Syntaxin 3 and SNAP-25 pairing, regulated by omega-3 docosahexaenoic acid, controls the delivery of rhodopsin for the biogenesis of cilia-derived sensory organelles, the rod outer segments

Retinal rod photoreceptor cells are modified neurons with primary cilia that elaborate a specialized light-sensing organelle, the rod outer segment (ROS). The ROS is filled with membranous disks housing the phototransduction machinery that converts photon absorption by rhodopsin into changes in neurotransmitter release from specialized ribbon synapses, thus transmitting photosensory information to the visual cortex (Burns and Arshavsky, 2005; Ridge et al., 2003; tom Dieck and Brandstatter, 2006). The ROS disk membrane proteins are embedded in a fluid milieu comprised of polyunsaturated phospholipids that are highly enriched in omega-3 docosahexaenoic acid [DHA, 22:6(n-3)]. DHA plays an important function in human health, including brain and retina development, function of sensory membranes and cell survival (Bazan, 2006; Hoffman et al., 2001; Marszalek and Lodish, 2005). However, the exceptionally high content of DHA phospholipids in ROS membranes renders them highly susceptible to light and oxidative damage (Anderson and Penn, 2004). Thus, ROS membranes are continuously removed through daily shedding and phagocytosis by retinal pigment epithelial (RPE) cells (Besharse, 1986). The renewal of light-sensitive ROS membranes is mediated by rhodopsin transport carriers (RTCs), which travel vectorially through the cell body, the rod inner segment (RIS), to the base of the cilium and fuse with the specialized domain that separates the ciliary membrane from the surrounding RIS plasma membrane (PM), delivering rhodopsin to the ROS (Deretic and Papermaster, 1991; Papermaster et al., 1986). Along with rhodopsin, DHA phospholipids are also co-transported on RTCs to the ROS (Rodriguez de Turco et al., 1997).

The incorporation of rhodopsin into RTCs at the trans-Golgi network (TGN) is regulated by the small GTPase Arf4, which binds to the conserved rhodopsin C-terminal VxPx ciliary-targeting signal (Deretic, 2006; Deretic et al., 1998; Deretic et al., 2005) and mediates the assembly of the ciliary-targeting complex, which is comprised of two small GTPases, Arf4 and Rab11, the Rab11/Arf effector FIP3, and the Arf GTPase-activating protein ASAP1 (Mazelova et al., 2009). Tethering and fusion of RTCs with the RIS PM at the base of the cilium is in turn regulated by the small GTPase Rab8 in conjunction with phosphatidylinositol (4,5)-bisphosphate [PtdIns(4,5)P₂], actin and the actin-binding proteins moesin and Rac1 (Deretic et al., 1995; Deretic et al., 2004; Moritz et al., 2001). Although the subsequent membrane-fusion event is crucial to replenish the light-sensing machinery, it remains poorly understood, as the soluble N-ethylmaleimide-sensitive factor attachment protein receptor (SNARE) proteins that drive the RTC fusion have yet to be identified.

Rab GTPases, Rab effectors and SNAREs are major components of the intracellular machinery that is responsible for targeted membrane delivery (Cai et al., 2007; Sudhof, 2007). In particular, Rabs recruit the effectors that promote membrane-tethering interactions (Seabra and Wasmeier, 2004; Zerial and McBride, 2001). Rabs also function by concentrating and activating SNAREs, accessory proteins and lipids at the sites of membrane fusion (Starai et al., 2007). Rab8 regulates polarized trafficking through cytoskeleton remodeling, which is necessary for membrane outgrowth and the formation of cellular protrusions (Ang et al., 2003; Peranen et al., 1996; Sato et al., 2007; Wandinger-Ness and Deretic, 2008). Moreover, Rab8 is essential for the formation of primary cilia, the highly conserved organelles that project from the surfaces of many cells (Nachury et al., 2007; Omori et al., 2008; Yoshimura et al., 2007). In retinal photoreceptors, Rab8 regulates RTC fusion and biogenesis of the cilia-derived light-sensing organelles, the ROS; Rab8 mutants cause defects in membrane tethering and accumulation of RTCs below the cilium, leading to rod cell death and rapid retinal degeneration (Deretic et al., 1995; Moritz et al., 2001). This suggests that the regulation of rhodopsin ciliary targeting by Rab8 might be part of a broad and more general role in the regulation of ciliogenesis.

Fusion of Rab8-positive carriers with the basolateral PM of epithelial cells is driven by the SNARE syntaxin 4 and occurs at the sites adjacent to the tight junctions, which are marked by the octameric Sec6/8 complex (Grindstaff et al., 1998; Kreitzer et al., 2003). The highly conserved membrane-tethering Sec6/8 complex, known as exocyst in yeast, consists of subcomplexes that are continuously assembled and disassembled during trafficking (Hsu et al., 2004; Munson and Novick, 2006; Novick and Guo, 2002). The Sec6/8 complex is also present in nerve terminals (Hsu et al., 1996), where it is not required for regulated exocytosis and neurotransmitter release, but is thought to target fusion for neurite outgrowth and receptor trafficking to the synapse (Murthy et al., 2003; Sans et al., 2003; Vega and Hsu, 2001). The Sec6/8 complex is also localized to the primary cilia in polarized epithelial cells (Rogers et al., 2004), and is therefore a candidate for a Rab8 effector at the RTC docking site.

The membrane-fusion event through which RTCs deliver rhodopsin to the cilium is likely to be mediated by a SNARE complex (Li et al., 2007; Malsam et al., 2008; Paumet et al., 2004; Rothman, 2002). SNARE complexes are generally composed of a four-helical bundle that bridges opposing membranes and brings them into close proximity to initiate fusion (Jahn et al., 2003; Sollner, 2003; Ungar and Hughson, 2003). Fusion with the PM requires the formation of a complex between syntaxins (Qa-SNAREs) and VAMPs (R-SNAREs) (Fasshauer et al., 1998), each contributing one helix to the four-helix SNARE bundle, and Qbc-SNAREs, either neuronal SNAP-25 or non-neuronal SNAP-23, which provide two helices to the central layer of the core complex (Jahn and Scheller, 2006). Although SNARE pairing alone is not sufficient to determine the specificity of organelle fusion in reconstituted systems, cognate SNAREs are correctly paired in biological membranes (Bethani et al., 2007; Brandhorst et al., 2006). SNAREs are targeted to appropriate membrane domains on the basis of specific sequences (ter Beest et al., 2005). The polarized distribution of Qa-SNAREs, such as syntaxin 3 and syntaxin 4, which localize to the apical and basolateral PM of epithelial cells (Low et al., 1996), is likely to contribute additional specificity of membrane targeting by promoting fusion with only certain target membranes. Recent evidence suggests that the local lipid environment, particularly phospholipids enriched in omega-3 and omega-6 fatty acids, also contributes to regulate SNARE function (Davletov et al., 2007).

In this study we examined the expression and distribution of the candidate Q-SNAREs and membrane-tethering factors, and identified additional components that are responsible for polarized trafficking of rhodopsin in retinal photoreceptors. We report that photoreceptor cells display a unique distribution of the Sec6/8 tethering complex and the SNAREs syntaxin 3 and SNAP-25. We also reveal a previously unrecognized role for omega-3 DHA in modulating signal transduction in retinal photoreceptors by enhancing syntaxin-3 incorporation into SNARE complexes at RTC fusion sites to promote expansion of the ciliary membrane and ROS biogenesis.

To establish the candidate SNAREs for RTC fusion in retinal photoreceptors, we examined the distributions of syntaxin 3 and syntaxin 4, which segregate in PM domains of polarized epithelial cells. We performed these experiments using frog retinas, because the large size of photoreceptor cells in the frog retina and the extensive turnover of components of its light-sensing membrane (Besharse, 1986), especially when compared with mammals, provides an ideal system to define the roles of individual proteins in this process. We first determined that anti-syntaxin antibodies that were directed to mammalian proteins did recognize proteins of the appropriate molecular weights in frog retinal post-nuclear supernatant (PNS) that was enriched in photoreceptor membranes (Deretic and Papermaster, 1991). Retinal PNS immunoblots that were probed with antibodies to mammalian syntaxin 3 and syntaxin 4 (Fig. 1A) revealed proteins of ∼32 kDa, as expected given the evolutionary conservation of SNARE proteins. Because retinal membrane proteins were not boiled for SDS-polyacrylamide gel electrophoresis (SDS-PAGE) analysis to avoid rhodopsin aggregation (Deretic and Papermaster, 1991), we also observed syntaxin-3-containing oligomers of higher molecular weight, which probably represent SDS-resistant SNARE complexes (Fig. 1A, asterisks) (also discussed later), as previously reported for similar preparations of rat brain extracts (Pellegrini et al., 1995).

We next examined the retinal distribution of syntaxins 3 and 4 by confocal microscopy. Syntaxin 3 exhibited a polarized distribution and was highly abundant in the RIS PM domain tightly juxtaposed to the ROS (Fig. 1B, arrows). The RIS, schematically presented in Fig. 1D, terminates at the adherens junctions (AJ) that form the outer limiting membrane (OLM; Fig. 1B, dotted lines). Syntaxin 3 displayed a concentration gradient along the RIS PM before reaching negligible levels at the OLM (Fig. 1B). By contrast, syntaxin 4 was undetectable in the photoreceptor PM (Fig. 1C). Immunoreactivity for syntaxin 3 was absent from the RPE cells (Fig. 1B), whereas a robust signal for syntaxin 4 was detected in their basal PM (Fig. 1C, arrows), as previously reported (Low et al., 2002), further confirming the antibody specificity.

Both SNAREs were absent from the ROS, but were enriched in the retinal outer plexiform layer (OPL) (Fig. 1B,C), as previously reported (Hirano et al., 2007; Morgans et al., 1996). Syntaxin 3 was also detected in the calycal processes (CPs) (Fig. 1E, see also Fig. 1K), the villous structures enveloped by the RIS PM that evaginate from the RIS and surround the base of the ROS. Although syntaxin 3 was highly enriched in the OPL, where synapses of photoreceptor cells are located (Fig. 1E), photoreceptor synaptic terminals appeared devoid of this SNARE, in contrast to a previously published report (Morgans et al., 1996). This was particularly evident when syntaxin-3-labeled retinas were compared with those labeled with an antibody to a synaptic-vesicle membrane protein, synaptophysin, which delineates synaptic terminals of rod photoreceptors (Fig. 1F). As schematically depicted in Fig. 1G, the synaptic terminals of rod and cone photoreceptors are located in the outer sublamina (o) of the OPL, whereas the inner sublamina (i) consists of the horizontal and rod and cone bipolar cell processes. Syntaxin 3 was detected only in the inner half (i) of the OPL (below the dashed line in Fig. 1E), but was absent from rod photoreceptor synapses in the outer half (o) of the OPL (above the dashed line in Fig. 1E).

The photoreceptor PM was syntaxin-3-positive, but showed no immunoreactivity for syntaxin 4 with either a polyclonal (Fig. 1H) or a monoclonal (Fig. 1I) antibody, suggesting that this SNARE is unlikely to regulate RTC fusion. Syntaxin-4 immunoreactivity was observed only in the ellipsoid area of the RIS (Fig. 1H,I; `E', square bracket). The ellipsoid, which is schematically presented in Fig. 1D, is densely packed with mitochondria, as shown in the electron microscopy (EM) image (Fig. 1K). Interestingly, mitochondrial localization of the R-SNARE VAMP-1B has been reported (Isenmann et al., 1998); our data suggest that, in photoreceptors, syntaxin 4, or a syntaxin-4-like protein, also displays this unusual mitochondrial localization. In double-labeled confocal sections, syntaxin 3 was also found to be abundant in the periciliary RIS PM (Fig. 1I, arrows). In amphibian photoreceptors, this area includes the periciliary ridge complex (PRC) (Fig. 1K), a highly specialized membrane domain consisting of symmetrical ridges and grooves that comprise the RTC fusion sites (Peters et al., 1983). Accordingly, we identified the PRC using an antibody to a scaffold protein, whirlin (USH2D) (Fig. 1J,L, arrows), a component of the protein network that is disrupted in human Usher syndrome (Liu et al., 2007b; Maerker et al., 2008). Whirlin immunoreactivity was localized within the syntaxin-3-enriched domain (compare Fig. 1J,L with Fig. 1I), demonstrating the predominant association of this SNARE with RTC fusion sites.

We next examined the distribution of syntaxins 3 and 4 among retinal subcellular fractions separated on sucrose density gradients. The distribution of syntaxin 3 closely paralleled that of the β2 subunit of Na,K-ATPase, a marker for the RIS PM (Schneider et al., 1991), suggesting that they were present in the same membrane (Fig. 1M). By contrast, ∼ 40% of syntaxin 4 was found in a gradient fraction with a higher buoyant density, consistent with its association with other membrane organelles. Because the fraction containing syntaxin 4 was previously shown to be enriched in mitochondria (Deretic and Papermaster, 1991), this further supports the mitochondrial association of syntaxin 4. To confirm the colocalization of syntaxin 3 and Na,K-ATPase in the RIS PM, we employed an antibody specific for the α-Na,K-ATPase subunit, which recognized a protein of ∼110 kDa in retinal PNS (Fig. 1N), consistent with the α3 isoform expressed in photoreceptors. By confocal microscopy, α-Na,K-ATPase was detected in the RIS PM, including in CPs (Fig. 1O), as well as in the inner retinal layers. Na,K-ATPase extensively colocalized with syntaxin 3 in the RIS PM in the proximity of the cilium (Fig. 1P, arrow). However, Na,K-ATPase was uniformly distributed between the base of the cilium and the OLM, whereas syntaxin-3 staining gradually diminished and reached negligible levels well above the OLM (Fig. 1P, arrowhead). Overall, the distribution of syntaxin 3 suggests that it might be selectively partitioned to serve as a SNARE for RTC fusion with the RIS PM.

We next investigated whether the observed concentration gradient of syntaxin 3 depends on the integrity of microtubules, because the disruption of microtubules in polarized epithelial cells causes depolarization of the apical SNARE syntaxin 3, but not of the basolateral syntaxin 4 (Kreitzer et al., 2003). In photoreceptors, an array of microtubules that radiates from the basal body below the cilium is tethered with their plus-end at the RIS PM (see Fig. 1D). These microtubules, which might be responsible for the restricted distribution of syntaxin 3, are known to be sensitive to anti-microtubule drugs (Vaughan et al., 1989). Upon nocodazole treatment to promote microtubule depolymerization, the PM in the proximity of the RTC fusion sites was depleted of syntaxin 3 (Fig. 2A-F). In nocodazole-treated retinas, syntaxin 3 delocalized and was detected even below the OLM (Fig. 2B, arrow), rather than in the periciliary area (Fig. 2A,C,E, arrows). Syntaxin 3 was also seen in intracellular vesicles (Fig. 2D,F, asterisk). The highest density of staining was detected in the lateral portion of the RIS PM (Fig. 2D,F, arrows), suggesting the microtubule-dependent redistribution of syntaxin 3. EM images of the cross-sections of nocodazole-treated retinas at the level of the Golgi confirmed that the treatment was effective in depolymerizing RIS microtubules (Fig. 2G,H). Golgi cisternae, which are probably supported by microtubules, extended away from the OLM in control retinas labeled with Rab6 (Fig. 2I). In the absence of microtubules, the photoreceptor Golgi, which was distinguished either by Rab6 staining (Fig. 2J) or GM130 staining (not shown), collapsed around the nucleus towards the OLM, nearly reaching the lateral RIS PM. Thus, microtubule-dependent change in Golgi organization accompanied the shift in syntaxin-3 polarity. By contrast, microtubule disruption had no effect on syntaxin-4 immunoreactivity in photoreceptors, although this SNARE did lose its polarized distribution in the RPE cells (not shown).

Consistent with previous reports in mammalian retinas (Greenlee et al., 2001; Greenlee et al., 2002), confocal microscopy with a monoclonal anti-N-terminal antibody showed abundant SNAP-25 in the RIS PM and the synapse, CPs, and around the cilium at the base of the ROS, but not in the ROS PM of the frog (Fig. 3A). Similarly, a polyclonal anti-C-terminal-SNAP-25 antibody (Oyler et al., 1989) displayed an identical distribution, albeit a more punctate pattern (not shown). These differences are probably due to the conformation-dependent differential accessibility of the two antigenic sites. SNAP-25 colocalized with whirlin at the base of the cilium (Fig. 3B, arrows), with synaptophysin in photoreceptor synapses (Fig. 3A,D-F) and with syntaxin 3 in the periciliary RIS PM (Fig. 3C, arrows). Subcellular fractionation confirmed the presence of SNAP-25 and syntaxin 3 in the RIS, and their absence from the isolated purified ROS (Fig. 3G). Notably, they displayed a nearly identical ladder of immunoreactivity in unboiled samples (Fig. 3G), suggesting that SNAP-25 is associated with syntaxin 3 in these SDS-resistant complexes in the RIS. Detailed analysis revealed that SNAP-25–syntaxin-3-positive domains encircled the base of the cilium (Fig. 3H-J, arrows), consistent with the formation of SNARE complexes specifically at the sites of RTC fusion. Although both SNAREs were highly concentrated in the periciliary area, SNAP-25 displayed uniform RIS PM localization that did not parallel the concentration gradient of syntaxin 3 (Fig. 3K). Nocodazole treatment resulted in extensive colocalization of SNAP-25 with syntaxin 3 in the lateral RIS PM (Fig. 3L, arrows). In control retinas, the pixel intensity plot (Fig. 3K, lower panel) along a line (yellow line in the upper panel) drawn approximately halfway between the ROS and the OLM (dotted line) showed that the lateral RIS PM is largely devoid of syntaxin 3. By contrast, upon nocodazole treatment, comparable pixel intensity for SNAP-25 and syntaxin 3 was observed in the lateral RIS PM (Fig. 3L, arrows, lower panel), consistent with ectopic SNARE pairing when microtubule-directed organization was disrupted.

To examine whether the Sec6/8 complex serves to tether RTCs at fusion sites, we performed confocal microscopy using an antibody to Sec8, as Sec8 is a representative component of this multimeric complex. Anti-Sec8 showed immunoreactivity along the RIS PM (Fig. 4A). Foci of high Sec8 immunoreactivity were also observed below the OLM (Fig. 4A). Importantly, Sec8 localized to the structures in the proximity of RTC fusion sites, tightly juxtaposed to the RIS PM labeled by syntaxin 3 (Fig. 4A, arrow in inset). By contrast, in nocodazole-treated retinas (Fig. 4B), extensive colocalization of Sec8 and syntaxin 3 was evident along the lateral RIS PM (Fig. 4B, arrowheads in inset). Immunolocalization of Sec6 was identical to that of Sec8 (not shown), in agreement with the conservation of the Sec6/8 tethering complex in the frog retina.

Sec8 localization was consistent with potential microfilament association. Actin filaments encircle photoreceptor inner segments, extending from the AJs outward to the CPs (Fig. 4C). In confocal sections labeled with phalloidin to reveal microfilaments (Fig. 4D), Sec8 did not colocalize with the filaments, but was concentrated at the sites of actin-cytoskeleton attachment to the RIS PM, including the AJs (Fig. 4D, asterisk). In the same section, Rab8, which is known to associate with photoreceptor actin cytoskeleton (Deretic et al., 1995; Moritz et al., 2001), colocalized with Sec8 at the sites of RTC fusion (Fig. 4D, arrow), as well as in the CPs (Fig. 4D, arrowhead). Colocalization of Rab8 and Sec8 was particularly evident at higher magnification, as Rab8-labeled RTCs were in the close proximity of, and appeared to be tethered to, the Sec8-labeled structures (Fig. 4E, arrows). Thus, the Sec6/8 complex might cooperate with syntaxin 3 and SNAP-25 to regulate the fusion of Rab8-positive RTCs to the membrane.

Syntaxin 3 was recently identified as the molecular target for omega-3 and omega-6 polyunsaturated fatty acids, which alter its conformation thereby promoting formation of the SNARE complex and membrane expansion at the growth cones of PC12 cells (Darios and Davletov, 2006). Given the similarity of the PM addition to ROS-membrane renewal and the unique importance of omega-3 DHA to photoreceptor physiology, we sought to define the role of DHA in the context of membrane trafficking required for effective photoreceptor function. Treatment with 100 μM DHA markedly increased the colocalization of syntaxin 3 and SNAP-25 in the periciliary domain (Fig. 5A). This appeared to be primarily due to the redistribution of syntaxin 3. Quantitative analysis of syntaxin 3 and SNAP-25 distribution performed in three separate experiments showed an additional ∼20% pixel colocalization in DHA-treated retinas (Fig. 5B). Moreover, co-immunoprecipitation of syntaxin 3 by anti-SNAP-25 antibody was also increased by ∼20% after DHA treatment (Fig. 5C), indicating a stimulatory effect of DHA on SNARE pairing in these preparations. Importantly, this increase in SNARE pairing was accompanied by a ∼15% increase in rhodopsin delivery to ROS, measured by the incorporation of newly synthesized rhodopsin into isolated purified ROS disk membranes (Fig. 5D). This significant effect was specific for DHA (P=0.01), as saturated palmitic acid [PA, 16:0] did not significantly impact rhodopsin delivery (P=0.17).

We next tested the ability of DHA to overcome the inhibition of fusion by propranolol, which causes membrane-tethering defects and inhibits RTC fusion (Deretic et al., 2004). As shown in Fig. 5E, consistent with our published data, propranolol nearly completely inhibited rhodopsin delivery to the ROS. DHA treatment specifically restored it to more than 20% of the control, whereas palmitic acid had no effect (Fig. 5E). Quantification of overexposed autoradiograms (to detect rhodopsin in the ROS of propranolol-treated retinas) showed that DHA significantly increased the amount of radiolabeled rhodopsin by restoring delivery to the ROS membrane in propranolol-treated retinas (15.6±0.8% vs 31.0±5.3% of control in DHA-treated, n=3, P=0.04). Propranolol also led to the redistribution of Sec8 and, by extension, of the tethering complex into the cytosol (Fig. 5F), most probably by causing the loss of PtdIns(4,5)P₂ from the PM fusion sites (Deretic et al., 2004). Similarly, syntaxin 3 and SNAP-25 also appeared greatly diminished at the base of the cilium (Fig. 5G, arrows) and colocalized only in the lateral RIS PM of propranolol-treated retinas (Fig. 5G, arrowheads). Treatment with DHA did, however, appear to prevent the loss of these SNAREs from the fusion sites (Fig. 5H, arrow). Accordingly, DHA, but not palmitic acid, increased co-immunoprecipitation of SNAP-25 by anti-syntaxin-3 antibody even in the presence of propranolol (Fig. 5I). These data suggest that DHA exerts its physiological role in membrane trafficking to ROS not only through lipid modulation, but also through activation of syntaxin 3, facilitation of SNARE pairing and increased availability of the fusion machinery at the base of the cilium, which together promote RTC fusion, ciliary membrane addition and delivery of rhodopsin to the sites of light capture.

In this study we provide evidence for the role of syntaxin 3 and SNAP-25 as the Q-SNAREs for the fusion of incoming carriers, RTCs, with the RIS PM, and therefore as regulators of ciliogenesis and ROS biogenesis in photoreceptors. Syntaxin 3 and SNAP-25 are highly concentrated in the proximity of the RTC fusion sites. The uniform distribution of SNAP-25 throughout the RIS PM and photoreceptor synapse indicates that, in addition to syntaxin 3, this SNARE is likely to also pair with other Qa-SNAREs. Interaction of SNAP-25 with all PM syntaxins has been demonstrated in PC12 cells (Bajohrs et al., 2005). One candidate for association in the RIS is the neuronal-specific syntaxin 1A, a prominent partner of SNAP-25 and a known regulator of synaptic vesicle fusion (Sudhof, 2004); syntaxin 1A is absent from ribbon synapses of photoreceptor cells (Brandstatter et al., 1996; Ullrich and Sudhof, 1994) but is present in the RIS of rat photoreceptors (Low et al., 2002). Quantitative proteome analysis of the mouse photoreceptor sensory cilium complex estimated 4×10⁵ molecules of SNAP-25, 1.9×10⁵ of syntaxin 3 and 1.2×10⁵ of syntaxin 1B per isolated photoreceptor RIS/ROS preparation (Liu et al., 2007a). This supports the notion that syntaxin 3 is the major partner for SNAP-25 in photoreceptor cells, and that SNAP-25 pairing with syntaxin 1A, or syntaxin 1B, might regulate a distinct trafficking pathway in the RIS.

Our findings suggest that microtubules promote the restricted distribution of syntaxin 3, which is consistent with the membrane cytoskeleton playing an essential role in concentrating regulators of RTC fusion around the cilium. In epithelial cells, syntaxins 3 and 4 are present in separate small clusters even before the establishment of cell polarity, and the integrity of syntaxin-3 clusters depends on intact microtubules and on the presence of cholesterol (Low et al., 2006). The enrichment of cholesterol in the periciliary region (Andrews and Cohen, 1983) and the involvement of microtubules in clustering syntaxin 3 in photoreceptors suggest that similar cellular mechanisms are likely to be crucial in developing and maintaining syntaxin-3 polarity. Re-distribution of syntaxin 3 to the lateral RIS PM of nocodazole-treated retinas, where SNAP-25 and the Sec6/8 tethering complex are already present in the proximity of the delocalized Golgi, suggests that this more peripheral membrane domain could become competent for fusion with RTCs under such circumstances. Involvement of microtubules in rhodopsin trafficking through the RIS has been unclear, because they are not absolutely required for delivery of rhodopsin to the ROS (Vaughan et al., 1989). The absence of microtubules does not prevent the delivery of post-Golgi carriers to the PM, indicating that they only provide fast and directional movement (Hirschberg et al., 1998). Thus, upon disruption of microtubule machinery in photoreceptor cells, if the Golgi and the cognate SNAREs responsible for RTC fusion are mislocalized, RTCs could be mistargeted and could fuse with inappropriate photoreceptor membranes such as the lateral RIS PM, or even the synapse. Although this would not abrogate delivery of rhodopsin to the ROS entirely (Vaughan et al., 1989), it could result in delivery of rhodopsin to RIS PM domains that are normally devoid of rhodopsin, which would be detrimental to cell polarity and retinal function. Consistent with this idea, the loss of photoreceptor polarity is highly correlated with the disease progression and blindness in patients with retinitis pigmentosa (Berson et al., 2002; Li et al., 1995).

The peripheral actin network at the base of the cilium plays a crucial role in the targeted delivery of ROS membranes and in photoreceptor morphogenesis (Chaitin, 1992; Williams et al., 1988). The capture of incoming Rab8-positive RTCs and their tethering at the fusion sites organized by PtdIns(4,5)P₂ is regulated by moesin and Rac1, through their cooperative regulation of the actin cytoskeleton (Deretic et al., 2004). The actin network and the actin-based motors are essential for Sec6/8 and the exocyst localization, and for polarized exocytosis (Hsu et al., 1999). Our finding that Rab8 and Sec8 are associated with microfilaments at RTC fusion sites at the base of photoreceptor cilium is consistent with emerging evidence that Rab8 is a master regulator of ciliogenesis (Nachury et al., 2007; Yoshimura et al., 2007). Because the Sec6/8 complex also localizes to the primary cilia in epithelial cells (Rogers et al., 2004), it is likely that it functions as one of the Rab8 effectors in ciliogenesis. In this process, Rab8 might cooperate with Rab11, which regulates budding of RTCs (Mazelova et al., 2009). In Drosophila, the Sec6/8 complex interacts with Rab11 and regulates rhodopsin transport to the rhabdomeres (Beronja et al., 2005). Our data suggest that this interaction might be conserved in vertebrates and that the formation of this tightly regulated signaling network, together with specific membrane-fusion proteins at the base of the cilium, ensures the rapid and directional membrane expansion for the sustained biogenesis of light-sensing organelles.

It is interesting that, when transfected into polarized epithelial cells (which express SNAP-23 in a non-polar fashion), SNAP-25 localizes both to the PM and, unlike SNAP-23, also to primary cilia (Low et al., 1998), similar to its localization in photoreceptor cells. Our data indicate that, together with SNAP-25, syntaxin 3 regulates ciliogenesis and ROS biogenesis, a conclusion that is in keeping with studies suggesting a role for syntaxin 3 in membrane expansion (Darios and Davletov, 2006). Photoreceptor connecting cilium corresponds to the transition zone of primary cilia, which is considered a gateway for the admission of specific proteins to this privileged intracellular compartment (Rosenbaum and Witman, 2002). Because SNARE proteins act as gatekeepers for the addition of the ciliary membrane material, it is remarkable that treatment with DHA is able to increase membrane delivery to the cilium and the ROS in a system that already operates at an extremely high level, synthesizing and delivering an equivalent of 3 μm² of membrane a minute (Besharse, 1986). Although the effect of DHA on the coupling of syntaxin 3 and SNAP-25 was relatively small, it was measurable and reproducible, which is consistent with physiological relevance in photoreceptor function. Unlike syntaxin 1, which readily pairs with SNAP-25, syntaxin 3 is unable to form a complex with SNAP-25, except in the presence of polyunsaturated fatty acids (Connell et al., 2007). Strikingly, even when bound to the SNARE regulator Munc18, syntaxin 3 is activated by DHA (Connell et al., 2007). Thus, the reported stoichiometry of Munc18-1 and syntaxin 3 (1.93×10⁵ vs 1.94×10⁵ per RIS/ROS) (Liu et al., 2007a) suggests that they might form a `dormant' syntaxin-3–Munc18 complex, the activation of which depends crucially on the lipid environment. In this context, the physiological role of DHA might be to selectively pair syntaxin 3 with SNAP-25, which has an access to primary cilia on its own, to open the gate for rhodopsin delivery and the addition of the ciliary membrane material.

Similar to SNARE overexpression (Starai et al., 2007), the increase in availability of activated SNAREs in retinas treated with propranolol and DHA appears to reduce the need for tethering factors because Sec6/8 complex dissociates from the membranes and yet fusion is enhanced. Free omega-3 and omega-6 fatty acids have been demonstrated to directly activate syntaxin 3 by increasing its α-helical content (Darios and Davletov, 2006). However, in photoreceptors, the total amount of unesterified DHA is small because of its rapid acylation into phospholipids. Only when retinas are incubated with high concentrations of DHA, similar to the concentrations used in this study, does the majority of DHA remain unesterified (Rodriguez de Turco et al., 1991). Even so, the physiological uptake of free DHA from the RPE by the photoreceptors (Rodriguez de Turco et al., 1994) leads to preferential incorporation into the RIS PM domains proximal to ROS that we now show are enriched in syntaxin 3. In addition, regulated localized release of DHA is likely to be required for SNARE pairing and ultimately for membrane delivery to the ROS. The mechanism that targets the localized production of free DHA in the RIS by phospholipase A2 thus becomes an important issue (Darios et al., 2007), especially because it also produces another fusion stimulator, lysophospholipid, that remains in the membrane. This mechanism also has to prevent rapid reacylation and excessive incorporation of DHA phospholipids into membranes that could destabilize cholesterol-rich domains, which provide diffusion barriers separating the ciliary membrane from the surrounding PM (Reiter and Mostov, 2006; Vieira et al., 2006). DHA metabolism and homeostasis could therefore profoundly affect ROS biogenesis and photoreceptor polarity, which are frequently severely compromised in retinal degenerative diseases (Bazan, 2006; Hoffman et al., 2001). The crucial role that DHA plays in retinal health is further underscored by the Age-Related Eye Disease Study 2 (AREDS2), which examines the effects of high supplemental doses of DHA on the development of Age-Related Macular Degeneration (http://www.nei.nih.gov/neitrials/index.aspx).

Although the precise molecular interactions remain to be delineated, identification of the candidate Sec6/8 tethering complex and the candidate PM SNAREs syntaxin 3 and SNAP-25, together with the ability to modulate their interaction with DHA, has provided a framework for future investigation to establish the function of these proteins in rhodopsin trafficking, ciliogenesis, photoreceptor polarity and health.

Confocal microscopy was performed on dark-adapted Rana berlandieri frog retinas as described (Deretic et al., 2004). Isolated eyecups were either fixed immediately with 4% paraformaldehyde or incubated for 5 hours at 22°C in oxygenated medium with, or without, 20 μM nocodazole (Vaughan et al., 1989), 0.5 mM propranolol (Deretic et al., 2004) or 100 μM DHA (Darios and Davletov, 2006). Following 4% paraformaldehyde fixation overnight, 100-μm sections were cut and labeled with rabbit polyclonal antibodies to: syntaxin 3 (1:100; Synaptic Systems), syntaxin 4 (1:100; Synaptic Systems), C-terminal SNAP-25 [371, 1:1000 (Oyler et al., 1989)], synaptophysin [1:100, a gift of F. Valtorta (Deretic and Papermaster, 1991; Valtorta et al., 1988)] and whirlin [1:100, antibody to frog whirlin, a kind gift of Tiansen Li (Tiansen Li and Jun Yang, Mass Eye and Ear Infirmary, Boston, MA, personal communication)]; and/or with mouse monoclonal antibodies to: syntaxin 4 (1:100; BD Biosciences), N-terminal SNAP-25 [SMI 81, 1:1000 (Philip Washbourne and M.C.W., unpublished)], Rab8 (1:100; BD Biosciences), Sec8 [8S2E12, 1:100, a gift of S. C. Hsu (Wang et al., 2004)] and α-Na,K-ATPase (M7-PN-E9, 1:100, GeneTex), and with Alexa-Fluor-633-conjugated phalloidin (1:30; Molecular Probes). Primary antibodies were detected with Cy2 and Cy3 goat anti-rabbit IgG or goat anti-mouse IgG (1:200; Jackson ImmunoResearch). All sections were counterstained with the nuclear stain TO-PRO-3 (1:1000; Molecular Probes). Confocal optical sections (0.7 μm) were obtained on a Zeiss 510 laser scanning confocal microscope (Carl Zeiss) using a 488 nm argon ion laser for Cy2, 543 nm HeNe laser for Cy3 and 633 nm HeNe laser for Alexa-Fluor-633 or TO-PRO-3 excitation. Digital images were prepared using Adobe Photoshop CS (Adobe Systems). Colocalization analysis (Pearson's coefficient) was calculated using SlideBook Image Analysis software (Intelligent Imaging Innovations).

Following 5 hours of incubation with 20 μM nocodazole or in the control media, retinas were fixed in 4% formaldehyde and 1% glutaraldehyde in 0.12 M cacodylate buffer pH 7.5 for 1 hour at 22°C, post-fixed in OsO₄ and embedded in Epon. Thin sections were examined in a Philips CM-100 transmission electron microscope.

Frogs were dark-adapted for 2 hours before the experiment. Isolated retinas were incubated for 1 hour at 22°C in oxygenated medium with [³⁵S]-Express protein labeling mixture (25 μCi/retina), followed by a 2-hour chase. In some experiments, incubation medium contained 0.5 mM propranolol (Deretic et al., 2004), 100 μM DHA or 100 μM palmitic acid (PA).

Retinal fractionation and preparation of PNS enriched in photoreceptor biosynthetic membranes were performed as described (Deretic and Papermaster, 1991). ROS were sheared using a 14-gauge needle, separated from the remainder of the retina by flotation on high density (34%) sucrose and purified on a step sucrose gradient (Papermaster and Dreyer, 1974). After gradient centrifugation at 100,000 g_av in a SW40 rotor (Beckman-Coulter) for 30 minutes, purified ROS membranes were collected from the 1.11-1.13 g/ml sucrose interface, diluted with 10 mM Tris acetate, pH 7.4, and sedimented for 20 minutes at 20,000 rpm in a JA25.5 rotor (Beckman-Coulter). Following ROS removal, retinal pellets were homogenized in 0.25 M sucrose, and retinal fragments and nuclei were sedimented at 4,000 rpm in a JA25.5 rotor for 4 minutes, generating PNS that was enriched in photoreceptor biosynthetic membranes. In some experiments, PNS was centrifuged at 80,000 rpm for 1 hour in a TLA 100.3 rotor (Beckman-Coulter) to sediment RIS membrane proteins, or at 17,500 g for 10 minutes at 4°C in a JA25.5 rotor to pellet large organelles, including 90% of photoreceptor RIS PM (Deretic et al., 2004). Pellets were resuspended in 0.25 M sucrose and, following fractionation on linear 20-39% (w/w) sucrose gradients, subcellular fraction pools were created as described (Deretic et al., 2004).

Membrane proteins were analyzed by SDS-PAGE as described (Deretic and Papermaster, 1991). Gels were stained by Phast Gel Blue R (Amersham Pharmacia Biotech) and imaged with a GS-700 Imaging Densitometer (BioRad). The images of the radiolabeled proteins were generated by autoradiography at –85°C using Kodak BioMax MR film. Gels were blotted onto Immobilon-P membranes and blots were probed with anti-Na,K-ATPase-β2 (1:100; BD Biosciences), or the antibodies listed above, followed by the secondary antibodies conjugated to HRP. Bound antibodies were detected using a chemiluminescent Western Lightning immunodetection system (Perkin Elmer Life Sciences). The distribution of detected antigens was quantified using Multianalyst software (BioRad).