Abstract
It is unclear how unconventional secretion interplays with conventional secretion for the normal maintenance and renewal of membrane structures. The photoreceptor sensory cilium is recognized for fast membrane renewal, for which rhodopsin and peripherin/rds (P/rds) play critical roles. Here, we provide evidence that P/rds is targeted to the cilia by an unconventional secretion pathway. When expressed in ciliated hTERT-RPE1 human cell line, P/rd is localized to cilia. Cilium trafficking of P/rds was sustained even when the Golgi functions, including trans-Golgi-mediated conventional secretion, were inhibited by the small molecules brefeldin A, 30N12, and monensin. The unconventional cilia targeting of P/rds is dependent on COPII-mediated exit from the ER, but appears to be independent of GRASP55-mediated secretion. The regions in the C-terminal tail of P/rds are essential for this unconventional trafficking. In the absence of the region required for cilia targeting, P/rds was prohibited from entering the secretory pathways and was retained in the Golgi apparatus. A region essential for this Golgi retention was also found in the C-terminal tail of P/rds and supported the cilia targeting of P/rds mediated by unconventional secretion. In ciliated cells, including bovine and Xenopus laevis rod photoreceptors, P/rds was robustly sensitive to endoglycosidase H, which is consistent with its bypassing the medial Golgi and traversing the unconventional secretory pathway. Because rhodopsin is known to traffic through conventional secretion, this study of P/rds suggests that both conventional secretion and unconventional secretion need to cooperate for the renewal of the photoreceptor sensory cilium.
Introduction
The sensory primary cilium of vertebrate photoreceptor cells houses thousands of photosensitive disk membranes that are renewed continuously throughout our lifespan. Cilia targeting of Peripherin/rds (P/rds; Arikawa et al., 1992) is assumed to be essential for disk membrane morphogenesis and renewal (Farjo et al., 2006). Rds/rds mice lacking functional P/rds are incapable of forming disks (van Nie et al., 1978; Travis et al., 1989). This disk membrane deficiency is one of the most severe among the retinal degeneration models and is comparable to rhodopsin knock-out mice that are also incapable of forming disks. One piece of evidence suggests that P/rds and rhodopsin are trafficked through different routes. In detached cat retinas undergoing degeneration, P/rds mislocalized to cytoplasmic vesicles, whereas rhodopsin mislocalized to the plasma membrane of photoreceptor cells (Fariss et al., 1997). Even though it was obtained under a pathological condition, this observation led to the hypothesis that P/rds is transported by a mechanism distinct from that of rhodopsin.
Most plasma membrane proteins are sorted by the trans-Golgi network (TGN) in the conventional secretory pathway. As glycoproteins pass through the Golgi apparatus, most receive a mature, complex oligosaccharide. P/rds, an N-linked glycoprotein, does not acquire a complex oligosaccharide and, accordingly, the oligosaccharide group of P/rds is endoglycosidase H (Endo H)-sensitive (Connell and Molday, 1990). This Endo H sensitivity indicates that P/rds was not processed by Golgi mannosidase II (GMII), which is localized to the medial- to trans-Golgi apparatus (Velasco et al., 1993; Rabouille et al., 1995) and is thus possibly a sign of P/rds trafficking through an unconventional secretory pathway. Unconventional secretory pathways are poorly defined secretory routes that do not require the TGN for sorting (Yoo et al., 2002; Schotman et al., 2008; Gee et al., 2011; Giuliani et al., 2011; Grieve and Rabouille, 2011; Hoffmeister et al., 2011). If unconventional secretion were to mediate the cilia trafficking of P/rds, then cooperation with the conventional secretory pathway that mediates the trafficking of rhodopsin would be required to ensure proper morphogenesis of the photoreceptive disk membranes.
Here, we tested the hypotheses that P/rds is trafficked through an unconventional secretory pathway bypassing the Golgi apparatus either partly or entirely and that such an unconventional pathway also plays an important role in the targeting of P/rds to the cilia and subsequently to the disk membrane. With the goal of elucidating the regions required for the specific localization of P/rds, we dissected the C-terminal (CT) tail, where the cilia-targeting signal is considered to reside (Tam et al., 2004; Salinas et al., 2013). These studies were conducted using mammalian cells that form primary cilia in culture. The roles of the CT tail region were further investigated using Xenopus laevis photoreceptor cells in vivo. These studies unveil a previously unknown role of an unconventional secretory pathway in the process of photoreceptor membrane morphogenesis and maintenance and also reveal the CT regions that act in a coordinated fashion to accomplish the unconventional cilia targeting of P/rds.
Materials and Methods
Constructs.
Full-length bovine P/rds (bP/rds) cDNA was a generous gift from Dr. Robert S. Molday (Department of Biochemistry and Molecular Biology, University of British Columbia, Vancouver, Canada). The coding region of bP/rds was subcloned into pMSCV vector with HA and FLAG tag following the CT tail for expression in cell culture. bP/rds-Dend2 was generated by fusing Dendra2 (Dend2; Clontech Laboratories) to the CT tail of bP/rds. pcDNA3.3-bP/rds was generated by inserting bP/rds-coding region into pcDNA3.3 using the XhoI site. X. laevis P/rds (xP/rds) cDNA (GeneBank ID#: AY062004, Rds38) was obtained from X. laevis messenger RNA by reverse transcription and further cloned into TOPO vector with X. laevis rhodopsin promoter and fused with Dend2 following the CT tail. Different deletions of bP/rds and xP/rds were generated by the QuikChange site-directed in vitro mutagenesis kit (Stratagene) and subcloning. They were either fused to HA and FLAG epitopes or Dend2 fluorescent protein. pEGFPn-SSTR3 was a generous gift from Dr. Kirk Mykytyn (Department of Pharmacology, Division of Human Genetics, and College of Medicine, The Ohio State University, Columbus, OH). pClneo-Sar1-Myc and pClneo-Sar1(H79G)-Myc vectors were generous gifts from Dr. Min Goo Lee (Yonsei University College of Medicine, Seoul, Korea). pmTurquoise2-Golgi was obtained from Addgene (plasmid 36205; Addgene), which includes the human trans-Golgi β-1, 4-galactosyltransferase 1(1–61) gene and the mTurquoise2 fluorescent protein (Goedhart et al., 2012). For labeling Golgi apparatus of X. laevis rods, glycosyltransferase and mTurquoise2 were subcloned into TOPO vector with the X. laevis arrestin promoter and the SV40 polyadenylation signal.
Cell culture.
hTERT-RPE1cells (ATCC) were grown in the medium recommended by ATCC including DMEM/F12 with 10% fetal bovine serum, 0.01 mg/ml hygromycin B, and 1% penicillin/streptomycin at 37°C with 5% CO2. Inner medullary collecting duct 3 (IMCD3) cells were cultured in DMEM/F12 with 10% FBS and 1% penicillin/streptomycin at 37°C with 5% CO2. To induce cilia, cells were incubated in DMEM/F12 without FBS along with 0.3 μm Cytochalasin D (Cyto D; Sigma-Aldrich) for 16 h to maximize the probability of ciliogenesis by a combination of serum starvation and mild perturbation of the actin cytoskeleton (Kim et al., 2010). For some of the experiments, cilia were induced by culturing hTERT-RPE1 cells in DMEM/F12 with 0.5% FBS for 48 h but without Cyto D to observe the pure effects of the following reagents without interference from Cyto D: brefeldin A (BFA; Sigma-Aldrich), 30N12 (ChemBridge), monensin (Sigma-Aldrich), cycloheximide (Sigma-Aldrich), GRASP siRNA, and Sar1(H79G). To suppress the expression of GRASP55, GRASP (GORASP2) siRNA (SMARTpool M-019045-01-0005; Thermo Scientific) was transfected into hTERT-RPE1cells stably expressing bP/rds-HA-FLAG using Lipofectamine RNAiMAX Transfection Reagent (Invitrogen) following the manufacturer's instructions. siGENOME-nontargeting siRNA pool #1 (D-001206-13-05; Thermo Scientific) was used as a control siRNA. pClneo-Sar-Myc and pClneo-Sar1(H79G)-Myc were transfected into hTERT-RPE1cells stably expressing bP/rds-HA-FLAG using Fugene 6 (Promega) following the manufacturer's instructions. pClneo-Sar1(H79G)-Myc carries one tyrosine to serine mutation at amino acid position 9, which did not affect the dominant-negative property. To generate cell lines stably expressing bP/rds or other genes, hTERT-RPE1or IMCD3 cells were transfected with cDNAs using Fugene 6 following the manufacturer's instructions. Twenty-four hours later, medium was replaced with fresh growth medium containing selection reagent (puromycin for pMSCV vector). Selection medium was replaced every 3 d until colonies formed 18–21 d later.
Inhibition of conventional secretion and protein synthesis.
After cilia induction by culturing in DMEM/F12 with 0.5% FBS for at least 48 h, hTERT-RPE1 cells stably expressing bP/rds-HA-FLAG were treated with 0.5 μg/ml BFA, 0.5 μm 30N12, or 1 μm monensin for 4 h before samples were prepared for Western blots or immunofluorescence microscopy. For fluorescence recovery after photobleaching (FRAP; see FRAP section below), hTERT-RPE1 cells stably expressing bP/rds-Dend2 or somatostatin receptor 3 (SSTR3)-GFP grown on glass-bottomed dishes were cultured in DMEM/F12 with 0.5% FBS for at least 48 h to induce cilia. For inhibition of conventional secretion, the cells were treated with 0.1 μg/ml BFA, 0.5 μm 30N12, 1 μm monensin, or 150 μm cycloheximide starting 0.5 h before photobleaching. Untreated cells were used for negative control experiments.
FRAP.
FRAP experiments were performed with a method similar to those described previously (Boehlke et al., 2010; Trivedi et al., 2012) with modifications optimized for our microscope system as follows. Cells were incubated in the DMEM/F12 medium with 0.5% FBS in the presence and absence of inhibitors. Cells were maintained inside a sealed chamber (DMIRB/E ONICS-D35; Tokai Hit) that maintained the temperature (37°C), humidity, and gas concentrations (5% CO2 in 95% air) on the microscope during the imaging procedures. FRAP was measured by a Leica TCS SP2 laser scanning confocal microscope using the FRAP module in Leica Confocal software version 2.61 Build 1537. To ensure the effective bleaching of ∼90% of the fluorescence signal, cilia were irradiated with an intense 488 nm laser. The laser power was set to 75%, the pixel dwell time to 4.9 μs, and the pixel size to 0.092 × 0.092 μm using the confocal software. Images were acquired before, immediately after, and 1 h after bleaching using a z-series scan with a step size of 0.5 μm. The imaging conditions did not cause significant photobleaching of fluorescent proteins. For qualitative assessment of cilia bleaching and recovery, maximum projection images were created for each condition. For quantitative assessment, single optical sections were analyzed using ImageJ software. Cilia fluorescence was quantified for prebleach and postbleach conditions and 1 h after photobleaching. The degrees of fluorescence recovery were normalized to prebleach fluorescence intensity after subtracting the measured background in the area with no objects. For BFA-treated cells, 50% of cilia did not show any appreciable recovery; therefore, we quantified the FRAP data only for the 50% of cilia for which P/rds recovered.
Cell surface biotinylation.
Cell surface proteins were labeled by biotinylation as described previously (Tian et al., 2009) with minor revisions. Briefly, cells were incubated with EZ-Link Sulfo-NHS-SS-Biotin (Proteochem) in PBS buffer containing the following (in mm): 137 NaCl, 10.1 Na2HPO4, 1.8 NaH2PO4, pH 7.4, along with 100 μm CaCl2 and 1 mm MgCl2, for 30 min at 4°C and the excess biotin was quenched by incubation with100 mm glycine for 30 min at 4°C. Cells were lysed in 1% Triton X-100 and the total lysate was fractioned to cell surface and cytoplasm parts by incubation with immobilized monomeric avidin agarose (Thermo Scientific) at 4°C overnight with end-over-end rotation. Loaded avidin beads were resuspended in 1× SDS-PAGE loading buffer.
Glycosylation analysis.
Peptide-N-Glycosidase F (PNGase F) and Endo H (New England Biolabs) treatments were done following the manufacturer's instructions. Briefly, 7 μl of protein sample was mixed with 1 μl of G7 (for PNGase F) or G5 (for Endo H) reaction buffer, 1 μl of 10× glycoprotein denaturing buffer, 1 μl of 10% Nonidet P-40 (for PNGase F), or H2O (for Endo H), and 1 μl of enzyme and the mixtures were incubated at 37°C for 1 h.
Western blots.
Cells or tadpole's eye balls were sonicated in PBS containing Complete Protease Inhibitor Mixture (Roche), assayed for total protein concentration, and then mixed with SDS sample buffer. The resulting samples were resolved by SDS-PAGE (12% polyacrylamide) and transferred onto PVDF membranes. The following primary antibodies were used: mouse mAb anti-HA (Covance), mouse mAb anti-bP/rds (a kind gift from Dr. Brian Kevany, Department of Pharmacology, Case Western Reserve University, Cleveland, Ohio), mouse mAb anti-X. laevis P/rds (Loewen et al., 2003; a kind gift from Dr. Robert S. Molday, The University of British Columbia, Vancouver, Canada), rabbit pAb anti-GFP (Novus Biologicals), mouse mAb anti-ATPase Na+/K+ α5 and mouse mAb anti-β-tubulin E7 (Developmental Studies Hybridoma Bank, University of Iowa, Iowa City, Iowa), mouse mAb anti-rhodopsin 1D4 (MacKenzie et al., 1984; the antibody originally generated by Dr. Robert S. Molday et al. was a kind gift from Dr. Vera Moiseenkova-Bell, Department of Pharmacology, Case Western Reserve University, Cleveland, OH), rabbit pAb anti-GORASP2 (GRASP55; Proteintech), and rabbit pAb anti-Dend2 (custom made by us). The immunoreactive bands were detected using HyGLO Quick Spray Chemiluminescent HRP antibody detection reagents (Denville Scientific) according to the manufacturer's directions. Band intensities were analyzed with ImageJ.
Immunofluorescence microscopy.
Cells were fixed in 4% paraformaldehyde for 30 min at room temperature. Tadpoles were anesthetized, decapitated, and their heads were fixed in 4% paraformaldehyde for 6 h at room temperature. Frozen heads were sectioned on a cryostat (CM1850; Leica) at −20°C. Fixed cells or eye sections were blocked in 1.5% normal goat serum diluted in PBS with v/v 0.1% Triton X-100 (PBST) for 1 h, incubated with primary antibody overnight at 4°C, and then secondary antibody for 1 h at room temperature. The following primary antibodies were used: rat mAb anti-HA (Invitrogen), rabbit anti-mannosidase II antibody (Abcam), mouse mAb anti-acetylated tubulin (Sigma-Aldrich), mouse mAb anti-GM130 and mouse mAb anti-P230 (BD Biosciences). Goat anti-mouse Cy3, donkey anti-rabbit Cy3 (Jackson ImmunoResearch), and donkey anti-rat Alexa Fluor 488 (Invitrogen) were used as secondary antibodies. To visualize the outer segment (OS), tadpole sections were incubated with 10 μg/ml Alexa Fluor 633-conjugated wheat germ agglutinin (Invitrogen) overnight at room temperature in PBS with v/v 0.5% Triton X-100. Samples were visualized by confocal microscopy.
Generation of transgenic X. laevis.
All animal experiments were performed in accordance with the Association for Research in Vision and Ophthalmology Statement for the Use of Animals in Ophthalmic and Vision Research based on the protocol approved by the institutional animal care and use committee at Case Western Reserve University.
Transgenic X. laevis expressing full-length or truncated xP/rds-Dend2 were generated using the sperm nuclear injection method (Smith et al., 2006). Injected eggs were housed at 16°C. Tadpoles of either sex were screened for the presence of green fluorescence in their eyes using a Leica MZ16F stereoscope when they were 7–10 d old and divided into three categories, “light,” “medium,” and “bright,” according to the intensity of green fluorescence in their eyes. To ensure that the animals had similar expression levels, only the bright tadpoles (representing the top 39% for each transgenic construct) were used for the experiments. Tadpoles were staged according to Nieuwkoop and Faber's normal table (Nieuwkoop and Faber, 1967).
Photoconversion of full-length or truncated xP/rds-Dend2 fusion protein and retina preparation for confocal imaging.
X. laevis transgenic tadpoles (10–14 d old, stage range of 41–46) raised in the dark were used for photoconversion at ∼10:00 A.M. The tadpoles were anesthetized with 0.026% tricaine (Sigma-Aldrich) and then placed in 6% methyl cellulose (Sigma-Aldrich). Photoconversion was performed by directing a 405 nm light source at their heads for 15–20 min with periodic breaks. For confocal imaging, tadpoles were killed immediately after photoconversion (10:00 A.M.), 4 h after photoconversion (2:00 P.M.), or 48 h after photoconversion (10:00 A.M.). Tadpoles were decapitated under anesthesia and retinas were dissected out and cultured in modified Wolf amphibian culture medium (55% MEM; Invitrogen), 31% Earle's sodium-free BBS (400 mg/L KCl, 200 mg/L CaCl2, 100 mg/L MgSO4, and 92 mg/L NaH2PO4•H2O, 10% FBS, 30 mm NaHCO3, 700 mg/L d-glucose; Lodowski et al., 2013).
Confocal microscopy.
Living retinas, fixed retinal sections, or fixed cells were imaged with a Leica TCS SP2 laser scanning confocal microscope. Leica Confocal Software Lite was used to construct maximum projection images from Z-stacks. All images used for statistical analysis were taken using the same laser power, zoom factor, image averaging, and resolution.
Image analysis, quantification, and statistics.
To analyze the cilia localization of bP/rds-HA-FLAG in cells quantitatively, fixed cells were colabeled with rat anti-HA for bP/rds-HA-FLAG and mouse anti-acetylated tubulin for cilia. Based on the relative intensity of bP/rds-HA-FLAG in cilia compared with other regions of the cell, cells positive for both acetylated tubulin and HA were divided into three categories: cells with bP/rds-HA-FLAG enriched in cilia were defined as strong; cells with bP/rds-HA-FLAG in cilia but with intensities lower than those of the other parts of the cell were defined as weak; and cells with no bP/rds-HA-FLAG in cilia were defined as NA. The percentage of each category was calculated by analyzing 100–1000 cells originated from at least three independent preparations.
To analyze the concentration of xP/rds in photoreceptor cells, green fluorescence intensity of Dend2 that fused to full-length xP/rds and CT truncations was quantified using ImageJ. Images were loaded and the shape tools were used to encompass an area of 15 × 15 pixel in both the OS and inner segment (IS) of photoreceptor cells. For IS, an area was selected inside the protein-dense region of xP/rds1-316-Dend2, xP/rds1-288-Dend2, and the corresponding relative region of xP/rds-Dend2. The concentration of fusion proteins was calculated by normalizing the measured intensity of Dend2 with the calibration curve of purified Dend2. The calibration curve of purified Dend2 was established as reported previously (Lodowski et al., 2013).
The Mann–Whitney rank sum test (MWRST) or the t test (SigmaPlot 12.5; Systat Software) was used to determine whether the difference in the median values between two group samples was statistically significant. For some of the experiments, variability of the protein expression levels lead to a distortion of normal distribution. Therefore, a nonparametric statistic test, the MWRST, was used to determine significance in those pairwise comparisons.
Electron microscopy.
Electron microscopy was performed as described previously (Lodowski et al., 2013). Briefly, hemisected tadpole heads were fixed in 4% w/v formaldehyde containing 0.1% w/v glutaraldehyde in 0.1 m HEPES buffer (Electron Microscopy Sciences) for 6 h at room temperature and then dehydrated in ethanol and embedded in LR White resin (Polysciences). Thin sections were blocked and incubated with antibody against Dend2 for 12 h at 4°C. Preimmune rabbit serum was used in place of the primary antibody for negative controls. After washing, samples were incubated for 1 h with 10 nm gold-conjugated goat anti-rabbit IgG (Ted Pella). The gold-labeled thin sections were examined by a JEOL 1200EX electron microscope.
Results
P/rds is not processed by Golgi mannosidase II
To study the cilia targeting of P/rds, we transiently expressed full-length bovine P/rds, tagged with HA and FLAG antibody epitopes (abbreviated as bP/rds-HA-FLAG), in hTERT-RPE1 cells for which primary cilia are inducible (Fig. 1). Upon induction of primary cilia, bP/rds-HA-FLAG localized to cilia (Fig. 1A, bP/rds-HA-FLAG, green). The cilia localization was confirmed by a colocalization with acetylated tubulin (Fig. 1A, red). As a control, a well established cilia marker, SSTR3 (Berbari et al., 2008), was delivered to the cilia (Fig. 1B) under the same conditions. We confirmed that bP/rds-HA-FLAG's cilia localization is not dependent on the fusion tags. bP/rds fused to the fluorescent protein Dend2 localized to the primary cilia appropriately (Fig. 1C). Furthermore, bP/rds without any fusion tags localized to the primary cilia (Fig. 1D). SSTR3 was observed in both cilia and somatic plasma membranes, whereas bP/rds was targeted to the primary cilia specifically and was barely observed in the somatic plasma membrane (a plasma membrane outside the primary cilia, defined in Hoffmeister et al., 2011). Therefore, the cilia-targeting mechanism of P/rds appears to be more specific than that of SSTR3.
Cilia targeting of P/rds in hTERT-RPE1 cells. A–D, bP/rds with HA and FLAG tags (A), SSTR3-GFP (B), bP/rds-Dend2 (C), or bP/rds (D; green) colocalized with acetylated tubulin (ATub, red) or α tubulin (αTub, red) in the primary cilia of hTERT-RPE1 cells. E, When expressed in hTERT-RPE1 cells, the majority bP/rds was Endo H (En] sensitive, whereas SSTR3 and rhodopsin were Endo H resistant. F, G, Endogenous bovine P/rds (F; detected by mAb anti-bovine P/rds) and endogenous X. laevis P/rds (G; detected by mAb anti-X. laevis P/rds) were sensitive to Endo H (En]. P/rds, SSTR3, and rhodopsin were all sensitive to PNGase F (P), confirming that they are glycoproteins. Arrow indicates the position of the Endo H-resistant form of P/rds; arrowhead indicates the position of EndoH-processed P/rds. The images are confocal images of a single x–y plane. Scale bar, 10 μm.
To understand the ER–Golgi-trafficking pathway of P/rds, we tested the glycosylation pattern of bP/rds expressed in hTERT-RPE1 cells. The full-length bP/rds acquired N-linked glycosylation in the ER, as demonstrated by its sensitivity to PNGase F (Fig. 1E, bP/rds-HA-FLAG, P). However, we found that the majority of bP/rds was not processed by GMII, as demonstrated by its robust sensitivity to Endo H (Fig. 1E, bP/rds-HA-FLAG, En, arrowhead). Consistently, P/rds was sensitive to Endo H in bovine and X. laevis retinas (Fig. 1F,G). Unlike bP/rds, rhodopsin and SSTR3 expressed heterologously in hTERT-RPE1cells were processed by GMII and acquired resistance to Endo H (Fig. 1E, SSTR3 and rhodopsin). These observations led us to hypothesize that P/rds was trafficked to the target cell membrane by an unconventional secretory pathway that bypasses at least the medial- and trans-Golgi, where glycosylated proteins acquire resistance to Endo H (Rabouille et al., 1995).
P/rds is trafficked through an unconventional trafficking route
To further investigate the trafficking pathway of P/rds, we established a stable hTERT-RPE1 cell line that expresses a relatively low level of bP/rds-HA-FLAG. In this cell line, the majority of P/rds was Endo H sensitive (Fig. 2A, Con, T, arrowhead). In the majority of the cells, P/rds localized to the primary cilia. To clarify the role of the Golgi apparatus in the cilia targeting of bP/rds, we used small molecule inhibitors of Golgi-mediated trafficking. BFA inhibits conventional secretion and Golgi function by facilitating resorption of the Golgi to the ER (Scheel et al., 1997; Nebenführ et al., 2002). Consistent with this resorption, BFA facilitated redistribution and dispersion of the cis-Golgi marker GM130 throughout the cell (Fig. 2B). In the absence of BFA treatment, we noted that a small amount of P/rds was Endo H resistant, although the majority was Endo H sensitive. However, this Endo H-resistant P/rds disappeared after BFA treatment (Fig. 2A, BFA). These results indicate that P/rds became inaccessible to GMII. Conventional secretion of P/rds, if any existed, was blocked nearly completely. Under this condition, P/rds was still maintained on the majority of the cilia (Fig. 2C), although an ∼10% reduction in the number of the P/rds-positive cilia was observed (69.67 ± 3.23% of the primary cilia were positive with P/rds-HA-FLAG in BFA treatment, 539 cilia from n = 3 preparations, compared with 78.81 ± 0.85% in untreated control, 448 cilia from n = 3 preparations, mean ± SD, p = 0.018 by t test).
Brefeldin A causes resorption of cis-Golgi structure to ER, but does not abolish the cilia targeting of P/rds. A, Western blot analysis of BFA (0.5 μg/ml for 4 h)-treated hTERT-RPE1 cells stably expressing bP/rds-HA-FLAG. Total cell lysate (T), and cell surface fraction (PM) isolated by biotinylation method were treated with Endo H (En). BFA treatment led to abolishment of the small amount of the Endo H-resistant form of P/rds on the cell surface. The arrowhead indicates the position of Endo H-processed P/rds. B, bP/rds (green) was colabeled with anti-GM130 (cis-Golgi, red) in the absence (Con) and presence (BFA) of BFA treatment. BFA treatment caused GM130 redistribution, which indicates the disruption of cis-Golgi structures. C, bP/rds (green) was colabeled with anti-acetylated tubulin (red) in the absence (Con) and presence (BFA) of BFA treatment. Cilia localization of bP/rds was maintained after BFA treatment. D, E, Trafficking of bP/rds-Dend2 or SSTR3, stably expressed in hTERT-RPE1 cells, was tested by FRAP in the absence (Con) and presence of (BFA) treatment. D, One hour after photobleaching, P/rds recovered in the cilia of control cells. P/rds recovered in 50% of cilia treated with BFA. Under the same conditions, SSTR3 recovery was blocked by BFA treatment. E, The amounts of cilia localized P/rds and SSTR3 immediately following and 1 h after photobleaching. The amount was normalized to the prebleach levels. Significant recovery of bP/rds-Dend2, but no significant recovery of SSTR3, was observed in BFA-treated cells. ***p < 0.001 by MWRST. The data are represented as mean ± SD. The numbers of cilia analyzed from at least three sample preparations are indicated in parentheses. F, Cells stably expressing bP/rds-Dend2 were treated with 150 μm cycloheximide for up to 24 h and then analyzed for the bP/rds-Dend2 protein level by Western blots. The half-life of bP/rds-Dend2 is estimated to be ∼1.8 h. The images are confocal images of a single x–y plane. Scale bars: B and C, 10 μm; D, 5 μm.
We then investigated whether P/rds is actively trafficked to the cilia by a Golgi-independent mechanism. FRAP allowed us to test whether the delivery of new P/rds can occur in the absence of the functional Golgi apparatus. To monitor the trafficking of P/rds to the cilia of living cells, P/rds was fused to the green fluorescent protein Dend2. After photobleaching of cilia-localized bP/rds-Dend2, ∼30% of the original level recovered within 1 h (Fig. 2D, E, P/rds-Dend2, Con), which closely coincides with the rate of degradation (∼30%/h) measured after stopping the protein synthesis (Fig. 2F). When a similar experiment was conducted for cilia-localized SSTR3-GFP, ∼30% of the original level (1 h − 0 h) recovered within 1 h (Fig. 2D,E, SSTR3-GFP, Con). To test the effect of BFA, cells were pretreated with BFA for a short time (∼0.5 h) before FRAP experiments. Because of the relatively slow replacement of proteins in the cilia (Fig. 2D,E), short-term BFA pretreatment (∼0.5 h) did not abolish SSTR3-GFP (Fig. 2D, SSTR3-GFP, BFA, Pre) or bP/rds-Dend2 (Fig. 2D, bP/rds-Dend2, BFA, Pre) in the cilia. To evaluate whether newly synthesized proteins were capable of entering the cilia in the presence of BFA treatment, we photobleached the preexisting bP/rds-Dend2 or SSTR3-GFP in the cilia. After photobleaching of cilia-localized bP/rds-Dend2, ∼20% of the original level recovered within 1 h (Fig. 2D,E, bP/rds-Dend2, BFA). We assessed the FRAP for 1 h, during which time the cilia structures were relatively intact. After 1 h, BFA treatment caused gradual cilia resorption and shortening, which might have occurred due to compromised cilia trafficking of structural components. Approximately half of cilia did not show appreciable recovery of P/rds. After BFA treatment, there was a 20% reduction in the number of cells with cilia, indicating the effect of BFA on the retraction of cilia. Under the same conditions, the cilia trafficking of SSTR3, which is dependent on the Golgi apparatus, was completely blocked (Fig. 2D,E, SSTR3, BFA). Because BFA, which blocked Golgi function, did not completely block the cilia targeting of P/rds, these observations are supportive of unconventional trafficking of P/rds. However, because of the partial blockage of P/rds trafficking in BFA-treated cells, we could not exclude the partial involvement of the Golgi apparatus in the trafficking of P/rds.
To further define the route of P/rds trafficking to the cilia, we tested the effects of 30N12, a small molecule screened by a chemical genetic approach that inhibits trans-Golgi to plasma membrane trafficking (Nieland et al., 2004). Unlike BFA, 30N12 does not disrupt ER to Golgi trafficking. Similar to its response after BFA treatment, Endo H-resistant P/rds, which was faintly visible, became hardly detectable after 30N12 treatment (the percentage of the Endo H resistant form of bP/rds-HA-FLAG dropped from 9.16 ± 1.07% in control cells to 1.30 ± 0.62% in 30N12-treated cells, mean ± SD, p < 0.001 by t test, n = 3 preparations; Fig. 3A). However, the population of P/rds-positive cilia was unaffected by 30N12 (78.19 ± 0.75% of the primary cilia were positive with bP/rds-HA-FLAG in 30N12 treatment, 615 cilia from n = 3 preparations, compared with 79.85 ± 2.75% in control, 529 cilia from n = 3 preparations, mean ± SD, p = 0.456 by t test; Fig. 3B). Despite the trans-Golgi-mediated trafficking being inhibited, P/rds was actively trafficked to the cilia, as assessed by FRAP analysis. One hour after photobleaching, recovery of bP/rds-Dend2 to 23.3% of the original level (Fig. 3C,D) was observed on average and all tested cilia showed recovery of bP/rds-Dend2. This degree of recovery in the presence of 30N12 was not significantly different from that observed in the absence of 30N12 (p = 0.054 by MWRST). This rate of recovery closely coincides with the stability of bP/rds-Dend2 (Fig. 2F), and thus suggests that the ciliary P/rds was maintained by active trafficking and decomposition of P/rds. We also confirmed that the recovery of P/rds requires the new synthesis of P/rds, because bP/rds-Dend2 trafficking was significantly inhibited after treatment with cycloheximide (Fig. 3C,D, CYH), which effectively blocked bP/rds-Dend2 synthesis (Fig. 2F). In the presence of 30N12, SSTR3-GFP did not recover in any of the cilia tested, indicating that conventional secretion was effectively blocked (Fig. 3C,D, SSTR3-GFP). 30N12 is a relatively new inhibitor of trans-Golgi to plasma membrane trafficking. More traditionally, monensin, an antibiotic, has been used to inhibit the Golgi-mediated trafficking events, especially at the late compartments of Golgi (Mollenhauer et al., 1990). We found that monensin significantly inhibited the cilia targeting of SSTR3, whereas it allowed the cilia targeting of P/rds in a manner similar to 30N12 (Fig. 3B–D). These experiments using BFA, 30N12, and monensin indicate that the conventional secretion from the TGN is not essential for P/rds trafficking. Therefore, P/rds is trafficked by an unconventional secretion mechanism.
Trans-Golgi exit is not essential for cilia targeting of P/rds. A, B, hTERT-RPE1 cells stably expressing bP/rds-HA-FLAG were treated with 0.5 μm 30N12 or 1 μm monensin, which block the function of trans-Golgi structures, for 4 h. A, The samples treated with 30N12 were analyzed by Western blots. Total lysate (T) and cell surface fraction (PM) were also treated with Endo H (En). After 30N12 treatment, the small amount of the Endo H-resistant form of P/rds on PM became nearly undetectable. B, bP/rds (green) was colabeled with anti-acetylated tubulin (red) with (30N12, Mon) or without (Con) treatment. Cilia localization of bP/rds was maintained after 30N12 or monensin treatment. C, D, The trafficking of bP/rds-Dend2 or SSTR3-GFP, stably expressed in hTERT-RPE1 cells, was tested by FRAP in the absence of (Con) and presence of 30N12 (30N12) or monensin (Mon). The effect of cycloheximide (CYH) treatment was also tested on the trafficking of bP/rds-Dend2. C, One hour after photobleaching, P/rds recovered in the cilia of 30N12- or monensin-treated cells. Under the same conditions, SSTR3 recovery was blocked by 30N12 or monensin treatment. P/rds recovery was blocked by CYH treatment, suggesting that protein synthesis is required for P/rds recovery in cilia. D, The amounts of cilia-localized P/rds and SSTR3 immediately following and 1 h after photobleaching. The amount was normalized to the prebleach levels. Significant recovery of bP/rds-Dend2 was observed in 30N12- or monensin-treated cells, but not in CYH-treated cells. SSTR3 recovered in untreated control cells; however, the recovery was blocked by 30N12 or monensin. Therefore, 30N12 and monensin blocks the conventional trafficking of SSTR3 effectively, but does not block the unconventional trafficking of P/rds. The data are represented as mean ± SD. ***p < 0.001 by MWRST. The numbers of cilia analyzed from at least three sample preparations are indicated in parentheses. The images are confocal images of a single x–y plane. Scale bars, 5 μm.
Unconventional trafficking pathway of P/rds is distinct from the previously described GRASP55-dependent route
Many membrane proteins use COPII-coated vesicles to exit the ER after their synthesis. We investigated whether the unconventional trafficking of P/rds is dependent on COPII. COPII assembly is mediated by Sar1 GTPase (Kuehn et al., 1998). We disrupted this assembly process by expressing the dominant-negative Sar1(H79G) (Aridor et al., 1995). The expression of Sar1(H79G) led to redistribution of Golgi apparatus, as indicated by dispersion of a cis-Golgi marker, GRASP55 (Fig. 4A, top). Because COPII-coated vesicles are essential for the maintenance of Golgi apparatus, this result indicates that COPII-mediated trafficking was effectively blocked by Sar1(H79G). A similar effect was observed when cells were treated with BFA (Fig. 2B), which blocks COPI-mediated trafficking. As a control, the expression of wild-type Sar1 did not affect the structure of the Golgi apparatus (Fig. 4A, bottom). The expression of Sar1(H79G) led to a dramatic decline in the number of cilia positive for bP/rds (20.01 ± 2.67% of the primary cilia were positive with bP/rds-HA-FLAG in the cells expressing Sar1(H79G), 205 cilia from n = 3 preparations, compared with 80.75 ± 1.02% in the cells expressing Sar1, 155 cilia from n = 3 preparations, mean ± SD, p < 0.001 by t test; Fig. 4B). Because the exit from the ER was blocked, P/rds accumulated in the ER (Fig. 4C). Therefore, this experiment indicates that the cilia targeting of P/rds is dependent on the COPII-coated vesicles that bud off from the ER.
Cilia targeting of P/rds in cilia is dependent on COPII, but not on the GRASP55-mediated mechanism. A, Cells stably expressing bP/rds that are also transiently expressing either wild-type or the H79G mutant of Sar1 (red) were colabeled with anti-GRASP55 (green). Expression of Sar1(H79G), but not wild-type Sar1, caused the disruption of cis-Golgi structure. B, P/rds (green) was colabeled with wild-type or the H79G mutant of Sar1 (red). The primary cilia is also labeled by anti-acetylated tubulin (red, the same color as Sar1). Cilia targeting of P/rds was inhibited by Sar1(H79G). C, P/rds (green) was colabeled for Sar1 (blue) and calreticulin (red, an ER marker). P/rds localized in ER in the cell expressing Sar1(H79G). D, E, Cells stably expressing bP/rds were transfected with GRASP55 siRNA and then analyzed for GRASP55 protein level (D) and the localization of P/rds (E; green) in primary cilia (acetylated tubulin, red). The siRNA effectively reduced the protein expression of GRASP55 by more than 10-fold, but did not affect the cilia targeting of P/rds. The images are confocal images of a single x–y plane. Scale bars, 10 μm.
Some of the unconventional trafficking pathways are dependent on GRASP55, a protein mainly localized to the cis-Golgi apparatus (Gee et al., 2011; Giuliani et al., 2011). We tested the effect of knocking down GRASP55 by siRNA, a treatment known to effectively block GRASP-dependent unconventional secretion (Gee et al., 2011). After siRNA knock-down, the protein expression of GRASP55 was suppressed to 6.48 ± 0.06% (mean ± SD) of the original level (Fig. 4D). This knock-down did not affect the cilia targeting of P/rds (75.61 ± 3.21% of the primary cilia were positive with bP/rds-HA-FLAG in GRASP55 siRNA-transfected cells, 489 cilia from n = 3 preparations compared with 77.76 ± 1.72% in nontargeting siRNA-transfected cells, 439 cilia from n = 3 preparations, mean ± SD, p = 0.45 by t test; Fig. 4E). Therefore, unconventional trafficking of P/rds is not dependent on GRASP55.
C-terminal tail of P/rds supports cilia targeting and enrichment
To understand the regions supporting the specific cilia targeting of P/rds, the CT tail region was dissected. For this dissection, the full-length and truncated P/rds were transiently expressed because transient expression allowed us to test the localizations of different constructs at various expression levels. The full-length bP/rds localized to ∼90% of acetylated tubulin-positive cilia (Fig. 5A, bP/rds-HA-FLAG, strong + weak). Among these positive cilia, the majority of them were highly concentrated with bP/rds (Fig. 5A, bP/rds-HA-FLAG, strong). The CT tail region of P/rds is highly homologous among species (Fig. 5B) and constitutes the largest domain exposed to the cytoplasmic side. The truncation mutant bP/rds1-342 localized to cilia similarly to bP/rds (Fig. 5A,C, bP/rds1-342-HA-FLAG), indicating that the last four amino acids of P/rds were dispensable for cilia localization. Truncation of an additional six amino acids (bP/rds1-336) led to a compromised localization of P/rds in cilia. bP/rds1-336 localized to 70% of acetylated tubulin-positive cilia (Fig. 5A,C, bP/rds1-336-HA-FLAG, strong + weak), a 20% decline compared with the cells expressing bP/rds. The amount of bP/rds1-336 present was also reduced in the majority of cilia (Fig. 5A, bP/rds1-336-HA-FLAG, weak and NA). Therefore, these results suggest that the six amino acids (EDAGQA) from 337 to 342 aa facilitate the passing of bP/rds through the transition zone and concentrating it to the primary cilia. However, because bP/rds1-336 still localized to the majority of cilia, the cilia localization of P/rds is contributed by another region. Further truncation of six amino acids (bP/rds1-330) led to total abolishment of cilia targeting (Fig. 5A,C, bP/rds1-330-HA-FLAG). These results indicate that the region spanning 331–342 aa contains the sequence necessary for cilia targeting and enrichment. Gradual compromise of the cilia localization suggests a modular nature of the cilia localization signals. The region within aa 337–342 is required for enrichment to the cilia and the collective regions at aa 331–342 are essential for cilia targeting.
The C-terminal tail region of P/rds is essential for the cilia targeting. A, hTERT-RPE1 cells transiently expressing full-length or CT-truncated bP/rds were analyzed for the percentage of the primary cilia positive to P/rds. Strong (black bar), bP/rds is enriched in cilia; weak (red bar), bP/rds is in cilia but not enriched; NA (green bar), P/rds is not observed in cilia. ***p < 0.001 compared with bP/rds by t test. Error bars represent SD. bP/rds (1060 cilia from n = 6 preparations), bP/rds1-342 (635 cilia from n = 3 preparations), bP/rds1-336 (416 cilia from n = 3 preparations), bP/rds1-330 (245 cilia from n = 3 preparations), bP/rdsΔ289-312 (501 cilia from n = 3 preparations). A gradual reduction in the cilia-targeting efficacy was observed by CT truncations. B, The alignment of the P/rds CT tail region indicates that this region is highly conserved among vertebrate species: identical (*) or homologous (:) residues are shown. C, bP/rds1-342 (green) was concentrated in cilia like full-length protein. bP/rds1-336 (green) and bP/rdsΔ289-312 (green) localized in cilia but were not enriched and bP/rds1-330 (green) did not localize in cilia. Acetylated tubulin (red) was used as a cilia marker. The images are confocal images of a single x–y plane. Scale bar, 10 μm.
Unconventional secretion prevents P/rds access to GMII
Although expressed in a stable fashion, some variation was observed in terms of the expression levels of P/rds in hTERT-RPE1 cells. In cells expressing at low levels, full-length bP/rds localized primarily to the cilia (Fig. 6A–C, bP/rds-HA-FLAG [low], 6D, bP/rds-HA-FLAG, arrow). However, in cells with overexpression, full-length bP/rds also localized to intracellular membranous structures (Fig. 6A–C, bP/rds-HA-FLAG [high], 6D, bP/rds-HA-FLAG, arrowhead). Truncation of the CT region from position 327–346 aa, which is essential for the cilia targeting, minimally affected the Endo H sensitivity of P/rds (Fig. 6E, bP/rds1-342-HA-FLAG, bP/rds1-336-HA-FLAG,bP/rds1-330-HA-FLAG, and bP/rds1-326-HA-FLAG). Because of the loss of the region essential for cilia targeting, bP/rds1-330 did not localize to the cilia and its intracellular localization was otherwise similar to that of the full-length P/rds (Fig. 6D, bP/rds1-330-HA-FLAG). In cells with overexpression, full-length bP/rds was also observed to partially localize to Golgi apparatus, as demonstrated by colocalization with the TGN marker P230, the cis-Golgi marker GM130, and the mid- and trans-Golgi marker GMII (Fig. 6A–C, bP/rds-HA-FLAG [high]). This partial colocalization could explain why a small fraction of bP/rds acquired Endo H resistance. However, in cells with optimum expression levels, bP/rds was able to localize to the cilia without appreciable localization to the Golgi apparatus, as demonstrated by poor colocalization with P230, GM130, and GMII (Fig. 6A–C, bP/rds-HA-FLAG, low).
C-terminal tail region of P/rds is essential for Golgi retention. A–C, hTERT-RPE1 cells transiently expressing full-length or CT-truncated bP/rds (green) were colabeled for GM130 (A; cis-Golgi), for GMII (B; medial and trans-Golgi), or for P230 (C; TGN). The images are confocal images of a single x–y plane. D, Maximum projection images of bP/rds and CT truncations. bP/rds is concentrated in cilia when the expression level is low (arrow) and localized to Golgi and other parts of cells except cilia when the expression level is high (arrowhead). CT truncation, down to the position 312, led to the accumulation of P/rds in the Golgi apparatus. Further truncation to position 288 led to exit of P/rds from the Golgi. Scale bars, 10 μm. E, CT-truncated bP/rds transiently expressed in hTERT-RPE1cells were treated with PNGase F (P) or Endo H (En). Truncation of the CT tail led to acquisition of Endo H resistance. Therefore, P/rds is capable of serving as a substrate for the GMII.
We found that further truncation of regions 317–330 or 313–330 aa from P/rds1-330 resulted in a dramatic increase in Endo H resistance (Fig. 6E, bP/rds1-316-HA-FLAG and bP/rds1-312-HA-FLAG). Unlike bP/rds, bP/rds1-312 localized predominantly to the Golgi apparatus (Fig. 6A–C, bP/rds1-312-HA-FLAG) in all cells tested. Within the Golgi, bP/rds1-312 was confined to the late Golgi compartments (Fig. 6A–C, bP/rds1-312-HA-FLAG), as demonstrated by colocalization with the TGN marker P230 and poor colocalization with the cis-Golgi marker GM130 and the mid- and trans-Golgi marker GMII. The confinement of bP/rds1-312 to the late Golgi compartments suggests that the majority of bP/rds1-312 passed through the medial- to trans-Golgi apparatus, where Endo H resistance was acquired by GMII. Therefore, P/rds can serve biochemically as a substrate for GMII, which confers Endo H resistance. However, a unique unconventional trafficking route made the majority of bP/rds inaccessible to GMII.
CT tail contains a region required for Golgi retention
Considering the unique localization of bP/rds1-312 to the Golgi apparatus (Fig. 6A–D, bP/rds1-312-HA-FLAG), we assumed that P/rds contains the information leading to the Golgi retention in its primary sequence. To test this assumption, the well conserved CT region of bP/rds (Fig. 5B) was further truncated to generate bP/rds1-288. We found that bP/rds1-288 was observed outside of the Golgi apparatus (Fig. 6A–D, bP/rds1-288-HA-FLAG), unlike bP/rds1-312, which was confined to the Golgi apparatus (Fig. 6A–D, bP/rds1-312-HA-FLAG). The majority of bP/rds1-288 was Endo H resistant (Fig. 6E, bP/rds1-288-HA-FLAG) and therefore processed by the Golgi resident enzyme GMII. Therefore, the removal of 289–312 aa allowed bP/rds1-288 to exit from the Golgi apparatus and this region is thus essential for Golgi retention.
A Golgi accumulation of bP/rds1-312 may be a consequence of inhibiting the protein's exit from the late-Golgi compartments for conventional secretion pathway and may contribute to cilia targeting through the unconventional route. To test this contribution, we studied the localization of bP/rdsΔ289-312, in which the region essential for the Golgi retention was removed but the region essential for the cilia targeting was retained. Of all of the transgene-positive cells with cilia, only 36% were strongly positive to bP/rdsΔ289-312, whereas 75% of cilia were strongly positive to bP/rds (Fig. 5A,C, bP/rds-HA-FLAG and bP/rdsΔ289-312-HA-FLAG). Concomitant with reduced cilia targeting, increased Endo H resistance of bP/rdsΔ289-312 was observed (Fig. 6E, bP/rdsΔ289-312-HA-FLAG), which is indicative of the access of P/rds to the later Golgi compartments. Therefore, the region from 289 to 312 aa supports the cilia targeting of P/rds through unconventional secretion, likely by inhibiting normal secretion through the conventional pathway. However, this region is not essential for cilia targeting, because bP/rdsΔ289-312 can still localize to the cilia at low efficiency.
C-terminal-deficient P/rds is incapable of reaching the cell surface
bP/rds1-288 lacks the entire CT region, which is essential for both unconventional cilia targeting and Golgi retention. We investigated whether bP/rds1-288 can take the conventional secretory pathway to reach the cell surface using EZ-Link Sulfo-NHS-Biotin cell surface labeling (Fig. 7A). We found that the cell surface expression of bP/rds1-288 was much less pronounced than that of full-length bP/rds in hTERT-RPE1 cells (Fig. 7A, compare PM in bP/rds-HA-FLAG with that in bP/rds1-288-HA-FLAG). We also stably expressed bP/rds1-288 in IMCD3 cells and confirmed that bP/rds1-288 is barely expressed at the cell surface (Fig. 7A, IMCD3, bP/rds1-288-HA-FLAG). Therefore, even though bP/rds1-288 can exit the Golgi apparatus, it cannot pass through a conventional secretory route.
Unconventional cilia targeting of P/rds in IMCD3 cells. A, hTERT-RPE1and IMCD3 cells stably expressing bP/rds or bP/rds1-288 were separated into cytoplasm (Cyto) and plasma membrane (PM) fractions by the biotinylation method and analyzed by Western blots. bP/rds1-288 is barely observed in the plasma membrane. Na+/K+ ATPase (bottom) was used to demonstrate that the plasma membrane proteins were pulled down effectively. B, PNGase F (P) and Endo H (En) treatments of IMCD3 cells stably expressing bP/rds or its truncation mutants. C, Maximum projection images of full-length and CT-truncated bP/rds expressed in IMCD3 (green). The arrow indicates the cilium. D, Top, IMCD3 and hTERT-RPE1 cells stably expressing bP/rds1-312 (HA, green) were labeled for a cis-Golgi marker GM130 (red). Bottom, Intensity of bP/rds1-312 and GM130 (arbitrary intensity units, AI) along the white bars shown in the top. Significant overlap between bP/rds1-312 and cis-Golgi was observed in IMCD3 cells, whereas the overlap is less significant in hTERT-RPE1 cells. Scale bars, 10 μm.
In IMCD3 cells, full-length bP/rds localized to 46.17 ± 1.99% (mean ± SD) of the primary cilia. The majority of bP/rds were sensitive to Endo H (Fig. 7B, bP/rds-HA-FLAG, En), suggesting that it can pass through the unconventional pathway to reach the cilia. As observed for hTERT-RPE1 cells, Golgi retention supports cilia targeting in IMCD3 cells (Fig. 7C). In the absence of the region for Golgi retention, but in the presence of the region for cilia targeting, the cilia targeting was 50% less efficient than in the presence of both regions (22.07 ± 1.75% of primary cilia were positive with bP/rdsΔ289-312, 268 cilia from n = 3 preparations, compared with 46.17 ± 1.99% with bP/rds, 154 cilia from n = 3 preparations, mean ± SD, p < 0.001 by t test). Therefore, when it is present, the Golgi retention supports the cilia targeting of P/rds and augments the unconventional secretory pathway. Unconventional cilia targeting thus operates both in IMCD3 and hTERT-RPE1 cells.
Although unconventional trafficking of full-length P/rds was observed in both hTERT-RPE1 and IMCD3 cells, cell-line-dependent variability was observed for Endo H resistance of CT-truncated P/rds. In the hTERT-RPE1cell line, P/rds1-312, which lacks the region essential for cilia targeting, was almost completely resistant to Endo H. However, in the IMCD3 cell line, bP/rds1-312 was only modestly resistant to Endo H, at a level ∼15% higher than that of bP/rds. This modest Endo H resistance in IMCD3 cells is consistent with the tendency of P/rds to localize to earlier Golgi compartments (Fig. 7D, IMCD3). In hTERT-RPE1 cells, bP/rds1-312 was more prone to localize to late Golgi compartments (Fig. 7D, hTERT-RPE1), where it can acquire Endo H resistance. Likewise, P/rds1-288, which lacks the entire CT tail, was highly resistant to Endo H in hTERT-RPE1 cells, whereas it was modestly resistant in IMCD3 cells. Despite acquiring Endo H resistance and exiting the Golgi apparatus, P/rds1-288 did not reach the cell surface effectively in hTERT-RPE1 cells (Fig. 7A, hTERT-RPE1). Similarly, P/rds1-288 did not reach the cell surface in IMCD3 cells (Fig. 7A, IMCD3). Therefore, despite the cell-line-dependent variability, neither conventional nor unconventional secretion operates for CT-truncated P/rds in hTERT-RPE1 and IMCD3 cells.
Role of the CT tail region in the unconventional trafficking of X. laevis P/rds
As tested in the ciliated mammalian cells, unconventional trafficking of P/rds is regulated by the CT regions responsible for cilia targeting and the Golgi retention. However, the information related to the Golgi localization of P/rds has not been defined in vivo. Therefore, we used X. laevis rod photoreceptors to test the functional interactions between the regions regulating the cilia targeting and Golgi retention. Previously, the roles of the CT region in cilia targeting were assessed using fusions of GFP, the palmitoylation site of rhodopsin, and variously truncated CT tails of P/rds in X. laevis rods (Tam et al., 2004). Although the previous study was informative in understanding cilia-targeting information, the fusion protein was a nonintegral membrane protein and may not model the unconventional trafficking of the four-transmembrane-protein P/rds. In addition, the GFP fused with the CT tails of P/rds is not glycosylated, so it is not feasible to assess the glycosylation pattern, one of the markers for unconventional trafficking. We studied the role of the CT tail region in context with the rest of the primary structure in X. laevis rod photoreceptor cells (Fig. 8). Full-length xP/rds fused to the fluorescent protein Dend2 (xP/rds-Dend2) localized to the disk membranes (Fig. 8A, xP/rds-Dend2) within the regions consistent with disk rims and incisures (Fig. 8B). We truncated the CT tail spanning 317–346 aa (xP/rds1-316-Dend2), which was essential for cilia targeting in cultured cells. This xP/rds1-316-Dend2 localized predominantly in the intracellular structures in the photoreceptor IS (Fig. 8A, xP/rds1-316-Dend2). These structures partially colocalized with a turquoise trans-Golgi marker (Fig. 8C), indicative of Golgi localization of xP/rds1-316-Dend2. To further understand the region required for Golgi retention, the additional region (289–316 aa) was removed from xP/rds1-316 to create xP/rds1-288. Similar to xP/rds1-316-Dend2, xP/rds1-288-Dend2 localized to IS punctate structures, as observed by fluorescence microscopy (Fig. 8A, xP/rds1-288-Dend2). However, unlike xP/rds1-316-Dend2, which localized to the Golgi apparatus (Fig. 8D, asterisk), xP/rds1-288-Dend2 was not enriched in the Golgi apparatus (Fig. 8E, asterisk), as revealed by immunoelectron microscopy. xP/rds1-288-Dend2 was concentrated in electron-dense vesicle structures (Fig. 8E, arrows). xP/rds-Dend2 did not accumulate in the Golgi or vesicular structures (Fig. 8F) and, as expected, was observed in the rim region of the disk membranes (Fig. 8G). The CT region required for Golgi retention is thus operational in X. laevis rods.
Mislocalization of CT-truncated xP/rds in rods. A, Localization of full-length or CT-truncated P/rds (green) in X. laevis rods. OS areas are labeled by wheat germ agglutinin (WGA, red). Nuclei are in blue. The bottom row of images are magnified views of the selected areas above. B, xP/rds-Dend2 (green) is localized to disk rims and incisures. C, xP/rds1-316-Dend2 (green) colocalizes with the turquoise Golgi marker (red). D–G, Immunoelectron microscopy localization of CT-truncated and full-length P/rds. D, xP/rds1-316-Dend2 mislocalizes in Golgi apparatus. E, xP/rds1-288-Dend2 mislocalizes in electron-dense vesicles of the IS. F, xP/rds-Dend2 does not localize in IS. G, xP/rds-Dend2 localizes in rim and incisures of OS. Asterisks indicate Golgi apparatus and arrows indicate electron dense vesicles. Animals were 14–15 d old.
CT-truncated xP/rds1-316 and xP/rds1-288 were also observed in the OS (Figs. 9, 10, xP/rds1-316-Dend2 and xP/rds1-288-Dend2). By oligomerizing with endogenous P/rds with an intact trafficking signal, CT-deficient P/rds can be brought to the OS (Ritter et al., 2011; Salinas et al., 2013). However, the qualitative localization studies cannot evaluate whether CT-tail-deficient P/rds are trafficked to the OS as efficiently as the full-length P/rds. To understand quantitatively the efficiency of OS targeting for xP/rds1-316 and xP/rds1-288 compared with xP/rds, we used a fluorescent protein-based quantification method (Fig. 9). Antibody-based fluorescence methods are not suitable for quantification due to poor penetration of antibody into the densely packed disk membranes of chemically fixed rods (Tam et al., 2006). Such poor penetration poses a difficulty in accurately calibrating the antibody signals to known molar quantities of standard proteins in nontissue environments. In this experiment, the fluorescence intensities were calibrated to known quantities of soluble Dend2 to obtain molar concentrations in the OS and IS of individual rods (Fig. 9A). OS targeting of xP/rds-Dend2 was efficient, as demonstrated by the majority of cells with OS concentrations >3 μm and the small fraction (23%) lower than 3 μm (Fig. 9B, xP/rds-Dend2). No appreciable mislocalization of xP/rds-Dend2 was observed in the IS (Fig. 9A,B, xP/rds-Dend2). We found that OS trafficking of xP/rds1-316-Dend2 was compromised, as demonstrated by the majority (94.9%) of cells with concentrations lower than 3 μm (Fig. 9B, xP/rds1-316-Dend2). The release from the Golgi apparatus resulted in the OS trafficking of xP/rds1-288-Dend2 being as efficient as that of xP/rds-Dend2 (22% lower than 3 μm; Fig. 9B, xP/rds1-288-Dend2).
Quantitative analysis of P/rds mislocalization in X. laevis rods. A, Localization of P/rds and its truncation mutants in unfixed retinas (maximum projections). Scale bar, 10 μm. B, Concentrations (Conc) of Dend2 fusion proteins were measured for OS and IS structures of individual rods. The numbers of rods analyzed for each transgene construct are shown in parentheses. Those rods originated from three individual tadpoles for each transgene construct. C, The IS/OS concentration ratio was analyzed for each transgene. FL, xP/rds-Dend2; 1–316, xP/rds1-316-Dend2; 1–288, xP/rds1-288-Dend2. The data are represented as mean ± SD. p < 0.001 compared with xP/rds-Dend2 by MWRST. Animals were 9–10 d old.
CT-truncated P/rds retained in IS are slow in renewal. A–C, Dend2 fluorescence of live retinal explants (maximum projections) 0 h (A), 4 h (B), or 48 h (C) after photoconversion. xP/rds1-316 in the IS does not exit the Golgi apparatus for >48 h. D, PNGase F (P) and Endo H (En) analysis of tadpole eyes expressing full-length and CT-truncated P/rds. CT-truncated P/rds are sensitive to Endo H similar to full-length P/rds. Animals were 10 d old in A–C and 14–15 d old in D.
The low OS concentrations of xP/rds1-316-Dend2 are therefore due to protein retention in the Golgi apparatus. Accordingly, the IS to OS concentration ratio, derived from the data points in the scatter plot (Fig. 9B), was significantly higher for xP/rds1-316 than for xP/rds1-288 (p = 0.039 by MWRST) and xP/rds (p < 0.001 by MWRST; Fig. 9C). In this analysis, high variability of expression levels was observed among individual cells, as described previously for this transient X. laevis transient transgenic system (Moritz et al., 2001). This variable cellular expression was advantageous when combined with single cell analysis, because this combination allowed us to inquire whether the IS mislocalizations of xP/rds1-316 and xP/rds1-288 are due to overexpression. We found that xP/rds1-316-Dend2 and xP/rds1-288-Dend2 were mislocalized in the IS without correlation with the different OS concentration and expression levels (Fig. 9B). Within the OS, the molar ratio of P/rds to rhodopsin is estimated to be ∼1:90 (Goldberg and Molday, 1996). Given that the rhodopsin concentration is ∼3 mm (Haeri and Knox, 2012), the P/rds concentration would be ∼30 μm. The OS concentrations of xP/rds-Dend2, xP/rds1-316-Dend2, and xP/rds1-288-Dend2 did not exceed 10 μm (Fig. 9B), indicating that the expression levels of xP/rds-Dend2 constructs were lower than that of endogenous P/rds. The observed IS localizations of xP/rds1-316-Dend2 and xP/rds1-288-Dend2 are thus not due to P/rds overexpression.
To understand the dynamics of P/rds trafficking, we applied a photoconversion technique that takes advantage of the photoconvertible fluorescent protein Dend2 (Gurskaya et al., 2006; Chudakov et al., 2007), which allowed us to discriminate newly synthesized proteins in green from old proteins in red (Lodowski et al., 2013; Fig. 10). We focused on the renewal of P/rds in the OS to understand whether OS targeting is compromised for CT-truncated P/rds. Right after photoconversion, the photoreceptor was occupied with red Dend2 fusion proteins (Fig. 10A). Four hours (Fig. 10B) and 48 h (Fig. 10C) after photoconversion, newly synthesized green xP/rds-Dend2 was observed in the basal part of the OS, whereas old xP/rds-Dend2 localized to a more distal portion of the OS (Fig. 10B,C, xP/rds-Dend2). These observations are consistent with the displacement of old xP/rds, with new xP/rds being synchronized with the process of disk morphogenesis and maintenance. Similar OS renewal was observed for xP/rds1-316 and xP/rds1-288. However, xP/rds1-316 demonstrated dimmer Dend2 fluorescence (Fig. 10A–C), which is consistent with the generally low OS concentrations of xP/rds1-316 (Fig. 9B,C). Therefore, the low OS concentration of xP/rds1-316 is not due to slowed disk membrane morphogenesis.
We then focused on the dynamics of protein trafficking and renewal in the IS. A high concentration of xP/rds1-316-Dend2 in the Golgi apparatus (Fig. 8C) leads to the question of whether the exit of proteins from the Golgi apparatus is slowed or if proteins are retained and unable to exit the Golgi apparatus. To determine which of these two mechanisms is correct, we studied how xP/rds1-316-Dend2 is renewed in the IS. Four hours after photoconversion, the IS structure was almost exclusively composed of old xP/rds1-316-Dend2 (Fig. 10B, xP/rds1-316-Dend2), suggesting that xP/rds1-316-Dend2 was retained in and unable to exit the Golgi apparatus. Forty-eight hours after photoconversion, old protein (red) still occupied the majority of the IS structures (Fig. 10C, xP/rds1-316-Dend2). New proteins were added to the periphery of the IS structures (Fig. 10C, xP/rds1-316-Dend2, arrow), which is suggestive of gradual inflation of the Golgi apparatus and the lack of protein renewal. Consistent with the inflation of the structure, an enlargement of the Golgi apparatus was observed in rods expressing xP/rds1-316-Dend2 by immunoelectron microscopy (Fig. 8D). These observations suggest that xP/rds1-316 causes the abnormal maintenance of the Golgi apparatus due to its aberrant Golgi accumulation. Unlike xP/rds1-316, xP/rds1-288 did not cause inflation of the Golgi apparatus. Within the IS structures, a subpopulation of xP/rds1-288 appeared to be renewed slowly from 4 to 48 h (Fig. 10B,C, xP/rds1-288-Dend2), as indicated by the coexistence of new and old proteins at 48 h after photoconversion (Fig. 10C, xP/rds1-288-Dend2, yellow). In conjunction with the high OS concentrations of xP/rds1-288, these photoconversion results suggest that xP/rds1-288 can leave the IS for OS targeting more readily than xP/rds1-316. Therefore, Golgi retention sequesters xP/rds1-316 in the IS and the loss of Golgi retention released xP/rds1-288 toward the OS.
As observed in IMCD3 cells, CT-truncated P/rds-Dend2 demonstrated robust Endo H sensitivity similar to xP/rds-Dend2 in X. laevis rods (Fig. 10D). Therefore, although full-length and CT-truncated P/rds-Dend2s were observed in the OS, it is unlikely that they passed through the conventional secretion. In summary, these studies on X. laevis are consistent with P/rds' passing through the unconventional secretion (Fig. 11). The removal of the region contributing to cilia targeting causes P/rds1-312 or P/rds1-316 to accumulate in the Golgi apparatus in both X. laevis and mammalian cultured cells (Fig. 11A). This is due to the Golgi retention information located in the CT region 289–312 aa. Further removal of the region allowed P/rds1-288 to exit the Golgi apparatus, but led to its mislocalization to intracellular vesicular structures (Fig. 11A,B). Therefore, in both mammalian cells and X. laevis rods, the regions regulating cilia targeting and Golgi retention are required for the proper subcellular localization of P/rds.
Model for unconventional cilia targeting of P/rds. A, The summarized localizations of P/rds and the CT truncations in photoreceptor cells and mammalian cells. ++, +, +/−, and − indicate that the transgenes' signals were strong, weak, very weak, and undetectable, respectively. Note: When overexpressed in mammalian cells, P/rds can also be detected in Golgi and other intracellular compartments. B, In both mammalian cells (left) and X. laevis rods (right), P/rds bypasses the Golgi apparatus, either the later portions or altogether, to reach the cilia, thus taking an unconventional secretory pathway (green arrow, P/rds). The Golgi retention signal prevents conventional secretion of P/rds1-312 or P/rds1-316 (red arrow, P/rds1-312 or P/rds1-316), but does not block unconventional cilia targeting in rods (red dash-dotted arrow), likely due to trafficking driven by oligomerization with endogenous P/rds. After release from Golgi (left, blue dotted arrow, P/rds1-288), P/rds1-288 is incapable of taking the secretory pathway in mammalian cells (left, blue arrow). Routes labeled 1 and 2 are the estimated routes explaining the cell-line-dependent variability observed between hTERT-RPE1 and IMCD3 cells. Neither 1 or 2 is a complete secretory pathway and cannot deliver P/rds1-288 to the cell surface. P/rds1-288 can reach the cilium through unconventional secretion in rods (right, blue dotted and dash-dotted arrows).
Discussion
P/rds passes through an unconventional trafficking pathway to be delivered to the primary cilia. This unconventional pathway is distinct from previously described GRASP55-mediated secretion (Gee et al., 2011). P/rds bypasses at least a part of the Golgi apparatus, where glycoproteins acquire a complex glycan and Endo H resistance, and also the TGN. Because of this bypass, P/rds targeted to the primary cilia lacks Endo H resistance in both photoreceptor cells in vivo and two ciliated mammalian cell lines, hTERT-RPE1 and IMCD3 (summarized in Fig. 11). Endo H sensitivity was documented previously in bovine retina (Connell and Molday, 1990) and was consistently observed when P/rds was expressed in COS-1 (Goldberg et al., 1995) and MDCK cells (Muller-Weeks et al., 2002; Stefano et al., 2002). P/rds was barely observed in the Golgi apparatus when expressed at low levels in hTERT-RPE1 cells. Similar low expression of P/rds was accomplished with AD293 cells, in which P/rds was barely observed in the Golgi apparatus (Khattree et al., 2013). Our study indicates that the observed Endo H sensitivity of P/rds is not due to its inability to serve as a substrate for the GMII enzyme. In hTERT-RPE1 cells, truncation of the CT tail led bP/rds1-312 to localize predominantly in the TGN, an indication that bP/rds1-312 passed through the mid- and trans-Golgi compartments, where bP/rds1-312 served as a substrate for GMII and acquired robust Endo H resistance. To exclude the possible involvement of conventional trafficking, we applied BFA, 30N12, and monensin, which are known to block conventional secretion. Because these treatments were highly toxic to the photoreceptor neurons and the organism, adequate assessment of P/rds trafficking has been challenging for in vivo studies. Therefore, we took advantage of ciliated hTERT-RPE1 cells in which P/rds was precisely targeted to the primary cilia and blockage of conventional secretion was well tolerated for a short period of time. In hTERT-RPE1 cells, a trace amount of full-length bP/rds was Endo H resistant. However, Endo H-resistant P/rds was abolished by BFA and 30N12, indicating that these treatments blocked the conventional secretion effectively. As assessed by FRAP, the 30N12 and monensin treatments, which inhibited the conventional trafficking of SSTR3, allowed the trafficking of bP/rds to the primary cilia. Therefore, the conventional secretory pathway is not essential for cilia targeting of P/rds.
P/rds and rhodopsin are trafficked in different ways: via unconventional and conventional secretion, respectively. Through conventional secretion, rhodopsin carrier cargoes bud from the trans-Golgi apparatus to be directed to the cilia (Deretic and Papermaster, 1991; Deretic et al., 2005; Mazelova et al., 2009; Wang et al., 2012). In detached cat retinas, rhodopsin localized to the somatic plasma membrane, whereas P/rds localized to intracellular membrane compartments (Fariss et al., 1997). Intracellular localization of P/rds is consistent with our observation that P/rds cannot take the conventional secretory route as a default pathway and accordingly accumulates in the intracellular compartments in the absence of active cilia targeting. Genetic evidence also suggests that P/rds and rhodopsin use different machineries at the late stage of trafficking to pass through the connecting cilia toward the OS. (Zhao and Malicki, 2011). The OS is mainly divided into two subcompartments: the disk membrane and plasma membrane. Rhodopsin is targeted to the disk membrane and plasma membrane of the OS (Basinger et al., 1976). P/rds is targeted to the disk membranes specifically (Fig. 8A,B). Such specific targeting of P/rds would be accomplished by an unconventional pathway that does not deliver protein to the somatic plasma membrane effectively.
The unconventional cilia targeting of P/rds is supported by at least two distinct regions that are essential for the cilia targeting and the Golgi retention, respectively. Both regions operate in photoreceptors, hTERT-RPE1, and IMCD3 cells. For the OS targeting of P/rds, the CT tail region from 317 to 336 aa plays an important role, as demonstrated by Tam et al. (2004). More recently, Salinas et al. (2013) demonstrated that the CT region from 327 to 336 aa is sufficient for this OS targeting, especially the valine at position 332. By removing the CT region from 331 to 336 aa, which also contains this critical valine residue, P/rds was no longer targeted to the primary cilia of hTERT-RPE1 cells. These results suggest that photoreceptors and hTERT-RPE1 cells share the same specific cilia-targeting mechanism, so hTERT-RPE1 is an appropriate model for the study of photoreceptor sensory cilia targeting. In addition to the specific mechanism, however, the photoreceptor may facilitate the cilia trafficking of proteins, also by a default mechanism (Baker et al., 2008), making the detailed assessment of trafficking information difficult. By taking advantage of hTERT-RPE1 cells, we found that a neighboring sequence from 337 to 342 aa plays a supportive role in enriching P/rds to the primary cilia. In the absence of this collective cilia-targeting/enrichment information, P/rds1-330 distributed to intracellular membrane compartments, with poor localization to the plasma membrane. It is unlikely that this poor plasma membrane localization is due to compromised fusogenic activity of P/rds conferred by the CT tail of P/rds. Poor plasma membrane localization was observed for bP/rds1-330, for which the fusogenic region 312–326 aa (Damek-Poprawa et al., 2005) is intact. A tendency to localize inside the cell, observed for bP/rds1-312, is also not consistent with the loss of fusogenic activity, because retention in the Golgi apparatus suggests compromised exit and membrane fission, not fusion activity. Because of the tendency to localize inside the cell, P/rds cannot be delivered to the plasma membrane by a default conventional secretory route in the absence of unconventional cilia targeting (Fig. 11B, left, blue arrows).
P/rds appears to be prevented from exiting the TGN (Fig. 11B, red arrows, P/rds1-312 or P/rds1-316). This prevention is important for cilia targeting because removal of the preventive region (289–312 aa) from P/rds led to compromised cilia targeting, with a concomitant increase in the Endo H resistance of P/rdsΔ289-312. This prevention was also observed in X. laevis rod photoreceptors (Fig. 11B, red arrows, P/rds1-312 or P/rds1-316). Using the photoconversion technique, we found that this Golgi retention mechanism can prevent P/rds from entering the secretory pathway and, accordingly, P/rds1-316 stayed in the Golgi compartment for 48 h. Release from this Golgi retention allowed P/rds1-288 for secretion, likely through an oligomerization-driven unconventional secretory pathway in rod photoreceptor cells (Fig. 11B, right, blue arrows); however, a subpopulation of P/rds1-288 was stuck in the IS electron-dense vesicles. One of the major advantages of using ciliated cells is the apparent absence of oligomerization-driven cilia targeting. Therefore, both P/rds1-312 and P/rds1-288 were incapable of passing through the secretory pathway in hTERT-RPE1 and IMCD3 cells and were not targeted to the cilia (Fig. 11B, left, red and blue arrows).
Golgi retention appears to be nonessential but supportive for unconventional secretion and cilia targeting of P/rds. Accordingly, BFA-treated cells in which the function of the entire Golgi was compromised still maintained P/rds in the cilia with a slight compromise in cilia trafficking, as assessed by FRAP. The nonessential but supportive role of Golgi retention was also indicated from the study of bP/rdsΔ289-312, which lacks the region essential for Golgi retention. bP/rdsΔ289-312 was trafficked to the cilia, indicating that the region from 289 to 312 aa is not essential for cilia targeting. However, some compromise in the cilia targeting was observed, suggesting the supportive role of this region. Golgi retention is a likely mechanism to locate P/rds proximal to the exit sites for unconventional secretion. For example, P/rds retained in the Golgi apparatus would be retrieved to the ER, where P/rds exits for unconventional secretion. The essential role of the ER is indicated from COPII dependency for the cilia targeting of P/rds. Alternatively, P/rds temporarily retained in the Golgi apparatus may exit from the cis-Golgi, as observed for polycystin 2 (Hoffmeister et al., 2011). Our observations are consistent with a partial involvement of early Golgi compartments in the cilia targeting of P/rds.
In summary, here, we provide evidence that P/rds is trafficked through an unconventional pathway. How the unconventional pathway cooperates with the conventional pathway is a question pertinent to the morphogenesis and function of the OS. The unconventional pathway may sequester P/rds from the trafficking routes of other photoreceptor proteins such as GARPs (Poetsch et al., 2001), ROM1(Bascom et al., 1992), and melanoregulin (Boesze-Battaglia et al., 2007), which interact with P/rds. The P/rds-GARP or P/rds-channel interaction is suggested to play important roles in disk morphogenesis and maintenance (Poetsch et al., 2001; Zhang et al., 2009). The P/rds-GARP interaction can start to occur at the place and timing of disk morphogenesis, as revealed by an in situ fluorescence complementation approach (Ritter et al., 2011). Using both conventional and unconventional routes for trafficking, photoreceptors may regulate protein–protein interactions in a spatiotemporal manner.
Footnotes
This work was supported by the National Institutes of Health (Grants EY020826, EY011373, and DK007319). We thank Robert S. Molday for full-length bovine peripherin/rds cDNA and mouse mAb anti-X. laevis peripherin/rds; Kirk Mykytyn for pEGFPn-SSTR3 vector and IMCD3 cells; Brian Kevany for mouse mAb anti-bovine peripherin/rds; Vera Moiseenkova-Bell for rhodopsin antibody; Theodorus W.J. Gadella, Jr. and Antoine Royant for the pmTurquoise2-Golgi vector; Min Goo Lee for pClneo-Sar1-Myc and pClneo-Sar1(H79G)-Myc vectors; and Hisashi Fujioka, director of the Electron Microscopy Core Facility at CWRU, for his support on immunoelectron microscopy. The mouse mAb anti-ATPase Na+/K+ α5 (developed by Douglas M. Fambrough) and the mouse mAb anti-beta-tubulin E7 (developed by Michael Klymkowsky) were obtained from the Developmental Studies Hybridoma Bank developed under the auspices of the Eunice Kennedy Shriver National Institute of Child Health and Human Development and maintained by The University of Iowa, Department of Biology, Iowa City, Iowa.
The authors declare no competing financial interests.
- Correspondence should be addressed to Yoshikazu Imanishi, Department of Pharmacology, School of Medicine, Case Western Reserve University, 2109 Adelbert Road, Wood Building, Cleveland, OH 44106-4965. yxi19{at}case.edu