Mutant ELOVL4 That Causes Autosomal Dominant Stargardt-3 Macular Dystrophy Is Misrouted to Rod Outer Segment Disks

Abstract

Purpose.: Autosomal dominant Stargardt macular dystrophy caused by mutations in the Elongation of Very Long Chain fatty acids (ELOVL4) gene results in macular degeneration, leading to early childhood blindness. Transgenic mice and pigs expressing mutant ELOVL4 develop progressive photoreceptor degeneration. The mechanism by which these mutations cause macular degeneration remains unclear, but have been hypothesized to involve the loss of an ER-retention dilysine motif located in the extreme C-terminus. Dominant negative mechanisms and reduction in retinal polyunsaturated fatty acids also have been suggested. To understand the molecular mechanisms involved in disease progression in vivo, we addressed the hypothesis that the disease-linked C-terminal truncation mutant of ELOVL4 exerts a dominant negative effect on wild-type (WT) ELOVL4, altering its subcellular localization and function, which subsequently induces retinal degeneration and loss of vision.

Methods.: We generated transgenic Xenopus laevis that overexpress HA-tagged murine ELOVL4 variants in rod photoreceptors.

Results.: Tagged or untagged WT ELOVL4 localized primarily to inner segments. However, the mutant protein lacking the dilysine motif was mislocalized to post-Golgi compartments and outer segment disks. Coexpression of mutant and WT ELOVL4 in rods did not result in mislocalization of the WT protein to outer segments or in the formation of aggregates. Full-length HA-tagged ELOVL4 lacking the dilysine motif (K308R/K310R) necessary for targeting the WT ELOVL4 protein to the endoplasmic reticulum was similarly mislocalized to outer segments.

Conclusions.: We propose that expression and outer segment mislocalization of the disease-linked 5–base-pair deletion mutant ELOVL4 protein alters photoreceptor structure and function, which subsequently results in retinal degeneration, and suggest three possible mechanisms by which mutant ELOVL4 may induce retinal degeneration in STGD3.

Introduction

Juvenile-onset autosomal dominant Stargardt macular dystrophy (STGD3) is caused by frame-shift mutations in elongation of very long chain fatty acids-4 (ELOVL4). ¹ ELOVL4 encodes a transmembrane protein that catalyzes the initial rate-limiting condensation reaction in very long chain fatty acid biosynthesis. ^2,3 The disease-causing mutations alter the c-terminal coding sequence, including deletion of a predicted dilysine endoplasmic reticulum (ER) retention motif. Tissue culture studies indicate that the mutant ELOVL4 can exert dominant negative effects on wild-type (WT) ELOVL4, misrouting it from the ER to Golgi membranes or aggresomes. ^4–7 Also, in cultured cells, mutant ELOVL4 cannot synthesize very long chain polyunsaturated fatty acids (VLC-PUFA) and inhibits the biosynthetic ability of WT ELOVL4.⁸  

Unlike cultured cells, photoreceptors sort nascent transmembrane proteins into at least four major membrane compartments: plasma, outer segment (OS) disk, inner segment (IS), and synaptic membranes. Furthermore, aggresomes have not been observed in photoreceptors and are only found in cells that express intermediate filament proteins. ⁹ Hence, it is unclear whether mislocalization of WT ELOVL4 by truncated mutant ELOVL4 occurs in photoreceptors. 

Several mouse models of STGD3 expressing mutant ELOVL4 develop slow, progressive retinal degeneration (RD) and have reduced VLC-PUFA. ^10–14 In vivo, the WT ELOVL4 protein is localized within photoreceptor outer nuclear layer and ISs without any OS or synaptic terminal labeling. ^7,15,16 Grayson and Molday ⁷ also reported expression of ELOVL4 in ganglion cells in human retina. However, photoreceptor-specific expression of fluorescent-tagged mutant ELOVL4 results in disorganized IS and OS membranes and subsequent photoreceptor dysfunction in pigs. ¹⁷ In mutant Elovl4 knockin mice, the mutant ELOVL4 was reportedly mislocalized to dendrites of bipolar cells. ¹⁸ Collectively, these studies suggest that RD is due to expression and mislocalization of mutant ELOVL4 protein or its inhibition of VLC-PUFA biosynthesis. Thus, dominant negative mechanisms and reduced VLC-PUFA are possible causes of RD. Heterozygous Elovl4 mice develop normal photoreceptors that do not degenerate, suggesting haploinsufficiency is not the cause of RD. ^19,20  

Here, we tested the hypothesis that the disease-linked C-terminal truncated ELOVL4 exerts a dominant negative effect on WT ELOVL4 to form aggregates that are misrouted to wrong intracellular compartments of photoreceptors. Expression and accumulation of such aggregates could eventually cause cell death. We generated Xenopus laevis that overexpressed hemagglutinin (HA)-tagged murine ELOVL4 variants in rods driven by the X. laevis rod opsin promoter and assessed expression, subcellular localization, and photoreceptor integrity. We also compared intracellular localizations of mislocalized mutant ELOVL4 with a mislocalized rhodopsin mutant (Q344ter) responsible for RP. ²¹  

We show that WT ELOVL4 localized to rod IS of X. laevis . Surprisingly, mutant ELOVL4 was misrouted from IS through the Golgi to rod OS membranes. However, unlike localization-defective mutants in rhodopsin, there was no mislocalization to other compartments, such as IS lateral plasma membrane or synaptic region plasma membranes. When coexpressed, mutant, but not WT ELOVL4, was misrouted to Golgi and OS disks. We established that the C-terminal dilysine motif is necessary for proper localization of ELOVL4 to IS. Thus, RD in STGD3 patients is likely due to mislocalization of the mutant ELOVL4 coupled with the mutant's inhibitory effect on the WT's ability to make VLC-PUFA. 

Materials and Methods

Transgene Expression Constructs

All transgenes used for expression of the various ELOVL4 fusion proteins in X. laevis photoreceptors were based on pXOP-2DI attb enhanced green florescent protein (eGFP) vector. ²² The green fluorescent protein (GFP) cDNA downstream of the X. laevis opsin promoter in the pXOP-2DI attb eGFP vector was excised with EcoRI and NotI and replaced with PCR-generated products of the various Elovl4 HA epitope tagged (YPYDVPDYA) at the N-terminus of the Elovl4, or untagged constructs using infusion cloning reactions (Clontech, Mountain View, CA, USA). Each construct contained Kozak translation initiation and stop codons to ensure proper expression. The rhodopsin mutant (Q344ter) and the bovine rhodopsin GFP-tagged protein construct variants have been described previously. ^21,23 All PCR products and site-directed mutagenesis were achieved using Phusion High-Fidelity PCR master mix with HF Buffer (New England Biolabs, Beverly, MA, USA). The transgene constructs were verified by sequencing. For transgenesis, expression vectors were linearized with FseI (New England Biolabs) and purified using Qiagen QIAquick gel extraction kit (Qiagen Sciences, Valencia, CA, USA). 

Generation and Rearing of Transgenic X. laevis

Transgenic X. laevis tadpoles were generated using methods described previously, ^{9,21,22,24,25} and the embryos were housed in 4-L tanks at 18°C in a 12-hour light/dark cycle incubator. Double-transgenic animals were generated by simultaneous injection of two transgene constructs. G418 selection of transgenic tadpoles was initiated 24 hours post fertilization for 96 hours. ^9,22,25 Normally developed tadpoles were selected and killed at 14 days postfertilization (dpf). One eye from each tadpole was enucleated and fixed overnight in 4% formaldehyde in 0.1 M sodium phosphate buffer (pH 7.4) for immunohistochemistry analyses, while the other eye was solubilized and used for Western and dot blot analyses. ^9,21,26,27 For dot blot analyses, 20 to 30 animals were generated and analyzed for each transgene. 

Immunohistochemistry and Confocal Microscopy

The fixed enucleated eyes were cryoprotected by infiltration with 20% sucrose solution in 0.1 M sodium phosphate buffer (pH 7.4), embedded in Tissue-Tek O.C.T. compound (Sakura Finetek, Torrance, CA, USA), and frozen. Retinal cryosections (12 μm) were cut, collected on Fisherbrand Superfrost/Plus microslides (Thermo Fisher Scientific, Waltham, MA, USA), and stored at −20°C until used. ^9,28 For confocal microscopy analyses, the retinal sections were labeled with HA monoclonal antibody (1:1000) (Cedarlane, Burlington, ON, Canada), anticalnexin at 1:50 dilution (Stressgen Biotechnologies, Victoria, BC, Canada), or previously characterized affinity-purified ELOVL4 (1:200) polyclonal antibodies ¹⁶ followed by a 1:750 dilution of Cy3-labeled secondary antibody (Jackson ImmunoResearch, West Grove, PA, USA). Confocal microscopy results are based on detailed observations of at least five transgenic animals per transgene, with at least four sections examined per animal. For some transgenes, additional sets of four or more sections were labeled with alternate labeling procedures. The total numbers of animals examined with each labeling procedure is given as “n” in figure legends. The affinity-purified custom rabbit ELOVL4 polyclonal antibodies were generated from synthetic 12-amino acid peptide (GKPQKNGKPKGE) corresponding to amino acids 301 to 312 of the mouse ELOVL4 protein. This antibody is able to recognize WT ELOVL4 but not the mutant ELOVL4 that loses these amino acids used in making the antibodies. The specificity of the ELOVL4 antibody to recognize the mouse ELOVL4 protein was determined by Western blot on protein lysates from mouse Elovl4 mini-gene–transfected human embryonic kidney (HEK) cells, mouse and rat retina, brain, and skin tissues that express the ELOVL4 protein. The affinity-purified ELOVL4 recognized approximately 32-kDa protein, which was specifically blocked by antigenic peptides used in making the antibodies ¹⁶ and is consistent with previously published ELOVL4 antibodies. ^7,15 To further confirm the specificity of the ELOVL4 antibodies, Elovl4 mini-gene tagged with triple HA at the N-terminus was transfected into HEK cells. Protein lysates from the HA-tagged Elovl4 transfected HEK cells were resolved on SDS-PAGE gels, transferred to nitrocellulose membrane, and probed with either HA or ELOVL4 antibodies. Both HA and ELOVL4 antibodies recognize the same molecular weight protein on the blots confirming the specificity of the ELOVL4 antibodies. ⁸ Monoclonal antibodies 2B2²⁹ and B630N ³⁰ were used to label mutant human Q344ter rhodopsin and endogenous X. laevis rhodopsin, respectively, as previously described. ³¹ Wheat germ agglutinin (WGA) conjugated to Alexa Fluor 488 or 555 diluted 1:200 (Invitrogen, Grand Island, NY, USA) was used to label photoreceptor membranes and Hoescht 33342 (10 μg/mL) (Sigma-Aldrich, St. Louis, MO, USA) was used to label nuclei. Confocal images were acquired using a Zeiss 510 Meta laser scanning confocal microscope (Carl Zeiss, Peabody, MA, USA) ^21,27,28 using a ×40 numerical aperture (NA) 1.4 water immersion objective. Using Zeiss imaging software, we collected red, green, blue (RGB) images in which each channel represented a different label. Each channel was then independently adjusted using Adobe Photoshop (Adobe Systems, Inc., San Jose, CA, USA). Antibody labeling was adjusted linearly. WGA labeling was adjusted nonlinearly to emphasize fainter labeling of IS membranes. Images were assembled into figures and annotated using Adobe Photoshop. 

Results

Differential Subcellular Localization of WT and Mutant ELOVL4 Proteins in X. laevis Rod Photoreceptors

To determine the relative localization of WT and mutant ELOVL4 independently in vivo, we took advantage of the large photoreceptors of X. laevis within which photoreceptor organelles and structures can be easily imaged and tracked by confocal microscopy. ³² We generated transgenic X. laevis that overexpressed murine untagged WT ELOVL4, HA-tagged WT ELOVL4 (HA-WT ELOVL4), and HA-tagged disease-linked 5-bp-deletion (797−801−del_AACTT) mutant ELOVL4 (HA-Mutant ELOVL4) proteins under control of the X. laevis rod opsin promoter. We used an HA tag rather than a GFP tag because of the potential for differences in protein biosynthesis and function in GFP-tagged fusion proteins. ^33,34 Also, previous studies from our laboratory had shown that HA-tagged murine ELOVL4 was enzymatically active. ^8,35 We assessed expression, subcellular protein localization, and photoreceptor integrity by Western and dot blots, immunolabeling, and confocal microscopy. Untagged WT ELOVL4 and HA-ELOVL4 localized to rod IS of X. laevis rod photoreceptors, as expected based on previously described ELOVL4 localization experiments in rodent and human photoreceptors (Figs. 1A, 1B). ^2,7,15 Within IS, ELOVL4 and HA-ELOVL4 labeling showed negative staining of membrane compartments that were positively stained with WGA. WGA staining is specific for an N-acetyl glucosamine modification of glycoproteins that occurs in the Golgi, and therefore stains Golgi and secretory pathway compartments beyond the Golgi. WGA labeling of internal membranes in ISs is largely coincident with the trans-Golgi marker rab6 (Fig. 2E), indicating that most of this labeling represents Golgi. ³⁶ HA-ELOVL4 labeling was coincident with calnexin labeling (Fig. 2A), indicating that it was localized to the ER, and did not colocalize with the trans-Golgi marker rab6 (Fig. 2C), indicating that it did not enter the Golgi. There was no difference in the localization of untagged WT ELOVL4 (Fig. 1A) and HA-tagged WT ELOVL4 (Fig. 1B), suggesting that the HA tag had no effect on the localization of the ELOVL4 protein. Within the retina, WT and HA-tagged ELOVL4 were observed only in rod photoreceptors, and not in other retinal cell layers. 

Figure 1

 

Confocal images of transgenic X. laevis rod photoreceptors expressing WT ELOVL4 (untagged), HA-tagged WT ELOVL4, HA tagged mutant ELOVL4, and rhodopsin-Q344Ter transgenes. (A, B) Merged confocal images of untagged WT-ELOVL4 signal ([A] ELOVL4: green, n = 5) or HA-tagged WT ELOVL4 ([B] HA: green, n = 8), Texas-Red WGA (red) and Hoechst 33342 dye (Nuclei, blue). WT-ELOVL4 and HA-ELOVL4 were identically exclusively localized to rod photoreceptor IS membranes, but excluded from WGA-positive internal membranes of the IS (white arrowhead). WT-ELOVL4 and HA-ELOVL4 were not found within OSs, inner nuclear layer (INL), or ganglion cell layer (GCL). (C) HA-Δ5-ELOVL4 (HA: green, n = 19) was misrouted from rod IS membranes to WGA-positive internal membranes of the IS (white arrowhead) and OS membranes. Colocalization of red and green signals appears yellow/orange. (D) Human rhodopsin-Q344Ter labeled with anti-rhodopsin monoclonal antibody 2B2 (hRho: green, n = 5) was found in rod OS, WGA-positive internal membranes (white arrowhead), lateral plasma membrane (blue arrowhead), and synaptic region plasma membrane (yellow arrowhead). (E) WT X. laevis rhodopsin labeled with monoclonal antibody B630N (xRho: green, n = 3) in X. laevis rod photoreceptors. Arrowheads as in (D). Grayscale images represent nonmerged WGA or antibody signals shown at higher magnification. White arrowheads = WGA-positive internal membranes. Blue arrowheads = plasma membrane. Yellow arrowheads = synaptic region plasma membranes. Scale bar: 50 μm (left), 10 μm (middle), 10 μm (grayscale).

Figure 2

 

Colocalization of the ER marker calnexin, and Golgi compartment marker rab6, with HA-ELOVL4 and HA-Δ5-ELOVL4 in rod photoreceptor cells. (A) Colocalization of HA-ELOVL4 and ER marker calnexin (HA: red, calnx: green, n = 3) within IS membranes of X. laevis rod photoreceptors. Note complete overlap of calnexin and HA-ELOVL4 signal. (B) Colocalization of HA-Δ5-ELOVL4 and calnexin (n = 3). Note partial overlap of calnexin and HA-Δ5-ELOVL4 signals. Arrowhead (B) indicates region of nonoverlap (HA-positive, calnexin negative). (C) Lack of colocalization of HA-ELOVL4 and Golgi marker, rab6 (HA: red, rab6: green, n = 3) within rod photoreceptor IS membranes. HA-ELOVL4 and rab6 do not colocalize. Arrowhead indicates HA-negative, rab6-positive membranes. (D) Colocalization of HA-Δ5-ELOVL4 and rab6 (HA: red, rab6: green, n = 3). HA-Δ5-ELOVL4 is misrouted to Golgi compartments resulting in significant overlap with rab6 within IS membranes of rod photoreceptor cells (arrowhead). (E) Colocalization of WGA and rab6 labeling (n = 3) in internal membranes of WT rod ISs. The signals are largely coincident within internal membranes of the IS (arrowhead), indicating that most of these membranes are Golgi. Grayscale images show the indicated isolated signals. Scale bars: 5 μm.

We then investigated the effect of the 5-bp deletion mutation responsible for Stargardt disease on the localization of HA-tagged ELOVL4. Instead of exclusive IS localization, HA-Δ5-ELOVL4 was also misrouted to OS membranes (Fig. 1C). Within IS, HA-Δ5-ELOVL4 labeling did not negatively stain WGA-positive membranes, and was not entirely coincident with calnexin labeling (Fig. 2B), indicating that it was localized to both ER and other membranes of the biosynthetic pathway, including Golgi, which was confirmed by colocalization of HA-Δ5-ELOVL4 and rab6 (Fig. 2D). Punctate aggregates similar to those previously observed in cell culture experiments were not present. ^4–7,37 Furthermore, the mislocalization of HA-Δ5-ELOVL4 (Fig. 1C) was distinct from mislocalization of a rhodopsin variant (Q344ter) that lacks any localization information. ^21,33 First, although substantial quantities of HA-Δ5-ELOVL4 were misrouted to OS, significant amounts of it were retained within IS membranes relative to transgenic Q344Ter rhodopsin (Figs. 1C–E). Second, whereas rhodopsin Q344ter was present in OS and IS membranes, including lateral IS plasma membranes, and synaptic region plasma membranes ²¹ (Fig. 1D), HA-Δ5-ELOVL4 was apparent only in OS membranes and internal IS membranes. 

To determine whether HA-Δ5-ELOVL4 protein was present in the OS plasma membrane, and to confirm its absence from lateral plasma membrane and synaptic region plasma membrane, we used detergent-free labeling to accentuate plasma membrane labeling relative to intracellular labeling. We labeled HA-Δ5-ELOVL4 and rhodopsin Q344Ter transgenic retinas in the presence and absence of TX-100 detergent, using monoclonal antibodies that bind predicted extracellular epitopes (Figs. 3A–C). In the absence of detergent, we did not observe any significant OS labeling of HA-Δ5-ELOVL4 (Figs. 3A, 3B), or significant labeling of plasma membrane in any region of the cells. In contrast, the absence of detergent dramatically accentuated Q344Ter rhodopsin plasma membrane labeling, including lateral plasma membrane and synaptic region plasma membrane (Fig. 3C). We concluded that HA-Δ5-ELOVL4 was not efficiently transported to photoreceptor plasma membranes. These results indicate that, although HA-Δ5-ELOVL4 was misrouted from IS ER membranes, it did not exhibit nonspecific mislocalization (Fig. 1D). Rather, HA-Δ5-ELOVL4 is largely confined to the ER, Golgi, and OS disks, and is excluded from all photoreceptor plasma membranes (Fig. 1C). 

Figure 3

 

HA-Δ5-ELOVL4 is partially delocalized to the OS, but not to plasma membrane and synaptic terminals. (A, B) Confocal micrographs of transgenic X. laevis rod photoreceptors expressing HA-Δ5-ELOVL4 ([A] HA: green, n = 3) labeled in the presence of detergent that reveals delocalization of HA-Δ5-ELOVL4 to OS without prominent plasma membrane labeling. (B) In absence of detergent, there is no observable OS or plasma membrane localization for HA-Δ5-ELOVL4 (HA: green, n = 3). (C) In contrast, in the absence of detergent, 2B2-labeling shows prominent delocalization of Rhodopsin-Q344Ter (hRho: green, n = 3) to OS and lateral plasma membranes (white arrowheads) and synaptic region plasma membrane (yellow arrowhead) of rod photoreceptors. In all images, nuclei were stained with Hoescht 33342 and photoreceptor membranes were labeled with WGA (red). Scale bar: 10 μm.

Effect of Mutant ELOVL4 on the Localization of WT ELOVL4 in Photoreceptors

In cell culture experiments, mutant ELOVL4 is reported to interact with WT ELOVL4 and misroute it from the ER to form mislocalized aggregates. ^4–7,37 We reasoned that if HA-Δ5-ELOVL4 interacted with WT ELOVL4, then it might carry WT ELOVL4 protein to the OS and possibly into the phagosomes of the RPE cells. To address this hypothesis, we coexpressed HA-Δ5-ELOVL4 and untagged WT ELOVL4 in transgenic X. laevis rods. The motif recognized by the antibody used to label untagged WT ELOVL4 binds within the C-terminus, and therefore does not bind HA-Δ5-ELOVL4. ¹⁶ When we coexpressed the HA-Δ5-ELOVL4 and untagged WT ELOVL4 in rods, the WT ELOVL4 localized properly to the IS (Figs. 4A, 4C), while a portion of the HA-Δ5-ELOVL4 protein was again misrouted from the IS to OS (Figs. 4A, 4B). Expression of HA-Δ5-ELOVL4 did not cause mislocalization of WT ELOVL4 protein to OS membranes (Fig. 4C). Again, we did not observe any protein aggregates as previously seen in cultured cells. ^6–8,37 Although we cannot rule out the possibility of undetectable levels of WT ELOVL4 mislocalization, our findings suggest that, in vivo, rod photoreceptors independently sort mutant and WT ELOVL4 to different cellular compartments. 

Figure 4

 

Effect of coexpression of HA-Δ5-ELOVL4 and WT ELOVL4 on mislocalization of WT ELOVL4 to photoreceptor OS. (A–C) Confocal micrographs of X. laevis rod photoreceptor cells coexpressing WT ELOVL4 (ELOVL4, red) and HA-Δ5-ELOVL4 (HA, green) with Hoescht 33342 (blue) (n = 5). WT ELOVL4 expression was restricted to IS without any OS localization (A, C), whereas HA-Δ5-ELOVL4 was distributed within IS and OS membranes (A, B). White arrowheads indicate internal IS membranes (likely Golgi) that are HA-positive and ELOVL4-negative. Left and center are from different transgenic retinas. Right shows higher magnification. Scale bars: 4 and 10 μm.

The Effect of the Carboxy-Terminal ER Retention/Retrieval Dilysine Motif (KXKXX) on IS Localization of the ELOVL4 Protein

We wished to confirm that OS localization of mutant ELOVL4 was due to absence of the dilysine motif and not some other change (such as a conformational change). We generated a construct consisting of full-length HA-WT ELOVL4 protein in which we mutated the lysine residues of the K P K GE motif to arginine (K308R/K310R). This HA-ELOVL4-KK-RR protein was misrouted to OS of X. laevis rod photoreceptors in a manner indistinguishable from the HA-Δ5-ELOVL4 protein (Fig. 5A). This result indicates that loss of the ER retention signal caused misrouting of the HA-Δ5-ELOVL4, including misrouting to OS. Although the HA-ELOVL4-KK-RR was delivered to the OS, unlike HA-Δ5-ELOVL4 (Fig. 2D), it was not prominent in internal WGA-positive IS membranes, possibly indicating some degree of misfolding. 

Figure 5

 

Role of the ELOVL4 KXKXX motif for inner segment targeting/retention of ELOVL4. (A) Confocal micrographs of transgenic X. laevis rod photoreceptors expressing HA-ELOVL4-KK-RR (HA: green, n = 7); note partial delocalization to OS membranes. Arrowhead indicates WGA-positive IS membranes. (B) Addition of KXKXX motif restores IS localization of HA-Δ5-ELOVL4 in transgenic X. laevis rod photoreceptors expressing HA-Δ5-ELOVL4-ERS (HA: green, n = 8). Grayscale shows the isolated WGA and HA signals. Scale bars: 10 μm.

To confirm the role of the dilysine motif in ELOVL4 localization, we generated a construct in which we restored the ER retention signal to the HA-Δ5-ELOVL4 protein. The last 12 amino acid residues (GKPQKNGKPKGE) from the carboxyl terminus of murine WT ELOVL4 protein containing the dilysine motif, KXKXX, were added to HA-Δ5-ELOVL4 (HA-Δ5-ELOVL4-ERS) and expressed in X. laevis rod photoreceptors. This 12 amino acid sequence, which contains the ER retention signal KXKXX, completely restored normal ER localization of the mutant ELOVL4 despite the absence of almost 39 amino acid residues normally found in the WT ELOVL4 protein (Fig. 5B). This further supports the hypothesis that OS mislocalization of HA-Δ5-ELOVL4 is due to loss of the ER retention signal KXKXX. 

The Effect of Carboxy-Terminal ER Retention/Retrieval Dilysine Motif (KXKXX) on ER Localization of Other Transmembrane Proteins

Many OS-targeted proteins such as rhodopsin, photoreceptor-specific retinol dehydrogenase, and RDS/peripherin have OS targeting signals. Others, such as GTPase-activating complex anchoring protein R9AP, are targeted to the OS by non–sequence-specific transmembrane domains. ^{21–23,26,38} These signals are required for exclusive OS localization, and in their absence, proteins are indiscriminately targeted to other membranes in addition to OS. Similarly, IS and cellular organelle–designated proteins, such as microsomal membrane resident cytochrome b5 and mitochondrial membrane–targeted antiapoptotic protein BclXL, contain C-terminal located single-pass transmembrane domains, flanked by positively charged residues (in the case of BclXL protein) that target them to ER or mitochondrial membranes, respectively. ³³ Modification of these domains or charged residues in these proteins result in default OS mislocalization of these non-OS–targeted proteins in frog rods. ³³ Furthermore, studies have demonstrated that juxtaposed dilysine residues at the carboxyl terminus of transmembrane proteins provide a signal to allow retrieval of ER-resident proteins. ³⁹ Our studies (above) have established that the KPKGE motif is necessary and sufficient for ER retention of ELOVL4, and that lack of this motif results in OS membrane targeting and probably contributes to subsequent RD. If so, this motif also should be able to redirect OS-localized transmembrane proteins, such as rhodopsin to the IS. To test this hypothesis, we generated four constructs based on rho-GFP, a fusion of bovine rhodopsin (rho) and GFP previously described. ³⁴ These included the following: 

1.  
   A Rho-GFP tagged protein lacking the C-terminal OS targeting signal of rhodopsin, VXPX (Rho-GFPΔCT),
2.  
   A construct in which we added the last 12 amino acids containing the ELOVL4 ER retention/retrieval signal GKPQKNGKPKGE (ERS) downstream of the Rho-GFPΔCT construct to generate Rho-GFPΔCT-ERS,
3.  
   A construct with ERS signal following the rhodopsin OS targeting signal, VXPX (CT) (Rho-GFP-CT-ERS), and
4.  
   A construct in which the ERS was flanked by rhodopsin OS targeting signals VXPX motif (Rho-GFP-CT-ERS-CT). ^23,38,40

We generated transgenic X. laevis tadpoles using each of these constructs and determined the localization of the GFP fusion proteins. In transgenic X. laevis , the bovine Rho-GFPΔCT protein lacking the OS targeting signal was not only localized to OS, but also mislocalized to IS membranes, including the lateral plasma membrane and synaptic region plasma membrane, as previously described, ³⁴ and was notably absent from ER membranes (Figs. 6A, 6B), similar to rhoQ344ter (Fig. 1). However, when we added the ELOVL4 ERS (KXKXX) membrane localization signal to the Rho-GFPΔCT to generate Rho-GFPΔCT-ERS, the resultant fusion protein was completely redirected to the ER membranes of the IS without any OS plasma membrane, synaptic localization, or localization to WGA-positive Golgi and post-Golgi membranes of the IS (Fig. 6C). Moritz et al. ³⁴ developed a Rho-GFP+VXPX (Rho-GFP-CT) construct that has many biochemical properties of rhodopsin and is sorted and targeted like native rhodopsin in X. laevis rod photoreceptors (Fig. 6A). We asked whether the ELOVL4 ERS sequence would redirect Rho-GFP-CT (Fig. 6A) from OS to the IS membranes when expressed as Rho-GFP-CT-ERS. As predicted, addition of the ELOVL4 ERS retention signal downstream of the VXPX motif completely overrode rhodopsin OS targeting information and caused retention of Rho-GFP-CT-ERS fusion protein in IS membranes (Fig. 6D) similar to that observed for Rho-GFPΔCT-ERS. We then asked if the ELOVL4 ERS motif is capable of functioning at positions other than the extreme carboxyl terminus of the protein. We therefore added the VXPX motif downstream of the ELOVL4 ERS motif. This was sufficient to redirect the resulting Rho-GFP-CT-ERS-CT fusion protein to OS membranes (Fig. 6E). These findings confirm that the ELOVL4 ERS signal KXKXX is a bonafide ER retention signal, and must be located at the extreme carboxyl terminal of the protein. 

Figure 6

 

The ELOVL4 KXKXX motif can override OS targeting signals, provided it is located at the extreme carboxyl terminus. (A) Confocal micrograph of rod photoreceptors expressing Rho-GFP-CT (GFP: green, n = 5), which is primarily targeted to rod OS membranes due to presence of the rhodopsin OS targeting motif VXPX. There was no significant IS lateral plasma membrane (blue arrow) or synaptic region plasma membrane (yellow arrow) associated with Rho-GFP-CT. (B) Deletion of the rhodopsin OS targeting motif VXPX caused Rho-GFP-ΔCT signal (GFP: green, n = 5) to be distributed not only to OS, but also to the plasma membrane of the IS, including the lateral plasma membrane (blue arrow) and synaptic region plasma membrane (yellow arrow). (C, D) Addition of the last 12 amino acids containing the ER retention/retrieval signal of ELOVL4 to Rho-GFP-ΔCT and Rho-GFP-CT to generate Rho-GFP-ΔCT-ERS and Rho-GFP-CT-ERS fusion proteins (GFP: green, n = 8) resulted in complete and efficient IS retention and targeting of the fusion proteins. The fusion proteins were excluded from WGA-positive internal membranes (white arrowheads). (E) Rhodopsin OS targeting signal VXPX overrides the KXKXX motif when placed downstream of the KXKXX motif and directs Rho-GFP-CT-ERS-CT fusion protein (GFP: green, n = 5) to photoreceptor OS. No Rho-GFP-CT-ERS-CT fusion protein is retained within IS, including the lateral plasma membrane (blue arrow) or synaptic region membranes (yellow arrow). Scale bars: 10 μm (color) and 5 μm (grayscale).

Transgenic Expression of Murine Elovl4 Variants and Retinal Degeneration in X. laevis Photoreceptors

Previous studies have shown that transgenic expression of mutant ELOVL4 induces RD ¹⁷ and accumulation of lipofuscin and A2E in mouse models of STGD3. ¹⁰ Therefore, we determined the effect of expression of ELOVL4 variants on retinal morphology in X. laevis rod photoreceptors. In confocal images, RD was observed in a small subset of animals expressing each ELOVL4 variant (Figs. 7C–J). We quantified rod loss in HA-ELOVL4 versus HA-Δ5-ELOVL4 animals using a dot blot assay for rod opsin as previously described ³¹ (Fig. 7A). The differences were not statistically significant, indicating that neither construct was dramatically more toxic than the other, although significant toxicity was observed using a previously characterized positive control (human N15S rhodopsin ³¹ ) (P = 0.008, n = 22 per group, Kruskal-Wallace followed by multiple comparisons according to Conover⁴¹). Relative to our previous studies involving rhodopsin mutants, ELOVL4 variants caused RD in a much smaller proportion of animals ^9,27,42–44 and RD was observed by histology in no more than three animals for any given construct. Retinas with RD always had intense HA signal, indicating RD was likely caused by high-level transgene expression. 

Figure 7

 

ELOVL4 transgenes are relatively nontoxic, but occasionally induce RD in X. laevis retina. (A, B) Quantification of rod loss in HA-ELOVL4 versus HA-Δ5-ELOVL4 animals using a dot blot assay for rod opsin. ³¹ There was no statistically significant reduction in rod opsin in X. laevis expressing HA-ELOVL4 versus HA-Δ5-ELOVL4 (A), indicating that neither construct was dramatically more toxic than the other. In contrast, expression of N15S rhodopsin causes a significant decrease in total rod opsin relative to HA-ELOVL4 (P = 0.008, n = 22 per group, Kruskal-Wallace followed by multiple comparisons according to Conover⁴¹). (B) Quantification of rod loss in animals expressing GFP, HA-Δ5ELOVL4, and HA-Δ5ELOVL4-3H3Q using dot blot for rod opsin. GFP is a relatively nontoxic transgene product that does not cause retinal degeneration. ³² Total rod opsin levels are not significantly different between groups, indicating that HA-Δ5ELOVL4 and HA-Δ5ELOVL4-3H3Q are not significantly more toxic than GFP, or than each other, although a trend toward RD is apparent for HA-Δ5-ELOVL4 (P = 0.45, n = 27 per group, Kruskal-Wallace). (C–J) Expression of transgenic GFP (GFP: green, n = 7) (C) did not cause retinal degeneration and retained photoreceptor integrity similar to nontransgenics (J). However, occasionally retinas expressing ELOVL4 variants (D–H) had RD apparent by histology, as indicated by both loss of rods and shortened rod outer segments (n as reported in Figs. 1, 5, 8). Scale bar: 50 μm.

Figure 8

 

Localization of histidine-deficient mutant ELOVL4. (A) Transgenic expressed GFP protein (GFP: green, n = 7) is present throughout the IS, and at lower concentrations in OS. (B) Transgenically expressed HA-Δ5-ELOVL4 (HA: green, n = 19) and (C) HA-Δ5-ELOVL4-3H3Q HA: green, n = 10) fusion proteins were both localized to IS membranes, including WGA-positive internal membranes (arrowheads), and to photoreceptor OSs. Scale bars: 10 μm (merged) or 5 μm (grayscale).

Effect of Outer Segment Mislocalization of Histidine-Deficient Mutant ELOVL4 on Retinal Degeneration

Although our mutant and WT ELOVL4 constructs each caused RD in a similar proportion of animals, it is possible that the underlying mechanisms differed between constructs. For example, it is possible that overexpression of HA-ELOVL4 caused cell death via ER stress, whereas overexpression of HA-Δ5-ELOVL4 caused cell death via a mechanism involving the enzymatic activity of ELOVL4 in inappropriate cellular compartments. Defects in fatty acid metabolism that result in generation of fatty acyl 3-keto intermediates (the product of ELOVL4) can cause cardiac and nervous system disorders. ^45–47 For example, deficiency in breakdown of 3-keto-acyl CoA intermediates in trifunctional protein patients causes peripheral neuropathy. ⁴⁵ Our cell culture studies showed that mutant ELOVL4 was not enzymatically active. ⁸ Although remote, it is still possible that HA-Δ5-ELOVL4 retains some condensation reaction activity in photoreceptor cells in the living retina. To address the possibility of generation of highly reactive 3-keto acyl-CoA intermediates within IS and OS membranes, we compared the effects of misrouting catalytically active and inactive mutant ELOVL4 on photoreceptors by mutating the critical dideoxy iron-binding motif HXXHH found within the catalytic core of most fatty acid elongase and desaturase enzymes to QXXQQ. ⁴⁸ Both HA-Δ5-ELOVL4 and HA-Δ5-ELOVL4-3H3Q fusion proteins were misrouted to OS (Figs. 8A–C). Compared with control GFP-expressing transgenic animals (Fig. 8C) that do not develop RD, we did not observe a statistically significant reduction in total rhodopsin levels with either construct (P = 0.45, n = 27 per group, Kruskal-Wallace), although some RD was observed in histological sections (Figs. 7A–G). There was no observable difference in retinal structure or morphology in photoreceptors expressing HA-Δ5-ELOVL4 versus HA-Δ5-ELOVL4-3H3Q (Figs. 8B, 8C). The absence of any phenotypic differences supports our conclusion from cell culture studies that 3-keto acyl intermediates are not a factor in RD in our model of STGD3. ⁸  

Discussion

Here, we demonstrate that WT ELOVL4 is localized to the ER membranes of rod IS while truncated mutant ELOVL4 is prominently misrouted via the Golgi to OS disk membranes due to the absence of the KXKXX motif located in the C-terminal region. However, the mutant ELOVL4 does not demonstrate prominent plasma membrane, synaptic membrane, and OPL layer mislocalization as reported in heterozygous-knockin mice (Elovl4^+/stgd3 ). ¹⁸ We further demonstrate the absence of misrouting of WT ELOVL4 to OS via interaction with mutant ELOVL4. Previous cell culture studies suggest that mutant ELOVL4 may cause RD via a dominant negative effect, in which the mutant ELOVL4 interacts with and causes mislocalization and aggregate formation of the WT ELOVL4. ^4–8,37 However, our results do not support this mechanism. In X. laevis photoreceptors, WT ELOVL4 was retained within the ER and IS membranes without any evidence of being mislocalized to the OS via interaction with mutant ELOVL4. Furthermore, neither mutant nor WT ELOVL4 formed aggregates equivalent to those observed in cultured cells. Our results do not rule out the possibility that WT and mutant ELOVL4 may interact at a functional level within ER membranes, which may impact the biosynthetic ability of WT ELOVL4. However, with regard to sorting within photoreceptor cells, the localizations of WT and mutant ELOVL4 are independent of each other. 

To understand the mechanism underlying the misrouting of the mutant ELOVL4 to photoreceptor OS, we expressed rhodopsin variants or mutant ELOVL4 fused to either the rhodopsin OS targeting signals or the ER targeting signal of ELOVL4. We demonstrate that the ELOVL4 ER retrieval/retention signal KXKXX is necessary and sufficient for constraining to the ER the localization of WT ELOVL4, mutant ELOVL4, and even OS-destined transmembrane proteins, such as a rhodopsin-GFP fusion protein. However, notably, the OS and post-Golgi compartments de-localization of mutant ELOVL4 does not precisely resemble the de-localization of mutant rhodopsin, lacking an OS localization signal in that some ER localization of mutant ELOVL4 persists. This is consistent with the KXKXX signal of ELOVL4 being a retention/retrieval signal recognized by receptors/coat proteins that reside/act in compartments distal to the ER, such as the ER-Golgi intermediate compartment and Golgi, where it promotes the recycling of ER-resident proteins that escape from the ER. Interestingly, unlike mislocalized rhodopsin Q344Ter, OS delocalized mutant ELOVL4 did not appear to be nonspecifically trafficked to the plasma membrane. This is in agreement with a previous study of ELOVL4 mutants in cultured cells in which mutant ELOVL4 accumulated in the Golgi or aggresomes of cultured cells without any plasma membrane labeling. ^4–7,37  

Although the HA-Δ5-ELOVL4 and ELOVL4-K308/310R lacking the ER localization signal escape the Golgi and are delocalized to OS membranes, significant amounts of both were retained within IS membranes, including the ER. This suggests that, although absence of the KXKXX signal accounts for mislocalization of the mutant ELOVL4, other signals or interactions within the ELOVL4 also may contribute to retaining it within the ER. These signals or interactions also may play a role in preventing the mutant ELOVL4 from reaching the plasma membrane. ⁴⁹ Indeed, ER retention motifs together with transmembrane-based retention signals have been described for several eukaryotic proteins. ^50–53 For example, the human UDP-glucuronosyltranferase 1A (UGT1A) protein has both ER retention motif and transmembrane hydrophobic cores that retain the UGT1A within the ER. ⁵² Similarly, polar transmembrane-based amino acids are involved in ER localization of presenilin 1. ⁵⁰ Hence, there may be multiple signals that contribute to the localization of ELOVL4 to ER. It has been previously reported that proteins found exclusively in rod disks, such as peripherin/rds, rom-1, and ABCA4, are localized intracellularly when expressed in nonphotoreceptor cell types. ^54,55 However, to our knowledge, this is the first report of a normally intracellular protein adopting a disk-specific localization. This suggests that mechanisms involved in exclusion of proteins from the plasma membrane in other cell types also may be involved in segregating OS disk membrane proteins in photoreceptors. The exact mechanism by which mutant ELOVL4 causes RD, however, remains elusive. We previously proposed three possibilities⁸: 

1.  
   Mutant protein generates toxic 3-keto-acyl intermediates that kill the cells,
2.  
   Some important cellular function(s) is compromised by reduced levels of VLC-PUFA, or
3.  
   Mislocalization of mutant protein causes cellular stress and/or interferes with important cellular functions.

The mutant ELOVL4 retains the histidine catalytic core motif (HXXHH) necessary for generating 3-keto-acyl fatty acids, the first and rate-limiting step in fatty acid elongation. The presence of this highly reactive fatty acid intermediate in the OS or other retinal membranes could potentially induce RD. However, cell culture studies in our laboratory show that the mutant ELOVL4 is not enzymatically active. ⁸ Also, in X. laevis retina, both HA-Δ5-ELOVL4-3H3Q and HA-Δ5-ELOVL4 constructs induced similar levels of RD. More importantly, coexpression of the HA-Δ5-ELOVL4 and WT ELOVL4 did not show an increase in RD, supporting the notion that the mutant ELOVL4 did not mislocalize the WT protein or, if it did, the WT did not synthesize toxic intermediates that destroyed retinal photoreceptors at 14 dpf. Thus, the X. laevis and cell culture studies do not support a role for toxic 3-keto-acyl intermediates generated by mutant ELOVL4 or by mislocalized WT ELOVL4 in the mechanism of RD in STGD3 macular dystrophy. ⁸ We do not, however, rule out the possibility of the mutant protein exerting a dominant negative effect on the VLC-PUFA biosynthetic activity of WT ELOVL4 within the photoreceptor IS as it transits through the ER, which would exacerbate the loss/reduction in retinal VLC-PUFA within photoreceptors, as reported in mice. ¹⁴  

A reduction in the function of WT ELOVL4 protein in the ER could result in a deficiency of VLC-PUFA, which may be required for the construction, function, and maintenance of healthy OS or other photoreceptor membranes; hence, the absence of sufficient quantities eventually results in retinal degeneration. This is supported by studies in knockin mice expressing one WT and one mutant copy of ELOVL4, showing that reduction of ELOVL4 in photoreceptors leads to reduction of retinal VLC-PUFA, which correlates with an age-dependent decline in photoreceptor function and structure. ^11,14,18 However, results from conditional depletion of retinal VLC-PUFA have not been consistent. Barabas et al. ⁵⁶ found no effect of VLC-PUFA reduction on rod or cone structure and function in mice, whereas Harkewicz et al. ⁵⁷ found a loss of cone but not rod function in mice conditionally depleted of ELOVL4. Haploinsufficiency of ELOVL4 reduces levels of retinal VLC-PUFA by half, ^19,20 as found in knockin mice, ¹⁴ but does not seem to cause RD. Thus, the studies to date using mouse models of STGD3 are inconclusive on the role of VLC-PUFA in the pathophysiology of this juvenile macular degeneration. 

Our results clearly show mislocalization of mutant ELOVL4 and thus support the possibility that OS and Golgi delocalization of mutant ELOVL4 contributes to RD in STGD3 patients. The integrity of a photoreceptor membrane compartment could be compromised by the abnormal presence of mutant ELOVL4. This is likely the OS disks, but also could be a membrane compartment located between the ER and OS disks in the biosynthetic pathway, such as the Golgi, or even the retinal pigment epithelium. This could be due to nonspecific effects, such as a corresponding reduction in the density of rhodopsin within OS disk membranes, or the formation of abnormal hetero-oligomers with other OS membrane proteins, leading to alterations in membrane ultrastructure or biochemistry. Alternatively, membrane biochemistry may be directly altered by the abnormal presence of the mislocalized mutant ELOVL4. Other support for mislocalization being involved in the pathogenesis of STGD3 comes from the studies showing that heterozygous-knockin (Elovl4^+/stgd3 ), but not the heterozygous-knockout (Elovl4^± ) have RD, albeit slow, ^11,13,14,18 which is consistent with our findings that expression of the mutant ELOVL4 did not induce widespread RD in tadpoles, suggesting age-dependent onset or RD in STGD3. Also, the early studies of Karan et al. ¹⁰ showed that transgenic mice expressing mutant human ELOVL4 had RD that correlates with the level of transgene expression. 

In conclusion, our studies clearly show that in X. laevis , mutant ELOVL4 is mislocalized to OS and Golgi membranes. WT ELOVL4 is confined to the IS membranes and is not mislocalized to OS or any IS membranes by interaction with mutant ELOVL4. A subset of the animals expressing mutant ELOVL4 transgenes developed RD, suggesting that expression of the mutant ELOVL4 contributes to RD; however, it is not clear that this occurred via mechanisms common to STGD3 patients. This question will require further investigation and comparison of transgenic animal phenotypes and STGD3 phenotypes. 

Acknowledgments

We thank Steve Brush for his useful comments and suggestions. 

Supported by grants from Hope for Vision and Knight Templar Eye Foundation, Inc. (MPA); the National Institutes of Health (EY04149, EY00871, EY021725, and RR17703), Research to Prevent Blindness, Inc., and the Foundation Fighting Blindness (REA); and Foundation Fighting Blindness-Canada and the Canadian Institutes of Health Research (MOP-64400) (OLM). 

Disclosure: M.-P. Agbaga, P; B.M. Tam, None; J.S. Wong, None; L.L. Yang, None; R.E. Anderson, P; O.L. Moritz, None