Intrinsic differences in rod and cone membrane composition: implications for cone degeneration

Abstract

Purpose

In many retinal pathological conditions, rod and cone degeneration differs. For example, the early-onset maculopathy Stargardts disease type 1 (STGD1) is typified by loss of cones while rods are often less affected. We wanted to examine whether there exist intrinsic membrane differences between rods and cones that might explain such features.

Methods

Abca4 mRNA and protein levels were quantified in rod- and cone-enriched samples from wild-type and Nrl^−/− mice retinas; rod- and cone-enriched outer segments (ROS and COS respectively) were prepared from pig retinas, and total lipids were analyzed by flame ionization, chromatography, and tandem mass spectrometry. Immunohistochemical staining of cone-rich rodent Arvicanthis ansorgei retinas was conducted, and ultra-high performance liquid chromatography of lipid species in porcine ROS and COS was performed.

Results

Abca4 mRNA and Abca4 protein content was significantly higher (50–300%) in cone compared to rod-enriched samples. ROS and COS displayed dramatic differences in several lipids, including very long chain poly-unsaturated fatty acids (VLC-PUFAs), especially docosahexaenoic acid (DHA, 22:6n-3): ROS 20.6% DHA, COS 3.3% (p < 0.001). VLC-PUFAs (> 50 total carbons) were virtually absent from COS. COS were impoverished (> 6× less) in phosphatidylethanolamine compared to ROS. ELOVL4 (“ELOngation of Very Long chain fatty acids 4”) antibody labelled Arvicanthis cones only very weakly compared to rods. Finally, there were large amounts (905 a.u.) of the bisretinoid A2PE in ROS, whereas it was much lower (121 a.u., ∼ 7.5-fold less) in COS fractions. In contrast, COS contained fivefold higher amounts of all-trans-retinal dimer (115 a.u. compared to 22 a.u. in rods).

Conclusions

Compared to rods, cones expressed higher levels of Abca4 mRNA and Abca4 protein, were highly impoverished in PUFA (especially DHA) and phosphatidylethanolamine, and contained significant amounts of all-trans-retinal dimer. Based on these and other data, we propose that in contrast to rods, cones are preferentially vulnerable to stress and may die through direct cellular toxicity in pathologies such as STGD1.

Introduction

Photoreceptor (PR) damage, malfunctioning and death occur in a large number of visual pathologies. Whether the underlying causes are genetic, chemical, or environmental, progressive loss of these cells leads to visual handicap and incapacitating blindness. The death of cones is particularly devastating since these cells are responsible for high acuity and chromatic daylight vision [1]. Cones reach maximal density within the fovea, and this region is especially vulnerable in widespread diseases such as age-related macular degeneration and autosomal recessive maculopathy Stargardt disease type 1 (STGD1). In the latter, mutations in the gene ABCA4 encoding a PR-specific flippase lead to retinal degeneration of variable phenotype, but in which early foveal damage and severe loss of central vision are commonly observed [2, 3]. Most ABCA4 mutations lead to a central-peripheral gradient of retinal damage, almost always with macular involvement [4, 5]. ABCA4 protein is localized to both rod and cone outer segment (ROS and COS respectively) disk edges/ membrane folds [6]. Even in the heterogeneous family of diseases known as retinitis pigmentosa, where genetic mutations are often exclusively in rods and initial symptoms are night blindness due to rod loss, there is secondary cone death leading to eventual loss of central vision [7]. It is not clear why such regional and cell type-preferred differences in PR loss occur in the face of generalized insults. Macular vulnerability could be related to specific structural and functional modifications in this area, such as differences in vascularization [8], Bruch’s membrane thickness [9], increased light exposure [10], cone density [11], or increased phagocytic load because of the higher PR/retinal pigmented epithelium (RPE) ratio [12]. But it has not been generally considered that some intrinsic feature of cones could render them differentially sensitive to generalized insults. Indeed, pan-retinal cone malfunction (while rod function remains normal) is observed in STGD1 patients [3], suggesting that this might be the case. Furthermore, with the G1961E mutation [13], quantitative fundus auto-fluorescence (AF, a non-invasive measure of bisretinoid) is not elevated when measured 7–9° outside the fovea while with other mutations there is a retina-wide increase [14].

With respect to STGD1, a hypothetical scheme has been proposed to account for PR death based upon seminal studies using genetically modified mouse strains [15, 16], biochemical analyses of retinal and RPE preparations [17] and rigorous in vitro experiments [18]. Under normal conditions, following light activation of rhodopsin, all-trans retinal (atR) is released within the lumen of disc membranes. There it reacts with phosphatidyl-ethanolamine (PE) (1:1 ratio) in disc membranes to form reversible N-retinylidene-PE (NRPE), which is rapidly transported into the OS cytosol by the flippase ABCA4 localized in the disc periphery. Both all-trans- and 11-cis retinal can form NRPE. In the cytosol, atR is reduced by retinol dehydrogenase, and then transported across the sub-retinal space to the adjacent RPE where it completes re-isomerization through the visual cycle. The active 11-cis retinal is finally shipped back to the PR to restore light sensitivity. In STGD1, mutations in the gene coding for ABCA4 lead to reduced transport efficiency; atR accumulates within the disc lumen where it undergoes reactions with PE in 2:1 ratio to form a complex mixture of irreversible bisretinoid adducts. PR undergo continuous membrane turnover involving shuttling of discs towards the apex and rhythmic phagocytosis by the adjacent RPE. This latter cell type hence accumulates these toxic bisretinoids which cannot be metabolized, giving rise to bisretinoid fluorophores that include but are not limited to A2E. Bisretinoids are toxic bi-products of the visual cycle, exhibiting both photoreactive and detergent properties [19]. Bisretinoids form AF lipofuscin deposits in RPE, and in STGD1, they are visible as conspicuous AF flecks by ophthalmological examination [20]. The accelerated build-up of lipofuscin is a hallmark of the disease in humans and mouse models [15–20]. Eventually the RPE can no longer assure critical maintenance of the PR, which then die. Experimental data for this model have been obtained from rods [15–17] but is tacitly considered to also be true for cones. However, direct experimental evidence for cones following the same pathogenic pathway is lacking, and some studies have suggested cone death may precede loss of RPE [5, 21, 22].

We decided to determine which biochemical characteristics relevant to the visual cycle might be similar or different between rods and cones, and may hence influence PR degeneration. We specifically explored the following questions, using three different animal models: (1) are there differences in Abca4 mRNA/protein levels between rods and cones? If true, this might predispose one or other cell type to damage from Abca4 mutations. (2) Are there differences in major lipid species (abundance and type) between rods and cones? If true, this could predispose one or other cell type to oxidative damage. (3) Are there differences in bisretinoid adducts between rods and cones? Based on the results and other available data, we present a novel hypothesis to suggest cone death in STGD1 may occur by direct toxicity, and perhaps in other pathological scenarios as well.

Materials and methods

Animals

Three animal models were used in the present study. All animal experimentation was done in accordance with the recommendations outlined in the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. Animal use in the present project was authorized by the local ethics committee (CREMEAS) and by APAFIS#8472–2016121318254073 v10 from the Ministry of Research. Breeding pairs of heterozygous Neural Retina Leucine zipper (Nrl^−/+) mice on a C57BL/6J wild-type (WT) background were used to generate WT and Nrl^−/− littermates [23]. Animals were housed at the institute animal facility, Chronobiotron CNRS UMR3415, in environmentally enriched cages and with access to food and water ad libitum. When ready for analysis (3 months), mice were genotyped from tail samples and euthanized by excess CO₂ inhalation according to institutional guidelines.

Arvicanthis ansorgei were originally imported from Mali and have since been maintained as a viable breeding colony in the Chronobiotron. Individual young adult (3 month) animals were euthanized as described above, their eyes rapidly enucleated and fixed for immunohistochemistry as described below.

Adult domestic pig (Sus scrufulus) eyes were obtained fresh from a local slaughterhouse (CopVial SA, Holtzheim) and kept on crushed ice during transport to the laboratory and dissection (∼ 2–3 h). Further details are given below.

Quantification of Abca4 mRNA

ABCA4 is known to be expressed by both rods and cones [6], yet there are no data on the relative expression levels between the two types. Any comparison of the two must control for differences such as sample purity, cell size, OS size, and sub-cellular distribution. In order to quantify Abca4 mRNA, we used two strains of mice, 3-month-old C57BL/6J wild-type (WT) littermates and 3-month-old Neural Retina Leucine (Nrl) homozygous null mice [23]. Whereas WT possess highly rod-dominated retinas (∼ 97% rods, < 3% cones) [24], the absence of the rodspecific transcription factor Nrl leads to differentiation of photoreceptor precursors into short wavelength-sensitive (“blue”) cones instead of rods [23], and hence a 100% cone phenotype. We used a vibratome sectioning technique previously described by us [25] to horizontally split retinas into two halves, outer retina (OR, containing the ONL, IS and OS) and inner retina (IR, containing the combined INL and GCL) (Supplementary Fig. 1). Initial controls for determining the cutting depth and judging the purity of respective fractions were performed by microscopic analysis and immunohistochemical staining for the presence of rhodopsin for WT OR (or presence of SW cone opsin for Nrl^−/− OR) and its absence from IR; and the presence of PKCa and neurofilament in IR and their absence from OR (Supplementary Fig. 1). Total protein values were also obtained for representative OR and IR fractions using the Bradford colorimetric assay. WT OR and IR (n = 10 each) and Nrl^−/− OR and IR (n = 8 each) were placed in individual Eppendorf tubes™ and snap frozen in liquid nitrogen until ready for use. For RT-PCR analysis, only OR fractions were used (IR samples were used for lipid measures, described below).

Retinal RNA was isolated with RNeasy mini kit (Qiagen), following the manufacturer’s instructions including DNAse treatment. We modified slightly the first step, as dissociation of tissues was performed using a pestle with 500 μL of TriReagent solution. One hundred microliters of chloroform was directly added, and the preparation was transferred to a phase-lock gel tube after 15-min incubation at room temperature. Then, this gel tube was centrifuged 15 min at 12,000g and 4 °C. To recover RNA, the supernatant was added to an equal volume of 70% ethanol. RT-PCR was performed from 500 ng of RNA with the high capacity RNA-to-cDNA kit (Applied Biosystems), following the manufacturer’s instructions. A second negative control was added, consisting of water, enzyme and RT buffer only, to confirm the absence of contamination. QPCR was performed in 96-well plates (Applied Biosystems), following the manufacturer’s instructions. Experiments were carried out on a 7300 Real-Time PCR System (Applied Biosystems) with a first step of denaturation for 5 min at 95 °C, followed by a cycle of 40 repetitions of 15 s at 95 °C, then 1 min at 60 °C. After estimation of the sample concentration to use (between 1/10 and 1/100) and validation of primer efficiency (> 80%), each sample was processed in duplicate and using a dilution range, for all studied genes: genes common to both PR types [Rom1, arrestin (Sag), rds/peripherin (Prph2), Abca4], rod-specific (Gnat1), cone-specific (Gnat2), and housekeeping genes (Hprt). See Supplementary Table 1 for TaqMan references (Applied Biosystems). Analysis was performed using the ΔΔCt method.

Quantification of Abca4 protein

The same mouse model was used to compare abca4 protein concentration by western blotting in ROS and COS. Protein extraction was performed by sonicating each retina in lysis buffer (20 mM Tris-HCl pH7.6, 150 mM NaCl, 1% Triton X-100, 1 mM EDTA, 0.2% SDS). The lysate obtained was centrifuged 20 min at 13,000 rpm and 4 °C, and the supernatant protein concentration was estimated by the Bradford method. Samples were then diluted to 1 mg/mL by addition of loading buffer containing 10% β-mercaptoethanol and denaturated 5 min at 95 °C. Samples (20 μg protein/lane for WT, 40 μg/lane for Nrl^−/−) were loaded and separated on SDS-PAGE gels, 90 min at 110 V, followed by wet transfer to PVDF membranes, 60 min at 90V. After protein transfer, membranes were incubated in blocking solution (20 mM Tris-HCl pH 7.6, 150 mM NaCl, 0.1% Tween 20, 5% fat-free skimmed milk), 1 h at room temperature before overnight incubation at 4 °C with the primary antibodies. To compensate for subcellular differences between ROS and COS (Nrl^−/− cones are short and stunted, abca4 protein is restricted to disc rims), we normalized using antibodies directed against two other OS-specific membrane proteins localized to this rim region: rds/Prph [26] and rom1 [27]. After extensive washing in buffer, 4 × 5 min, membranes were incubated with secondary antibodies (rabbit anti-mouse IgG-HRP for monoclonal primaries, goat anti-rabbit IgG-HRP for polyclonal primaries), diluted 1:10,000 for 2 h at room temperature. After washing again, 4 × 5 min, membranes were developed in the dark with the ECL Western Blotting Detection kit (Pierce), following the manufacturer’s instructions. Exposure times for each blot ranged between 5 and 30 s to obtain linear densitometric readings.

Photoreceptor outer segment fractionation

In order to compare rod and cone membrane composition, it was necessary to obtain purified or enriched preparations of each. Procedures were developed previously, based upon homogenization and flotation of retinal samples on sucrose gradients, which enable the preparation of highly enriched ROS fractions from bovine retinas [28]. The availability of such pure fractions has allowed detailed analysis of membrane composition in terms of lipids [29] and proteins [30]. In contrast to rods, similar approaches have not been explored for COS, the reasons being they are much less numerous in many mammalian retinas (∼ 20-fold difference in humans) [31], and are also shorter, more deeply embedded and surrounded by an insoluble extracellular matrix [32]. We modified the protocols used for ROS preparation to permit enrichment of COS membranes, using an animal model better suited to such objectives. Pigs (Sus scrufulus) have large retinas roughly equivalent in size to those of humans, are available in large quantities from slaughterhouses, and have a relatively elevated percentage of cones (15–20%, even higher in the visual streak) [33] in comparison to the cow (∼ 5%). For each individual experiment, 60 adult pig eyes were used. All the following steps were done at 4 °C: after discarding surrounding connective tissues, eyes were opened by circumferential cuts around the ora serrata to remove the cornea and vitreous. Retinas were gently detached from the back of the eyecup with a spatula, cut at the optic nerve and placed in cold sodium chloride (0.9%). Tissues were stirred gently using a magnetic stirrer for 20 min, allowed to settle for 5 min, and then, the crude supernatant was collected and filtered through a 70 μm mesh cell strainer. Twenty milliliters cold NaCl was added to the remaining retinas, and they were stirred again for 15 min and the filtration repeated as above. The supernatant was pooled with the first one, then poured into six ultracentrifugation tubes (Ultra-Clear Centrifuge Tubes 17 ml, Beckman). The suspension was centrifuged at 25,000 rpm for 1 h using a Beckman ultracentrifuge equipped with swinging bucket rotor (Beckman Coulter, SW32Ti). The supernatant was discarded, and the pellets were resuspended in a total of 6 ml 20% sucrose buffer (20 mM Tris-HCl, 2 mM MgCl₂, 0.13 NaCl, pH 7.2). Using clear ultracentrifuge tubes, 1 ml of the suspension was loaded on top of a discontinuous sucrose gradient (27%, 2 ml; 40%, 4 ml; 45%, 4 ml; 47%, 4 ml; 50%, 1 ml: total 15 ml, each prepared in the same Tris buffer; these concentrations were determined empirically during pilot experiments) and centrifuged at 32,000 rpm for 2 h at 4 °C. After centrifugation, material in the tubes was separated into four bands, denoted fractions (F)1–4 counting from the surface, as well as a pellet. All four bands were aspirated individually and placed in Oak Ridge centrifuge tubes 50 ml (Nalgene). The tubes were filled with phosphate buffered saline (PBS) and centrifuged for 20 min at 12,000 rpm with a fixed angle rotor (Sorvall, SA-600). Pellets were resuspended in 1 ml PBS. 50 μl were collected and mixed with 50 μl of extraction buffer containing protease and phosphatase inhibitors before freezing at − 20 °C for western blotting analysis. Western blotting was done as described above. One to 15 μg total ROS/COS protein were loaded on each lane of a 10% SDS-PAGE and allowed to migrate for 1–1.5 h. Proteins were electro-transferred to PVDF membranes, and then blocked in 0.1 M Tris-HCl pH 7.6 containing 5% non-fat milk powder prior to incubation with primary antibodies: rho-4D2 (anti-rhodopsin, specific for ROS: 28) or cone MW opsin antibody (AB5405, Merck Millipore; specific for COS: 29). Fractions were also probed with mitochondrial antibody (MTC02, Abcam, Paris, France). The remaining samples were centrifuged in 1.5 ml Eppendorf tubes for 20 min at 13,200 rpm; the supernatants were discarded and the pellets kept at − 80 °C.

A mitochondria isolation kit (Thermo Scientific, Waltham, MA, USA) was used on F3 to reduce contamination. According to the manufacturer’s instructions, we chose the reagent-based method for soft tissues, omitting the tissue disruption steps. The pellet was resuspended in 500 μl BupH™ PBS and centrifuged at 1000g for 3 min. Supernatant was discarded and 800 μl of bovine serum albumin (BSA)/reagent A solution was added. After vortexing, the solution was left on ice for 2 min. Ten microliters of mitochondria isolation reagent B was added and incubated for 5 min, vortexing at maximum speed every minute. Eight hundred microliters of reagent C was added and gently homogenized before centrifuging at 700g for 10 min. The pellet was designated “F3 mitochondria-depleted fraction” (F3’). After resuspending in 500 μl PBS, a small aliquot was removed for western blotting analysis and the rest was stored at − 20 °C. The supernatant was transferred to a new Eppendorf tube and centrifuged at 13,200 rpm for 15 min to obtain an enriched mitochondrial pellet.

Electron microscopy

Retinal samples (RET, F2 and F3’ fractions) were fixed in 2% paraformaldehyde and 0.1% glutaraldehyde in PBS for 2h at 4 °C, rinsed in buffer, dehydrated and prepared for colloidal gold immuno-electron microscopy using LR White™ resin [31]. Ultra-thin sections were obtained and processed for double immunolabelling with primary antibodies (monoclonal rho-4D2 and polyclonal cone MW opsin) [34, 35] followed by secondary antibodies (goat anti-mouse IgG and goat anti-rabbit IgG coupled respectively to 5 and 15 nm colloidal gold. After rinsing, sections were briefly refixed in 2.5% glutaraldehyde for 10 min prior to observation by transmission electron microscopy (Hitachi TEM).

General lipid analysis

The distribution of lipids into phospholipids (PL), free FA, and cholesterol esters (CL) in the two fractions, mitochondrial samples (M) and whole pig retina (RET) was determined using a combination of thin-layer chromatography on silica gel-coated quartz rods and flame ionization detection (Iatroscan® system, Iatron, Tokyo, Japan) [33]. The values obtained for each compound were corrected according to their response factor using specific calibration curves, as published by our laboratory [36]. Data are given as the percentage of total lipids per sample. Similar analyses were performed for WT and Nrl^−/− mice retinas divided into OR and IR fractions.

Fatty acid analysis

Lipids were extracted from the different fractions (F2, F3/F3’, M, RET) exactly as according to our previous publications [36, 37] based on the original Folch method [38]. PL were purified from total lipid extracts using silica cartridges, and total PL were transmethylated using boron trifluoride in methanol according to Morrison and Smith [39]. FA methyl esters (FAMEs) and dimethylacetals (DMAs) were identified by comparison with commercial and synthetic standards. The data were processed using the EZChrom Elite software (Agilent Technologies, Massy, France) and reported as the percentage of total FA. The structural analysis of PL was performed according to previously described procedures [36, 37 and references therein]. Similar analyses were performed for WT and Nrl^−/− mice retinas divided into OR and IR fractions.

Immunohistochemistry

Immunohistochemical staining of frozen retinal sections of cone-rich diurnal rodents Arvicanthis ansorgei was performed as described previously by us [40]. After blocking non-specific binding sites with PBS containing 0.1% BSA and 0.1% Tween20 for 30 min, sections were incubated in a mixture of rabbit polyclonal anti-ELOVL4 (ELOngation of Very Long chain fatty acids 4) antibody (generous gift of Dr. M-P. Agbaga, Dept. Ophthalmology, Univ. Oklahoma Health Sciences Center, Oklahoma City, OK, USA) [41] diluted 1:1000, and mouse monoclonal anti-phosducin (BD Transduction Labs., Erembodegem, Belgium) diluted 1:100, overnight at 4°C. Parallel sections were incubated in anti-rhodopsin monoclonal antibody rho-4D2 [34] and anti-cone mid-wavelength sensitive (red/green) opsin polyclonal antibody (AB5405, Merck Millipore, Molsheim, France) [35], 1 μg/ml. Sections were rinsed thoroughly in PBS and incubated in a cocktail of goat anti-rabbit IgG-Alexa594 (for ELOVL4 and cone arrestin antibodies), rabbit anti-mouse IgG-Alexa488 (for phosducin and rhodopsin antibodies) and 4′,6-diamidino-2-phenylindole (DAPI) (each 1 μg/ml in blocking buffer) for 2 h at room temperature. Slides were washed thoroughly, coverslipped and examined under an Optiphot 2 fluorescence microscope equipped with image analysis software (BIR, Nikon).

UPLC-MS analysis of bisretinoids in ROS and COS

Representative samples of porcine ROS and COS were sent for analysis by ultra-high performance liquid chromatography (UPLC) coupled with mass spectrometry (MS). The recipient (JRS) was blinded to the identity of the samples. Using previously published methods [42], fractions were run on a C4 BEH300 column (2.1 × 100 mm), solvent A water/acetonitrile 1:1 + 0.1% formic acid; solvent B isopropanol:acetonitrile 9:1 + 0.1% formic acid using a time gradient (in min) of 0 (99% A), 5 (99% A), 15 (85% A), 35 (65% A), and 55 (40% A). Components were identifed and quantifed as the absorbance peak area at 450 nm divided by total protein mass.

Results

Mouse cones contain higher amounts of Abca4 mRNA and protein than rods

Expression levels of abca4 mRNA in WT and Nrl^−/− retinas, quantified by qPCR, were plotted against expression levels of other relevant rod and/or cone genes. As predicted Gnat1 mRNA expression was exclusive to WT samples (Nrl^−/− retinas contain no rods), and Gnat2 mRNA was ∼ 6-fold greater in Nrl^−/− retinas, the latter reflecting the presence of small numbers of cones in WT retinas. When quantified individually, all photoreceptor genes showed significantly higher values in WT compared to Nrl^−/− retinas: abca4, 2.4-fold greater, Prph2 4-fold greater, Rom1 4-fold greater, SAg 0.7-fold greater (Fig. 1); only the housekeeping gene Hprt was equal in both (Supplemental data 1). But cross-comparison between abca4 and either Prph2 or Rom1 showed that relative expression of abca4 mRNA was significantly higher in Nrl^−/− retinas: abca4 versus Prph2, 1.8-fold higher, abca4 versus Rom1, 1.6-fold higher. On the contrary, relative expression of abca4 versus SAg was slightly higher in WT retinas (∼ 1.2-fold higher) (Fig. 1). In summary, compared to the other two rim proteins, abca4 mRNA was more abundant in cones. Detailed data of RT-PCR analyses are shown in Supplementary Table 2.

Fig. 1.

Quantification of relative mRNA expression levels in rod- and cone-enriched retinal samples. mRNA expression levels of candidate photoreceptor genes from rod and/or cone PR were quantified by qPCR according to the Methods section. Gnat1 (A) and Gnat2 (B) being expressed either uniquely in rods (Gnat1) or cones (Gnat2) were used to ensure the validity of the approach,. As predicted, Gnat1 was only present in isolated WT mouse PR, whereas Gnat2 was very highly enriched in Nrl^−/− mouse PR. All four genes shared by rods and cones (C Abca4, D Rds/prph2, E Rom1, and F SAg) were more highly expressed in WT compared to Nrl^−/−, reflecting the larger ONL. But when abca4 values were normalized against either Rds/Prph2 (G) or Rom1 (H), levels were significantly higher in cone specimens. On the other hand, abca4 compared to SAg (I) remained preponderant in rods, arguing against the profiles being artefactual

Similar results were echoed in analyses of protein samples isolated from WT and Nrl^−/− retinas. Immunoblots of samples from WT and Nrl^−/− retinas run side by side and probed with anti-abca4, anti-rds/Prph or anti-Rom1 all showed higher absolute levels in WT due to the greater proportion of PR. But when abca4 expression levels, measured by densitometry, were normalized to the other two rim proteins, again Nrl^−/− retinas contained relatively higher levels of abca4: abca4 versus rds/Prph, 3.93-fold increase; abca4 versus Rom1, 1.68-fold increase (Fig. 2). In summary, compared to the other two rim proteins, abca4 protein was more abundant in cones.

Fig.2.

Quantification of relative protein expression levels in rod- and cone-enriched retinal samples. Western blots were probed with specific antibodies against the disk rim-specific proteins Abca4 (A), Rds/Prph2 (B), and Rom1 (C), along with β-actin as loading control, for both WT and Nrl^−/− fractions. Blots were exposed for different times to obtain densitometric values within the linear range for each antibody and each specimen, and calculated as a ratio of staining intensity. When adjusted in this way, abca4 levels in cones always exceeded those for Rds/Prph2 and Rom1 (D). Independent blots performed a minimum of n = 3

Verification of porcine rod and cone membrane fraction purity

A representative image of the bands separated by sucrose density gradients is shown in Fig. 3A. Initial experiments showed that the ROS and COS were concentrated in F2 and F3/F3’, which were subjected to western blotting and immunogold electron microscopy to estimate purity. We initially verified that the rhodopsin and cone opsin antibodies cross-reacted and were specific for porcine ROS and COS respectively. By immunoelectron microscopy, rho-4D2 was highly specific for ROS (Fig. 3B), whereas anti-cone MW opsin stained uniquely COS (Fig. 3C). Western blotting of F2 with rho-4D2 showed prominent immunoreactive bands at ∼ 34, 70, and 100 kDa and higher, corresponding to the monomeric and multimeric forms of rhodopsin (Fig. 3D). Rho-4D2 staining of F3 revealed only a very faint band. On the contrary, immunoblotting with anti-cone MW opsin showed an intense single band at ∼ 43 kDa in F3 (and F3’), but very faint immunoreactivity in F2 (Fig. 3E). Immunoblotting with anti-mitochondria antibody MTC02 showed very weak staining of the F2 fraction, and moderate label of the F3 fraction (Fig. 3F). Hence, we further depleted F3 of mitochondria using a commercial kit: such fractions (F3’) no longer displayed MTC02 immunoreactivity. We performed immunogold electron microscopy of the fractions with the respective antibody markers. F2 fractions were essentially composed of ROS, heavily decorated with colloidal gold (Fig. 3G). These samples were not stained with anti-MW cone opsin immunogold, or following omission of the primary antibody (not shown). Analogous experiments on F3’ fractions revealed a more heterogeneous profile, with large numbers of disrupted membrane profiles. Many of these membranous fragments (∼ 30–50%) were decorated by anticone MW opsin immunogold staining (Fig. 3H). Control sections lacked binding of gold particles. Based on these data, the terms “ROS” will designate F2, and “COS” will designate F3’.

Fig. 3.

Characterization of pig retinal rod and cone OS fractions by fluorescence immunohistochemistry, western blotting and immunogold electron microscopy. A Immediately following ultracentrifugation on discontinuous sucrose gradients, tubes contained a series of fractions (F) designated 1–4 from top to bottom. Each band was individually collected and analyzed by western blotting (see below) to determine rhodopsin/cone opsin content. Fractions 2 and 3 were processed further, fractions 1 and * (the latter contained mostly mitochondria, lysosomes, and nuclei) were discarded. B, C Validation of rod- or cone-specific antibody markers by immunogold electron microscopy. Prior to their use as ROS and COS markers, rho-4D2 (anti-rhodopsin) [28] and anti-cone MW opsin [29] were screened on ultra-thin sections of LR White-embedded pig retina. B Anti-rhodopsin-immunogold staining of in situ pig retina showed strong specific binding to ROS (arrow), with no labeling of rod inner segments (RIS) or COS (not seen). C Immunogold analyses using cone MW opsin antibody showed strong specific binding to COS and no labeling of cone IS (CIS) or ROS in in situ retina. Scale bars = 0.5 μm. D Western blots probed with anti-rhodopsin antibody (D) showed heavy staining of F2, visible as the monomeric (35 kDa) and dimeric (70 kDa) forms, but only very light staining of F3 and F4 fractions. The panel shows two independent experiments. E When similar staining was performed using anti-cone MW opsin, F2 was weakly labelled whereas F3 and F4 showed clear immunoreactivity visible as a single band ∼ 40 kDa. The panel shows two independent experiments. F F4 was discarded due to heavy mitochondrial contamination, but F3 was treated to reduce mitochondrial contamination (F3’), shown by immunoblotting using a mitochondrial antibody marker MCT02. A lightly stained band at 60 kDa is seen for F2, a much heavier one for F3, but following treatment (F3’) staining was not seen. G, H Immunoelectron microscopy of F2 and F3’ fractions: F2 fractions were heavily decorated by rho-4D2-immunogold particles (arrow) (G), whereas sections lacking primary antibody were devoid of label (not shown). H F3’ fractions showed strong staining of membranous profiles by cone opsin-immunogold, whereas other more electron dense material was unlabelled. Omission of primary antibody led to complete absence of staining (not shown). Scale bar in G = 0.5 μm

Pig cone membranes contain significantly lower levels of several major lipid species compared to rods

Analysis of total lipids from ROS and COS, as well as isolated retinal mitochondria and whole pig retina, revealed many highly significant differences. Cholesterol (CL) levels were over 3-fold higher in ROS compared to COS, expressed as both % total lipid and as a ratio of CL: PL (Fig. 4A, B, Supplementary Table 3). We concentrated our subsequent analyses on FA, and especially striking was the large difference between ROS and COS in the levels of the principal PUFA, 22:6n-3 (DHA): while this value was 20.6% total FA in ROS, it was only 3.3% in COS, almost seven fold lower (Fig. 4C, Table 1; Supplementary Table 4). This reduction was not due to mitochondrial contamination, since this fraction (MITO) contained 16.3% DHA, and whole retina (RET) had 16.2% DHA. There were also highly significant differences in the second most abundant PUFA, 20:4n-6 (arachidonic acid): 7.27% in ROS, 2.03% in COS, 8.3% in MITO and 8.96% in RET (Table 1; Supplementary Table 4). In accordance with these differences in PUFA, saturated and mono-unsaturated FA were both significantly higher in COS compared to ROS: saturated FA = 44.83% for ROS, 65.42% for COS (43.68% for MITO and 43.62% for RET); and mono-unsaturated FA = 16.18% for ROS, 23.66% for COS (20.69% for MITO and 28.33% for RET). We also calculated the mean ω3 and ω6 FA values, as well as the ratio of ω6/ω3: these were 21.02%, 12.97%, and 0.62 for ROS, and 3.47%, 4.48%, and 1.41 for COS, respectively, i.e., ω3 FA were the prevalent form in ROS, while ω6 FA were in excess in COS (Table 1; Supplementary Table 4).

Fig.4.

Cholesterol and fatty acid composition of pig retina fractions. Total lipids were extracted as described in the “Materials and methods” section. A Relative percentage of cholesterol (CL) in ROS, COS, mitochondria (M), and whole pig retina (RET). B CL/PL ration in ROS and COS. C DHA as a percentage of total FA in ROS, COS, M, and RET. ***p < 0.001 ROS compared to COS

Table 1.

The fatty acid composition of the four sample groups (ROS, COS, MITO, and RET) was determined by gas chromatography. FA species, individual fatty acid; the mean percentage of each fatty acid was determined from individual experiments (n = 9 for ROS and COS, 3 for MITO, and 8 for RET). The data for DHA (22:6n-3) are italicized. At the foot of the table are combined values for different FA groups: DMA, dimethyl acetals; MUFA, mono-unsaturated FA; PUFa, poly-unsaturated FA; SFA, saturated FA. Also given are Ω3 and Ω6 class FA. Data from this table were used to construct Fig. 4

FA species	ROS mean %	COS mean %	MITO mean %	RET mean %
12:0	0.02	0.09	0.02	0.00
14:0	0.28	0.34	0.30	0.31
15:0	0.09	0.15	0.10	0.11
dma16:0	1.44	0.86	1.44	2.07
16:0	19.01	26.90	19.37	21.63
16:1n-9	0.56	0.75	0.65	0.66
16:1n-7	0.36	0.74	0.58	0.48
17:0	0.28	0.45	0.25	0.28
dma18:0	3.13	1.73	2.72	3.27
dma18:1n-9	0.22	0.28	0.25	0.28
dma18:1n-7	0.17	0.12	0.19	0.23
18:0	24.06	35.31	23.27	20.98
18:1t	0.15	0.22	0.16	0.52
18:1n-9	11.81	16.62	14.72	13.82
18:1n-7	3.00	5.70	4.24	3.68
18:2n-6	0.64	1.16	1.04	0.98
20:0	0.23	0.25	0.18	0.20
18:3n-6	0.08	0.11	0.08	0.09
20:1n-9	0.21	0.29	0.27	0.22
18:3n-3	0.12	0.21	0.19	0.13
20:2n-6	0.12	0.19	0.13	0.12
20:3n-9	0.20	0.10	0.28	0.31
22:0	0.05	0.08	0.05	0.06
20:3n-6	0.22	0.15	0.27	0.25
22:1n-9	0.04	0.14	0.07	8.96
20:4n-6	7.36	2.43	8.30	0.12
24:0	0.11	0.09	0.14	0.05
22:4n-6	2.46	0.60	2.14	1.95
22:5n-6	2.24	0.43	1.76	1.52
22:5n-3	0.71	0.18	0.53	0.56
U22:6n-3	U20.61	U3.31	U16.31	U16.21
Sum SFA	44.13	63.66	43.68	43.62
Sum MUFA	16.14	24.47	20.69	28.33
Sum PUFA	34.79	8.87	31.03	22.23
Sum DMA	4.95	2.99	4.60	5.84
Sum Ω3	21.46	3.71	17.03	16.90
Sum Ω6	13.06	5.07	13.72	5.02
Ω6 / Ω3	0.62	1.45	0.84	0.30

To see whether these differences were specific to the species in question, or due to an experimental artifact, analysis of FA was performed in a second entirely separate model, namely WT C57BL/6J and Nrl^−/− mice. As described in the “Materials and methods” section, retinas from each phenotype were divided into two halves, OR and IR, isolated by precision vibratome sectioning (Supplementary Fig. 1). Hence, comparison of the OR from the two strains provides virtually rod-pure and cone-pure samples. Here as well there was a large reduction in DHA in the Nrl^−/− mouse: WT OR (rods), 41.86% DHA; Nrl^−/− OR (cones), 24.41% (p < 0.001). In contrast, DHA content of the IR from either strain was identical: 19.63% in WT, 18.87% in Nrl^−/− (Fig. 5). These data are given in more complete form in Supplementary Table 5, where overall similarity with ROS and COS data can be seen. Notably, ω3 and ω6 levels showed the same dependence, with ω3 members prevalent in WT (rods) and ω6 members prevalent in Nrl^−/− (cones) retinas.

Fig. 5.

DHA as a percentage of total FA in ONL and INL isolated from WT and Nrl^−/− mice retinas. ***p < 0.001 WT OR compared to Nrl^−/− OR (and Nrl^−/− compared to IR from either genotype)

When (PU)FA composition was examined as a function of PL group, a further fundamental difference was seen between the two membrane fractions: ROS contained significant amounts of VLC species associated with PC. The longest combined chains (≥ 50 total carbon atoms) were always composed of DHA in the sn-2 position, and VLC-PUFA (28–36 carbons) in the sn-1 position (Fig. 6A, Supplementary Table 6). Remarkably, similar analyses of COS showed that there was an almost total lack of VLC-PUFA of 30–36 carbons (Fig. 6A and Supplementary Table 6). The drastic reduction of such VLC-PUFA suggested that maybe cones would be deficient in the enzymes necessary for elaboration of such molecules. As ELOVL4 has been shown to synthesize VLC PUFA from shorter FA chains [38], we performed immunohistochemical staining against retinal sections from a third independent animal model, the cone-rich rodent Arvicanthis ansorgei. This species has the advantage of possessing elevated cone numbers (∼ 30%) [37] precisely organized into the two sclera-most rows within the ONL, which facilitates their identification both by nuclear morphology (Fig. 7A) and by differential staining using anti-rhodopsin and anti-cone arrestin antibodies (Fig. 7B). Immunolabelling with anti-ELOVL4 immunostaining showed that whereas rod cell bodies and IS were strongly labeled, cone cell bodies and IS appeared virtually unstained (Fig. 7C). This was confirmed by double immunolabelling with ELOVL4 and phosducin (another rod cell body marker) (Fig. 7D, E)

Fig.6.

VLC-PUFA are very reduced in COS. Histogram showing amounts of VLC-PUFA from ROS (black bars) and COS (white bars). Data is extracted from Supplemental Table 2. Notice the almost complete absence of VLC-PUFA of > 28 carbons

Fig. 7.

Immunohistochemical staining of Arvicanthis ansorgei retina. This species was chosen because of its high cone number and clear separation of rod (R) and cone (C) cell body layers within the outer nuclear layer (ONL). A DAPI staining of ONL (composed of upper rows of cones and lower rows of rods) and INL. The intermediate outer plexiform layer (OPL) is unstained. B Double immunostaining using anti-rhodopsin to label rods (green) and anti-cone arrestin (CARR) to label cones (red). The latter antibody is especially strong in cone pedicles lying in the upper OPL. Staining is very strong also in the inner and outer segments. C Immunolabelling for ELOVL4 immunoreactivity shows strong staining of rod cell bodies (R) but very weak staining in cone cell bodies (C) and absence from INL. Cone IS are also less stained than rod IS (seen as dark profiles, examples indicated by arrows). D Phosducin immunoreactivity is only seen in rod cell bodies and IS, clearly outlining the unstained cone IS (arrows). E Merge of C and D shows that ELOVL4 immunostaining is much more abundant in rod cell bodies and IS compared to cones. Scale bar in E = 25 μm for all panels

Pig COS membranes contain much less PE than ROS

Total phospholipid analyses showed that unexpectedly, phosphatidylethanolamine (PE) was very under-represented in COS fractions. Whereas phosphatidylcholine (PC) was present in large amounts in both ROS and COS (∼ 380 μg/mg P for ROS, 265 μg/mg P for COS), PE was still abundant in ROS (∼ 280 μg/mg P, thus PC:PE ratio of distinct forms with carbon chain lengths between 28 and 58:42), but very reduced in COS (43 μg/mg P, or PC:PE 50) than PE-related species (59 distinct forms with carbon ratio of 86:14) (Fig. 8). In terms of individual PL species, chain lengths between 28 and 46), with the great majority for both ROS and COS there were more PC-related (76 of lipid mass between chain lengths 30–40 (Fig. 8). The PL profiles as a function of chain length are quite similar for both ROS and COS, with most PC being present in shorter chain lengths (< 36 carbons), while PE was very poorly represented at these shorter lengths. ROS PE was mostly represented in species > 38 carbons; indeed, a single PE species (PE 18:0/22:6) accounted for almost 60% of total ROS PE. It was noted that PL species possessing DHA tails were very enriched in PE, since these forms represented 16% total PC for ROS and 7% total PC for COS, but 71% total PE for ROS and 42% total PE for COS.

Fig. 8.

Relative levels of PC and PE species in ROS and COS fractions. x-Axis shows total chain length in number of carbons; y-axis shows amount of individual phospholipid (PL). Data extracted from Supplemental Table 3

Pig cone membranes present a different bisretinoid profile compared to rods

UPLC analysis of ROS and COS fractions revealed distinctly different profiles: whereas ROS elution profiles were consistent with previous reports, with a large peak corresponding to the bisretinoid pyridinium compound A2PE [39], COS elution profiles had a much smaller A2PE peak (905 a.u. in rods, 121 a.u. in cones; 13% ROS values). In contrast, COS fractions exhibited the notable presence of a peak corresponding to the bisretinoid all-trans-retinal dimer (atRAL-di); this peak was barely detectable in ROS samples (115 a.u. in cones, 22 a.u. in rods; 19% cone levels) (Fig. 9).

Fig. 9.

UPLC read-out of eluted fractions from ROS and COS samples. A Actual elution profiles of ROS (upper trace) and COS (lower trace), with expansion of traces at positions corresponding to A2PE and atRAL-di (right boxes). B Quantification of values under the curve for A2PE and atRALdi in ROS and COS

Discussion

The present study evaluated key potential differences between rods and cones, characteristics potentially important in degeneration of the two types. We showed that (1) cones possess greater levels of abca4 mRNA and abca4 protein compared to rods; (2) COS membranes have a very different lipid composition compared to those of rods: COS contain far less CL (> 3-fold difference) and VLC-PUFA (∼ 13-fold less), and are especially impoverished in DHA (∼ 7-fold less) compared to ROS; (3) cones express much lower levels of ELOVL4 protein compared to rods; (4) the two populations have very different PL compositions, with COS having low levels of PE; and (5) UPLC analysis of the two membranes reveals significant differences in bisretinoid profiles. These data were obtained from three different mammalian species using multiple technical approaches and analytical procedures.

The data showing relatively higher Abca4 mRNA and protein expression levels in cones relative to rods was obtained by comparing values from WT and Nrl^−/− mice. These two mouse strains provide an almost perfect scenario to compare virtually pure rod and cone samples. We reasoned that by normalizing abca4 to two other proteins expressed uniquely by both rods and cones and sharing the same subcellular distribution, i.e., Rds/Prph2 [26] and Rom1 [27], this would adjust for differences in cell numbers and ROS and COS size, and allow us to draw meaningful conclusions. Both Rds/Prph2 and Rom1 are reported to localize exclusively to rim regions of discs and membrane folds in rods and cones [43, 44], as is the case for abca4 [6, 45]. Furthermore, Rds/Prph2 and Rom1 interactions are similar in rods and cones [44]. COS are small and stunted in Nrl^−/− compared to WT mice, so accurate normalizations are critical. Abca4, Rds/Prph2, and Rom1 all showed far lower expression levels in the knockout strain compared to WT (∼ 10%), both for mRNA and proteins. Similar results for western blotting have been shown previously [44]. We were very careful to replicate each analysis many times by either RT-PCR or quantitative western blotting, and are confident the data are robust. The data show that Abca4 levels systematically exceed those of both Rds/Prph2 and Rom1 when plotted as a ratio, suggesting that abca4 function is more critical for cones. We speculate that mutations and decreased transport efficiency in abca4 could be expected to impact cones more significantly. The retina, particularly the PR, contain the highest levels of DHA in the body [46, 47, 48]. This PUFA has been shown to improve neurological and ophthalmological development in children [46, 47, 48], and numerous animal studies have demonstrated beneficial effects upon PR degeneration [9, 49, 50, 51]. DHA is a major source of neuroprotectin D1, which has been shown to enhance cell survival in a number of experimental models [52, 53, 54]. Furthermore, DHA is also a precursor for resolvins, another group of active substances which possess anti-inflammatory properties [55]. It was thus very surprising to observe that COS membranes were greatly impoverished in this major PUFA compared to ROS. The analyses were performed in nine independent experiments with similar findings. The principal contaminant of the COS fraction, mitochondria, could be excluded as underlying this difference since preparations of isolated mitochondria displayed DHA content almost equivalent to ROS and whole retina; and published DHA levels in mitochondria isolated from other tissues (brain) also report elevated values [56]. Thus, mitochondrial contamination would have actually increased DHA values. An important caveat that needs underlining is that the COS fractions used here were not as pure as ROS fractions. The protocol used was developed for ROS isolation, and even with the modifications reported here is probably sub-optimal for obtaining COS material. The major contaminant appears to be mitochondrial membranes, which are very abundant in cone IS. Indeed, parallel proteomic studies indicate large numbers of mitochondrial proteins in the COS fractions (unpublished data). Notwithstanding, immunological screening showed cone opsin was restricted to this fraction, while very little rhodopsin could be detected. Furthermore, lipidomic analysis of isolated OR slices from WT and Nrl^−/− mice echoed these data, with a twofold greater concentration of DHA in rods than in cones. Very similar results for the mouse retina have been shown previously [57]. The smaller difference compared to pig ROS/COS may come from the inclusion of DHA-rich mitochondria in the vibratome fractions. Finally, DHA-enriched dietary treatment of Elovl4 transgenic mice revealed a greater effect on cone than rod function in older mice [58].

Previous data on natural and/or artificial membranes showed that increased PUFA content in ROS would enhance rod sensitivity to light [59, 60, 61, 62]. The higher levels of CL in these cells would have opposing effects, since CL is known to decrease membrane fluidity [63]. But since the properties of CL on lowering membrane fluidity are dramatically restricted in the presence of high levels of PUFAs [64], overall visual transduction should be facilitated in PUFA-rich ROS membranes compared to COS. It is difficult to predict how these membrane properties would influence cone cell behavior, since analogous studies with purified cone opsin are not possible. The large CL differences seen between ROS and COS preparations do not convey the whole complexity: Boesze-Battaglia and colleagues showed that CL content in ROS discs varies as a function of position, with newly formed proximal disks (and plasma membrane) being CL-rich, but during apical displacement the CL content decreases markedly [65, 66]. Recently, it has been shown that differences in PUFA and PL even exist between the central lamellar regions and edges of ROS discs [67]. Our findings are mean values and do not provide any information about possible basal to apical changes in CL or PUFA content. In terms of retinoid trafficking, membrane turnover and the visual cycle, perhaps less fluid COS might sequester more retinal and facilitate remodeling of the OS after phagocytic clearance. The increased stiffness could also affect phagocytosis rates: although precise comparative measures of phagocytic efficiency are difficult to obtain, our published estimates from Arvicanthis indicate a tenfold slower uptake rate for COS than for ROS membrane debris [40].

An additional unexpected observation was the almost complete absence of VLC-PUFA (≥ 50 total carbons) in COS fractions. Such differences led us to predict a deficiency in elongase enzyme in cones. In Stargardt’s disease type 3 (autosomal dominant), the causal mutation is a deletion in the Elovl4 gene leading to truncated mis-directed protein [68]. Previous reports have shown ELOVL4 mRNA and protein localization within cone (and rod) IS [69, 70], whereas the antibody used in the present study [41] showed robust staining of Arvicanthis rod cell bodies and IS that was clearly reduced in cone cell bodies and IS. The reasons for these discrepancies with the published studies are unknown, although the antibodies cited were directed against very different epitopes within the ELOVL4 sequence. If the low levels of VLC-PUFA also exist in human cones, it raises implications in the pathogenesis of Stargardt’s disease type 3, which possibly cone degeneration could be a “bystander effect” linked to initial rod death. Previously published analysis of VLC-PUFA in human macula and peripheral retina also showed significantly lower values in the macula [71], fully in line with our results. The same authors show much higher levels of VLC-PUFA in mouse compared to human retina (∼ 10-fold greater in mouse), indicating that there are also large differences between species. Also, work showing the absence of either structural or functional effects upon cones in mice in which the Elovl4 gene was conditionally deleted, fit with the idea that cones are largely lacking this enzyme and are unaffected under such experimental conditions [72].

Another intriguing difference between ROS and COS concerns the much lower levels of PE in the latter. Previous studies have shown roughly equal amounts of PC and PE in ROS isolated from a range of vertebrate species [73], although purified ROS plasma membrane has much less PE than ROS discs (11% vs > 40%) [74]. We could find no published data on PL values for COS. A widely accepted model of ABCA4 transport function, based on rigorous experimental data, indicates that atR reacts with PE to form NRPE which constitutes the actual physiological ligand recognized by the Abca4 substrate [18]. However, this model is based on estimates from ROS lipid composition; in vivo mouse studies [15] are also only informative only for ROS. Thus, a prediction from membranes relatively depleted in PE would be that the condensation reaction between atR and PE is less likely to occur, and formation of downstream bisretinoids like A2PE would be reduced. This was exactly what we observed, with UPLC analysis of COS fractions containing only 13% of the A2PE levels observed in ROS. Intriguingly, COS exhibited notable levels of unconjugated atRAL-di, a bisretinoid distinct from A2PE/A2E but even more photoreactive [75]. Since atRAL-di might also not be transported efficiently by abca4, and also readily reacts with singlet oxygen, the intracellular presence of this bisretinoid could provide a clear pathway for direct cone poisoning. It should be noted that these data are in discrepancy with published reports showing high levels of A2E in the Nrl/abca4 double knockout mouse [22].

In conclusion, compared to rods, cones display several important differences in protein and lipid composition: higher levels of abca4, lower levels of DHA, CL, and PE, and notable amounts of the bisretinoid atRAL-di (schematized in Fig. 10). We postulate that these differences would render cones more vulnerable than rods to mutations in Abca4. This has important implications for cone pathophysiology, since it would influence the way they respond to noxious conditions present during multiple retinal diseases or environmental insults. We hypothesize that these differences, combined with reduced clearance of “toxic” membranes through the slower rates of cone phagocytosis [40] (Fig. 10), lead to direct poisoning of cones in STGD1. Secondary death of rods would follow RPE malfunctioning. This has implications for disease treatment: transplantation of RPE (or stem cell-derived RPE) [76] would preferentially protect rods but not cones if these latter actually die early in some forms of STGD1 [21]. Not to be forgotten, it is also clear that multiple other factors affect cone survival such as oxidative stress [77] or availability of rod-derived cone viability factor [78]. We propose that inherent cone vulnerability might alter the balance between survival and degeneration depending on the particular physiological or pathological context. We are currently testing the idea that cones are more vulnerable than rods to abca4 loss in an experimental animal model.

Fig. 10.

Schematic representation of rod-cone differences and their possible role in cone survival/death. In A, rod disks and plasma membrane (PM) are represented with equal amounts of phosphatidylcholine (blue circles) and phosphatidylethanolamine (PE, red circles). In physiological conditions (1), light activates rhodopsin to liberate atR which reacts with PE in the disk lumen to form N-retinylidene-PE (NRPE), which exits rapidly via Abca4 flippase activity (stippled ovals at disc edge) to enter the cytosol (blue arrow). In pathological STGD1 states (2), following rhodopsin light activation, NRPE transport is hampered by mutated abca4 (deaths-head on stippled oval) leading to chemical reactions within the disc membrane and formation of bisretinoids such as A2PE the immediate precursor of A2E. In B, a similar scenario is shown for cones, but in this case, there is far less PE (red circles) and DHA within the membrane, and twice the amount of Abca4 protein (double stippled ovals). In physiological conditions (3), after cone opsin light activation, there is also free atR but the low levels of PE lead to less NRPE product and higher levels of the bisretinoid all-trans-retinal dimer (atRAL-di). In pathological states like STGD1 (4), the higher intradiscal amounts of the toxic bisretinoid atRAL-di foster extensive photo-oxidative damage. C These rod-cone differences are to be further seen in the context of ROS and COS membrane topology and partitioning. In the case of ROS, the formation of discrete discs at the basal surface and their shuttling towards the apical surface leads to a gradient of young to old discs (depicted as gradual blackening of disks). The continuously folded membrane of cones does not permit such a gradient to occur. In pathological situations like STGD1, clearance of A2PE-laden disks (blue circles) is relatively efficient since the older discs will carry more bisretinoid. This “purifies” the rods but leads to poisoning of the adjacent RPE. However, free diffusion of membrane components around the COS (illustrated with curved arrow) will lead to constant increases in levels of toxic compounds, which reach a threshold and cause cones to die

Key message.

• Rod and cone degeneration in pathological scenarios often differ. For example, in Stargardt’s Disease (STGD1) patients typically show macular degeneration, sometimes with pan-retinal cone failure but retained rod function. If cones exhibit structural and biochemical differences to rods, this would likely affect pathogenic processes.

• We show that the expression levels of Abca4 mRNA and protein, the causal gene in STGD1, are significantly higher in cones than rods. Furthermore, cone membrane lipid composition is very different from that of rods, with large reductions in amounts of docohexaenoic acid (DHA), cholesterol and phosphatidyl-ethanolamine. The profiles of toxic lipid adducts (bisretinoids) differ between rod and cone membranes, with rods containing much more A2PE and cones containing much more all-transretinal dimer.

• These fundamental differences could explain the preponderant cone degeneration phenotype in diseases such as STGD1 and suggest cones may die by direct poisoning.