, and the polyglutamylated microtubules appear disrupted and significantly reduced. At P10, ret Arl2 −/− ONL appears to stabilize to a degree, but the ONLs thickness is reduced, the connecting cilia are not forming (Figure 6D), and the photoreceptors are degenerating. The microtubules are severely reduced at P6 ( Figure 6B) and P10 ( Figure 6D), consistent with the reduced levels of tubulin heterodimers in the absence of ARL2 [22,31,34]. The results suggest that the ret ARL2 −/− heterodimer output and MT formation is decreased but not As ARL2 and rootletin are known interactants [14], we investigated whether ARL2 is involved in the centriole cohesion. The centrosome linker proteins CEP250 (C-Nap1), rootletin (CROCC1), and CEP68 connect the mother and daughter centrioles during ciliogenesis [46,47]. We probed P10 Arl2 F/F ; Egfp-Cetn2 + ( Figure 5C) and ret Arl2 −/− ; Egfp-Cetn2 + cryosections ( Figure 5D) with anti-CEP250. The results show that C-NAP1 localizes between the mother and daughter centrioles of both the control and ret Arl2 −/− photoreceptors ( Figure 5C',D'), suggesting that the absence of ARL2 does not affect centriole cohesion.

Microtubules Are in Disarray in the ARL2 Knockout
The results shown in Figure 3C are consistent with an MTC defect, generated by the absence of ARL2. The photoreceptor MTC is controlled by the microtubule organizing center (MTOC) consisting of the basal body and daughter centriole. The microtubule minus ends are anchored in the pericentriolar matrix and the basal body [48]. To test the status and stability of the ret Arl2 −/− MTC in the retina ONL and INL, we used antiacetylated α-tubulin (Ac-Tub) and anti-polyglutamylated tubulin (polyE-Tub) antibodies to probe the P6 and P10 retina cryosections ( Figure 6). The acetylation of lysine 40 of α-tubulin and the addition of glutamate and glycine to both αand β-tubulin, referred to as polyglutamylation and polyglycylation, are known to stabilize the microtubules (reviewed in [49]. Ac-Tub (white) and polyE-Tub (red) are labeled strongly at P6 and traverse the ONL of the Arl2 F/F retina ( Figure 6A,C). At this age, Mueller glia, which penetrate the ONL at later ages, are not detectable in the ONL/INL [41] (see discussion). At P10, photoreceptor OS are emerging, but the labeling of the connecting cilia and axonemes by Ac-Tub and PolyE is relatively weak ( Figure 6C, right panel, inset), in contrast to a previous report where polyglutamylated microtubules are enriched at the photoreceptor connecting cilium [50]. At P6 (Figure 6B), the ret Arl2 −/− ONL is strongly distorted, acetylated, and the polyglutamylated microtubules appear disrupted and significantly reduced. At P10, ret Arl2 −/− ONL appears to stabilize to a degree, but the ONLs thickness is reduced, the connecting cilia are not forming (Figure 6D), and the photoreceptors are degenerating. The microtubules are severely reduced at P6 ( Figure 6B) and P10 ( Figure 6D), consistent with the reduced levels of tubulin heterodimers in the absence of ARL2 [22,31,34]. The results suggest that the ret ARL2 −/− heterodimer output and MT formation is decreased but not abolished.

Microtubule Cytoskeleton Is Unstable in Both ret ARL2 −/− and ret Dync1h1 −/− Retina
The results suggest that the absence of ARL2 early in the postnatal development lowers the tubulin heterodimer levels and affects the MTC maturation, which in turns affects the dynein function. We investigated whether the deletion of DYNC1H1, the key subunit of the cytoplasmic dynein, would affect the MTC similarly, as observed with the ablation of ARL2. We compared the status of the P10 MTC in ret ARL2 −/− and P6 ret Dync1h1 −/− photoreceptors compared to the controls by immunohistochemistry using anti-TUBA1A (α-1A tubulin, red) and anti-TUBB3 (β3-tubulin, green) antibodies (Figure 7). In the control Arl2 F/F and Dync1h1 F/F sections, α-and β-tubulin form a normal photoreceptor cytoskeleton ( Figure 7A,C). In the ret Arl2 −/− ( Figure 7B) and ret Dync1h1 −/− sections ( Figure 7D), the MTC is disorganized and the microtubules are severely reduced. The ret Dync1h1 −/− sections show a reduced ONL thickness and the near-complete absence of microtubules ( Figure  7D). This result indicates that dynein and ARL2 are essential for the microtubule and MTC stabilization.

Microtubule Cytoskeleton Is Unstable in Both ret ARL2 −/− and ret Dync1h1 −/− Retina
The results suggest that the absence of ARL2 early in the postnatal development lowers the tubulin heterodimer levels and affects the MTC maturation, which in turns affects the dynein function. We investigated whether the deletion of DYNC1H1, the key subunit of the cytoplasmic dynein, would affect the MTC similarly, as observed with the ablation of ARL2. We compared the status of the P10 MTC in ret ARL2 −/− and P6 ret Dync1h1 −/− photoreceptors compared to the controls by immunohistochemistry using anti-TUBA1A (α-1A tubulin, red) and anti-TUBB3 (β3-tubulin, green) antibodies (Figure 7). In the control Arl2 F/F and Dync1h1 F/F sections, αand β-tubulin form a normal photoreceptor cytoskeleton ( Figure 7A,C). In the ret Arl2 −/− ( Figure 7B) and ret Dync1h1 −/− sections ( Figure 7D), the MTC is disorganized and the microtubules are severely reduced. The ret Dync1h1 −/− sections show a reduced ONL thickness and the near-complete absence of microtubules ( Figure 7D). This result indicates that dynein and ARL2 are essential for the microtubule and MTC stabilization.

Discussion
To assess the status of the photoreceptor microtubules, we generated the conditional retina and rod knockouts deleting ARL2 before and after ciliogenesis ( Figure 1). The absence of ARL2 during the early retina development affected the stability of ONL and INL (Figures 2 and 3) and the MTC in the photoreceptors as early as P6 (Figures 6 and 7). OSs did not form in the central retina and the ERG responses were diminished ( Figure 4). The ONL/INL distortion also affected the formation of the OPL and synapses ( Figure 3D). The function of the CEP164 (promoting the docking of the basal body to the apical membrane and generating the connecting cilium) and CEP250 (required for the centriole cohesion) were not affected, but the connecting cilia did not localize to the cortex of the cell and were found to be misplaced within the ONL ( Figure 5).
Apart from its function as a release factor for prenylated proteins bound to PDE6D [51], ARL2 is involved in the biosynthesis of αβ-tubulin heterodimers, the building blocks of the microtubules [22,52,53]. The history of ARL2 as a factor in tubulin heterodimer biosynthesis is well documented [30,34,54,55]. The assembly of the αβ-tubulin dimer occurs in a folding cycle following the biosynthesis of α-and β-tubulins, a highly controlled process (see graphical abstract). The assembly requires the orchestrated action of a set of proteins of the chaperonin-containing TCP1 complex (CCT) [56], as well as tubulin-specific co-factors, i.e., TBCA-D [31,32,52]. The in vitro biochemical studies revealed that α-tubulin binds to TBCB and β-tubulin to TBCA that are replaced by TBCD and TBCE [32,52]. ARL2, when bound to TBCD, exchanges GDP with GTP and plays a critical role in the formation of a super complex consisting of TBCC/TBCE/α-tubulin and TBCD/β-tubulin/ARL2-GTP. Following the release of TBCD and TBCE, triggered by GTP hydrolysis, α-tubulin and βtubulin are released as the heterodimers. The consequence Is that, in the absence of ARL2, the tubulin heterodimer levels are reduced, thereby affecting the microtubule's assembly (see graphical abstract). Low levels of heterodimers during the early development of the

Discussion
To assess the status of the photoreceptor microtubules, we generated the conditional retina and rod knockouts deleting ARL2 before and after ciliogenesis ( Figure 1). The absence of ARL2 during the early retina development affected the stability of ONL and INL (Figures 2 and 3) and the MTC in the photoreceptors as early as P6 (Figures 6 and 7). OSs did not form in the central retina and the ERG responses were diminished ( Figure 4). The ONL/INL distortion also affected the formation of the OPL and synapses ( Figure 3D). The function of the CEP164 (promoting the docking of the basal body to the apical membrane and generating the connecting cilium) and CEP250 (required for the centriole cohesion) were not affected, but the connecting cilia did not localize to the cortex of the cell and were found to be misplaced within the ONL ( Figure 5).
Apart from its function as a release factor for prenylated proteins bound to PDE6D [51], ARL2 is involved in the biosynthesis of αβ-tubulin heterodimers, the building blocks of the microtubules [22,52,53]. The history of ARL2 as a factor in tubulin heterodimer biosynthesis is well documented [30,34,54,55]. The assembly of the αβ-tubulin dimer occurs in a folding cycle following the biosynthesis of αand β-tubulins, a highly controlled process (see graphical abstract). The assembly requires the orchestrated action of a set of proteins of the chaperonin-containing TCP1 complex (CCT) [56], as well as tubulin-specific co-factors, i.e., TBCA-D [31,32,52]. The in vitro biochemical studies revealed that α-tubulin binds to TBCB and β-tubulin to TBCA that are replaced by TBCD and TBCE [32,52]. ARL2, when bound to TBCD, exchanges GDP with GTP and plays a critical role in the formation of a super complex consisting of TBCC/TBCE/α-tubulin and TBCD/β-tubulin/ARL2-GTP. Following the release of TBCD and TBCE, triggered by GTP hydrolysis, α-tubulin and β-tubulin are released as the heterodimers. The consequence Is that, in the absence of ARL2, the tubulin heterodimer levels are reduced, thereby affecting the microtubule's assembly (see graphical abstract). Low levels of heterodimers during the early development of the photoreceptor effectively curtail the production of MT filaments which are necessary to stabilize dynein.
In Drosophila melanogaster neuroblasts, Arl2 physically associates with tubulin cofactors C, D, and E. An Arl2 RNA interference, Arl2-GDP expression, or Arl2 deletions caused microtubule abnormalities, suggesting that ARL2 regulates the microtubule's growth [57]. Chen et al., conclude that Arl2 and its cofactors are probably responsible for exquisitely regulating the free tubulin heterodimer levels in the cell and the localization of dynein at the microtubules. However, the knockdown of ARL2 by siRNA in a tissue culture produced conflicting results. The deletion of ARL2 in immortal HeLa cells did not produce a phenotype [33], but silencing ARL2 in human neural progenitor cells provoked an apoptotic phenotype [58]. The depletion of Arl2 by siRNA in human cell lines resulted in a TBCD-mediated microtubule disruption [31]. ARL2 siRNA significantly reduced the cilia lengths of ARPE19 cells [15].
We observed that the cytoplasmic dynein levels in the photoreceptor IS were reduced in ret Arl2 −/− photoreceptors compared to the controls (Figure 3), suggesting that cytoplasmic dynein may become unstable in the absence of microtubule tracks. We previously observed that the ablation of the dynein-heavy chain in the retina negatively affected the MTCs stability [41]. We therefore tested whether the microtubules and dynein were interdependent for their stability. Floxed Dync1h1 P6 retina cryosections probed with anti-TUBB3 and anti-TUBA1A revealed a stable MTC comparable to the Arl2 controls ( Figure 7C). By contrast, the Dync1h1 −/− sections revealed a severely attenuated ONL with an unstable MTC with few intact microtubules ( Figure 7D). The effect on the microtubules in the ret Arl2 −/− sections is less dramatic, presumably because the removal of ARL2 reduces but not obliterates the tubulin heterodimer levels. The MTC instability in the ONL could be in part caused by the degenerating Mueller glia. However, at P6, Mueller glia processes are not detectable in the ONL/INL and begin to penetrate the ONL after P10, as we have shown using anti-glutamine synthase as a marker [41]. Therefore, the instability of the ONL MTC is most likely caused by the reduced availability of tubulin heterodimers in photoreceptor axons. The direct interaction of ARL2 with cytoplasmic dynein has, to the best of our knowledge, not previously been reported. A link between the microtubule's stability and dynein-2 (IFT dynein) has been provided in Tetrahymena. The microtubules of a microtubule-dynein-2 complex appear to be intact under conditions that result in a microtubule depolymerization. When dynein was dissociated from the complex with the addition of ATP, no microtubules were found in specimens under the same depolymerizing conditions, suggesting a microtubule catastrophe [59].
Surprisingly, the rod Arl2 −/− retina (rod knockout) is stable and functional in the first six postnatal weeks ( Figure 4E-H). The tubulin heterodimer biosynthesis is expected to slowly decrease in rod Arl2 −/− mice in the first two postnatal weeks before the Cre expression under the control of the rhodopsin promoter begins. An MTC can be established, and the photoreceptors are mature. After the onset of the expression of Cre, this is visible by immunohistochemistry at around P16 [40]. ARL2 is knocked out but the tubulin heterodimer apparently is still produced at levels sufficient to maintain the MTC. These findings suggest that the effect of the inactivation of ARL2 is defined by a developmental switch, ending the rod maturation phase (P6-P16). In the early developmental phase, photoreceptors are building the MTC, requiring high levels of tubulin, establish nuclear layers, and IS and OS. In the mature phase (>P16), rods are more stable and can tolerate the loss of ARL2 and reductions in the heterodimers. An example for a developmental switch is the PKD1 (polycystin-1) knockout. In a mouse Pkd1 conditional knockout model, the inactivation of PKD1 before postnatal day 13 results in the formation of severe cystic kidney cysts within 3 weeks, whereas an inactivation at day 14 and later results in cysts only after 5 months. These findings suggest that the effects of a Pkd1 inactivation are defined by a developmental switch that signals the end of the terminal renal maturation process [60].

Conclusions
ARL2 is a multifunctional protein expressed in vertebrates, invertebrates, yeast, and plants. Our results suggest that during the early development of the photoreceptor, the photoreceptor microtubule cytoskeleton is unstable due to the insufficient production of tubulin heterodimers. We suggest that the ret Arl2 −/− phenotype appears to be a consequence of a triple insult: an impaired tubulin heterodimer synthesis, reduced microtubule assembly, giving rise to a nonfunctional MTC, and the downregulation of dynein. By contrast, the rod Arl2 −/− photoreceptors are stable and can tolerate the loss of ARL2 and reductions in the heterodimers.

Institutional Review Board Statement:
The study was conducted according to guidelines of the Declaration of Helsinki and was approved by the University of Utah's Institutional Animal Care and Use Committee (Protocol 18-11005).

Informed Consent Statement: Not applicable.
Data Availability Statement: The data supporting the reported results are shown in Figures 1-7. Additional data from our lab concerning DYNC1H1 and CEP164 can be found in [40,41,61] and [39], respectively.