Farnesylation of the Transducin G Protein Gamma Subunit Is a Prerequisite for Its Ciliary Targeting in Rod Photoreceptors

Primary cilia are microtubule-based organelles, which protrude from the plasma membrane and receive a wide range of extracellular signals. Various cilia use G protein-coupled receptors (GPCRs) for the detection of these signals. For instance, vertebrate rod photoreceptors use their cilia (also called outer segments) as antennae detecting photons by GPCR rhodopsin. Rhodopsin recognizes incoming light and activates its G protein, transducin, which is composed of three subunits α, β, and γ. Similar to all G protein γ subunits, the transducin Gγ1 subunit undergoes C-terminal prenylation resulting in the addition of an isoprenoid farnesyl; however, the significance of this posttranslational modification is unclear. To study the role of the farnesyl group, we genetically introduced a mutant Gγ1 that lacked the prenylation site into the retinal photoreceptors of mice. The biochemical and physiological analyses of these mice revealed that mutant Gγ1 dimerizes with the endogenous transducin Gβ1 subunit and that the resulting Gβγ dimers display reduced hydrophobicity. Although mutant Gβγ dimers could form a heterotrimeric G protein, they could not mediate phototransduction. This deficiency was due to a strong exclusion of non-farnesylated Gβγ complexes from the cilia (rod outer segments). Our results provide the first evidence that farnesylation is required for trafficking of G-protein βγ subunits to the cilium of rod photoreceptors.


INTRODUCTION
Ciliary signaling is commonly mediated by G protein-coupled receptors (GPCRs), which are integral membrane proteins (Schou et al., 2015). Their coupled heterotrimeric G proteins are soluble proteins attached to the lipid bilayer through posttranslational lipid modification. The difference in their lipid bilayer interaction underlies distinct cilia targeting mechanisms for GPCRs and G proteins, which is exemplified in retinal rod photoreceptors of vertebrates.
Rods use their cilia, called outer segments, as antennae specialized for photon detection. The outer segment is packed with flat lamellar membranes, or discs, containing photosensory GPCR, rhodopsin, coupled to heterotrimeric G protein, transducin (Pearring et al., 2013). When activated by light, rhodopsin signals transducin triggering a chain of signaling events leading to rod hyperpolarization and synaptic response. It is generally accepted that rhodopsin is added to the disc membrane through a distinct trafficking pathway regulated by Arf and Rab GTPases (Wang and Deretic, 2014). Nascent discs originate from lamellae, which grow outward from the outer segment and either evaginate or invaginate from the ciliary plasma membrane (Steinberg et al., 1980;Burgoyne et al., 2015;Ding et al., 2015;Volland et al., 2015). Fusion of the leading edges of adjacent lamellae results in formation of discrete discs, which are no longer continuous with the ciliary plasma membrane. Then the concentration of any integral membrane protein present on the disc, including rhodopsin, does not change. At the same time, the amount of transducin in the rod outer segment is variable. In resting rods, transducin accumulates in outer segment discs, which maximizes the rate of transducin activation by rhodopsin, hence, the sensitivity of rod responses. However, during sustained exposure to bright ambient light which saturates rods, a major fraction of transducin undergoes translocation from the rod outer segment to the cell body. This results in a drastic reduction in the concentration of transducin in the outer segment (Brann and Cohen, 1987;Philp et al., 1987;Whelan and McGinnis, 1988;Sokolov et al., 2002;Nair et al., 2005;Lobanova et al., 2007;Rosenzweig et al., 2007). This response is reversed in the dark, when the displaced transducin returns to the outer segment and re-populates discs (Sokolov et al., 2002;Belcastro et al., 2012). Previous research indicates that transducin exits outer segments by diffusion, after losing its affinity to disc membrane, which highlights the central role of lipid groups in this process (reviewed in Calvert et al., 2006;Artemyev, 2008;Slepak and Hurley, 2008;Pearring et al., 2013).
Rod transducin is a prototypical heterotrimeric G protein comprised of α subunit and a complex of tightly associated β and γ subunits. The α and γ subunits are lipid-modified, which is required for their membrane association and protein interaction (Wedegaertner et al., 1995). Similar to other G-protein γ subunits, rod transducin γ (Gγ 1 ) carries a covalently-attached isoprenoid, farnesyl (Fukada et al., 1990). The prenylation reaction, the addition of an isoprenoid group either farnesol or geranylgeraniol, is directed at cysteine residue 71 within a carboxyl-terminal CAAX box (C 71 VIS) displayed by the nascent Gγ 1 (Lai et al., 1990). The isoprenoid attachment is followed by the removal of the last three amino acids (VIS) and methylation of the carboxyl group of the new c-terminus (Higgins and Casey, 1994). Thus, mature Gγ 1 is three amino acid residues shorter than its precursor polypeptide and its c-terminal C 71 is farnesylated and methylated. While the prenylation of the G-protein γ subunits is not required for the assembly of the βγ dimer, it is indispensable for the membrane targeting of Gβγ (Simonds et al., 1991). In addition to membrane targeting, prenylation enhances the binding of Gβγ to G-protein α subunits (Iñiguez-Lluhi et al., 1992). Currently there is a lack of in vivo experiments, assessing why farnesylation of Gγ is important for visual function. Some evidence suggests that posttranslational lipid modification controls transducin compartmentalization in retinal photoreceptors. For example, replacing the farnesyl group of Gγ 1 with the more hydrophobic isoprenoid group, geranylgeranyl renders the transducin Gβ 1 γ 1 dimer incapable of undergoing light-driven translocation from rod outer segments (Kassai et al., 2005). In Drosophila, farnesylation of transducin Gγe was critical for the formation of membrane-associated Gαβγ heterotrimer competent to interact with its GPCR (Schillo et al., 2004). However, it is not known if this role is also true for vertebrates or if there are other roles. Our data provide the first evidence that farnesylation of transducin Gγ 1 is also necessary for targeting of the Gβ 1 γ 1 dimer to the sensory cilium of vertebrate rod photoreceptors in vivo.

MATERIALS AND METHODS
Generation of HA Gγ 1 and HA Gγ 1 C71S Transgenic Mice A transgene expressing the γ subunit of rod-specific transducin (protein: Gγ 1 , gene: Gngt1) was synthetized by Integrated DNA Technologies (IDT, Coralville, IA, USA). This construct designated as HA Gγ 1 contained HA epitope tag, LMA linker, mouse Gngt1 coding sequence, and a 4.4 kb mouse rhodopsin promoter (Lem et al., 1991). The cysteine-to-serine substitution at position 71, resulting in HA Gγ 1 C71S mutant lacking the prenylation site, was introduced by a PCR-based strategy and the QuikChange Lightning Site-Directed Mutagenesis Kit (Agilent Technologies, Santa Clara, CA, USA) with the following primers: forward primer 5 -AAG GAA CTC AAA GGA GGC TCT GTG ATT TCA TAG TAG G and reverse primer 5 -CCT ACT ATG AAA TCA CAG AGC CTC CTT TGA GTT CCT T (with the underlined base indicating the change from wild type (WT) sequence). The integrity of both constructs was confirmed by sequence analysis. Then they were purified and injected into the pronuclei of zygotes from superovulated FVB females at the WVU Transgenic Animal Core Facility. Transgene integration was determined for both groups by PCR genotyping of tail DNA using the following primers: forward primer 5 -TAC CCA TAC GAT GTT CCA GAT TAC GCT and reverse primer 5 -TCA CAC AGC CTC CTT TGA GTT CCT. The colonies were established by crossing transgenic HA Gγ +/− 1 and HA Gγ 1 C71S +/− heterozygotes with WT partners of 129-E background (Charles River). To move both transgenes to Gγ 1null background, HA Gγ +/− 1 and HA Gγ 1 C71S +/− heterozygotes were subjected to several round of crossing with Gγ −/− 1 mice (Kolesnikov et al., 2011) to obtain HA Gγ mice. An identical breeding strategy using Gα t1 knockout mice (Calvert et al., 2000) was used to generate HA Gγ strain. All experiments involving mice were performed according to procedures approved by the Animal Care and Use Committee of West Virginia University.

Dark Adaptation and Light Conditioning of Mice
For dark adaptation, mice were kept in their original cages in the dark room for at least 12 h, and from then on, all procedures were performed under dim red light. For light conditioning, animal's pupils were dilated with a mixture of 1.25% phenylephrine hydrochloride and 0.5% tropicamide ophthalmic solution for 20 min, after which mice were exposed to diffused 5000 lux white light, while free running in a white box for 10 min. Subsequently, mice were euthanized and their eyes were harvested and fixed.

Quantification of Proteins by Western Blotting
To quantify protein levels in isolated retinas, retina was dissected from the eye, gently cleaned from the contaminating tissues, and frozen on dry ice. Frozen retina was thawed and immediately homogenized by short ultrasonic pulses in 0.2 ml of urea sample buffer (USB) containing 125 mM Tris/HCl, pH 6.8, 4% SDS, and 6M urea. The extract was cleared by centrifugation. Total protein concentration was determined on a Nanodrop ND-1000 spectrophotometer, and all tested samples were adjusted to the lowest value. Prior to analysis by SDS PAGE, bromophenol blue tracking dye and 5% β-mercaptoethanol were added to each sample. To quantify Gα t1 , a whole eye was enucleated and frozen on dry ice. Frozen eyes were homogenized in USB supplemented with bromophenol blue tracking dye and 5% β-mercaptoethanol by short ultrasonic pulses. The extracts were cleared by centrifugation. Equal aliquots of the compared samples were separated next to each other on a 10%-20% SDS PAGE, transferred to Immobilon FL membrane (Immunobilon-FL, Millipore, Billerica, MA, USA), and analyzed by Western blotting, using an Odyssey Infrared Imaging System (LI-COR Biosciences, Lincoln, NE, USA) according to the manufacturer's protocol.
Pull Down of Epitope-Tagged HA Gγ 1 and HA Gγ 1 C71S Retinas were dissected and homogenized in RIPA buffer (R0278, Sigma, St. Louis, MO, USA) by short ultrasonic pulses. Resulting retinal extracts were cleared by centrifugation, and the supernatant was incubated with washed Pierce anti-HA magnetic beads (88836, Thermo Fisher Scientific, Waltham, MA, USA) for 30 min at room temperature. Beads were washed four times for 3 min with RIPA buffer, after which the captured proteins were eluted with 3% ammonium hydroxide solution and vacuum-dried. For Western blot analysis, lyophilized samples were reconstituted in USB buffer. In phosducin co-precipitation assay, instead of RIPA buffer the pull downs were conducted in 10 mM Hepes/HCl, pH 7.0, 180 mM NaCl, 2% IGEPAL CA-630 (56741, Sigma, St. Louis, MO, USA) supplemented with Protease inhibitors cocktail set 1 (539131, Calbiochem).

Triton X-114 Phase Partitioning
Method was modified from Justice et al. (1995). Retinal extract was prepared by homogenizing four retinas in 0.2 ml of Buffer A (PBS, protein inhibitor cocktail (Roche), bromophenol blue tracking dye), and clearing insoluble parts by centrifugation. 20 µl of 10% Triton X-114 (648468, Calbiochem) was added to 180 µl of the retinal extract, mixed by gentle inversion and pre-warmed to 37 • C for 5 min. The sample was centrifuged (300× g, 10 min) at 37-40 • C leading to a separation of aqueous and Triton X-114 layers, with the later becoming blue-colored due to migration of bromophenol blue into the detergent phase. The upper aqueous layer was collected in a new test tube and mixed with 20 µl of 10% Triton X-114. The lower Triton X-114 layer was mixed with 0.2 ml of Buffer A, to equalize the compositions and volumes of the two fractions. Then, 0.8 ml of RIPA buffer was added to each fraction, and the epitope-tagged Gγ 1 was captured with anti-HA magnetic beads and analyzed by Western blotting, as described above.

Immunofluorescence Microscopy
Frozen Retinal Cross-sections Enucleated eye was immersed in freshly prepared 4% paraformaldehyde in PBS for 5 min at room temperature, and then the cornea was removed. The eye was fixed for additional 55 min, washed in PBS three times for 10 min, and incubated in 20% sucrose in PBS overnight at 4 • C. The eye was then incubated in 1:1 mixture of 20% sucrose in PBS:OCT (Cryo Optimal Cutting Temperature Compound, Sakura) for 1 h. The lens was removed, and the resulting eyecup was positioned in a plastic tray with OCT, and flash frozen on a dry ice/ethanol bath. Sixteen micrometer thick cross-sections were cut on a Leica CM1850 Cryostat, and placed on Superfrost Plus slides (Fisher Scientific). Retinal sections mounted on slides were washed with PBS to remove OCT, and then blocked for 1 h in PBS containing 5% goat sera and 0.5% Triton X-100. Primary and secondary antibodies were diluted in PBS containing 2.5% goat sera and 0.5% Triton X-100. Incubation with primary antibodies typically lasted overnight at room temperature. Slides were washed two times for 15 min with PBS containing 0.1% Triton X-100, prior to the incubation with secondary antibodies, Alexa Fluor 568 goat anti-rat (A11077, Life Technologies) and Alexa Fluor 488 donkey anti-rabbit (A21206, Invitrogen), and 4 ,6-diamindino-2-phenylindole dihydrocholide (DAPI; Roche, Indianapolis, IN, USA) nuclear stain, for 1 h. Slides were washed three times for 10 min with PBS containing 0.1% Triton X-100, mounted with ProLong Gold (Life Technologies), and cover slipped. Images were acquired on a Zeiss LSM 510 and Nikon C2 confocal microscopes, and processed using a NIC Elements Imaging Software.

Flat-Mounted Retina
The whole eye was fixed in 4% paraformaldehyde in PBS for 5 min at room temperature, the anterior segment was removed, and the retina was extracted. The retina was cleaned from contaminating tissues and fixed in the same solution for 6 h. The fixed retina was washed three times for 30 min with PBS, blocked for 4 h in PBS containing 5% goat sera and 0.5% Triton X-100, and probed with primary and secondary antibodies, which were prepared as described above, for at least 14 h. The washes after applying each antibody were three times for 30 min with PBS containing 0.1% Triton X-100. The washed retina was positioned on Superfrost Plus slides (Fisher Scientific) with the outer segments of the photoreceptors facing up. Four equally placed radial cuts were made to flatten the retina, and then the retina was mounted, and cover slipped for imaging as described above.

Electroretinography (ERG)
Mice were dark-adapted overnight prior to testing, and all procedures were performed under dim red light. During recording, mice were anesthetized by breathing 1.5% isoflurane with 2.5 liters per minute (lpm) oxygen through a nose cone, while lying on a heated platform. The animal's pupils were dilated with a mixture of 1.25% phenylephrine hydrochloride and 0.5% tropicamide ophthalmic solution. A reference needle electrode was inserted under the loose skin between the ears. Flash Electroretinography (ERG) responses were recorded from both eyes with custom-made silver wire electrodes positioned on each cornea, with contact being made with a drop of hypromellose solution (2% hypromellose in PBS; Gonioscopic Prism Solution, Wilson Ophthalmic, Mustang, OK, USA). The recordings were performed on the UTAS Visual Diagnostic System with BigShot Ganzfeld, UBA-4200 amplifier and interface, and EMWIN 9.0.0 software (LKC Technologies, Gaithersburg, MD, USA).

Experimental Design and Statistical Analyses
Each confocal microscopy experiment was repeated at least three times to ensure that the results were reproducible from animal to animal and also between the slides. The representative images with the most commonly observed staining patterns and unperturbed morphology, usually originating from central retina in immediate vicinity to the optic nerve, were selected for the figures. In all quantifications (except for Figure 2A), the significance level was determined using the independent two-tailed Student's t-test, and the values were expressed as mean ± SEM. The light-sensitivity curves of ERG a-wave (Figure 2A) were compared using Friedman repeated measures analysis of variance (ANOVA) on ranks.
HA Gγ 1 band and a 35 kD band ( Figure 1B). The 35 kD band was identified as Gβ 1 by Western blotting (Figure 1C). When this assay was conducted in a milder non-ionic detergent, the pull downs also contained phosducin (Figure 1D), an abundant rod phosphoprotein that forms a complex with Gβγ dimers (Gaudet et al., 1996). Thus, each transgenic protein formed a dimer with endogenous Gβ 1 and the resulting Gβγ dimers were capable of binding to phosducin in vivo. The hydrophobicity of these chimeric Gβγ dimers was assessed based on their partitioning between the aqueous and detergent Triton X-114 phases ( Figure 1E; Supplementary Figure S1). We found that Gβ 1 HA Gγ 1 C71S predominantly partitioned to the aqueous phase, while a larger fraction of the Gβ 1 HA Gγ 1 was retained by the detergent phase. Gβ 1 HA Gγ 1 partitioning was similar to the partitioning of endogenous Gβ 1 γ 1 . We attributed the Frontiers in Molecular Neuroscience | www.frontiersin.org Rescue of Gγ 1 Knockout Mice by HA Gγ 1 and HA Gγ 1 C71S To evaluate the functional efficiency of HA Gγ 1 and HA Gγ 1 C71S, the corresponding transgenic lines were backcrossed into a Gγ 1 knockout background. Gγ 1 knockout mice have been shown to display diminished rod responses measured by ERG (Kolesnikov et al., 2011). Therefore, we comparatively analyzed HA Gγ 1 and HA Gγ 1 C71S mice of Gγ −/− 1 background vs. their transgene-negative littermates by ERG. We focused on the amplitude of the a-wave, since it is directly generated by mass rod response (Figure 2A; Supplementary Figure S2). Our analysis revealed that Gγ −/− 1 mice expressing HA Gγ 1 produced significantly larger ERG a-waves than their transgenenegative siblings across a wide range of stimuli (Figure 2A, compare white and black circles). This rescue of the a-wave appeared to be partial, because its maximum amplitude of 250 µV was about half of an average WT value (Supplementary Figure S2). No ERG a-wave increase was observed in HA Gγ 1 C71S-expressing mice (Figure 2A, compare gray and black circles).
Rod photoreceptors of Gγ 1 knockout mice fail to elicit normal ERG responses due to a major destabilization of transducin, a heterotrimeric G protein that mediates visual signaling. As such, the level of transducin Gα t1 subunit in Gγ 1 -null rods declines by more than 5-fold (Lobanova et al., 2008;Kolesnikov et al., 2011), and was 6.1 ± 0.9% (n = 3, SEM) of the WT level in the used Gγ −/− 1 strain (data not shown). To test the ability of HA Gγ 1 and HA Gγ 1 C71S to rescue this phenotype, we compared the level of Gα t1 in HA Gγ 1 and HA Gγ 1 C71S mice of Gγ −/− 1 background to that of their transgene-negative littermates by Western blotting. Given that Gα t1 is exclusively expressed in rod photoreceptors, the analysis was conducted in whole eye extracts. We found that both transgenes, HA Gγ 1 and HA Gγ 1 C71S increased the level of Gα t1 in the retina of Gγ 1 knockout mice by ∼25% of the base level of Gα t1 in our Gγ −/− 1 strain (Figure 2B), which corresponds to 1.5% increase compared to the WT level of Gα t1 .
While breeding these mice, we estimated how much of HA Gγ 1 is expressed from the transgene compared to the endogenous Gγ 1 . For that, whole retinal extracts were prepared from HA Gγ 1 littermates of Gγ −/− 1 and Gγ +/− 1 backgrounds. Aliquots of these extracts, containing equal amounts of total protein, were analyzed by Western blotting using antibody against all three transducin subunits with β-Tubulin, as a loading control (Figure 2C). The amount of transgenic HA Gγ 1 in the retina was much lower than the endogenous Gγ 1 level, which made a direct comparison of both bands inaccurate due to the non-linearity of Western blotting. To overcome this limitation, we compared retinal extracts of HA Gγ +/− 1 ; Gγ −/− 1 mice with WT retinal extracts, which was diluted with Gγ 1null extracts. Our goal was to find a dilution at which both signals would match. Each specific band was detected with anti-Gγ 1 , which recognizes the same epitope within HA Gγ 1 and Gγ 1 . We found that when a WT retinal extract was diluted by 100-fold, both signals began to match, which thus provided an accurate estimation that transgenic HA Gγ 1 was expressed at about 1.2 ± 0.3% (n = 3, SEM) of the endogenous Gγ 1 . The level of HA Gγ 1 remained the same in HA Gγ 1 mice of WT background.
FIGURE 3 | Mosaic expression of the transgene in the retina. Flat-mounted retina from HA Gγ 1 mouse was analyzed by immunofluorescence confocal microscopy. Rod outer segments (OSs) were visualized with antibody against peripherin/rds (Per, green), and epitope-tagged HA Gγ 1 was visualized with antibody against HA (red). In the shown 100 µm × 100 µm × 20 µm z-stack, about 9% of rods are estimated to express HA Gγ 1 .
The low abundance of HA Gγ 1 in the retina was generally consistent with the mosaic expression of the transgene. To determine the percentage of rods that expressed our transgene, we visualized HA Gγ 1 -expressing rods in flat-mounted retinal preparation with antibody against HA by immunofluorescent confocal microscopy (Figure 3, red). Their number was compared with the total number of rod cells visualized with peripherin/rds (Figure 3, green). According to our estimation the transgene was expressed in 9% of rods. Thus, the level of HA Gγ 1 in each transgene-positive rod was about 10% of the endogenous Gγ 1 level.

Subcellular Localization of HA Gγ 1 and HA Gγ 1 C71S under Different Conditions of Illumination
Vertebrate rods display a unique physiological response to saturating levels of light including the translocation of transducin from the rod outer segment to other compartments of these cells (Artemyev, 2008). To test whether HA Gγ 1 re-distributes along with transducin, we examined its subcellular localization under different conditions of illumination (Figure 4). The epitope-tagged HA Gγ 1 was readily detectable in retinal crosssections by immunofluorescence confocal microscopy with the antibody against HA tag. We confirmed that this protein was only expressed in a fraction of rods, however this mosaicism often provided an unobstructed view of individual rod cells. This view would have been impossible otherwise due to a high density and significant overlap of these neurons in the retina.
Our analyses revealed that in dark-adapted mice HA Gγ 1 appeared predominantly in rod OSs, where it co-localized with the outer segment marker peripherin/rds (Figure 4). In mice exposed to bright ambient light for 10 min, HA Gγ 1 immunoreactivity declined in rod outer segments and increased in other cellular compartments, including the inner segment, cytoplasm around the nucleus, and the synapse (Figure 4). In agreement with previous report (McGinnis et al., 2002), the base of the outer segment adjacent to the inner segment appeared to lose the HA Gγ 1 signal first. Such subcellular distribution of HA Gγ 1 supported a notion that its chimeric Gβγ dimers accumulated in the rod outer segment in FIGURE 4 | Subcellular localization of HA Gγ 1 and HA Gγ 1 C71S under different conditions of illumination. HA Gγ 1 and HA Gγ 1 C71S mice were either dark-adapted overnight (Dark) or light-conditioned for 10 min (Light). Protein localization was determined in frozen retinal cross-sections by immunofluorescence confocal microscopy. Peripherin/rds (Per, green) was utilized as the OS marker, HA Gγ 1 and HA Gγ 1 C71S were visualized with antibody against HA-tag (red), and photoreceptors nuclei (blue) were stained with DAPI. The indicated retinal layers are: OS, outer segments layer; IS, inner segments layer; ONL, outer nuclear layer containing rod nuclei; OPL, outer plexiform layer containing rod synapses.
FIGURE 5 | Exclusion of HA Gγ 1 C71S and syntaxin 3 from rod OSs. Retina of dark-adapted HA Gγ 1 C71S mice were analyzed by immunofluorescence confocal microscopy. The immunostaining of HA Gγ 1 C71S (red) and syntaxin 3 (green) are shown together with DIC image. Photoreceptors nuclei are stained with DAPI. The indicated retinal layers are: OS, outer segments layer; IS, inner segments layer; ONL, outer nuclear layer containing rod nuclei; OPL, outer plexiform layer containing rod synapses. the dark, and moved away from this compartment during sustained light exposure. This data further strengthened our conclusion that epitope-tagged HA Gγ 1 expressed from a transgene became fully integrated into rod photoreceptors' transducin pool.
Similar to HA Gγ 1 , transgenic HA Gγ 1 C71S was clearly visible in rods. In contrast to HA Gγ 1 , this protein was found to be excluded from rod outer segment under all conditions of illumination. For example, even after 12 h of dark adaptation rod outer segments contained virtually no HA Gγ 1 C71S (Figure 4). The distribution of HA Gγ 1 C71S in rods was identical to that of syntaxin 3 (Protein: syntaxin 3, gene: Stx3/Syn-3; Figure 5), whose segregation from the outer segments has been attributed to the gating function of photoreceptor's cilium (Datta et al., 2015). For that reason, we first hypothesized that Gβγ dimers comprised of non-farnesylated HA Gγ 1 C71S may be excluded from the rod outer segment due to the gating function of the cilium. However, we also monitored subcellular localization of arrestin (protein: visual arrestin 1, gene: Arr1/Sag), a phototransduction protein that undergoes translocation in the opposite direction of transducin, i.e., arrestin withdraws from the cell body and accumulates in the rod outer segment upon light stimulation (Broekhuyse et al., 1985;Philp et al., 1987;Mangini and Pepperberg, 1988;Whelan and McGinnis, 1988; also reviewed in Pearring et al., 2013). We observed unobstructed translocation of arrestin in all examined rods, indicating that this non-lipidated soluble protein comparable to a Gβγ dimer in size moves freely through connecting cilium, while HA Gγ 1 C71S remains blocked (Figure 6). This result is generally incompatible with a notion the photoreceptor's cilium indiscriminately obstructs the diffusion of all soluble proteins, including non-farnesylated Gβγ.

Transducin α Subunit Is Required for the Accumulation of HA Gγ 1 in Rod Outer Segments
To determine the role of the transducin α subunit (protein: Gα t1 , gene: Gnat) in the trafficking of HA Gγ 1 to the rod OS, HA Gγ 1 mice were backcrossed into Gα t1 -null background, which eliminated the expression of transducin α subunit (Calvert et al., 2000). In the absence of Gα t1 , endogenous Gβ 1 γ 1 spreads throughout the rod photoreceptor and does not undergo lightdriven translocation (Zhang et al., 2003;Lobanova et al., 2008;Belcastro et al., 2012). When we analyzed subcellular localization of HA Gγ 1 in the dark-adapted Gα t1 -null rods, we found this FIGURE 7 | Subcellular localization of HA Gγ 1 in the absence of transducin α subunit. HA Gγ 1 mice of Gα t1 −/− background were either dark-adapted overnight (DARK) or light-conditioned for 10 min (LIGHT). Protein localization was determined in frozen retinal cross-sections by immunofluorescence confocal microscopy. Peripherin/rds (Per, green) was utilized as the OS marker, HA Gγ 1 (red) was visualized with antibody against HA-tag, and photoreceptors nuclei (blue) were stained with DAPI. The indicated retinal layers are: OS, outer segments layer; IS, inner segments layer; ONL, outer nuclear layer containing rod nuclei; OPL, outer plexiform layer containing rod synapses.
protein in all compartments of rod cells, including the inner segment, cytoplasm around the nuclei, and the synapse. At the same time, a significant amount of HA Gγ 1 remained in the rod outer segment regardless of the different conditions of illumination (Figure 7). These data further demonstrate that Gβγ dimers comprised of transgenic HA Gγ 1 fully emulate the properties of endogenous Gβ 1 γ 1 and that they could serve as a surrogate for endogenous Gβ 1 γ 1 .

HA Gγ 1 Enhances Targeting of the Transducin α Subunit in the Rod Outer Segment
Back-crossing HA Gγ 1 and HA Gγ 1 C71S mice into a Gγ 1 -null background allowed us to evaluate whether farnesylated and non-farnesylated Gγ could target endogenous transducin Gα to the rod OS. For each strain, we compared subcellular localization of Gα t1 in the dark-adapted transgene-negative and transgenepositive mice. In rods of transgene-negative Gγ 1 -null mice that do not express any Gγ, and therefore cannot assemble Gβγ expressing indicating transgene (Tg) were dark-adapted overnight and protein localization was determined in frozen retinal cross-sections by immunofluorescence confocal microscopy. Transducin α subunit (green) was detected with antibodies against Gα t1 , epitope-tagged HA Gγ 1 and HA Gγ 1 C71S (red) were visualized with antibody against HA, and photoreceptors nuclei (blue) were stained with DAPI. The indicated retinal layers are: OS, outer segments layer; IS, inner segments layer; ONL, outer nuclear layer containing rod nuclei; OPL, outer plexiform layer containing rod synapses.
dimer (Kolesnikov et al., 2011), Gα t1 was present in all rod cellular compartments (Figure 8). This result is consistent with previous work that found a similar subcellular distribution of Gα t1 in rods of Gγ t1 -null mice, which was determined by Western blot analysis of serial tangential sections of the retina (Lobanova et al., 2008). In contrast to Gγ 1 -null mice, Gα t1 in HA Gγ 1 -expressing rods was found predominantly in rod outer segments where it co-localized with HA Gγ 1 (Figure 8). However, in HA Gγ 1 C71S-expressing rods the localization of Gα t1 remained the same as in the transgene-negative mice (Figure 8). These results demonstrate that farnesylated HA Gγ 1 enhances targeting of Gα t1 to the outer segment of Gγ 1 -null rods, the same as WT Gγ 1 (Lobanova et al., 2008), whereas non-farnesylated HA Gγ 1 C71S appeared to lack this ability. They also support a notion that Gα t1 is unable to shuttle non-farnesylated Gβγ dimers into the cilium, which is evident from a complete exclusion of HA Gγ 1 C71S from rod outer segments (Figure 8).

Farnesylation Controls Ciliary Localization of Gβγ Dimer
We found that epitope-tagged HA Gγ 1 and HA Gγ 1 C71S readily dimerizes with endogenous transducin Gβ 1 subunit in vivo, which further supports the idea that prenylation of a Gγ is not required for its dimerization with Gβ (Simonds et al., 1991;Pronin and Gautam, 1993;Higgins and Casey, 1994). Interestingly, HA Gγ 1 and HA Gγ 1 C71S were equally competent to rescue the expression of Gα t1 in rod photoreceptors of Gγ 1 knockout mice ( Figure 2B). Given that Gα t1 becomes destabilized in this mouse model due to a posttranslational mechanism (Lobanova et al., 2008;Kolesnikov et al., 2011), our result suggests that either Gβ 1 HA Gγ 1 or Gβ 1 HA Gγ 1 C71S could form a transducin heterotrimer. This result is consistent with the ability of the purified Gβ 1 γ 1 that was subjected to endo-Lys-C proteolysis, which removes the farnesyl modification, to form a heterotrimeric G protein in vitro (Lambright et al., 1996). However, only HA Gγ 1 restored rod visual responses of Gγ 1null mice, whereas HA Gγ 1 C71S did not have a physiological effect (Figure 2A). The functional deficiency of HA Gγ 1 C71S was explained by its exclusion from rod outer segments (Figures 5, 6), indicating that any G protein comprised of HA Gγ 1 C71S remained spatially separated from its coupled receptor, rhodopsin. The exclusion of HA Gγ 1 C71S was likely caused by the inability of HA Gγ 1 C71S to undergo farnesylation. This conclusion is supported by the observation that the N-acylation-deficient Gα t1 predominantly localizes in the inner compartments of rod instead of the outer segment (Kerov et al., 2007). Furthermore, eliminating endoproteolysis and c-terminal methylation by a knockout of RAS-converting enzyme 1 in rods has no effect on the accumulation of Gγ 1 in rod OSs, although farnesylation is a prerequisite for both steps (Christiansen et al., 2011). However, the last three amino acids (VIS) that are removed by endoproteolysis in HA Gγ 1 and endogenous Gγ 1 are present on HA Gγ 1 C71S. Nevertheless, we attribute the mislocalization of HA Gγ 1 C71S and the lack of rescue in Gγ 1null mice to the lack of farnesylation, since the farnesyl group contributes more to the hydrophobicity of Gγ 1 than the last three amino acids.

Rod Outer Segment Is Accessible for the Individual Gα and Gβγ Subunits
We have observed the reciprocal requirement of Gα t1 and Gβ 1 γ 1 subunits for their efficient ciliary sequestration in resting rods. Using two knockout mouse models in which Gα t1 or Gβ 1 γ 1 was ablated, we have determined subcellular localization of each individual subunit when its partner was absent. Epitopetagged Gβ 1 HA Gγ 1 , which was used as a surrogate for Gβ 1 γ 1 , was clearly visible in rod outer segments of Gα t1 knockout mice (Figure 7). Gα t1 was present in outer segments of Gγ 1 -null rods lacking Gβ 1 γ 1 dimer (Figure 8). Our results further confirm the conclusion of previous studies that the accumulation of transducin in outer segments of dark-adapted rods is based on G protein heterotrimer assembly (Zhang et al., 2003;Lobanova et al., 2008). Yet, it is important to note that together these results disagree with the notion that assembly of heterotrimeric G protein is a prerequisite for ciliary trafficking of transducin in rods (Zhang et al., 2003). Instead these results suggest that the Gα t1 subunit and Gβ 1 γ 1 dimer could each access the rod outer segment as an individual entity.

What Drives Ciliary Exclusion of Non-farnesylated Gβγ Dimer?
Our most surprising observation from these studies was the strong exclusion of non-farnesylated Gβ 1 HA Gγ 1 C71S from rod OSs, where this protein remained below the level of detectability for immunofluorescence confocal microscopy under all tested conditions (Figures 4, 5). This conclusion was further supported by non-farnesylated Gβ 1 HA Gγ 1 C71S's failure to rescue light responses elicited by the outer segments of Gγ 1 -null rods (Figure 2A). However, non-farnesylated Gβ 1 HA Gγ 1 C71S was capable of increasing the levels of Gα t1 likely by forming heterotrimeric transducin elsewhere in the cells (Figure 2B). Below, we describe two mechanisms that could explain why non-farnesylated Gβ 1 HA Gγ 1 C71S was excluded from rod OSs. Some evidence demonstrates that subcellular translocation of transducin is aided by two ''trafficking chaperones'' that bind transducin α and βγ subunits and conceal their lipid groups. Those include Unc119 that acts a sheath for the N-terminal myristoyl group of Gα t1 (Gopalakrishna et al., 2011), and phosducin that binds free Gβ 1 γ 1 . When phosducin forms a complex with Gβ 1 γ 1 , the farnesyl moiety of Gγ 1 becomes buried in between the β-propeller blades of Gβ 1 (Loew et al., 1998). Deletion of Unc119 or phosducin was shown to affect the accumulation of transducin in rod outer segments (Sokolov et al., 2004;Zhang et al., 2011), which suggests that transducin's subunits are likely to diffuse in this compartment as soluble Unc119/Gα t1 and Gβ 1 γ 1 /phosducin complexes. In light of these data, the inability of soluble non-farnesylated Gβ 1 HA Gγ 1 C71S to enter the outer segment seems rather counterintuitive. In fact, this result is more consistent with a different mechanism proposed by Baehr and colleagues (Karan et al., 2008) that is based on intraflagellar transport (IFT). According to this mechanism, transducin passes through the connecting cilium and enters the outer segment as a membrane cargo pulled along axonemal microtubules by motor proteins. The IFT mechanism would explain a prerequisite for the Gβγ lipid modification-without a farnesyl group Gβγ would not be able to anchor itself to the ciliary membrane. It is also consistent with the observation that swapping the farnesyl group of Gγ 1 to a more hydrophobic geranylgeranyl does not alter the targeting of Gβ 1 γ 1 to the rod outer segment (Kassai et al., 2005). Yet, it does not explain why non-farnesylated Gβγ remains excluded from the OS. Two mechanisms of exclusion are plausible.
The first mechanism is based on the premise that the connecting cilium comprises of a gate that hinders diffusion of soluble proteins. This is exemplified by the alleged gatekeeping action of centrins, which have been proposed to form dynamic oligomers with Gβ 1 γ 1 , regulated by Ca 2+ and phosphorylation by casein kinase 2 in the connecting cilium of rod photoreceptors (Giessl et al., 2006;Trojan et al., 2008). This mechanism predicts that soluble proteins cannot pass through the gate if closed, however visual arrestin, a soluble protein lacking any lipid modification, moves to rod OSs, while Gβ 1 HA Gγ 1 C71S remains blocked (Figure 6). Therefore, we favor an alternative mechanism, in which soluble Gβ 1 HA Gγ 1 C71S does access the OS, and then becomes expelled from this compartment due to the steric volume exclusion effect demonstrated by Calvert and colleagues .
According to it, the presence of a soluble protein in the rod outer segment inversely depends on its size due to highly constrained cytoplasmic space between outer segment disc membranes. Furthermore, the authors specifically proposed that transducin Gβ 1 γ 1 could be expelled from rod outer segments upon binding to phosducin, because the Gβ 1 γ 1 /phosducin complex is two times larger than Gβ 1 γ 1 alone . It is plausible that non-farnesylated Gβ 1 HA Gγ 1 C71S cannot efficiently dissociate from a complex with phosducin, and therefore remains excluded from the OS. Some evidence support this hypothesis. For example, it was demonstrated that eliminating S54 and S71 phosphorylation sites of phosducin that promote dissociation of the Gβ 1 γ 1 /phosducin complex significantly hindered trafficking of transducin to the rod outer segments (Belcastro et al., 2012). Empirical testing of this role of phosducin remains the goal of subsequent investigation.

AUTHOR CONTRIBUTIONS
CB and MS: designed the studies, generated mouse models, conducted experiments, analyzed data and wrote the article. MB: generated mouse models. JM, DH and SK: conducted experiments. OK: generated mouse model and wrote the article.

FUNDING
This work was supported by WVU Health Sciences Center Bridge Award. The Transgenic Animal Core Facility at WVU was supported by CoBRE grants, RR031155 and RR016440. The Microscope Imaging Facility at WVU was supported by the Mary Babb Randolph Cancer Center and NIH grants P20 RR016440, P30 RR032138/GM103488 and P20 RR016477.

ACKNOWLEDGMENTS
We would like to thank Drs. Kathleen Boesze-Battaglia and Wesley Clay Smith for generously providing antibodies against peripherin/rds and arrestin.