Abstract
Contrary to Alzheimer's disease (AD), the APOE2 allele increases and the APOE4 allele reduces the risk to develop age-related macular degeneration (AMD) compared with the most common APOE3 allele. The underlying mechanism for this association with AMD and the reason for the puzzling difference with AD are unknown. We previously demonstrated that pathogenic subretinal mononuclear phagocytes (MPs) accumulate in Cx3cr1-deficient mice due to the overexpression of APOE, interleukin-6, and CC chemokine ligand 2 (CCL2). We here show using targeted replacement mice expressing the human APOE isoforms (TRE2, TRE3, and TRE4) that MPs of TRE2 mice express increased levels of APOE, interleukin-6, and CCL2 and develop subretinal MP accumulation, photoreceptor degeneration, and exaggerated choroidal neovascularization similar to AMD. Pharmacological inhibition of the cytokine induction inhibited the pathogenic subretinal inflammation. In the context of APOE-dependent subretinal inflammation in Cx3cr1GFP/GFP mice, the APOE4 allele led to diminished APOE and CCL2 levels and protected Cx3cr1GFP/GFP mice against harmful subretinal MP accumulation observed in Cx3cr1GFP/GFPTRE3 mice. Our study shows that pathogenic subretinal inflammation is APOE isoform-dependent and provides the rationale for the previously unexplained implication of the APOE2 isoform as a risk factor and the APOE4 isoform as a protective factor in AMD pathogenesis.
SIGNIFICANCE STATEMENT The understanding of how genetic predisposing factors, which play a major role in age-related macular degeneration (AMD), participate in its pathogenesis is an important clue to decipher the pathomechanism and develop efficient therapies. In this study, we used transgenic, targeted replacement mice that carry the three human APOE isoform-defining sequences at the mouse APOE chromosomal location and express the human APOE isoforms. Our study is the first to show how APOE2 provokes and APOE4 inhibits the cardinal AMD features, inflammation, degeneration, and exaggerated neovascularization. Our findings reflect the clinical association of the genetic predisposition that was recently confirmed in a major pooled analysis. They emphasize the role of APOE in inflammation and inflammation in AMD.
Introduction
In humans, the APOE gene has three common genetic variants (APOE2, APOE3, and APOE4), due to two polymorphisms rs7412 and rs429358 that are imbedded in a well-defined CpG island, and lead to two cysteine-arginine interchanges at residues 112 and 158 (Yu et al., 2013). The APOE2 allele is associated with higher APOE concentrations in plasma, CSF, and brain tissue (Riddell et al., 2008; Bales et al., 2009) due to impaired clearance caused by APOE2's decreased affinity for the low-density lipoprotein receptor (Mahley and Rall, 2000). Its transcription can also be increased in certain cell types (astrocytes, neurons) due to the loss of CpG sites associated with APOE3 and APOE4 alleles (Yu et al., 2013). Compared with the APOE3 allele, the APOE4 allele is transcribed similarly in neurons and astrocytes (Yu et al., 2013), but its protein concentrations in plasma, CSF, and brain parenchyma are decreased (Riddell et al., 2008; Bales et al., 2009; Sullivan et al., 2011). The structural changes in the APOE4 protein also lead to diminished association with high-density lipoprotein (Dong and Weisgraber, 1996) and impaired reverse cholesterol transport (Heeren et al., 2004; Mahley et al., 2009).
APOE2 allele carriers are at increased risk for developing late age-related macular degeneration (AMD) [odds ratio (OR) = 1.83 for homozygote APOE2 allele carriers] and are protected against Alzheimer's disease (AD), whereas the APOE4 allele protects against AMD (OR = 0.72 per haplotype) and is a risk factor for AD compared with the most common APOE3 allele (Mahley and Rall, 2000; McKay et al., 2011). This association was recently confirmed in a pooled study of >20,000 subjects (McKay et al., 2011). It is found for both clinical forms of late AMD: wet AMD, which is defined by choroidal neovascularization (CNV) and geographic atrophy, which is characterized by an extending lesion of both the retinal pigment epithelium (RPE) and photoreceptors. In AD, the APOE4 allele is associated with greater β-amyloid burden, possibly due to reduced efficacy in clearance of β-amyloid via multiple pathways (Bales et al., 2009; Mahley et al., 2009). The mechanism underlying the associations of the APOE isoforms with AMD remains unexplained.
APOE is the main lipoprotein of the brain and the retina (Mahley and Rall, 2000; Anderson et al., 2001). It is strongly expressed in mononuclear phagocytes (MPs), such as macrophages and microglial cells (Peri and Nüsslein-Volhard, 2008; Levy et al., 2015), and plays a major role in macrophage lipid efflux and reverse cholesterol transport in conjunction with APOA-I (Mahley and Rall, 2000; Mahley et al., 2009). APOE and APOA-I can also induce interleukin-6 (IL-6) and CC chemokine ligand 2 (CCL2) in MPs in the absence of pathogen-derived ligands (Smoak et al., 2010; Levy et al., 2015).
We recently showed that subretinal MPs that accumulate in AMD strongly express APOE (Levy et al., 2015). The subretinal MPs of Cx3cr1GFP/GFP mice that develop subretinal inflammation and cardinal features of AMD (Combadière et al., 2007) express similar high levels of APOE (Levy et al., 2015), but also IL-6 (Levy et al., 2015) and CCL2 (Sennlaub et al., 2013). We showed that APOE-induced IL-6 release from MPs represses RPE immune suppression, prolongs subretinal MP survival, and promotes subretinal inflammation (Levy et al., 2015). Furthermore, we demonstrated that increased levels of CCL2 in Cx3cr1GFP/GFP mice recruit pathogenic inflammatory CCR2+ monocytes to the subretinal space (Sennlaub et al., 2013). In consequence, subretinal pathogenic MPs accumulate in Cx3cr1GFP/GFP mice due to increased MP recruitment and decreased MP elimination. ApoE deletion in Cx3cr1GFP/GFP mice prevented age- and stress-induced subretinal MP accumulation and reduced associated CNV (Levy et al., 2015).
We here investigated the influence of the APOE alleles and isoforms on subretinal inflammation and associated photoreceptor degeneration and choroidal neovascularization, major hallmarks of AMD.
Materials and Methods
Animals.
Targeted replacement mice that express human APOE isoforms (TRE2, TRE3, and TRE4) were engineered as previously described (Sullivan et al., 1997) and provided as a generous gift by Dr. Patrick Sullivan, backcrossed with C57BL/6 mice to eliminate the Crb1rd8 contamination in the three strains and crossed to Cx3cr1GFP/GFPmice (Charles River). Mice were housed in the animal facility under specific pathogen-free condition, in a 12/12 h light/dark (100–500 lux) cycle with water and normal diet food available ad libitum. All experimental protocols and procedures were approved by the local animal care ethics committee “Comité d'éthique en expérimentation animale Charles Darwin” (Ce5/2010/013; Ce5/2011/033; Ce5/2010/044). We used male mice for choroidal neovascularization experiments, whereas experiments on aged and light challenged mice were performed on mice of either sex, as we did not observe differences between the sexes in these conditions.
Light challenge and laser injury model.
Two-month-old mice of either sex were adapted to darkness for 6 h, pupils dilated and exposed to constant green LED light (starting at 2 A.M., 4500 lux, JP Vezon equipements) for 4 d as previously described (Sennlaub et al., 2013). Laser coagulations were performed on male mice with a 532 nm ophthalmological laser mounted on an operating microscope (Vitra Laser, 532 nm, 450 mW, 50 ms, 250 μm) as previously described (Levy et al., 2015). Intravitreal injections of 2 μl of PBS, isotype control rat IgG1, and rat anti-mouse CD14 (BD Biosciences) were performed using glass capillaries (Eppendorf) and a microinjector. The 2 μl solution of the antibodies was injected at 50 μg/ml, corresponding to an intraocular concentration of 5 μg/ml assuming their dilution by ∼1/10th in the intraocular volume.
Immunohistochemistry, CNV, and MP quantification and histology.
RPE and retinal flatmounts were stained and quantified as previously described (Sennlaub et al., 2013) using polyclonal rabbit anti-IBA-1 (Wako) and rat anti-mouse CD102 (clone 3C4, BD Biosciences) appropriate secondary antibodies and counterstained with Hoechst if indicated. Preparations were observed with fluorescence microscope (DM5500, Leica). Histology of mice eyes and photoreceptor quantification were performed as previously described (Sennlaub et al., 2013).
Cell preparations and cell culture.
In accordance with the Declaration of Helsinki, volunteers provided written and informed consent for the human monocyte expression studies, which were approved by the Centre national d'ophthalmologie des Quinze-Vingt hospital (Paris) ethics committees (no. 913572). Peripheral blood mononuclear cells were isolated from heparinized venous blood from healthy volunteer individuals by 1-step centrifugation on a Ficoll Paque layer (GE Healthcare) and sorted with EasySep Human Monocyte Enrichment Cocktail without CD16 Depletion Kit (Stem Cell Technology). Mouse peritoneal macrophages, bone marrow-derived monocytes, and photoreceptor outer segment (POS) isolation (all in serum-free X-Vivo 15 medium) were performed as previously described (Sennlaub et al., 2013). In specific experiments, cells were stimulated with the different recombinant human APOE isoforms (5 μg/ml, Leinco Technologies), recombinant human APOE3 90 min heat-denatured (5 μg/ml, Leinco Technologies), APOE3 (5 μg/ml) with LPS inhibitor polymyxin B (25 μg/ml, Calbiochem), rat anti-IgG isotype control (25 μg/ml, R&D Systems), rat anti-mouse CD14 (25 μg/ml, R&D Systems), mouse anti-IgG isotype control (25 μg/ml), mouse anti-human TLR2 (25 μg/ml, Invivogen), human IgA2 isotype control (25 μg/ml, Invivogen), human anti-human TLR4 (25 μg/ml, Invivogen), and POS prepared as previously described (Sennlaub et al., 2013).
Reverse transcription and real-time PCR and ELISA.
IL-6, CCL2, and IL-1β RT-PCRs using Sybr Green (Invitrogen) and ELISAs using mouse or human IL-6 DuoSet (R&D Systems), mouse or human CCL2 Duoset (R&D Systems), and human APOE Pro kit (Mabtech) were performed as previously described (Sennlaub et al., 2013; Hu et al., 2015; Levy et al., 2015).
Statistical analysis.
GraphPad Prism 5 (GraphPad Software) was used for data analysis and graphic representation. All values are reported as mean ± SEM. Statistical analysis was performed by one-way or two-way ANOVA followed by Dunnett's post test or Mann–Whitney test for comparison among means depending on the experimental design. The p values are indicated in the figure legends.
Results
The APOE2 allele leads to age- and stress-related subretinal MP accumulation, retinal degeneration, and exacerbated choroidal neovascularization
The subretinal space, located between the RPE and the POS, does not contain significant numbers of MPs under normal conditions (Penfold et al., 2001; Gupta et al., 2003; Combadière et al., 2007; Levy et al., 2015). This is likely the result of physiologically low levels of chemoattractants along with strong immunosuppressive RPE signals that quickly eliminate infiltrating MPs (Sennlaub et al., 2013; Levy et al., 2015). We have previously shown that the lack of the tonic inhibitory CX3CL1/CX3CR1 signal, observed in Cx3cr1-deficient mice, is sufficient to induce pathogenic chronic subretinal MP accumulation as a consequence of increased recruitment and decreased elimination (Combadière et al., 2007; Sennlaub et al., 2013; Levy et al., 2015). We showed that this accumulation is dependent on the overexpression of APOE in Cx3cr1-deficient MPs (Levy et al., 2015). To evaluate a potential role of the human APOE isoforms in subretinal inflammation, we used targeted replacement mice expressing human isoforms (TRE2, TRE3, and TRE4) (Sullivan et al., 1997). We first backcrossed the strains with C57BL/6J mice to eliminate the Crb1rd8 contamination in the three strains, which can lead to AMD-like features (Mattapallil et al., 2012). The mice were raised under 12 h light/12 h dark cycles at 100–500 lux at the cage level, with no additional cover in the cage, the conditions that induce MP accumulation in Cx3cr1-deficient mice with age (Combadière et al., 2007). Quantification of subretinal IBA-1+ MPs on retinal and RPE/choroidal flatmounts of 2- and 12-month-old TRE2, TRE3, and TRE4 mice revealed that TRE2 mice develop age-dependent subretinal MP accumulation compared with TRE3 and TRE4 mice (Fig. 1A). Similarly, TRE2 mice accumulated significantly more subretinal MPs after a 4 d light challenge, and the MPs continued to accumulate after return for 10 additional days in normal light conditions (Fig. 1B; the intensity of our light challenge model used herein was calibrated to induce subretinal inflammation in inflammation-prone Cx3cr1GFP/GFP mice but not in WT mice) (Sennlaub et al., 2013).
We also observed a thinning of the outer nuclear layer that contains the photoreceptor nuclei on histological retinal sections from 12-month-old TRE2 mice compared with TRE3 mice (Fig. 1C; micrographs taken at equal distance from the optic nerve). Photoreceptor nuclei row counts (Fig. 1C′) and calculation of the area under the curve (Fig. 1C″) revealed that the age-related accumulation of subretinal MPs in TRE2 mice is associated with significant photoreceptor cell loss compared with TRE3 and TRE4 mice.
In addition, in laser-induced CNV, subretinal IBA-1+ MPs (green staining, counted on the RPE at a distance of 0–500 μm to CD102+ CNVs, red staining) were significantly more numerous in TRE2 mice 7 d after a laser impact (Fig. 1D) and had developed significantly greater CNV lesions (Fig. 1E) compared with the other strains.
Together, our data demonstrate that TRE2 mice, expressing the APOE2 AMD risk allele, develop age-related subretinal inflammation and photoreceptor degeneration and exaggerated inflammation and CNV after laser injury similar to late AMD.
The APOE2 allele increases APOE levels in the eye and APOE transcription and innate immunity receptor cluster (IIRC) activation in MPs
We previously showed that the levels of soluble APOE are elevated in adult TRE2 mouse brains and diminished in TRE4 brains compared with TRE3 mice in a model of AD (Bales et al., 2009). Similarly, ELISA of APOE levels of homogenates of PBS-perfused posterior segments (retina and RPE/choroid plexus) of 12-month-old mice revealed significantly higher levels of APOE in TRE2 mice compared with TRE3 and TRE4 mice (Fig. 2A). Furthermore, immunohistochemical localization of APOE on retinal flatmounts of TRE2 mice (Fig. 2B, red staining) revealed strong APOE expression in subretinal IBA-1+ MPs (Fig. 2B, green staining).
The polymorphism rs7412 that defines the APOE2 isoform also leads to the loss of a CpG site in the APOE2 allele that has been shown to moderately, but significantly, increase APOE transcription in brain astrocytes, but not in hepatocytes (Yu et al., 2013). Our data confirm that APOE transcription in hepatocytes does not differ between genotypes (Yu et al., 2013) (Fig. 2C, RT-PCR) and that APOE concentrations in the blood were significantly increased in TRE2 mice (Fig. 2D, blood plasma ELISA), shown to be due to its decreased clearance rate (Mahley and Rall, 2000). However, bone marrow-derived monocytes (cultured with POS for 3 d to mimic subretinal macrophage differentiation) (Fig. 2E, RT-PCR) and peritoneal macrophages (Fig. 2F, RT-PCR) from TRE2 mice transcribed significantly higher levels of ApoE mRNA compared with MPs of the other mouse strains. Accordingly, the APOE secretion of TRE2 mice macrophages was robustly increased (Fig. 2G, ELISA of supernatant) compared with the other groups.
APOE and APOA-I have been shown to activate the TLR2-TLR4 CD14-dependent IIRC in mouse peritoneal macrophages in the absence of pathogen-derived ligands and to induce inflammatory cytokines, such as IL-6 (Smoak et al., 2010; Levy et al., 2015), but also CCL2 (shown for APOA-I) (Smoak et al., 2010). We here show that human blood-derived CD14+ monocytes significantly secrete IL-6 after 24 h of recombinant lipid-free APOE3 stimulation (Fig. 2H) similar to mouse macrophages. Ninety minute heat denaturation completely abolished the induction, whereas the LPS inhibitor polymyxin B did not, confirming that LPS contamination of APOE3 is not accountable for the effect, as shown for APOA-I using multiple approaches (Smoak et al., 2010). This induction was due to the activation of the TLR2-TLR4 CD14-dependent IIRC, as neutralizing antibodies to CD14, TLR2, and TLR4 inhibited this effect, compared with control antibodies (Fig. 2H). Accordingly, peritoneal macrophages from TRE2 mice that express increased amounts of APOE (Fig. 2C,D) also transcribed significantly more IL-6, CCL2, and IL-1β compared with macrophages of the other isoforms (Fig. 2I).
Together, our results show that the APOE2 allele increases APOE levels in the tissue and APOE expression in MPs. We confirm that APOE activates the IIRC and show that the excessive APOE expression in macrophages from TRE2 mice is associated with increased production of inflammatory cytokines in vitro.
IIRC inhibition reduces subretinal MP accumulation and choroidal neovascularization in TRE2 mice in vivo
To evaluate whether increased IIRC activation is implicated in subretinal MP accumulation observed in TRE2 mice in vivo, we inhibited the IIRC by an intravitreal injection of a CD14-neutralizing antibody in the laser-induced CNV model (Figure 3). The antibody, which blocks APOE-dependent cytokine induction (Levy et al., 2015), inhibited subretinal MP accumulation around the laser injury, quantified on IBA1-stained RPE/choroidal flatmounts (Fig. 3A) and CD102+ CNV formation (Fig. 3B) at day 7 after laser injury of TRE2 mice compared with control IgG.
These results confirm that the CD14-dependent inflammatory cytokine induction participates in subretinal MP accumulation in TRE2 mice in vivo, similar to Cx3cr1GFP/GFP mice (Levy et al., 2015).
The APOE4 allele protects APOE-overexpressing Cx3cr1GFP/GFP mice from subretinal MP accumulation, retinal degeneration, and exacerbated choroidal neovascularization
Cx3cr1-deficient mice lack the tonic inhibitory signal of neuronal CX3CL1 and develop subretinal MP accumulation and concomitant photoreceptor degeneration with age when raised in cyclic light at 100–500 lux (Combadière et al., 2007; Chinnery et al., 2012; Sennlaub et al., 2013; Hu et al., 2015; Levy et al., 2015). The accumulation can be prevented by raising the animals in darkness (Combadière et al., 2007) or in dim light conditions (Luhmann et al., 2013) and be accelerated by a light challenge (Sennlaub et al., 2013; Hu et al., 2015; Levy et al., 2015) (for more details, see Sennlaub et al., 2013, mini review in supplemental data). Although these features do not mimic all the aspects of AMD (Drusen formation and RPE atrophy), they do model subretinal inflammation and associated photoreceptor degeneration, two hallmarks of AMD (Gupta et al., 2003). Cx3cr1 deletion also increases subretinal MP accumulation in diabetes (Kezic et al., 2013), in a paraquat-induced retinopathy model (Chen et al., 2013), and in a retinitis pigmentosa model (Peng et al., 2014). We previously demonstrated that pathogenic MPs accumulate in Cx3cr1-deficient mice due to the overexpression of APOE, IL-6, and CCL2 (Sennlaub et al., 2013; Levy et al., 2015). To evaluate a possible influence of the APOE4 isoform in a model of pathological subretinal inflammation, we crossed TRE3 and TRE4 mice to Cx3cr1GFP/GFP mice (Levy et al., 2015).
Quantification of subretinal IBA-1+ MPs on retinal and RPE/choroidal flatmounts of 2- and 12-month-old Cx3cr1GFP/GFPTRE3 mice and Cx3cr1GFP/GFPTRE4 mice revealed that the age-dependent subretinal MP accumulation observed in Cx3cr1GFP/GFPTRE3 mice was prevented in Cx3cr1GFP/GFPTRE4 mice (Fig. 4A). A 4 d light challenge led to similar initial subretinal MP accumulation, but the increase of MPs after return to normal light conditions was significantly blunted in Cx3cr1GFP/GFPTRE4 mice compared with Cx3cr1GFP/GFPTRE3 mice (Fig. 4B).
Furthermore, micrographs of histological sections of 12-month-old mice revealed a thicker outer nuclear layer in Cx3cr1GFP/GFPTRE4 mice compared with the thinned Cx3cr1GFP/GFPTRE3 mice (Fig. 4C). Photoreceptor nuclei row counts (Fig. 4C′) and calculation of the area under the curve (Fig. 4C″) show that the inhibition of the age-related accumulation of subretinal MPs in Cx3cr1GFP/GFPTRE4 mice compared with Cx3cr1GFP/GFPTRE3 mice significantly inhibited the associated photoreceptor cell loss. The age-related subretinal MP accumulation and photoreceptor degeneration observed in Cx3cr1GFP/GFPTRE3 mice are significantly increased compared with TRE3 mice presented in Figure 1, similar to Cx3cr1GFP/GFP mice expressing mouse APOE (Sennlaub et al., 2013) (MP/mm2 of 12-month-old mice: TRE3 mice: 7.315 ± 1.72 SEM; Cx3cr1GFP/GFPTRE3 mice: 16.6 ± 2.22 SEM; area under the curve: TRE3 mice: 148.3 ± 1.52 SEM; Cx3cr1GFP/GFPTRE3 mice: 137.3 ± 2.12 SEM).
Moreover, laser-induced subretinal IBA-1+ MPs in 2-month-old mice (green staining) adjacent to CD102+ CNVs (red staining) was again significantly inhibited in Cx3cr1GFP/GFPTRE4 mice compared with Cx3cr1GFP/GFPTRE3 mice (Fig. 4D) and had developed significantly greater CNV lesions at 14 d after laser injury (Fig. 4E) compared with the other strains.
In summary, our data demonstrate that the APOE4 allele, which is protective for AMD, inhibits subretinal inflammation and concomitant degeneration and CNV in Cx3cr1 deficiency compared with APOE3.
The APOE4 allele decreases ocular APOE levels in Cx3cr1GFP/GFP mice and activates the IIRC inefficiently
To investigate whether the APOE3 and APOE4 allele influences the APOE level in the eyes of Cx3cr1GFP/GFPTRE mice, we analyzed APOE levels of homogenates of PBS-perfused posterior segments (retina and RPE/choroid plexus) of 12-month-old mice. APOE levels were significantly lower in homogenates of 12-month-old Cx3cr1GFP/GFPTRE4 mice compared with Cx3cr1GFP/GFPTRE3 mice (Fig. 5A), similar to APOE levels in the eyes of TRE mice (see above) and brains of PDAPP mice expressing human APOE isoforms (Bales et al., 2009).
APOA-I and APOE likely activate the IIRC by modifying the cholesterol content of the lipid rafts in which they are located (Smoak et al., 2010). As the APOE4 isoform has an impaired capacity to promote cholesterol efflux and transport (Heeren et al., 2004; Mahley et al., 2009), we next tested its ability to activate the IIRC of blood-derived human monocytes in culture. Interestingly, stimulation of monocytes for 24 h by recombinant APOE4 induced significantly less IL-6 and CCL2 secretion compared with the induction of the cytokines by equimolar concentrations of APOE3 (Fig. 5B).
APOE transcription (Fig. 5C, RT-PCR) and APOE secretion (Fig. 5D, ELISA of supernatant) in peritoneal macrophages from Cx3cr1GFP/GFPTRE3 mice and Cx3cr1GFP/GFPTRE4 mice were comparable, similar to previous reports from astrocytes: the APOE4 allele did not diminish APOE production (Yu et al., 2013). However, in accordance with a decreased ability to activate the IIRC, peritoneal macrophages from Cx3cr1GFP/GFPTRE4 mice transcribed significantly less CCL2 compared with macrophages from Cx3cr1GFP/GFPTRE3 mice, whereas the IL-6 transcription was variable, but not significantly different (Fig. 5E).
Together, our results show that the APOE4 allele leads to decreased APOE tissue levels and to a reduced capacity to activate the IIRC and induce CCL2 in MPs.
Discussion
We have previously shown that the lack of the tonic inhibitory CX3CL1/CX3CR1 signal, observed in Cx3cr1-deficient mice, is sufficient to induce pathogenic chronic subretinal MP accumulation due to increased CCL2-dependent monocyte recruitment and IL-6-dependent decrease of subretinal MP elimination (Combadière et al., 2007; Sennlaub et al., 2013; Hu et al., 2015; Levy et al., 2015). We showed that this accumulation is dependent on the overexpression of APOE in Cx3cr1-deficient MPs (Levy et al., 2015). As the APOE isoforms are associated with significant differences in APOE levels in humans (Mahley and Rall, 2000) and in humanized transgenic mice expressing APOE isoforms (Bales et al., 2009; Yu et al., 2013), we here evaluated the consequences of the three isoforms on chorioretinal homeostasis.
Our study shows that TRE2 mice, carrying the AMD risk allele, develop age-, light-, and laser-induced subretinal MP accumulation associated with photoreceptor degeneration and excessive CNV. TRE2 mice displayed increased tissue levels of APOE measured in whole retinal/choroidal protein extracts compared with TRE3 and TRE4 mice. This increase is likely due to reduced LDLR-dependent APOE2 uptake (Fryer et al., 2005) of APOE that is produced in the RPE, the inner retina and by subretinal MPs (Anderson et al., 2001; Levy et al., 2015). Similar to CX3CR1GFP/GFP mice and AMD patients (Levy et al., 2015), subretinal MPs in TRE2 mice stained strongly positive for APOE. In vitro, we show that the APOE2 allele is associated with increased APOE transcription and secretion in macrophages from TRE2 mice, as previously shown for astrocytes (Yu et al., 2013). The APOE levels in and around subretinal MPs are therefore likely elevated because of increased APOE transcription and decreased LDLR-dependent clearance in the tissue. It is not yet clear to what extent APOE from nonmyeloid cells participates in the subretinal inflammation and whether extracellular or intracellular APOE within the MPs is the determining factor.
APOE is capable of activating the CD14/TLR2/TLR4-dependent IIRC and inducing IL-6, as previously shown for mouse macrophages (Smoak et al., 2010; Levy et al., 2015). We here confirm that APOE can induce inflammatory cytokines in a similar manner in human monocytes. Moreover, macrophages from TRE2 mice that express significantly higher levels of APOE also transcribed higher levels of inflammatory cytokines, such as IL-6, CCL2, and IL-1β, in accordance with an APOE activation of the IIRC and similar to APOE-overexpressing Cx3cr1-deficient macrophages (Sennlaub et al., 2013; Hu et al., 2015; Levy et al., 2015). We previously showed that CCL2 (Sennlaub et al., 2013) and IL-6 (Levy et al., 2015) promote subretinal MP accumulation by increasing monocyte recruitment and decreasing MP clearance, respectively. Indeed, inhibition of the IIRC by a CD14-blocking antibody in laser-injured TRE2 mice decreased subretinal MP accumulation and neovascularization. These results demonstrate that IIRC activation is significantly involved in the subretinal MP accumulation in TRE2 mice in vivo. Together, Cx3cr1GFP/GFP mice and TRE2 mice both overexpress APOE in mononuclear phagocytes, although for different reasons. In both mouse strains, the increased APOE is associated with IIRC activation, CCL2 and IL-6 induction, and pathogenic subretinal inflammation. Although these features do not mimic all the aspects of AMD (Drusen formation and RPE atrophy), they do model subretinal inflammation and associated photoreceptor degeneration, two hallmarks of AMD (Gupta et al., 2003). We previously demonstrated the importance of APOE in this process, as APOE deletion protected Cx3cr1GFP/GFP mice against the inflammation (Levy et al., 2015). Interestingly, increased levels of CCL2 and IL-6 are also observed in late AMD (Seddon et al., 2005; Jonas et al., 2010; Sennlaub et al., 2013; Chalam et al., 2014), where chronic MP accumulation is observed (Penfold et al., 2001; Gupta et al., 2003; Combadière et al., 2007; Sennlaub et al., 2013; Levy et al., 2015). The observation that inhibition of subretinal MP accumulation in a variety of animal models represses CNV (Sakurai et al., 2003; Tsutsumi et al., 2003; Liu et al., 2013) and degeneration (Guo et al., 2012; Rutar et al., 2012; Suzuki et al., 2012; Kohno et al., 2013) strongly suggests that chronic subretinal inflammation partakes in AMD pathogenesis.
The chronic nonresolving inflammation in AMD is associated with an increase in APOE (Klaver et al., 1998; Anderson et al., 2001; Levy et al., 2015) similar to other inflammatory conditions (Rosenfeld et al., 1993). To evaluate whether a potential influence of the protective APOE4 allele would become apparent in a situation of increased inflammation and APOE abundance, we crossed TRE3 and TRE4 mice to the APOE-overexpressing Cx3cr1GFP/GFP mice. In the inflammatory context of Cx3cr1GFP/GFP mice, the APOE4 allele led to diminished APOE levels and the APOE4 allele protected Cx3cr1GFP/GFP mice against harmful subretinal MP accumulation observed in APOE3-carrying Cx3cr1GFP/GFP mice. APOE4 is characterized, among others, by its decreased capacity to transport cholesterol compared with APOE3 (Heeren et al., 2004) and might thereby be less capable of modifying the cholesterol content of the lipid rafts and activating the IIRC (Smoak et al., 2010). Indeed, our results show that recombinant APOE4 induced less IL-6 and CCL2 compared with equimolar APOE3 concentrations in human monocytes in vitro. Similarly, macrophages from Cx3cr1GFP/GFPTRE4 mice transcribed less CCL2 compared with macrophages from Cx3cr1GFP/GFPTRE3 mice, although their APOE expression was comparable. It is not clear why Cx3cr1GFP/GFPTRE4 macrophages did not differ from Cx3cr1GFP/GFPTRE3 macrophages in terms of IL-6 transcription levels. Other unknown regulatory elements likely influence the transcription of the individual cytokines in macrophages, and the interplay of Cx3cr1 deficiency and the human APOE3 and APOE4 isoform in the mouse macrophages might affect these pathways. We previously showed that CCL2 inhibition in Cx3cr1GFP/GFP mice significantly inhibited age-, laser-, and light-induced subretinal MP accumulation (Sennlaub et al., 2013) and diminished production of CCL2, as a result of reduced APOE concentrations and APOE4's impaired capacity to induce cytokines might explain the reduced inflammation and inhibition of degeneration and CNV observed in Cx3cr1GFP/GFP TRE4 mice.
To our knowledge, this is the first study to describe a comprehensive pathomechanism of the involvement of APOE isoforms in AMD that is in accordance with the clinical observation of the APOE2 allele being an AMD risk factor and the APOE4 allele an AMD-protective genetic factor. One previous study demonstrated that TRE4 mice on high-fat diet develop lipid accumulations in the Bruch's membrane, proposed as similar to early AMD (Malek et al., 2005), which are also observed in ApoE−/− mice (Ong et al., 2001). Although these observations might apply to early AMD, they are unlikely to play a role in late AMD in which increased APOE immunoreactivity is observed (Klaver et al., 1998; Anderson et al., 2001; Levy et al., 2015) and in which the APOE4 allele plays a protective role (McKay et al., 2011). The involvement of increased reverse cholesterol transport in AMD might also be supported by the observation that APOA-I levels are elevated in the vitreous of AMD patients (Koss et al., 2014). Furthermore, a polymorphism of the ATP-binding cassette transporter 1 (associated with low high-density lipoprotein and therefore possibly impaired reverse cholesterol transport) has recently been shown to be protective against advanced AMD (Chen et al., 2010).
Our study also sheds an interesting light on the puzzling differences of the APOE isoform association with AMD (McKay et al., 2011) and AD (Mahley and Rall, 2000), two major age-related neurodegenerative diseases. In AD, the APOE4 allele is associated with greater β-amyloid burden, possibly due to decreased APOE tissue concentrations and reduced efficacy in clearance of β-amyloid clearance via multiple pathways (Bales et al., 2009; Mahley et al., 2009). Cx3cr1GFP/GFP mice that express increased amounts of APOE in all MPs, including MCs (Levy et al., 2015), are protected against β-amyloid deposition in AD mouse models (Lee et al., 2010). In AMD, we show that excessive APOE expression associated with the AMD-risk APOE2 allele leads to the induction of inflammatory cytokines that promote pathogenic subretinal inflammation (Figs. 1, 2, 3) (Levy et al., 2015), similar to Cx3cr1-deficient mice (Combadière et al., 2007; Sennlaub et al., 2013; Peng et al., 2014; Levy et al., 2015). On the other hand, we show that the APOE4 allele is protective in the context of APOE overexpression of Cx3cr1GFP/GFP mice (Levy et al., 2015), due to decreased APOE tissue concentrations (Riddell et al., 2008; Bales et al., 2009; Sullivan et al., 2011) and its reduced capacity to induce inflammatory cytokines (Figs. 4, 5).
Together, our study shows that the APOE2 allele leads to increased APOE expression, IIRC activation, and subretinal inflammation, whereas the APOE4 allele diminishes IIRC activation and inflammation. Our study provides the rationale for the previously unexplained implication of the APOE genotype in AMD and opens avenues toward therapies inhibiting pathogenic chronic inflammation in late AMD.
Footnotes
This work was supported by Institut National de la Santé et de la Recherche Médicale, ANR Maladies Neurologiques et Psychiatriques Grant ANR-08-MNPS-003, Grant ANR Geno 2009 R09099DS, Labex Lifesenses, Carnot, ERC starting Grant ERC-2007 St.G. 210345, and Association de Prévoyance Santé de ALLIANZ.
The authors declare no competing financial interests.
- Correspondence should be addressed to Dr. Florian Sennlaub, Institut National de la Santé et de la Recherche Médicale, Unité Mixte de Recherche S 968, Institut de la Vision, Paris, F-75012, France. florian.sennlaub{at}inserm.fr