Molecular Vision 2014; 20:307-324
Received 19 November 2013 | Accepted 11 March 2014 | Published 14 March 2014
1Lab for Retinal Cell Biology, Department of Ophthalmology, University of Zurich, Switzerland; 2Zurich Center for Integrative Human Physiology, University of Zurich, Switzerland; 3Neuroscience Center Zurich, University of Zurich and ETH Zurich, Switzerland
Correspondence to: C. Grimm, University of Zurich, Ophthalmology, Laboratory for Retinal Cell Biology, Wagistrasse 14, Schlieren, 8952, Switzerland; Phone: +41 44 556 3001; FAX: +41 44 556 3999; email: firstname.lastname@example.org
Purpose: Erythropoietin (EPO) was originally described for its antiapoptotic effects on erythroid progenitor cells in bone marrow. In recent years, however, EPO has also been shown to be cytoprotective in several tissues, including the retina. There, exogenous application of EPO was reported to exert neuro- and vasoprotection in several models of retinal injury. EPO and the erythropoietin receptor (EPOR) are expressed in the retina, but the role of endogenous EPO-EPOR signaling in this tissue remains elusive. Here, we investigated the consequences for cell physiology and survival when EpoR is ablated in rod photoreceptors or in the peripheral retina.
Methods: Two mouse lines were generated harboring a cyclization recombinase (CRE)-mediated knockdown of EpoR in rod photoreceptors (EpoRflox/flox;Opn-Cre) or in a heterogeneous cell population of the retinal periphery (EpoRflox/flox;α-Cre). The function of the retina was measured with electroretinography. Retinal morphology was analyzed in tissue sections. The vasculature of the retina was investigated on flatmount preparations, cryosections, and fluorescein angiography. Retinal nuclear layers were isolated by laser capture microdissection to test for EpoR expression. Gene expression analysis was performed with semiquantitative real-time PCR. To test if the absence of EPOR potentially increases retinal susceptibility to hypoxic stress, the knockdown mice were exposed to hypoxia.
Results: Newborn mice had lower retinal expression levels of EpoR and soluble EpoR (sEpoR) than the adult wild-type mice. In the adult mice, the EpoR transcripts were elevated in the inner retinal layers, while expression in the photoreceptors was low. CRE-mediated deletion in the EpoRflox/flox;Opn-Cre mice led to a decrease in EpoR mRNA expression in the outer nuclear layer. A significant decrease in EpoR expression was measured in the retina of the EpoRflox/flox;α-Cre mice, accompanied by a strong and significant decrease in sEpoR expression. Analysis of the retinal morphology in the two knockdown lines did not reveal any developmental defects or signs of accelerated degeneration in the senescent tissue. Similarly, retinal function was not altered under scotopic and photopic conditions. In addition, EpoR knockdown had no influence on cell viability under acute hypoxic conditions. Retinal angiogenesis and vasculature were normal in the absence of EPOR. However, expression of some EPOR-signaling target genes was significantly altered in the retinas of the EpoRflox/flox;α-Cre mice.
Conclusions: Our data suggest that expression of EPOR in rod photoreceptors, Müller cells, and amacrine, horizontal, and ganglion cells of the peripheral retina is not required for the maturation, function, and survival of these cells in aging tissue. Based on the expression of the EPOR-signaling target genes, we postulate that expression of soluble EPOR in the retina may modulate endogenous EPO-EPOR signaling.
The cytokine erythropoietin (EPO) has long been recognized as the principal hormonal regulator of erythropoiesis, stimulating the growth and promoting the differentiation of early erythroid progenitor cells . In the adult, this cytokine is produced mainly by the kidney, and is secreted into the blood circulation to reach the bone marrow . There, EPO binds to the cognate EPO receptor (EPOR) on erythroid progenitor cells, thus preventing apoptosis and stimulating their differentiation and maturation into erythrocytes . The expression of Epo is oxygen-regulated, and is induced by hypoxia-inducible factors when tissue oxygenation is reduced . Therefore, EPO secretion increases under hypoxic conditions, eventually resulting in an increase in the hematocrit .
The source of Epo expression is not limited to the kidney. In fact, about 10% of EPO found in the bloodstream is of non-renal origin . Numerous tissues, including the brain, have been identified as secreting EPO (reviewed in ). Similarly, EpoR expression is broader than originally reported, being present in, among others, the brain, heart, and liver . Expression of Epo and EpoR has also been found in the retina [9,10]. The widespread tissue distribution of EPOR proposes that the antiapoptotic effects of EPO may go well beyond the prosurvival effects on early erythroid progenitors. In fact, the tissue-protective abilities of this cytokine have been demonstrated in various experimental injury models over the last few years. For example, exogenous application of EPO protected kidney and heart cells against injury provoked by ischemia reperfusion [11-13], and was neuroprotective in different brain injury models . Neuroprotective effects of EPO have also been reported in the retina, including protection of retinal ganglion cells (RGCs) in experimental degenerative models [15-18], as well as preservation of photoreceptor survival after light exposure . Much research effort has been put into elucidating the ability of exogenous applications of EPO to prevent cell death in view of the potential use of this cytokine as a therapeutic agent against degenerative diseases. However, the function of endogenous EPO-EPOR signaling in extrahematopoietic tissues, including the retina, has not yet been fully clarified. This is in part due to the embryonic lethal phenotype of Epo null and EpoR null mice, which die in utero because of impaired production of mature red blood cells [20,21].
In the brain, EPOR is thought to be required for neural progenitor cell (NPC) proliferation and thus for correct brain development. In fact, lack of EpoR results in severe impairment in embryonic neurogenesis [22,23]. A similar mechanism might be present in the developing retina; that is, EPO-EPOR signaling may be essential for retinal progenitor cell (RPC) proliferation, in particular during the period of physiologic hypoxia in postnatal retinal development . In addition to neuroprotection, EPO-EPOR signaling has been linked to developmental angiogenesis through angiogenic and vasoprotective properties . Deletion of Epo or EpoR in mice severely affected angiogenesis and resulted in reduced complexity of the vessel networks, characterized by narrower vessel diameter and reduced vascular branching . Consequently, the absence of EPO-EPOR signaling in the early postnatal mouse retina could potentially impair angiogenesis. To address these questions, we generated a conditional knockdown mouse line lacking the EpoR gene in a heterogeneous cell population of the retinal periphery. We show that the absence of EPOR did not result in overt defects in maturation, angiogenesis, and function of the peripheral retina in these mice, and that the tissue survived acute hypoxic exposure.
EPO-EPOR signaling may be implicated not only in retinal development but also in the ability of retinal cells to cope with chronic stress, which could impair long-term survival. In the adult retina, EPO and EPOR expression follows a circadian rhythm, with increased protein levels before the onset of daily light . This expression pattern hints at a possible role of EPO-EPOR signaling in coping with increased phototoxic stress during the day. To test whether rod-specific EpoR expression is necessary for long-term photoreceptor survival, we generated a conditional knockdown mouse line with a rod-specific ablation of EpoR. We found that expression of EPOR is not necessary for the function and survival of rod photoreceptors.
All animal experiments were conducted according to the regulations of the Cantonal Veterinary Authority of Zurich and the statement of the Association for Research in Vision and Ophthalmology (ARVO) for the use of animals in ophthalmic and vision research. All mice were kept at the animal facility of the University Hospital Zurich in a 12 h:12 h light-dark cycle with food and water ad libitum. Light was maintained at 60 lux at cage level during the light period. Mice were euthanized by CO2 inhalation followed by cervical dislocation. All experiments were conducted with a minimum of n = 3 mice for each time point and condition.
Conditional EpoR knockdowns were generated by breeding EpoRflox/flox mice  carrying loxP sites in exons 1 and 4 of EpoR either to mice expressing Cre recombinase under control of the α-element of the Pax6 promoter (α-Cre)  resulting in EpoRflox/flox;α-Cre double mutant animals or to mice expressing Cre recombinase under control of the rhodopsin promoter (Opn-Cre)  resulting in EpoRflox/flox;Opn-Cre double mutant mice. EpoRflox/flox littermates were used as controls. All mice were on a mixed BL/6;129S6 background and homozygous for the Rpe65450Leu allele. C57BL/6 mice were used to analyze postnatal gene expression. To investigate the spatial expression pattern of Cre recombinase in the α-Cre retina, α-Cre mice were crossed to Ai6 reporter mice (stock number: 007,906, Jackson Laboratory, Bar Harbor, ME) that express ZSGREEN upon CRE-mediated recombination of a floxed STOP cassette . The retinas were cryosectioned and analyzed for native ZSGREEN expression with fluorescent microscopy. Spatial activity of CRE recombinase in the Opn-Cre mice has been previously reported . The mice were genotyped using genomic DNA isolated from ear biopsies with conventional PCR using the following conditions: initial denaturation (95 °C, 5 min); 35 cycles of denaturation (95 °C, 45 s), annealing (temperature see Table 1, 45 s) and elongation (72 °C, 45 s); and final extension (72 °C, 10 min). Primer pairs and annealing conditions are shown in Table 1. PCR products were run on a 1.5% agarose gel for size detection.
Retinas were isolated through a corneal incision and immediately snap frozen in liquid nitrogen. Total RNA was isolated using the RNeasy isolation kit (RNeasy, catalogue number: 74,104; Qiagen, Hilden, Germany) or the High Pure RNA isolation kit (catalogue number: 11,828,665,001; Roche Diagnostics, Basel, Switzerland). Residual genomic DNA was removed by an incubation step with DNase. RNA (650 ng or 1000 ng) was reverse transcribed using oligo(dT) and M-MLV reverse transcriptase (catalogue number: M1701; Promega, Madison, WI). Ten nanograms of cDNA were used for gene expression analysis with semiquantitative real-time PCR using a LightCycler 480 instrument (Roche Diagnostics), the LightCycler 480 SYBR Green I Master kit (catalogue number: 04,887,352,001; Roche Diagnostics), and specific primer pairs (Table 2). The primer pairs were designed to span large intronic sequences or to cover exon-exon boundaries. Gene expression was normalized to actin beta (Actb), and relative quantification was calculated using the comparative threshold method (ΔΔCT) and Light Cycler 480 software (Roche Diagnostics).
To analyze retinal morphology with light microscopy, the eyes were enucleated and fixed in 2.5% glutaraldehyde in 0.1 M cacodylate buffer (pH 7.3) at 4 °C overnight. After fixation, the cornea and the lens were removed, and the eyecup was separated into a superior half and an inferior half by cutting through the optic nerve head. Trimmed tissue was washed in cacodylate buffer, contrasted with osmium tetroxide (1%, 1 h, room temperature), dehydrated by incubation in increasing ethanol concentrations, and embedded in Epon 812. Semithin sections (0.5 μm) were prepared and counterstained with toluidine blue. An Axioplan digitalized microscope (Zeiss Meditec, Jena, Germany) was used to examine the slides. Retinal thickness from the nerve fiber layer to the RPE was measured at fixed distances from the optic nerve head (ONH) using ImageJ (software version 1.43; developed by Wayne Rasband, National Institutes of Health, Bethesda, MD; available NIH).
For immunofluorescence, the eyes were enucleated and fixed in 4% paraformaldehyde (PFA) at 4 °C overnight. The cornea and the lens were removed, and the eyecups were incubated in 30% sucrose in phosphate buffered saline (PBS; 1X; 137 mM NaCl, 2.7 mM KCl, 81 mM Na2HPO4, 19 mM NaH2PO4, pH 7.4) for cryoprotection until the tissue sunk to the bottom of the tube. The tissue was then embedded in tissue cryoprotective medium (catalogue number: 14,020,108,926 Leica Microsystems Nussloch, Nussloch, Germany) and rapidly frozen in a 2-methylbutane bath cooled with liquid nitrogen. Twelve-micrometer-thick sections were cut through the optic nerve head on a Cryostat (Leica), air-dried, and stored at −80 °C until further use. Sections on slides were incubated with blocking solution (3% horse serum in PBS/0.3% Triton X-100) for 1 h at room temperature. Primary antibodies were applied in blocking solution overnight at 4 °C (Table 3). The slides were washed three times with PBS and incubated with Cy2- or Cy3-conjugated secondary antibodies (Jackson ImmunoResesarch, Soham, UK) in blocking solution (dilution 1:500) for 1 h at room temperature. To stain the blood vessels, cryosections were incubated overnight at 4 °C with G. simplicifolia isolectin IB4-Alexa 594 (1:200; catalogue number: I21413; Invitrogen, Basel, Switzerland) in blocking solution. After three washing steps in PBS, the cell nuclei were stained with 4',6-diamidino-2-phenylindole dihydrochloride (DAPI), and the slides were mounted with antifade medium (10% Mowiol 4–88 (vol/vol); Calbiochem, San Diego, CA) in 100 mM Tris (pH 8.5), 25% glycerol (wt/vol), and 0.1% 1,4-diazabicyclo [2.2.2] octane [DABCO]). Immunofluorescently labeled proteins were visualized using an Axioplan fluorescence microscope (Zeiss).
Eyes were enucleated and incubated in 2% PFA in PBS for 5 min. Subsequently, the cornea and the lens were removed, and the retina was carefully separated from the eyecup. The retina was cut into a “clover-leaf” shape, fixed in methanol (−20 °C), and post-fixed in 4% PFA in PBS for 10 min. The flatmounted retinas were washed briefly with PBS and placed in blocking solution (1% fetal calf serum, 0.1% Triton X-100 in PBS) for 1 h. The retinas were incubated overnight at 4 °C with G. simplicifolia isolectin IB4-Alexa 594 (1:100, catalogue number: I21413; Invitrogen) in blocking solution. The retinal flatmounts were washed with PBS and mounted using antifade medium (see above). Imaging was performed using a digitalized light microscope (Axiovision, Zeiss).
The mice pupils were dilated with 1% Cyclogyl (Alcon, Cham, Switzerland) and 5% phenylephrine (Ciba Vision, Niederwangen, Switzerland) 1 h before angiography. Mice were anaesthetized with 2.2 µl/g bodyweight of a ketamine/xylazine mix (51/6 mix, subcutaneous injection) 5–10 min before angiography. Eyes were kept moisturized with hydroxypropylmethylcellulose (Methocel 2%, Omnivision, Puchheim, Germany). For fluorescein angiography, 50 µl of 0.2% fluorescein solution (AK Fluor, Akorn, Lake Forest, IL) were injected intraperitoneally, and the retinal vasculature was analyzed with the Micron III retinal imaging system (Phoenix Research Labs, Pleasanton, CA).
Electroretinograms (ERGs) were recorded according to a previously described procedure [32,33]. The ERG equipment consisted of a Ganzfeld bowl, a direct current amplifier, and a PC-based control and recording unit (LKC UTAS-E 3000, LKC Technologies, Hellendorn, the Netherlands). Mice were dark adapted overnight and anesthetized with ketamine (66.7 mg/kg bodyweight) and xylazine (11.7 mg/kg bodyweight). The pupils were dilated, and single-flash ERG responses were obtained under dark-adapted (scotopic) and light-adapted (photopic) conditions. Light adaptation was accomplished with a background illumination of 30 cd/m2 starting 5 min before recording. Single white-flash stimulus intensity ranged from –3.7 to 1.9 log cd · s/m2 under scotopic and from –0.6 to 2.9 log cd · s/m2 under photopic conditions, divided into ten and eight steps, respectively. Ten responses were averaged with an interstimulus interval (ISI) of either 4.95 s (for –3.7, –3, –2.6, –2, −1.6, –1, –0.6, −0.02, 0.4, and 0.9 log (cd · s/m2)) or 16.95 s (for 1.4, 1.9, 2.4, and 2.9 log (cd · s/m2)).
The eyes were enucleated, immediately embedded in tissue freezing medium (catalogue number: 14,020,108,926; Leica Microsystems Nussloch), and frozen in a 2-methylbutane bath cooled with liquid nitrogen. Retinal sections (20 μm) were fixed (5 min in acetone), air dried (5 min), and dehydrated (30 s in 100% ethanol, 5 min in xylene). Microdissection was performed using an Arcturus XT Laser capture device (Molecular Devices, Sunnyvale, CA). RNA was isolated using the Arcturus kit for RNA isolation (catalogue number: KIT0204; Molecular Devices) according to the manufacturer's directions including a DNase treatment to digest residual genomic DNA. Equal amounts of RNA were used for reverse transcription using oligo(dT) and M-MLV reverse transcriptase (Promega).
The mice were exposed to reduced oxygen levels in a hypoxic chamber (In Vivo cabinet Model 30; Coy Laboratory Products, Grass Lake, MI) equipped with an oxygen controller (Coy Laboratory Products). During hypoxia, mice had access to food and water ad libitum. By altering the O2:N2 ratio, oxygen levels were reduced to the desired concentration in 2% steps over 1 h. Mice were kept at 7% O2 for 6 h. To analyze the gene and protein expression, the retinas were isolated immediately after the hypoxic period and processed for further analysis. To investigate the retinal morphology, the eyes were enucleated 12 days after hypoxic exposure.
Statistical analysis was performed using Prism 4.0 software (GraphPad, San Diego, CA). All data are presented as mean ± standard deviation (SD). The number of samples (n) used for individual experiments is given in the figure legends. Statistical differences of means were calculated using the Student t test. Differences with p values below 0.05 were considered significant.
EpoR expression is elevated in the embryonic brain and decreases into adulthood . A similar situation may be present in the retina. Thus, we analyzed the expression profile of Epo and EpoR in the newborn and adult retinas of C57BL/6 wild-type mice with semiquantitative real-time PCR. Epo was expressed in the newborn retina as early as P3 (Figure 1A). Starting at P10, the expression levels increased by about two- to threefold, reaching more or less steady-state levels in the adult retina (Figure 1A). Similarly, the mRNA levels of EpoR were maintained at moderately low levels up to P10, the time point when expression increased by sevenfold to reach a 14-fold increase at 6 months of age (Figure 1B). The EpoR gene can be expressed through alternative splicing as a membrane-bound (referred to as “EpoR” in the text) and a soluble isoform (soluble EpoR (sEpoR)), which encodes for the extracellular domain of EPOR [34-36]. The expression of this isoform has not yet been reported in the retina. To measure the expression of EpoR and sEpoR, we designed primers specific to each isoform (Appendix 1). Interestingly, the expression profile of sEpoR was similar to the membrane-bound isoform; that is, it was detectable at low levels up to P10, the time point when expression increased about tenfold and reached approximately 20-fold higher levels in the adult retina (Figure 1C). The EPOR not only forms homo-oligomers but has also been suggested to form hetero-oligomers with the beta common receptor (Csf2rb (βCR)) [37,38]. In contrast to EpoR, expression of Csf2rb (βCR) remained at steady-state levels in the retina throughout the analyzed period (Figure 1D).
In addition to elucidating the temporal expression profile of EpoR, we also analyzed the spatial expression pattern in the retinal tissue to test EpoR expression in specific retinal layers. For this purpose, we isolated individual layers (the outer nuclear layer [ONL], the inner nuclear layer [INL], the ganglion cell layer [GCL]) of 12-week-old EpoRflox(flox mice by laser capture microdissection. Analysis of the expression of marker genes for the individual retinal layers showed high purity of the isolated tissue with only a little cross contamination (Appendix 2). EpoR transcripts were detected mainly in the GCL and the INL, with markedly lower levels in the ONL, thus suggesting low expression of EpoR in photoreceptors (Figure 2).
To explore the possible functions of endogenous EPO-EPOR signaling in the retina, we generated two mouse strains bearing a knockdown of EpoR in specific retinal cell populations. We ablated loxP-flanked exons 1–4 of the EpoR genomic sequence either in rod photoreceptors, by generating EpoRflox/flox;Opn-Cre mice, or in an assorted population of neurons and glia of the retinal periphery by generating EpoRflox/flox;α-Cre mice. Note that the lack of exons 1–4 affects the expression of not only EpoR but also sEpoR (Appendix 1).
By using reporter mouse lines [30,39], we assessed the specificity and localization of CRE-mediated recombination. Although Opn-Cre-mediated recombination was specific for the ONL, expression of CRE was patchy and detected in approximately 40% of the total rod population (qualitative estimation) as previously reported . CRE-mediated recombination in the α-Cre mouse was detected in a mixed cell population of the retinal periphery and a smaller subset of cells in the central retina (Appendix 3 ). Immunostaining with antibodies specific for different retinal cells suggested CRE activity in the Müller glia cells, rod photoreceptors, and horizontal, amacrine, and a subset of ganglion cells (Appendix 3). Whereas in Opn-Cre mice, recombination of floxed sequences was maximal only around 10 weeks of age (not shown), CRE activity in α-Cre mice showed early onset at E10.5 .
At the genomic level, Opn-Cre and α-Cre successfully deleted loxP-flanked exons 1–4 in the EpoR gene as evidenced by the appearance of a 220 bp fragment (Figure 3A) that was amplified from recombined genomic DNA when the respective specific primers for PCR were used (Table 1) . To assess the knockdown at the mRNA level, we performed semiquantitative real-time PCR on the total retinal RNA. A statistically significant (p<0.05) 35% decrease in EpoR mRNA expression was measured in the retinas of the EpoRflox/flox;α-Cre mice (Figure 3B). However, the decrease in EpoR mRNA was not significant in the EpoRflox/flox;Opn-Cre mice at P84. This was unexpected, but since the total retinal RNA was used for analysis, the effect of a cell-specific knockout may have been masked by cells that did not undergo recombination, especially if those cells (e.g., in the INL and the GCL) expressed the gene at much higher levels (Figure 2). Consequently, we isolated RNA from individual retinal nuclear layers (the ONL, INL, GCL) separated by laser capture microdissection. Amplification of specific marker genes for each nuclear layer suggested high purity of the isolated layers (Appendix 2). Gene expression analysis showed an EpoR expression level in the ONL of EpoRflox/flox;Opn-Cre mice that was reduced by about 35% compared to the control littermates, even though the knockdown did not reach statistical significance (Figure 3C). This efficiency of EpoR knockdown was similar to the fraction of photoreceptors that express Cre recombinase in the Opn-Cre mouse retina [29,31]. Although the knockdown efficiency was not high when calculated on the level of the total retina or the ONL (cells without active CRE still express EpoR), most Cre-expressing cells (e.g., 30–40% of rods) should have recombined floxed sequences on both alleles leading to many cells harboring a complete knockout and therefore lacking the EPOR protein. Due to the lack of reliable antibodies [42,43], the EPOR protein levels and tissue distribution were not analyzed with western blotting or immunofluorescence.
The absence of EPOR results in increased apoptosis of NPCs and thus affects brain development . Since α-Cre mice recombine floxed sequences in embryonic retinal progenitor cells, the absence of or reduced levels of EPOR may lead to developmental defects and thus to morphological alterations in the adult retina. However, histological analysis did not reveal tangible signs of developmental defects, and normal morphology was observed up to 1 year of age in the retinal periphery of the EpoRflox/flox;α-Cre mice (Figure 4A). Similarly, the retina and in particular the photoreceptors of the EpoRflox/flox;Opn-Cre mice showed normal morphology up to 1 year of age (Figure 4A). Analysis of the expression of rhodopsin (Rho), guanine nucleotide binding protein, alpha transducin 2 (Gnat2), visual system homeobox 2 Vsx2 (Chx10), and POU domain, class 4, transcription factor 1 Pou4f1 (Brn3a) indicated that the number of rods, cones, bipolar cells, and RGCs, respectively, was not affected in the EpoR knockdown mice up to 7.5 months of age (Figure 4B,C). The absence of retinal degeneration in both EpoR knockdown strains was confirmed with morphometric analysis of retinal thickness in 7.5-month-old mice (Figure 4D). Furthermore, ERG recordings pointed to normal retinal function (Figure 5A,B). Neither the EpoRflox/flox;Opn-Cre mice nor the EpoRflox/flox;α-Cre mice showed significant changes in a- and b-wave amplitudes under scotopic or photopic conditions when compared to the control littermates (Figure 5C–E). Taken together, these results demonstrate that expression of EpoR was not necessary for the correct maturation and function of rod photoreceptors and cells in the retinal periphery. In addition, long-term survival of the retinal neurons, including photoreceptors, was not impaired by the absence of EPOR in our knockdown mice.
The induction of Epo expression under hypoxic conditions suggested that this cytokine may be particularly important for preserving cell viability when oxygen is limited [9,44]. In fact, EPO was shown to support neuronal survival against hypoxia-induced cell death in vitro , and in an in vitro model of cerebral ischemia . These findings have been corroborated in vivo in a model of brain ischemia/hypoxia . In line with this notion, neural cells lacking EpoR showed increased sensitivity to hypoxia in vitro . To investigate the ability of EPOR to mediate the resistance of retinal cells against hypoxia, we exposed 12-week-old EpoR knockdown mice to hypoxia (7% O2; 6 h) and analyzed the retinal morphology 12 days after the hypoxic exposure to detect potential signs of tissue damage. Qualitative investigation of retinal morphology did not reveal signs of cell loss upon hypoxic exposure in both strains of EpoR knockdown mice when compared to the control littermates (Figure 6A, compare to Figure 4A for the retinal morphology of normoxic mice). These qualitative findings were confirmed with morphometric measurements of the retinal thickness, which was comparable to that of the normoxic mice (Figure 6B, compare to Figure 4D for retinal thickness of normoxic mice). Thus, ablation of EpoR in the retinal periphery or in rod photoreceptors does not increase retinal susceptibility to acute hypoxia.
EPOR downstream signaling activates several signal transduction pathways. They may ultimately modulate the transcription of target genes involved in various biologic processes, including cell growth or apoptosis . Genes shown to be influenced by EPO signaling include the antiapoptotic genes B-cell lymphoma 2 (Bcl2) and Bcl-2-like 1 (Bcl2l1 [BclXL]) [49,50], the proapoptotic genes caspase 1 (Casp1)  and apoptotic protease activating-factor 1 (Apaf1) [51,52], the neurotrophic factor brain-derived neurotrophic factor (Bdnf) , and the gliotic response gene glial fibrillary acidic protein (Gfap) .
To test the influence of EpoR knockdowns on EPO-EPOR signaling in the retina, we exposed mice at P84 to a short period of hypoxia (7% O2, 6 h) and analyzed the gene expression immediately thereafter (Figure 7). As expected, expression of Epo was increased in all mouse retinas after this treatment. Expression of EpoR was not significantly affected after hypoxia in both knockdown strains. CRE-mediated deletion of exons 1–4 of the EpoR gene prevented expression of not only EpoR but also sEpoR (Appendix 1) [34-36,55]. The expression levels of this isoform were not reduced in the EpoRflox/flox;Opn-Cre retinas when the total retinal RNA was measured. However, sEpoR transcripts were significantly (p<0.001) reduced by approximately two thirds in the retinas of the EpoRflox/flox;α-Cre mice at P84. In both knockdown mice strains, hypoxia did not influence the expression of sEpoR. Importantly, knockdown of EpoR did not lead to compensatory upregulation of Csf2rb (βCR). Whereas expression of Bcl2l1 and Bcl2 was not affected by the lack of EPOR, Apaf1, Casp1, and Gfap were downregulated and the prosurvival factor Bdnf was upregulated specifically in the hypoxic retinas (and thus in the presence of the increased EPO) of EpoRflox/flox;α-Cre mice (Figure 7B). This indicates that lack of EPOR influenced the gene expression profile in the hypoxic retinas in ways that suggested reduced expression of stress-related genes and increased expression of prosurvival genes. These results were unexpected since EPO-EPOR signaling is generally accepted to be prosurvival, and thus, lack of EPOR should reduce such signaling (see the Discussion section). In contrast to the EpoRflox/flox;α-Cre mice, the lack of EPOR in the rods of the EpoRflox/flox;Opn-Cre mice did not lead to detectable changes in the gene expression profile (Figure 7A), suggesting that EPOR signaling may affect the inner retinal layers rather than the photoreceptor layer.
Since evidence suggests that EPOR signaling is involved in angiogenesis , we investigated whether ablation of EPOR in retinal progenitors during development might affect retinal vascular development. Using retinal flatmounts and cross-sections, we did not detect any overt vascular abnormalities in the EpoRflox/flox;α-Cre mice. In particular, the capillary coverage in the retinal periphery was not decreased (Figure 8A), and all three retinal capillary plexi developed properly at their correct location (Figure 8B). Perfusion and stability of the retinal vasculature were also not grossly affected as revealed with fluorescein angiography and the absence of signs of vascular leakage (Figure 8C). As expected, ablation of EPOR in mature rod photoreceptors did not lead to any detectable differences from the wild-type (Figure 8A–C).
Elucidating the expression pattern of EPOR in the retina has been the scope of many vision scientists over the past few years, and various studies have used immunohistochemistry to investigate the expression of EPOR. Unfortunately, the commercially available anti-EPOR antibodies have recently been shown to give unreliable results, and thus, caution should be taken when interpreting these studies [42,43]. In the absence of reliable antibodies, the expression of EpoR should be analyzed with alternative methods. Colella et al. showed with in situ hybridization that EpoR transcripts accumulate at elevated levels in the INL and the GCL . Our data on the expression of EpoR mRNA in the different retinal nuclear layers isolated with laser capture microdissection confirm this observation, and most importantly suggest that photoreceptors do not express high levels of EpoR. Instead, EpoR may be expressed predominantly in the neurons and presumably Müller cells of the inner retina (Figure 2). Interestingly, Chen and coworkers reported that neural progenitors express higher levels of EPOR than mature neurons in the brain . In spite of this, we measured increased retinal expression of EpoR starting at P10, a time point when differentiation of RPCs is completed  (Figure 1B). Thus, in the retina, mature neurons and glia seemed to express higher levels of EpoR compared to the RPCs.
Yu and coworkers showed that EpoR null mice have fewer NPCs . Chen et al. made a similar observation in EpoR null mice that were rescued with selective EpoR expression driven by the endogenous EpoR promoter in hematopoietic tissue but not in the brain . This indicates that endogenous EPO-EPOR signaling supports cell viability in the embryonic brain. However, we did not detect any morphological or functional defects that would hint at developmental abnormalities in the retinal periphery of EpoRflox/flox;α-Cre mice (Figure 4A, Figure 5). Together, these observations indicated that endogenous EPO-EPOR signaling is not crucial for RPC survival and development of retinal neurons (RGCs, amacrine, horizontal, and photoreceptor cells) and glia (Müller cells), at least not in the retinal periphery. Furthermore, the absence of tangible signs of retinal degeneration in the retinal periphery of 1-year-old EpoRflox/flox;α-Cre retinas suggested that EPOR signaling is also not necessary for the survival of retinal cells under non-pathological conditions.
The absence of any detectable consequence of EpoR knockdown in rod photoreceptors may be explained by the low expression level of EpoR in the ONL (Figure 2). This also suggests that EPOR may not be directly involved in photoreceptor physiology and survival, a conclusion supported by the integrity of the retinal tissue in aging EpoRflox/flox;Opn-Cre mice (Figure 4A).
In addition to the membrane-bound isoform, we also detected retinal expression of sEpoR, as has been previously reported for the brain . Expression of sEpoR in the postnatal retinas of wild-type mice followed a profile remarkably similar to the membrane-bound isoform of EpoR (Figure 1C). This is interesting regarding the proposed function of sEPOR. Based on a study published by Soliz et al., sEPOR may act as a negative regulator of EPO signaling. The soluble isoform consists of the extracellular domain and binds EPO, therefore sequestering the cytokine and preventing it from binding to the membrane-bound EPOR, and thus from activating intracellular signaling cascades [60,61]. Soluble receptors frequently modulate cytokine signaling by stabilizing the cytokine, changing its tissue concentration, or modifying its interaction with the membrane-bound receptor [62,63]. For example, intravitreal injection of sEPOR in a rat model of retinal detachment led to increased photoreceptor apoptosis , and coapplication of EPO and sEPOR blocked the protective effects of EPO in an in vitro model of cerebral ischemia . The soluble form of EPOR could have a similar function in the retina and abrogate EPO-EPOR signaling by sequestering endogenous EPO. As a consequence, EPO-EPOR signaling in the retina may occur only under conditions where EPO protein levels are elevated, such as under hypoxia or upon exogenous EPO applications. Intriguingly, the expression of sEpoR in the brain and plasma has been reported to be reduced in hypoxia [60,65], thus allowing advanced modulation of the EPO response under these conditions. However, similar downregulation of sEpoR expression was not observed in the retina after acute hypoxia (Figure 7). Either the hypoxic period was too short to cause downregulation of sEpoR, or regulation of sEPOR synthesis is not controlled at the transcriptional level in the retina.
To understand the regulatory function of sEPOR, the retinal cell populations that express each isoform must be identified. Ablation of EpoR in the retinal periphery of the EpoRflox/flox;α-Cre mice caused a strong and significant decrease in the sEpoR transcript levels of about two thirds, whereas EpoR expression decreased only by roughly one third (Figures 3B, 7B). Based on this observation, we deduced that the population of cells that was CRE-positive in the EpoRflox/flox;α-Cre retina was expressing elevated levels of sEpoR in the wild-type retina under normal conditions. Due to the heterogeneous population of cells undergoing CRE-mediated recombination in the α-Cre mouse retina, it is difficult to make any assumptions about the identity of the retinal cell types that express sEpoR. Nevertheless, based on the unaltered expression of sEpoR in the retinas of the EpoRflox/flox;Opn-Cre mice (Figure 7A), we suggest that rod photoreceptors do not significantly participate in the secretion of sEPOR.
We did not measure significant changes in gene expression in the normoxic or hypoxic retinas of the EpoRflox/flox;Opn-Cre mice (Figure 7A). The relatively low expression of EpoR in the ONL of the EpoRflox/flox;Opn-Cre mice could explain this observation. The gene expression profile in the EpoRflox/flox;α-Cre knockdown mice, however, gave unexpected results. Several genes (Apaf1, Casp1, Gfap) that had been expected to show increased expression in the absence of EPOR were in fact expressed at lower levels in the hypoxic retinas of these mice. The opposite was true for genes that had been expected to be downregulated in EpoRflox/flox;α-Cre mice (Bdnf; Figure 7B). The reduced expression of sEpoR in the retinas of the EpoRflox/flox;α-Cre mice may explain these surprising results. Cells not affected by EpoR ablation (cells that do not express CRE) could show increased EPO-EPOR signaling due to potentially reduced levels of extracellular sEPOR. The diminished presence of sEPOR would result in reduced sequestration of EPO, thus allowing augmented activation of the membrane-bound EPOR and downstream target genes. However, this hypothesis is speculative and clearly needs additional support with experimental data. Analysis of gene and protein expression in mice with efficient knockdown of EpoR in a single retinal cell type would be needed to specifically determine the influence of EPO-EPOR signaling on retinal physiology. Alternatively, the consequences of the absence of EPOR in the retina could be studied in EpoR null mice that are rescued by selective EpoR expression in hematopoietic tissue but not in the neural cells .
The angiogenic activity of EPO during development may be related to the stimulation of endothelial progenitor cell mobilization or proliferation . Expression of EpoR has been detected in various types of vascular endothelial cells , and it is therefore assumed that the angiogenic activities of EPO are directly mediated through binding to EPOR expressed on the surface of vascular endothelial cells. In the α-Cre mouse, Cre recombinase is not expressed in retinal vascular endothelial cells . Therefore, it may not be surprising that the retinal vasculature did not show any signs of developmental defects. Nevertheless, we also ruled out indirect effects suggesting that the EPO-EPOR system in RGCs, amacrine, horizontal, and Müller cells does not significantly contribute to the development of the vasculature in the retinal periphery. However, it would be of great interest to specifically ablate EpoR in vascular cells and analyze the retinal vasculature under these conditions.
Based on data presented in this study, we conclude that EPOR is not required for the development and long-term survival of retinal neurons (RGCs, amacrine, horizontal, and rod photoreceptor cells) and glia (Müller cells) in the retinal periphery. The parallel longitudinal expression profile of EpoR and sEpoR in the postnatal retina suggests that EPO-EPOR signaling may not be constantly activated in this neural tissue.
Appendix 1. Specificity of primers for erythropoietin receptor (EpoR) isoforms.
Appendix 2. Assessment of cross-contamination by semiquantitative real-time PCR of retinal layers isolated by laser capture microdissection.
Appendix 3. Localization of CRE activity in the retina of Ai6;α-Cre reporter mice.
The authors thank Coni Imsand for outstanding technical assistance with retinal histology, Christel Beck and Andrea Gubler for genotyping. This work was supported by the Swiss National Science Foundation (31003A_133043).