Truncation Mutation in HRG4 (UNC119) Leads to Mitochondrial ANT-1–Mediated Photoreceptor Synaptic and Retinal Degeneration by Apoptosis

Abstract

purpose. To characterize the time course of apoptosis and degeneration in a transgenic mouse model of retinal degeneration based on truncated mutant HRG4; to investigate the nature of binding of the mutant HRG4 to its target, ADP-ribosylation factor-like (ARL)2; to study its effects on the downstream molecules Binder-of-ARL2 (BART) and adenine nucleotide transporter (ANT)-1 and on the induction of apoptosis.

methods. Saturation binding, microscopic morphometric, Western blot, immunofluorescence, and TUNEL analyses were used.

results. Increased apoptosis did not occur until 20 months in the transgenic retina, consistent with the delayed-onset degeneration in this model. The truncated HRG4 protein exhibited approximately threefold greater affinity for ARL2 than the wild-type HRG4, likely resulting in nonfunctional sequestration of ARL2. A significant decrease in ARL2 was present by 20 months, accompanied by a 50% decrease in ANT-1 in the photoreceptor synaptic mitochondria, with evidence of mitochondrial dysfunction. Preapoptotic degeneration in the photoreceptor synapse was demonstrated with cytochrome c release and caspase 3 activation within the synapse—without evidence of TUNEL-positive apoptosis in the photoreceptor cell body—indicating an initial event in the synapse leading to apoptosis. Caspase 3 was activated in the accompanying secondary neuron, consistent with transsynaptic degeneration.

conclusions. The results support a novel mechanism of retinal degeneration in which preapoptotic degeneration starts in the photoreceptor synapse because of a deficiency in ANT-1 and spreads to the secondary neuron transsynaptically, followed by apoptosis and degeneration in the cell body of the photoreceptor.

Aphotoreceptor protein present in the rod and cone photoreceptor synapses, HRG4 was identified by a subtractive cloning strategy to isolate retina-enriched transcripts. ¹ HRG4 has been shown to be one of the most abundant proteins in the retina and is predominantly associated with synaptic vesicles in the photoreceptor synapses. ^{2 3} HRG4 is homologous to the Caenorhabditis elegans protein UNC119, a mutation that causes problems in coordination, feeding, and chemosensation in the worm. ⁴ To date, HRG4 (UNC119) has been described in various species, including human, rat, mouse, C. elegans, and Drosophila; however, its precise function is unknown. ^{1 5 6}  

The 240-amino acid HRG4 protein consists of two domains. The proximal quarter is rich in proline and glycine and is moderately conserved, and the distal three quarters are highly conserved among species. ¹ A premature termination codon mutation precisely at the border between the proximal and distal domains of HRG4 was demonstrated in a patient with late-onset cone–rod dystrophy. ⁷ At age 57, the patient—who had symptoms of poor night vision, defective color vision, and light sensitivity from age 40—had reduced visual acuity (20/40), myopia, macular atrophy, pericentral ring scotoma, and electroretinographic findings consistent with cone–rod dystrophy. This mutation results in the expression of a 56-amino acid truncated protein. A transgenic mouse model expressing the same mutation in the retina was shown similarly to develop late-onset retinal degeneration, confirming the pathogenic nature of the truncated mutant protein. ⁷ Consistent with the association of HRG4 with synaptic vesicles, severe degeneration of the photoreceptor synapses and specific decreases in some synaptic vesicle proteins were seen in the transgenic retina. ^{7 8}  

In a first step toward elucidation of the function of HRG4, adenosine diphosphate (ADP)–ribosylation factor-like protein 2 (ARL2) was demonstrated to interact with HRG4 by the yeast two-hybrid strategy. ⁹ ADP-ribosylation factors (ARFs) are small, ras-like, guanine nucleotide–binding proteins that participate in a number of important biologic functions, including the activation of phospholipase D and vesicular trafficking. ^{10 11 12 13} ARF-related proteins (ARLs) are 40% to 60% homologous to ARFs. At present, their functions remain generally unknown. ^{14 15} ARL2 is unique among the ARLs in its high affinity for guanosine triphosphate (GTP) and guanosine diphosphate (GDP), its lack of N-myristoylation, and its abundance in neural tissue. ¹⁴ Recently, ARL2 was shown to interact with the Binder-of-ARL2 (BART) protein, to enter the mitochondria, and to bind adenine nucleotide transporter (ANT)-1 in the inner mitochondrial membrane. ^{16 17} Although the precise functional significance of this interaction is not yet known, ANT-1 has been shown to control the level of ATP in the cytoplasm by exchanging cytoplasmic ADP for mitochondrial adenosine triphosphate (ATP). ¹⁸ ANT-1 has also been thought to be involved in apoptosis as a component of the permeability transition pore (PTP), ^{19 20} though recently its role in PTP has been questioned. ^{21 22}  

The truncated HRG4 mutant protein is expressed in the transgenic retina. ⁷ Given the nature of the mutant protein, the interaction between HRG4 and ARL2 in the photoreceptor synapses of the transgenic model of the patient with late-onset cone–rod dystrophy is likely to be abnormal and appears to involve a dominant-negative mechanism. The transgenic model offers an important opportunity to investigate this hypothesis as the molecular basis of the pathogenesis. In this study, we used this model to show that the mutant truncated HRG4 protein has approximately threefold greater affinity for ARL2 than wild-type HRG4, which may result in sequestration of the ARL2 effector in a nonfunctional manner in the transgenic retina, consistent with the postulated dominant-negative mechanism. The long-term consequences of the abnormal HRG4-ARL2 interaction in the photoreceptor synapses of the outer plexiform layer (OPL) included a late-onset decrease in the level of ARL2 in the retina, accompanied by a significant decrease in mitochondrial ANT-1. Other evidence of stress in the mitochondria of the OPL included increased cytochrome c levels consistent with mitochondrial proliferation and preapoptotic degeneration with release of cytochrome c into the photoreceptor synapses and activation of caspase 3 in this location. The observed changes in the synaptic mitochondria were accompanied by a generalized degeneration characterized by apoptosis of photoreceptor cell bodies and transsynaptic degeneration of bipolar cells. 

These findings in the transgenic model of human cone–rod dystrophy strongly suggest that HRG4, an abundant protein in the synaptoplasm of rod and cone photoreceptor synapses, plays a central role in the regulation of levels of ANT-1, a critical protein localized to the inner membrane of mitochondria that is abundant in photoreceptor synapses. Disturbance of this HRG4-mediated regulation can lead to caspase 3–associated apoptosis of photoreceptors and inner retinal neurons, presumably through an abnormal interaction of HRG4 with the ARL2/BART complex. 

Methods

Production of Recombinant Proteins

Recombinant full-length human ARL2 and truncated full-length rat HRG4 (RRG4) proteins were prepared. DNA fragments containing the full coding sequence of human ARL2, the coding sequence of rat HRG4 (RRG4) from the ATG initiation codon to a termination codon at position 57 (truncated proximal), and codons 1 through 240 followed by a termination codon (full-length) were inserted into the bacterial expression vector pGEX4T-2 (Pharmacia Biotech, Piscataway NJ). Each bacterial expression clone was expressed (Gene Fusion System; Pharmacia Biotech). One liter of bacterial culture with an OD₆₀₀ of 0.8 was incubated at 30°C for 5 hours with 0.15 mM isopropyl β-D-thiogalactopyranoside (IPTG) to induce the expression of a fusion protein of glutathione-S-transferase (GST) and the desired protein. The bacterial pellet was suspended in 50 mL phosphate-buffered saline (PBS; 150 mM NaCl, 16 mM Na₂HPO₄, and 4 mM NaH₂PO₄ [pH 7.3]) containing 2 mM EDTA, 0.1% β-mercaptoethanol, and 0.2 mM phenylmethylsulfonyl fluoride (PMSF) and was lysed on ice by mild sonication. After solubilization with 1% Triton X-100, the fusion proteins were adsorbed onto 1 mL glutathione–Sepharose beads and were washed for purification. Recombinant proteins were used in the form of fusion proteins attached to glutathione–Sepharose for some experiments, including pull-down analysis. To obtain nonfusion recombinant proteins, the beads were washed with PBS and incubated with 10 NIH units biotinylated thrombin (Novagen, San Diego, CA) at room temperature for 2 hours to release the recombinant protein from GST bound to glutathione–Sepharose. The amounts of recombinant proteins were quantitated by optical density, protein assay (Protein Determination; Sigma, St. Louis, MO), and Western blotting. 

Preparation of Rat and Mouse Retinal Protein Extract

All procedures using animals were conducted in accordance with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. Adult rat and mouse retinas were homogenized in PBS containing 1 μg/mL aprotinin and 100 μg/mL PMSF and were centrifuged at 15,000g. Protein concentration of the supernatant was measured (Protein Determination Kit; Sigma). 

Saturation Binding Experiments and Western Blot Analysis

For saturation binding with recombinant proteins, recombinant ARL2 protein was combined with increasing amounts of full-length or proximal recombinant RRG4-GST protein in 150 μL PBS. For saturation binding with retinal proteins, increasing amounts of full-length or proximal recombinant RRG4-GST protein was combined with 3 to 5 mg rat retinal proteins. Saturation binding with recombinant GST was also run as control. Each mixture was incubated at 4°C overnight with agitation, and the pelleted glutathione–Sepharose beads with bound proteins were washed with PBS. Equal amounts of bound proteins were boiled and analyzed by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE), and the proteins were electroblotted onto a nitrocellulose membrane (Immobilon-P; Millipore, Bedford, MA). The blot was incubated at 4°C overnight in a blocking buffer containing 5% (wt/vol) nonfat dried milk in 50 mM Tris-HCl (pH 7.5), 200 mM NaCl, and 0.05% Tween 20. After the blocking of nonspecific binding sites, the blot was incubated with 1:200 affinity-purified ARL2 antibody, prepared as previously described, ⁹ at room temperature for 1 hour. After three washes with TBS-T (0.1% Tween 20 in Tris-buffered saline [TBS]), the membranes were incubated with the secondary antibody (peroxidase-conjugated anti–rabbit IgG antibody; Amersham Pharmacia Biotech, Piscataway, NJ) at a 1:1000 dilution in TBS-T at room temperature for 1 hour. Finally, the membranes were washed five times in TBS-T and subjected to detection by enhanced chemiluminescence (ECL; Amersham Pharmacia Biotech). Signals on exposed film were quantitated by densitometry, and B_max and K _d of binding were determined by nonlinear regression curve fitting of the data using commercial software (Prism 4; Graphpad, San Diego, CA). 

Western Blot Analysis of Retinal Proteins

Retinal proteins (20–50 μg), prepared as described from transgenic and normal control retinas of mice of varying ages, were subjected to SDS-PAGE, electroblotted onto nitrocellulose membranes, reacted with polyclonal antibody against ARL2 (1:200), ⁹ BART (1:200; Protein Tech Group, Inc., Chicago, IL), ANT-1 (1:40; Oncogene Research Products, San Diego, CA), S-antigen (1:500; A9C6, gift of Larry Donoso), or rhodopsin (1:100; R2 to R12, gift of Paul Hargrave) and were detected by ECL, also as described. Signals on exposed film were quantitated by densitometry, and levels of ARL2, BART, and ANT-1 were standardized against nonsynaptic photoreceptor proteins, S-antigen, or rhodopsin for each sample. Statistical significance of differences in signals was determined by the Student’s t-test. 

Histopathology

Frozen and paraffin sections of retina from normal and transgenic mice of different ages were stained with hematoxylin and eosin (H&E) and examined microscopically. For the determination of outer and inner nuclear layer thickness, retinal sections were obtained along the vertical meridian through the optic nerve head (ONH) from the superior to the inferior periphery. The entire span was divided into zones 3 to 4 mm (peripheral), 1 to 3 mm (mid), and 0 to 1 mm (central) away from the ONH, and the thicknesses of the outer nuclear layer (ONL) and the inner nuclear layer (INL) were measured in each zone in the digitized images. Ten readings were taken for each zone, and averages were calculated. Statistical significance of differences in nuclear layer thickness in each zone of the retina for a given age in normal and transgenic mice was determined by the Student’s t-test. Photomicrographs were taken to document the histopathologic changes. 

TUNEL Staining

Fragmentation of DNA in apoptotic cells was detected by terminal deoxynucleotidyl transferase (TdT)–mediated dUTP nick-end labeling (TUNEL) assay, with the use of an assay kit (ApoAlert DNA Fragmentation Assay Kit; Clontech Laboratories, Palo Alto, CA) with fluorescence microscopy or a TUNEL system (DeadEnd Colorimetric TUNEL System; Promega, Madison, WI) with light microscopy according to the manufacturer’s protocol. For quantitative assessment of apoptosis, to increase the signal-to-noise ratio, the TUNEL system was converted to a fluorescence-based detection system by using the biotinylated dUTP for streptavidin-conjugated binding (Alexa Fluor 488; Molecular Probes/Invitrogen, Carlsbad, CA). Numbers of apoptotic cells were determined by counting TUNEL-positive cells in sections spanning the entire retina. Statistical significance was determined by the Student’s t-test. 

Preparation of Frozen Sections

Age-matched transgenic and nontransgenic control mice were examined. Mouse eyes were enucleated, fixed in 4% paraformaldehyde in 0.1 M phosphate buffer for 2 hours, and immersed in 20% sucrose in 0.1 M phosphate buffer overnight. Before sectioning, the eyes were embedded in frozen tissue matrix (OCT; Miles, Elkhart, IN) and frozen in liquid nitrogen. Six-micrometer–thick sections were cut on a cryostat and stored at –80°C until use. 

Immunofluorescence

Frozen sections were analyzed by immunofluorescence using polyclonal antibodies to ANT-1 (1:10; Oncogene Research, San Diego, CA), HRG4 (1:100), ³ active caspase 3 (Abcam, Cambridgeshire, UK), or monoclonal antibody to cytochrome c (1:100–200; eBioscience, San Diego, CA), singly or in various combinations. In general, the frozen sections were warmed to room temperature, blocked for 60 minutes with blocking buffer (10% goat serum, 1% BSA, 0.1% Triton X-100 in PBS, pH 7.4), and incubated with the primary antibody in PBS at the appropriate dilution for 90 minutes. After three washes in PBS, sections reacted with the primary polyclonal or monoclonal antibody were incubated with the secondary antibody (FITC-labeled goat anti–rabbit IgG [Life Technologies, Grand Island, NY] or Alexa Fluor 488 or 568 goat anti–rabbit IgG [Molecular Probes/Invitrogen]) for polyclonal antibody and goat anti–mouse IgG (Alexa Fluor 488; Molecular Probes/Invitrogen) for monoclonal antibody at a dilution of 1:50 for 30 minutes, followed by three final washes in PBS. The sections were mounted (Fluoromount-G; Southern Biotechnology Associates, Inc, Birmingham, AL), analyzed, and digitally photographed (photomicroscope III or Zeiss Laser Scanning System LSM 510; Carl Zeiss, Oberkochen, Germany). Double immunofluorescence analyses for HRG4/ANT-1 and caspase 3/cytochrome c were performed. For double immunofluorescence with polyclonal antibodies, the section was treated with cold methanol after the first primary and secondary antibody reactions to block epitopes, then treated with the second primary and secondary antibody using secondary antibodies conjugated with a different fluorescent dye, as previously described. ³ Double immunofluorescence for cytochrome c and caspase 3 was also combined with TUNEL in triple analysis. In this procedure, after the double immunofluorescence staining, the section was treated with increasing concentrations of ethanol and chloroform and then with decreasing concentrations of ethanol, as described, ²³ before the TUNEL staining by a colorimetric system. 

Results

Enhanced Affinity of Truncated HRG4 for the Target Protein ARL2

To investigate the interaction of the truncated proximal HRG4 protein with ARL2, recombinant full-length wild-type HRG4, truncated mutant HRG4, and ARL2 proteins were prepared as described in Methods. Both HRG4 proteins were expressed as GST hybrid proteins bound to glutathione–Sepharose. A saturation-binding experiment using recombinant ARL2 and increasing amounts of recombinant full-length or proximal HRG4 was performed. Analysis of the results using nonlinear regression curve fitting demonstrated approximately threefold higher affinity of proximal HRG4 for ARL2 compared with full-length HRG4, with a K _d of 34.5 ± 11.1 nM for proximal HRG4 and a K _d of 101.6 ± 34.5 nM for full-length HRG4 (Fig. 1) . The binding capacities of full-length and proximal HRG4 did not appear to differ significantly, with a B_max of 33 ± 5.3 nmol/mg protein for full-length HRG4 and a B_max of 40 ± 3.7 nmol/mg protein for proximal HRG4. There was minimal binding of recombinant GST, the control, to ARL2. 

We have previously demonstrated the presence of ARL2 in extracts of retinal proteins. ⁹ In the retina, ARL2 bound to HRG4 should be in equilibrium with the unbound form. To test the ability of the exogenously added wild-type and truncated mutant HRG4 to compete for binding to ARL2 with native HRG4 in the retina, total retinal protein extract was used in a saturation-binding experiment with increasing amounts of recombinant full-length wild-type and truncated mutant HRG4-GST bound to glutathione–Sepharose. As seen with the recombinant ARL2, the result demonstrated an approximately threefold increase in the ability of the truncated HRG4 to compete for binding to the native ARL2 compared with full-length HRG4, reflecting increased affinity, with a K _d of 36.4 ± 3.5 nM for truncated HRG4 and a K _d of 97.4 ± 2.0 nM for full-length HRG4 (Fig. 2) . Again, the binding capacities of the proximal truncated and full-length HRG4 for native ARL2 did not seem to differ significantly, with a B_max of 5.0 ± 0.5 pmol/mg protein for full-length HRG4 and a B_max of 4.0 ± 0.2 pmol/mg protein for proximal HRG4. There was minimal binding of the control recombinant GST to retinal ARL2. 

Late-Onset Apoptosis and Degeneration Affecting the ONL and the INL and the Architecture of the ONL in the Transgenic Model

The HRG4 transgenic mouse is a valuable model of human disease because, as do many human retinopathies, it manifests slowly progressive, late-onset retinal degeneration caused by a single-gene etiology. We have previously described the early- and late-stage histologic appearance of these retinas, the light and electron microscopic features of the dramatic photoreceptor synaptic degeneration, and electrophysiological changes that accompany the synaptic degeneration. ⁷ In this study, the time course of retinal degeneration and apoptosis in the transgenic model were examined in further detail. 

The extent to which different regions of the retina degenerated over time was first examined in peripheral, mid, and central zones of vertical meridian pan-retinal sections of transgenic mice and age-matched controls by determining the thickness of the nuclear layer in the ONL and INL. The result demonstrated little difference between normal and transgenic mice at 6 and 12 months of age but up to an approximately 40% decrease in nuclear layer thickness in both the ONL and the INL by 20 months of age (Fig. 3) . At 20 months, the severity of the degeneration observed in the ONL was periphery > central ≫ mid, whereas all three regions were approximately equally affected in the INL (Fig. 3) . No significant difference was seen between the superior and inferior halves of each zone. 

As previously shown, ⁷ the histopathology of the transgenic mouse retina at 12 months of age consisted of relatively normal-appearing ONL and INL except for migrating photoreceptor nuclei (Fig. 4A , 12 month, Tg, arrows). At 17 months of age in the transgenic retina, there was evidence of disorganization of the ONL architecture with localized protrusions of the ONL into the inner segment (IS) zone, suggestive of localized collapse of the ONL attributed to loss of synaptic anchorage of photoreceptors in the OPL (Fig. 4A , 17 month, Tg, arrow). The 20-month-old transgenic retina demonstrated significant thinning of ONL and INL compared with the normal control (Fig. 4A , 20 month, N and Tg). 

Quantitation of TUNEL-positive apoptotic cell bodies per retinal section was compared in normal and transgenic animals at 6, 12, 17, and 20 months of age. This result demonstrated that despite the profound degeneration of the synapses, a significant increase in the number of retinal cells undergoing apoptosis did not occur until 20 months of age in the transgenic animals, confirming the late onset of the end stage of the retinal degeneration (i.e., photoreceptor cell death) in this model (Fig. 4B) . 

Changes in ARL2, BART, and ANT-1 by Mutant HRG4

We have previously demonstrated that the truncated proximal HRG4 protein is expressed in the transgenic retina. ⁷ To investigate the long-term effect of the demonstrated preferential binding of the truncated mutant HRG4 to ARL2 on the level of ARL2 in the transgenic retina, proteins were extracted from the retinas of transgenic and control mice at 6, 12, and 20 months of age and were subjected to Western blotting. To demonstrate accurately specific changes in ARL2, the level of ARL2 for each sample was standardized against that of S-antigen or rhodopsin, nonsynaptic photoreceptor proteins in the outer segment (this was also done for analyses of BART and ANT-1). The result demonstrated a decrease in the level of ARL2 with age in the transgenic compared with the normal retina (Figs. 5A 5B) . By 20 months of age, the ARL2 level in the transgenic retina was reduced to approximately 30% to 40% of the level of control retina. Immunofluorescence of 20-month-old normal and transgenic retinas (6-μm sections, mid retina, identical antibody concentration and photographic exposure) for ARL2 showed strong signals in the OPL and weaker signals in the INL, IPL, GC, and IS in the normal retina; the signal was significantly decreased in the OPL, INL, IPL, and GC but not in the IS in the transgenic retina (Fig. 5C 5Nand Tg). The transgenic retina showed significant thinning because of degeneration. 

ARL2 has been recently shown to bind the cytoplasmic BART protein. ¹⁶ To investigate whether the observed decrease in ARL2 had an effect on BART, BART levels were compared in the transgenic and normal retina at 6, 12, and 20 months of age. The S-antigen–standardized result showed no difference in BART levels at 6 and 12 months but did show a significantly increased BART level at 20 months of age (Fig. 5D) , coincident with the time when the ARL2 level was reduced to approximately 30% of the normal level. 

A complex of BART and ARL2-GTP has been shown to enter the mitochondria and to bind ANT-1 in the inner mitochondrial membrane. ¹⁷ ANT-1 is a mitochondrial protein involved in ATP/ADP exchange between the mitochondria and the cytoplasm and in apoptosis, though this role has been recently challenged. ^{18 19 20 21 22} The sequestration of ARL2 by the truncated HRG4 in the cytoplasm and the eventual decrease in ARL2, despite a possible compensatory increase in BART, might be predicted to result in less of the ARL2–BART complex available to form and enter the mitochondria to interact with ANT-1. To determine whether ANT-1 was affected in this model, ANT-1 levels, standardized against S-antigen and rhodopsin, were examined at 6, 12, and 20 months of age. As with ARL2 and BART, no differences in the level of ANT-1 were observed between normal and transgenic retinas in 6- and 12-month-old mice. Significantly, however, the level of ANT-1 was decreased to approximately half that of normal in the transgenic retinas at 20 months of age (Figs. 5E 5F) . 

Localization of ANT-1 in normal retina, particularly in the synapses of the OPL (the site of HRG4 action), was determined by double immunofluorescence using antibodies against ANT-1 and HRG4. In the double-stained sections, HRG4 was localized to the OPL and IS of the photoreceptors, as previously shown ³ (Fig. 6A , HRG4, green). Consistent with the widespread distribution of mitochondria throughout many layers of the retina, immunoreactivity for ANT-1 was observed in the GC and the IPL and in parts of the INL, OPL, and IS layers of the retina (Fig. 6A , ANT-1, red). The combined fluorescence image demonstrated prominent ANT-1 staining in the mitochondria-rich synapses in the OPL (yellow) and at the outer margin of the INL, where bipolar and horizontal cells and their dendrites were present (red), whereas HRG4 staining occurred only in the OPL in this region (yellow; Fig. 6ACombined). The staining pattern is more clearly demonstrated in the magnified images (Fig. 6A , lower panels). 

ANT-1 immunofluorescence levels were compared in 20-month-old normal and transgenic retinas (6-μm sections, mid retina; identical antibody concentration and photographic exposure). In agreement with the result of the Western blot analysis (Figs. 5E 5F) , ANT-1 immunofluorescence appeared markedly decreased in the 20-month-old transgenic retina compared with the normal retina (Fig. 6B , top panels). Similar to the decrease observed for ARL2, the decrease was most prominent in the OPL but was also present in the inner retina, and no decrease was seen in the IS. As shown in the magnified images, the entire OPL and the adjacent outer INL regions were uniformly affected, consistent with ANT-1 decreases in the photoreceptor synapses and the mitochondria-rich bipolar/horizontal cell, dendrite, and perinuclear regions (Fig. 6B , bottom panels). 

Cytochrome c Release from Mitochondria, Caspase 3 Activation, and Apoptosis

It has been demonstrated in certain myopathies that a decrease in ANT-1 can result in significant disturbance of mitochondrial-coupled respiration and ATP/ADP exchange between the mitochondria and the cytoplasm. ^{24 25} Such dysfunction is known to induce a compensatory increase in the size and number of mitochondria. ²⁶ To investigate the effect of the observed decrease in ANT-1 on mitochondria in the OPL/INL of the transgenic retinas, the status of mitochondria was examined by immunofluorescent staining of cytochrome c. The result showed a marked increase, relative to control, in cytochrome c immunofluorescent in the OPL and in perinuclear regions in the outer INL, consistent with the increased number of mitochondria (Fig. 7A) . Significantly, photoreceptor synapses that appeared to be filled with cytochrome c were visible, suggestive of leakage of cytochrome c from the mitochondria into the synaptoplasm (Fig. 7A , arrows). Release of cytochrome c from the mitochondria into the cytoplasm is a hallmark of preapoptotic phenomena. If the observed staining pattern were truly reflective of cytochrome c release from the synaptic mitochondria, evidence of apoptotic activity in the photoreceptor might be expected. To test this, the presence of apoptosis was determined by TUNEL staining in the same retinal sections used for cytochrome c immunofluorescence. Using this analysis, photoreceptors were confirmed to be undergoing apoptosis in the ONL of the transgenic retina (Fig. 7B) . 

To further confirm the coexistence of photoreceptor synapses showing apparent increased cytochrome c in the synaptoplasm with apoptosis of photoreceptors, a triple analysis of TUNEL staining, cytochrome c, and caspase 3 immunofluorescence was performed. Activation of caspase 3 occurs with release of cytochrome c and other preapoptotic proteins from the mitochondria and formation of the apoptosome. The results demonstrated TUNEL-positive, caspase 3–positive photoreceptors undergoing apoptosis in the transgenic ONL in the same retinal regions with photoreceptor synapses that were strongly positive for cytochrome c immunofluorescence (Fig. 8) . The location of some of the apoptotic photoreceptors adjacent to the external limiting membrane suggested that at least some of the dying photoreceptors were cones. 

Photoreceptor Synapse as First Site of Preapoptotic Degeneration

We have previously demonstrated that in addition to photoreceptor degeneration, the truncation mutation of HRG4 results in a significant degree of cell death in the INL by transsynaptic degeneration ⁷ (see also the quantitation of INL thickness in Fig. 3 ). In retinal sections double immunostained for cytochrome c and caspase 3, we observed direct evidence of how this might occur. A photoreceptor synapse in the transgenic OPL that is filled with cytochrome c and undergoing preapoptotic degeneration with caspase 3 activation is seen in Figures 9A 9B 9C 9D . This synapse is intimately associated with an INL cell, most likely a bipolar cell, also undergoing apoptotic degeneration with caspase 3 activation. Such an image is highly suggestive of transsynaptic degeneration in action, whereby a dying photoreceptor synapse kills its postsynaptic neuronal partner (Figs. 9E 9F 9G) . Despite analysis of sections encompassing approximately 17 μm in thickness by confocal microscopy, there was no evidence in the adjacent ONL of apoptosis in a photoreceptor cell body corresponding to this synapse. This finding was consistent with the synapse as the first site of preapoptotic degeneration in this model, subsequently leading to apoptosis in the cell body of the photoreceptor. 

Discussion

Biochemical analysis of the interaction of the truncated mutant HRG4 with ARL2 and pursuit of its consequences in this transgenic model of HRG4-based human cone–rod dystrophy led to several significant new insights into the molecular mechanism of slowly progressive, late-onset retinal degeneration. First, a decreased level of ARL2 was observed as a long-term effect of expression of the truncated mutant HRG4 in the transgenic retina. Interestingly, in addition to a prominent decrease of ARL2 in the OPL in which HRG4 is localized, a decreased level also appeared in the transgenic retina internal to the OPL—that is, the INL, the IPL, and the GC layers. A decrease in ARL2, however, was not present in the IS, in which ARL2 and HRG4 are also present. The ARL2 decrease in the whole internal retina starting with the OPL, combined with the significant thinning of the retina from degeneration, may explain the approximately 70% decrease in ARL2 seen in the Western blot analysis of total retinal protein. A specific decrease in ARL2 was supported by the standardization against two other photoreceptor proteins, rhodopsin and S-antigen. The mechanism by which this occurs, while not entirely clear, may be attributed to the significant transsynaptic degeneration that has already been documented in this model. ⁷ It is unknown why a decrease in ARL2 was not present in the IS, but this may reflect a spatial difference in mechanisms. 

The function of ARLs is still largely unknown. Recently, it has been shown that GTP-bound ARL2 binds the protein BART. The ARL2–BART complex enters the mitochondria and interacts with ANT-1, a mitochondrial inner membrane protein. ^{16 17} Analysis of BART and ANT-1 in the transgenic model further revealed changes in BART and ANT-1 in the older transgenic mice. By 20 months of age, when ARL2 was significantly decreased, the level of BART in the transgenic retina was significantly increased, suggesting a compensatory response of BART to the decreasing level of ARL2. Most significantly, there was a substantial decrease in ANT-1, a protein of the inner mitochondrial membrane, in the retinas of older (20 months) transgenic mice. Interestingly again, in addition to the prominent decrease in the OPL, a decrease in ANT-1 was also present in the whole transgenic inner retina but not in IS. The same explanation given for ARL2 most likely also applies for ANT-1, and the underlying mechanism may be transsynaptic degeneration. Further studies will be required to determine how a disturbance in the homeostasis of the HRG4-ARL2-BART-ANT-1 interactions could result in the observed ANT-1 decrease, though it is reasonable to predict that less of the ARL2-BART complex might be formed and therefore available to enter the mitochondria and interact with ANT-1 as a result of interference by the truncated HRG4 protein, with its threefold affinity for ARL2. 

Regardless of its mechanism of occurrence, the reduction in ANT-1 revealed by the model provided the impetus to turn our attention to mitochondria and the mechanism of apoptosis in this model. ANT-1 is a key mitochondrial protein that plays at least a dual role in these organelles, first by exchanging ATP for ADP between the mitochondria and the cytoplasm, thereby controlling the level of ATP in the cytoplasm and second by participating in apoptosis as a component of the mitochondrial permeability transition pore, ^{18 19 20} though this role of ANT-1 has been questioned recently. ^{21 22} The functional significance of the interaction between the ARL2–BART complex and ANT-1 is unknown, ¹⁷ but the observed decrease in ANT-1 of approximately 50% in the transgenic retina is significant. An ANT-1 defect is known to cause myopathy in humans. ²⁴ In an animal model of myopathy, a knockout of ANT-1 has been shown to result in uncoupling of mitochondrial respiration, depletion of cytoplasmic ATP, increase in reactive oxygen species, and serious problems in cellular energetics. ^{25 26 27} In the knockout model, the ARL2 level was increased, confirming a functional relationship between these proteins, as also demonstrated in our transgenic model. Interestingly, in our HRG4-mediated transgenic model, we observed the opposite phenomenon—a decrease in ANT-1 in the face of a decrease in ARL2. Human ANT-1 mutants are known to cause another problem in the eye, progressive external ophthalmoplegia. Analysis of expression of homologues of these ANT-1 mutants in yeast revealed marked growth defects, reduced amounts of various mitochondrial respiratory proteins and cellular respiration, defects in ADP compared with ATP transport, and mitochondrial DNA damage. ²⁸ In one of the mutants, a marked reduction in ATP transport was shown to be caused by a decrease in the amount of homologous ANT-1 protein. Thus, the observed decrease in ANT-1 of approximately 50% in the older transgenic retina would be expected to lead to significant problems involving mitochondrial respiratory uncoupling and disturbance in ATP–ADP exchange. This was consistent with the apparent proliferation of mitochondria, a known reaction to mitochondrial dysfunction, ²⁶ observed in the photoreceptor synapses as assessed by cytochrome c immunofluorescence. Mitochondrial dysfunction was shown to be profound and was accompanied by the release of cytochrome c into the photoreceptor synaptoplasm, as confirmed by the activation of caspases, ²⁹ including caspase 3, and ultimately by the apoptosis of photoreceptors and some INL cells, as demonstrated by TUNEL analysis, an end-stage marker of apoptosis revealing fragmentation of DNA in a dying nucleus. 

Thus, a likely mechanism of apoptosis in the model includes the disturbance in ATP–ADP exchange, disruption in the mitochondrial respiratory chain, and production of reactive oxygen species, all of which are supported by the demonstrated mitochondrial proliferation and release of cytochrome c from the mitochondria, confirmed by caspase 3 activation, and all likely consequences of the decrease in ANT-1. Such changes have been shown to lead to apoptosis. ^{27 30 31 32 33 34 35} Alternatively, though the role of ANT-1 in mitochondrial permeability transition has become controversial, ^{21 22} the decrease in ANT-1 may directly result in release of proapoptotic proteins, including cytochrome c, into the synaptoplasm by an unknown effect on the permeability transition pore. ^{29 36} In support, ANT-1 deficiency was recently reported to increase the sensitivity of hepatocytes to calcium-mediated mitochondrial permeability transition. ²¹ ANT-1 depletion and exposure to high calcium concentration in the transgenic photoreceptor synapse may result in permeability transition, release of cytochrome c into the cytoplasm, and apoptosis, as observed. 

This transgenic model represents an example of a localized degeneration that first occurs in a neuronal synapse, with release of pro-apoptotic proteins that subsequently find their way to the cell body, ultimately killing the whole neuron by apoptosis, as previously described. ³⁷ A photoreceptor synaptic defect was considered a possibility in the patient with the cone–rod dystrophy with the HRG4 mutation in light of supernormal findings on scotopic electroretinography that extended into the lower intensity stimuli and that was found in the daughter carrying the mutation, along with severe photophobia in both patient and daughter (Weleber R, personal communication, October 2005). In the HRG4 transgenic model, the apoptosis-inducing pathogenic mutant protein HRG4 is uniquely enriched in the photoreceptor synapses and executes its action in the synapses. Apoptosis was observed in the transgenic photoreceptor cell bodies accompanied by photoreceptor synapses that were filled with cytochrome c and were undergoing degeneration themselves. Initial preapoptotic degeneration in the synapse was confirmed by the demonstration of cytochrome c leakage and caspase 3 activation in a synapse without evidence of a TUNEL-positive apoptotic photoreceptor cell body nearby. The time interval between synaptic degeneration and cell body apoptosis must have been short because few degenerating synapses without accompanying apoptosis in the cell body could be found. Transsynaptic degeneration was demonstrated by the activation of caspase 3 in the INL cell associated with this synapse. The observed transsynaptic effect might have occurred because of an excitotoxic effect of release of a large amount of glutamate from the degenerating photoreceptor synapses or because of a more long-term effect of lack of afferent stimulation or important neurotrophins originating from the synapses. ³⁸  

The pattern of retinal degeneration observed in the older transgenics, consisting of structural abnormalities of the ONL without much thinning of the nuclear layers, was characteristic of this model. What appeared to be localized collapses of the ONL into the inner segment zone might have been the result of untethering of photoreceptor synapses in the OPL caused by degeneration of the synapses, an early step in pathogenesis. The late-onset nature of actual neuronal death in this model was clearly demonstrated by the significant loss of nuclei in the ONL and the INL at 20 months, but not at younger ages (6 and 12 months), and by the ability to detect measurable numbers of TUNEL-positive nuclei only at 20 months. The pathogenic mechanism, presumably involving an abnormal truncated HRG4-ARL2 interaction, is believed to be operative from birth. This prompts the question why it takes so long for the retinal degeneration—and the observed imbalances in the ARL2, BART, and ANT-1 levels—to occur. One possible explanation, demonstrated by the expression analysis of ANT-1 mutants in yeast, is that mitochondrial DNA deletions induced by reactive oxygen species, one of the consequences of the ANT-1 defect, accumulate slowly over time. ²⁸ As postulated to explain the slowly progressive nature of progressive external ophthalmoplegia, a mechanism involving mitochondrial DNA damage might also explain the late onset of degeneration in the mutant HRG4 transgenic mouse and the patient. 

³  These authors contributed equally to the work presented here and should therefore be regarded as equivalent authors.

 Supported by National Institutes of Health Grant EY10848, Foundation Fighting Blindness, and Research to Prevent Blindness.

 Submitted for publication April 21, 2005; revised September 26 and November 23, 2005; accepted February 22, 2006.

 Disclosure: N. Mori, None; Y. Ishiba, None; S. Kubota, None; A. Kobayashi, None; T. Higashide, None; M.J. McLaren, None; G. Inana, None

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Figure 1.

 

Saturation binding of recombinant full-length and truncated HRG4 to ARL2. Binding of increasing amounts of recombinant full-length (full hrg4) and proximal (prox hrg4) HRG4 to recombinant ARL2 is shown after nonlinear regression curve fitting of the results. For full-length HRG4, K _d = 101.6 ± 34.5 nM, B_max = 33.0 ± 5.3 nmol/mg protein. For proximal HRG4, K _d = 34.5 ± 11.1 nM, B_max = 40.0 ± 3.7 nmol/mg protein. Triplicates were run for each sample. Bar, SEM. Control recombinant GST showed minimal binding to ARL2.

Figure 2.

 

Saturation binding of recombinant full-length and truncated HRG4 to native ARL2 in retinal extract. Binding of increasing amounts of recombinant full-length (full hrg4) and proximal (prox hrg4) HRG4 to ARL2 in rat retinal extract is shown after nonlinear regression curve fitting of the results. For full-length HRG4, K _d = 97.4 ± 2.0 nM, B_max = 5.0 ± 0.5 pmol/mg protein. For proximal HRG4, K _d = 36.4 ± 3.5 nM, B_max = 4.0 ± 0.2 pmol/mg protein. Triplicates were run for each sample. Bar, SEM. Control recombinant GST showed minimal binding to ARL2.

Figure 3.

 

Inner and outer nuclear layer thickness of normal and transgenic retinas at different ages. Averages of INL and ONL thickness (in μm) in the peripheral (3 to 4 mm from ONH), mid (1 to 3 mm from ONH), and central (0 to 1 mm from ONH) zones of vertical meridian sections from 6 different 6-, 12-, and 20-month-old normal (N) and transgenic (Tg) retinas are shown with SEM. Significant decreases in nuclear layer thickness are observed at 20 months of age for ONL and INL, with the magnitude of decrease being periphery>central≫mid in the ONL and periphery≈mid≈central in the INL. No differences in the decreases in thickness were present between the superior and inferior regions of the vertical meridian sections examined. P values were calculated by the Student’s t-test.

Figure 4.

 

Histopathology and timing of apoptosis in retinal degeneration in the transgenic retina. (A) Histopathology of retinal degeneration in the 12-, 17-, and 20-month-old transgenic retina. Hematoxylin and eosin–stained sections from the mid retina of 12-, 17-, and 20-month-old normal (N) and transgenic (Tg) animals are shown. Arrows in the 12-month-old transgenic retina point to migrating photoreceptor nuclei. The 17-month-old transgenic retina shows focal protrusions of the ONL into the IS (arrow), consistent with collapse of the photoreceptors due to loss of anchoring at the synapses because of synaptic degeneration. The 20-month-old transgenic retina shows significant thinning of ONL and INL. Bar, 10 μm. (B) Frequency of apoptosis in the normal and transgenic retina at different ages. Average numbers of TUNEL-positive cells in a whole retinal section (vertical meridian section) from 4 different normal (N) and transgenic (Tg) animals of 6, 12, 17, and 20 months of age are shown with SEM. Significant increase in the number of TUNEL-positive cells was observed in the 20-month-old transgenic retinas. P value was calculated by the Student’s t-test.

Figure 5.

 

Levels of ARL2, BART, and ANT-1, standardized against S-antigen and rhodopsin, in normal and transgenic retinas of different ages. (A, B) ARL2 levels in normal and transgenic retinas of different ages. A representative result of a Western blot analysis of ARL2 in normal and transgenic retinas of 6-, 12-, and 20-month-old mice is shown at the top. Results were densitometrically quantitated and standardized against S-antigen (also done for BART and ANT-1) and rhodopsin (also done for ANT-1) levels for each sample, and averages of ARL2/S-antigen and ARL2/rhodopsin (arbitrary units) from 3 experiments are shown in bar graphs below with SEM. P value was calculated by the Student’s t-test. N, normal; Tg, transgenic. (C) Immunofluorescence of ARL2 in normal and transgenic retina. Immunofluorescence (FITC, green) of ARL2 in 20-month-old normal (N) and transgenic (Tg) retinas are shown. In the normal retina, the strongest immunofluorescence is seen in the OPL and weaker immunofluorescence is seen in the GC, IPL, INL, and IS. A significant decrease in immunofluorescence is seen in all the layers except the IS in the transgenic retina. ONL and INL in the transgenic retina are thinner because of retinal degeneration. Bar, 10 μm. (D) BART level in 20-month-old normal and transgenic retina. A representative result of a Western blot analysis of BART in normal and transgenic retinas of 20-month-old mice is shown at the top. Results were densitometrically quantitated, and averages of BART/S-antigen (arbitrary units) from 3 experiments are shown in a bar graph below with SEM. P value was calculated by the Student’s t-test. No difference was observed in BART level between normal and transgenic at 6 and 12 months of age. (E, F) ANT-1 level in 20-month-old normal and transgenic retina. A representative result of a Western blot analysis of ANT-1 in normal and transgenic retinas of 20-month-old mice is shown at the top. Results were densitometrically quantitated, and averages of ANT-1/S-antigen and ANT-1/rhodopsin (arbitrary units) from three experiments are shown in bar graphs below with SEM. P value was calculated by the Student’s t-test. No difference was observed in ANT-1 level between normal and transgenic at 6 and 12 months of age.

Figure 6.

 

Immunofluorescence of ANT-1 and HRG4 in mouse retina. (A) Double immunofluorescence of HRG4 and ANT-1 in normal mouse retina. HRG4 immunofluorescence (Alexa Fluor 488, green) is present in the OPL and IS, and ANT-1 immunofluorescence (Alexa Fluor 568, red) is present in the GC, IPL, parts of INL, OPL, and IS. In the OPL and the adjacent INL, the combined image shows colocalization of HRG4 and ANT-1 in the OPL (yellow) but localization of only ANT-1 in the outer region of INL (red) (combined image, upper panel; magnified images, lower panel). DIC, Nomarsky confocal image. Scale bar, 10 μm. (B) Immunofluorescence of ANT-1 in 20-month-old transgenic and normal retina. ANT-1 immunofluorescence (FITC, green) in the normal retina is as described in A, with the strongest signal in the OPL where the photoreceptor synapses are present and adjacent INL containing bipolar/horizontal cells and their dendrites (N, upper panel; magnified image, lower panel). In the transgenic retina, a significant decrease in immunofluorescence is seen in the OPL and adjacent INL in addition to the rest of the internal retina, but not in IS (Tg, upper panel; magnified image, lower panel). Scale bar, 10 μm. Control is without primary antibody.

Figure 7.

 

Cytochrome c and TUNEL staining in normal and transgenic retina. (A) Cytochrome c immunofluorescence in 20-month-old normal and transgenic retina. Immunofluorescence is significantly increased in the transgenic (Tg) IPL, the outer region of INL, and OPL compared to normal (N). Arrows point to photoreceptor synapses that appear to be filled with cytochrome c in the transgenic retina. (B) TUNEL staining in the 20-month-old normal and transgenic retina. Arrows point to photoreceptors undergoing apoptosis and positive for TUNEL staining (diaminobenzidine) in the transgenic (Tg) retina.

Figure 8.

 

Triple analysis of TUNEL staining and double immunofluorescence for cytocohrome c and activated caspase 3 in the 20-month-old normal and transgenic retina. DIC, Nomarsky confocal microscopy image with arrows pointing to darkened photoreceptor nuclei undergoing apoptosis and positive for TUNEL staining (diaminobenzidene) in the transgenic (Tg) retina. Caspase 3, immunofluorescence (Alexa Fluor 568, red) for activated caspase 3 with arrows showing signal in the photoreceptors indicated by arrows in the DIC image in the transgenic retina. Cytochrome c, immunofluorescence (Alexa Fluor 488, green) for cytochrome c showing increased signal in the transgenic (Tg) OPL. Arrowheads point to photoreceptor synapses that appear to be filled with cytochrome c. Combined, merged photograph of cytochrome c and activated caspase 3 immunofluorescence with arrows and arrowheads pointing to the described structures. Two of the four photoreceptors undergoing apoptosis (arrows) appear to be cone photoreceptors because of their location adjacent to the outer limiting membrane. Scale bar, 10 μm.

Figure 9.

 

Cytochrome c and activated caspase 3 double immunofluorescence in 20-month-old transgenic retina showing preapoptotic degeneration in photoreceptor synapse and adjacent INL cell. (A) Merged image of double immunofluorescence for cytochrome c (Alexa Fluor 488, green) and activated caspase 3 (Alexa Fluor 568, red). Arrow points to signal in a photoreceptor synapse (triangular) and INL cell (circular) associated with it. (B–D) Photoreceptor synapse (triangular) positive for cytochrome c (cyto C) and activated caspase 3 (casp 3) immunofluorescence and merged image (combined), consistent with cytochrome c release from mitochondria and preapoptotic degeneration in this synapse. (E–G) INL cell (circular), negative for cytochrome c immunofluorescence (cyto C), positive for activated caspase 3 immunofluorescence (casp 3), and merged image (combined), consistent with apoptosis in this cell. Arrows in E point to the edge of the photoreceptor synapse, positive for cytochrome c immunofluorescence and shown in B–D, which appears to be associated with this INL cell. Confocal microscopy images B–D and E–G are 2 μm apart, consistent with the images showing the same synapse and INL cell. Images are consistent with a direct demonstration of transsynaptic degeneration with the synapse undergoing preapoptotic degeneration, inducing apoptosis in the INL cell associated with it.