Background
Age-related macular degeneration (AMD) is the leading cause of visual impairment in the elderly population in industrialized nations [
1‐
3]. Due to the degeneration of photoreceptors in the cone-rich macula and/or the ingrowth of blood vessels, patients suffering from AMD lose central, high acuity vision [
4‐
6]. While there is no therapy available for geographic atrophy (dry AMD), the neovascular (wet) form of AMD is treated by vascular endothelial growth factor (VEGF)-targeting therapies to slow disease progression [
7‐
9]. Choroidal neovascularization (CNV) defines the classic form of neovascular AMD and is characterized by the ingrowth of blood vessels from the choroid to the subretinal space [
6,
10]. Retinal angiomatous proliferation (RAP) has been described as an additional, distinct form of neovascular AMD [
11]. In RAP, also known as deep retinal vascular anomalous complexes, vessels originate not from the choroid but from the deep retinal plexus in the inner retina and extend into the photoreceptor layer and the subretinal space [
12,
13].
AMD is a multifactorial disease. Besides genetic and environmental risk factors [
5,
14,
15], tissue hypoxia and changes in retinal blood flow have been implicated in its etiology [
16‐
20]. Oxygen supply to photoreceptors in the eyes of elderly people may be impaired due to an age-dependent reduction of choroidal blood flow [
21,
22] and accumulation of drusen [
23]. Choroidal ischemia in dry AMD [
24,
25] and decreased choroidal blood volume in AMD [
26] further support the hypothesis that hypoxia might be implicated in disease development and progression. The retina is considered as one of the most metabolically active tissues and is therefore highly vulnerable to changes in oxygen tension [
17,
27]. In conditions of reduced oxygen supply (hypoxia), molecular responses are activated with hypoxia-inducible factor 1 (HIF1) playing a key regulatory role for adapting the cell/tissue to the new condition. Heterodimeric HIF1 proteins are composed of an oxygen-labile α-subunit and a constitutively expressed β-subunit [
28]. Under normoxic conditions, HIF1A is hydroxylated by prolyl hydroxylase domain (PHD) proteins. This promotes the interaction with the von Hippel-Lindau (VHL) ubiquitin E3 ligase complex leading to ubiquitination and rapid degradation of hydroxylated HIF1A by proteasomes. Under hypoxic conditions, hydroxylation of HIF1A is reduced. Hence, HIF1A accumulates, enters the nucleus and drives transcription of a multitude of target genes [
29‐
31].
In this study, we investigated the consequences of a chronic hypoxia-like response triggered in cone photoreceptors to elucidate the mechanisms of cell death in cone degenerative diseases such as AMD. To this end, we used
R91W;Nrl
-/-
double-mutant mice which express only cone photoreceptors in a well-layered, functional retina [
32]. To induce the hypoxia-like response, we ablated the VHL protein specifically from cones using the Cre-loxP system. We analyzed the effects of the cone-specific activation of the hypoxic response and validated the contribution of HIF1A to the resulting retinal pathology.
Methods
Mice
All experimental procedures were performed according to ‘The Association for Research in Vision and Ophthalmology’ statement on animal use in ophthalmic and vision research and the regulation of the veterinary authorities of Kanton Zurich, Switzerland.
R91W;Nrl
-/-
mice were generated by crossing
Rpe65
R91W
(R91W) [
33] to
Nrl
-/-
mice [
34], and were described recently [
32].
BPCre;R91W;Nrl
-/-
;Vhl
f/f
(=
cone
ΔVhl
) mice were generated by breeding
R91W;Nrl
-/-
mice to
Vhl
f/f
mice [
35] and mice expressing the
Cre recombinase under the transcriptional control of the blue cone opsin (BP) promoter [
36]. To generate
BPCre;R91W;Nrl
-/-
;Vhl
f/f
;Hif1a
f/f
(
=cone
ΔVhlHif1a
) mice,
BPCre;R91W;Nrl
-/-
;Vhl
f/f
were bred to
Hif1a
f/f
mice [
37].
R91W;Nrl
-/-
;Vhl
f/f
and
R91W;Nrl
-/-
;Vhl
f/f
;Hif1a
f/f
littermates without
Cre recombinase served as respective control mice (=
ctrl). Genotyping was performed by PCR using DNA isolated from ear clips and primer pairs as described previously [
38,
39]. Presence of
BPCre was tested using the following primer pair: forward (5’-GGACATGTTCAGGGATCGCCAGGCG-3’) and reverse (5’-GCATAACCAGTGAAACAGCATTGCTG-3’). The amplification reaction resulted in a 268 bp fragment in the presence of the transgene. To test for deletion of floxed sequences, genomic DNA was isolated from retinal tissues and tested by PCR using appropriate primer pairs as described [
38,
39].
129S6 (Taconic, Ejby, Denmark) mice were used as wild-type controls. To test expression of the
Cre recombinase
BPCre;R91W;Nrl
-/-
mice were bred to a ZsGreen reporter line (
Ai6 mice,
Gt(ROSA)26Sor
tm6(CAG-ZsGreen1)Hze
, [
40]). Mice of both sexes were used for experiments and were housed at the animal facility of the University Zurich under a 14 h : 10 h light/dark cycle with lights on at 6 am and lights off at 8 pm. Food and water were provided
ad libitum.
Morphology/quantification
To evaluate retinal morphology, eyes were enucleated and fixed in 2.5% glutaraldehyde in cacodylate buffer (pH 7.2, 0.1 M), according to the previously described procedure [
41]. By cutting through the optic nerve head, nasal and temporal halves of the eyecups were separated and embedded in epon plastic. Semi-thin cross sections (0.5 μm) were counterstained with toluidine blue and analyzed by light microscopy (Axioplan; Zeiss, Jena, Germany). Thickness of the outer nuclear layer was measured at indicated distances from the optic nerve head using the Adobe Photoshop CS6 ruler tool (Adobe Systems, Inc., San Jose, CA, USA) on reconstructed retinal panorama images.
Immunofluorescence
After euthanasia, eyes were marked nasally, enucleated and fixed in 4% paraformaldehyde (PFA) in phosphate buffer for 1 h at 4°C. Cornea and lens were removed and the dissected eyecups postfixed for 2 h in 4% PFA. After immersion in 30% sucrose (in PBS 0.1 M) the eyecups were embedded in tissue freezing medium (O.C.T., Leica Biosystems Nussloch GmbH, Nussloch, Germany), frozen in a 2-methylbutane bath cooled by liquid nitrogen and stored at -80°C. Cryosections (12 μm) were blocked (3% normal goat serum (Sigma-Aldrich, St. Louis, MO, USA), 0.3% Triton X-100 (Sigma) in PBS) and incubated with the following primary antibodies overnight at 4°C: Isolectin GS-IB4-Alexa594 from Griffonia simplicifolia (1:300, I21413; Thermo Fisher Scientific, Waltham, MA, USA), rabbit anti-allograft inflammatory factor 1 (alias IBA1, 1:1000, 019-19741; Wako, Neuss, Germany), rabbit anti-albumin (ALB, 1:500, RARaAlb; Nordic Immunology, Tilburg, Netherlands). Sections were washed with PBS, incubated with secondary antibodies (Cy3-labeled, Jackson ImmunoResearch Laboratories, Westgrove, PA, USA) for 1 h at room temperature (RT), counterstained with DAPI (4',6-Diamidine-2'-phenylindole dihydrochloride, Roche, Basel, Switzerland) and analyzed by fluorescence microscopy (Axioplan; Zeiss).
Analysis of retinal vasculature in whole mounted retinas
Eyes were isolated and incubated for 5 to 10 minutes in 2% PFA in PBS as described recently [
42]. After removal of cornea and lens, the retina was dissected and flat-mounted in PBS. After postfixation in 4% PFA for 1 h at RT, flat-mounts were blocked (3% normal goat serum, 0.3% Triton X-100 in PBS, 1 h) and incubated with isolectin GS-IB
4-Alexa594 (1:300, Thermo Fisher Scientific) at 4°C overnight. Retinas were washed in PBS, mounted on glass slides and analyzed by fluorescence microscopy (Axioplan/ApoTome; Zeiss). Blood vessels were reconstructed in three dimensions using Imaris software (Versions 7.7.2/8.3.0, Bitplane AG, Zurich, Switzerland). For better recognition and distinction of the vascular plexi, the z-value of the z-stacks was increased five times.
RNA isolation and semi-quantitative real-time PCR
Retinas were isolated through a slit in the cornea, frozen in liquid nitrogen and stored at -80°C. RNA was extracted using an RNA isolation kit (RNeasy; Qiagen, Hilden, Germany) including a DNAse treatment. 1 μg of RNA, oligo (dT) and M-MLV reverse transcriptase (Promega, Fitchburg, WI, USA) were used to prepare cDNA. To analyze gene expression by real-time PCR, 10 ng of cDNA template was amplified using a PCR polymerase ready mix (LightCycler 480 SYBR Green I Master, Roche Diagnostics, Rotkreuz, Switzerland), specific primer pairs (Table
1) and a thermocycler (LightCycler 480, Roche Diagnostics). Expression levels were normalized to β-actin (
Actb) and relative expression was calculated using the comparative threshold cycle method (∆∆C
T) of the LightCycler480 software (Roche Diagnostics). At least 3 mice per strain and time point were used.
Table 1
Primers used for real-time PCR
Actb
| CAACGGCTCCGGCATGTGC | CTCTTGCTCTGGGCCTCG | 153 |
Adm
| TCCTGGTTTCTCGGCTTCTC | ATTCTGTGGCGATGCTCTGA | 133 |
Bnip3
| CCTGTCGCAGTTGGGTTC | GAAGTGCAGTTCTACCCAGGAG | 93 |
Casp1
| GGCAGGAATTCTGGAGCTTCAA | GTCAGTCCTGGAAATGTGCC | 138 |
Fgf2
| TGTGTCTATCAAGGGAGTGTGTGC | ACCAACTGGAGTATTTCCGTGACCG | 158 |
Gfap
| CCACCAAACTGGCTGATGTCTAC | TTCTCTCCAAATCCACACGAGC | 240 |
Glut1
| CAGTGTATCCTGTTGCCCTTCTG | GCCGACCCTCTTCTTTCATCTC | 151 |
Pdgfb
| GCTGCTGCAATAACCGCAAT | GTGGTCCTCCAAGGTCACTG | 131 |
Pdgfrb
| CTTGCCCTTCAAAGTGGTGG | CCAGGTGGAGTCGTAAGGC | 199 |
Egln1
| GCAGCATGGACGACCTGAT | CAACGTGACGGACATAGCCT | 123 |
Sema3f
| CGTCGCGCACAGGATTA | GGAAAATGGCTGCATCGGTA | 166 |
Timp3
| GCCTCAAGCTAGAAGTCAACAAA | TGTACATCTTGCCTTCATACACG | 69 |
Vegf
| ACTTGTGTTGGGAGGAGGATGTC | AATGGGTTTGTCGTGTTTCTGG | 171 |
vWF
| CCCTGGACAACTTGACAGCAG | ACAAGCAGGCAGATCTCATACC | 192 |
Protein isolation and Western blotting
Proteins were isolated by homogenizing the retinas in ice-cold Tris-HCl (100 mM, pH 7.5) using ultrasound at 4°C. Protein concentrations were determined spectrophotometrically by using Bradford reagent (Bio-Rad, Hercules, CA, USA). SDS-PAGE and Western blotting were performed as described [
41] and proteins detected by rabbit anti-HIF1A (1:2000, NB100-479, Novus Biologicals, Littleton, CO, USA) and mouse anti-β-actin (1:10’000, A5441, Sigma-Aldrich) antibodies. HRP-conjugated secondary antibodies were applied for 1 h at RT and signals were visualized using Western lightning chemiluminescence reagent (PerkinElmer, Waltham, MA, USA) and X-ray films.
Electroretinography (ERG)
Mice were dark-adapted overnight and pupils were dilated under dim red light with 1% cyclogyl (Alcon Switzerland SA, Rotkreuz, Switzerland) and 5% neosynephrin-POS (Ursapharm Schweiz GmbH, Roggwil, Switzerland) 30 minutes prior to recording. Mice were anesthetized with a subcutaneous injection of ketamine (85 mg/kg, Pfizer PFE Switzerland GmbH, Zurich, Switzerland) and xylazine (Rompun 2%, 4mg/kg, Bayer, Leverkusen, Germany). To keep the cornea moist, a drop of mydriaticum dispersa (Omnivision AG, Neuhausen, Switzerland) was applied to each eye. Low background illumination for 5 min was used for light adaptation. Electroretinograms were recorded simultaneously from both eyes with an LKC UTAS Bigshot unit (LKC Technologies, Inc. Gaithersburg, MD, USA) as described [
32]. Flashes of 8 different light intensities ranging from -10 to 25 dB (0.25–790.5694 cd*s/m
2) were applied under photopic conditions. 10 responses were averaged per light intensity. Traces from n≥4 mice were averaged for each light intensity.
Fundus imaging and fluorescein angiography
Fundus photographs were taken with a Micron IV system (Phoenix Research Labs, Pleasanton, CA, USA). The pupils were dilated and the mice were anesthetized as described above. Methocel 2% (Omnivision) was applied to lubricate the eyes. Fluorescein images were captured 1-5 minutes after intraperitoneal injection of 20 μL of 2% fluorescein solution (Akorn, Lake Forest, IL, USA).
Experimental design and statistical analysis
Two-way ANOVA with Šídák’s multiple comparison test was performed using GraphPad Prism (version 7.02, GraphPad Software, San Diego, CA, USA) for statistical analysis of gene expression levels and ERG traces. All data are shown as means ± SD.
Discussion
Previously, it has been demonstrated that the knockdown of
Vhl in rod cells causes HIF1A stabilization under normoxic conditions [
39]. Here, we particularly investigated the effects of a chronic hypoxia-like response in cone photoreceptors. We show that ablation of
Vhl in cones resulted in a chronic hypoxia-like situation with accumulation of HIF1A and induction of HIF1A target genes. This led to pathological vessel growth into the photoreceptor layer, reduced retinal function and severe, progressive cone degeneration. These consequences resemble some of the features of AMD pathology, particularly of a subset of neovascular AMD known as RAP. In human patients, HIF1A and HIF2A were detected in macrophages and endothelial cells of neovascular membranes associated with AMD [
62]. Additionally, the correlation of drusen density and decreased choroidal blood flow [
26] as well as choroidal ischemia in AMD patients [
23‐
25] suggest reduced oxygen transport from the choroid to the inner retina [
63]. Therefore, tissue hypoxia and HIFs might play an important role in disease development and/or progression.
Expression of numerous genes involved in metabolism, stress, cell survival and angiogenesis was increased in
cone
∆Vhl
mice (Fig.
1d,
3c,
5c), as was
Timp3 (Fig.
3c). TIMP3 is an important regulator of extracellular matrix (ECM) remodeling through inhibition of matrix metalloproteases and has been suggested to be a senescence-related protein [
64] that is potentially regulated by HIF2 in rods [
38]. Importantly, TIMP3 inhibits angiogenesis by interacting with VEGF [
65] and is relevant for Bruch’s membrane integrity [
64,
66]. Furthermore, mutations in
Timp3 have been associated with Sorsby’s fundus dystrophy, an autosomal dominant maculopathy with submacular choroidal neovascularization [
67,
68]. We can only speculate about the reasons for elevated
Timp3 during the aging process in
cone
∆Vhl
mice but it seems plausible that retinas of
cone
∆Vhl
mice require extensive ECM remodeling due to progressive cone degeneration (Fig.
3).
We detected early and prominent vascular defects with abnormal vessel growth into the ONL, subretinal space and RPE layer in
cone
∆Vhl
mice. Increased VEGF levels in transgenic mice expressing
Vegf under the transcriptional control of the rhodopsin promoter (
rho/VEGF mice) were shown to cause subretinal neovascularization [
69‐
71]. In our model,
Vegf was not upregulated at PND14, at a time when vessels had already started growing into the ONL (Fig.
5). However, transient
Vegf expression in the INL regulates the formation of the deep plexus during development [
49], i.e. already minor changes in the local
Vegf concentration gradient, which are not detectable by RT-PCR, may misguide retinal vessels in
cone
∆Vhl
mice. To date, it remains unclear which pro-angiogenic factor(s) are responsible for the observed neovascularization. Nonetheless, our findings demonstrate that this is a
Hif1a-mediated mechanism, as additional deletion of
Hif1a fully rescued the vascular defects (Fig.
6).
Interestingly, cone-specific deletion of
Hif1a did not affect formation of the three vascular plexi (Fig.
6e), while it has been shown that knockdown of
Hif1a in most cells of the retinal periphery prevented formation of the intermediate plexus [
42]. Thus, our results suggest that
Hif1a expression in cones, as opposed to other cells presumably in the INL, is not essential for the development of the intermediate plexus.
Our data do not define whether cone degeneration in
cone
∆Vhl
mice is due to the effects of the chronic intrinsic activation of the hypoxic response or to early development of vascular defects. Another mouse model shows, however, that a chronic hypoxia-like response in rods leads to slow photoreceptor degeneration in the absence of reported vascular defects (
rod
∆Vhl
mice [
39]). This suggests that the long-term activation of HIF1 may reduce cell survival by an intrinsic mechanism. During hypoxia, HIF1 regulates mitochondrial respiration and thus metabolic adaptation towards glycolysis [
72] leading to reduced energy (ATP) levels and potentially starvation. It has been shown that starvation of cones may lead to cone cell death [
73]. Furthermore, it has been suggested that cones might be more sensitive to reduced oxygen and nutrient levels compared to rods, for reasons not fully understood [
74,
75]. This could possibly explain the progressive degeneration and the accelerated phenotype in
cone
∆Vhl
compared to
rod
∆Vhl
mice.
It might also be possible, however, that photoreceptor degeneration is secondary to the early pathological vascular defects. Inactivation of
Vhl in the retinal periphery during development leads to vessel growth into the ONL and severe retinal degeneration [
52,
76]. Whereas the retinal periphery was strongly degenerated at 10 weeks of age, photoreceptor loss in the normal vascularized central retina was less pronounced [
52]. We observed microglia activation and impaired function of the blood-retina-barrier in
cone
∆Vhl
mice, as shown by IBA1 and ALB staining, respectively (Fig.
4h-k). It seems likely that retinal hemorrhages exacerbate cone photoreceptor degeneration in our model, as suggested for other models [
77,
78]. Photoreceptor degeneration was also observed in very low density lipoprotein receptor knockout mice (
Vldlr
-/-
; [
79]) where early retinal neovascularization causes vessels to extend from the deep plexus into the subretinal space [
80,
81]. Interestingly, it has been suggested that the
Vldlr
-/-
phenotype is linked to HIF1A. Joyal and colleagues proposed that starved
Vldlr
-/-
photoreceptors have reduced amounts of the Krebs cycle metabolite α-ketoglutarate, which decreases PHD activity and thus promotes stabilization of HIF1A. Consequently,
Vegf is secreted, leading to RAP-like neovascularization [
82]. We hypothesize that the combination of early vascular defects and the activation of a chronic hypoxia-like response, which affects expression of genes involved in cellular metabolism and retinal stress, leads to the progressive cone degeneration observed in
cone
ΔVhl
mice.
Our findings identify HIF1 in cones as the factor causing cone degeneration, pathological neovascularization and loss of function. Similarly, chronic activation of the hypoxic response in rods resulted in an HIF1- and age-dependent retinal degeneration (Barben et al., submitted). Whereas this provides strong evidence that activation of HIF1 in rod and cone photoreceptors leads to retinal degeneration, the activation of HIF2 but not of HIF1 is responsible for metabolic stress in RPE upon RPE-specific
Vhl inactivation [
75]. Thus, our data and data published by others suggest cell type-specific roles for HIF1A and HIF2A in retinal pathology.