Long-Term Tolerance to Retinal Ischemia by Repetitive Hypoxic Preconditioning: Role of HIF-1α and Heme Oxygenase-1

Abstract

purpose. To determine whether the duration of ischemic tolerance in the retina could be extended by repetitive presentations of the preconditioning stimulus and to begin to elucidate the mechanistic underpinnings of the resultant novel phenotype.

methods. Adult male Swiss-Webster ND4 mice were repeatedly preconditioned with systemic hypoxia (RHP) over 12 days; 4 weeks later, the mice were subjected to 30 minutes of unilateral retinal ischemia. Protection was quantified morphologically and functionally 1 week after ischemia by histologic analyses and scotopic electroretinography, respectively. Temporal expression patterns of hypoxia-inducible factor (HIF)-1α and heme oxygenase (HO)-1 were measured in response to RHP and after retinal ischemia by immunoblot analysis and immunohistochemistry.

results. Morphologic and functional protection against ischemia-induced reductions in retinal layer thicknesses and layer cell counts, and a- and b-wave amplitudes, was documented for at least 4 weeks after RHP. There was no evidence of tissue injury or dysfunction by RHP alone. Temporally associated with this period of long-term tolerance (LTT) to retinal ischemia were sustained increases in retinal levels of HIF-1α and HO-1 protein lasting at least 1 and 4 weeks, respectively, after the last RHP stimulus.

conclusions. A novel form of sustained retinal ischemic tolerance is described, wherein endogenous adaptive responses triggered by repeated hypoxia afford protection against injury many weeks after the preconditioning stimulus. HIF-1α -mediated, long-lasting increases in retinal HO-1 expression may contribute to the LTT phenotype. Further elucidation of the genetic and molecular basis of such adaptive plasticity could provide therapeutic targets for preventing and/or treating a variety of ischemic retinopathies.

Ischemic tolerance, defined as a transient increase in resistance to acute ischemic injury as the result of a preceding exposure to a nondamaging “preconditioning” stimulus, is now well established in the retina, ¹ as in other parts of the CNS. ² Brief retinal ischemia ^{3 4 5} and systemic hypoxia ⁵ can serve as ischemic tolerance-inducing preconditioning stimuli, in both rats ^{3 4} and mice, ⁵ and in cell culture models ⁶ of retinal ischemia. Although providing robust protection, the duration of the endogenous adaptive response induced by ischemic and hypoxic preconditioning in vivo is brief, typically only lasting from 1 to 3 days after the preconditioning stimulus. To date, the possibility that this endogenous neuroprotective phenotype could be extended for longer periods of time, and, in turn, that identifying and documenting such a phenomenon could provide therapeutic targets for protecting against disorders characterized by the more progressive retinal neurodegeneration seen in glaucoma and other ischemic retinopathies, remains unexplored. 

The present study was undertaken based on our hypothesis that repeated presentations of the preconditioning stimulus, if appropriately administered in a noninjurious manner, would extend the duration of the ischemia-tolerant state. Morphologic and functional evidence is provided herein that increasing the frequency of hypoxic preconditioning, without changing its magnitude or duration, is effective in establishing what we have termed long-term tolerance (LTT), wherein a neuroprotective phenotype is exhibited in retina that lasts many weeks after the last of a series of repetitive hypoxic preconditioning (RHP) stimuli. 

We also began to examine the induction and expression mechanisms whereby this protracted protective response could be triggered and maintained after RHP. Spatiotemporal changes in the retinal expression of the transcriptional activator hypoxia-inducible factor (HIF)-1α were measured in response to RHP. As detailed in recent reviews, ^{7 8} once stabilized by hypoxia, HIF-1α binds to hypoxia response elements in the promoter region of dozens of genes that promote cell survival and other adaptive responses to stress. ⁹ Growing evidence indicates that hypoxia-driven changes in many of these HIF-1α target genes may be integral to establishing ischemic tolerance in the CNS. ⁸ Moreover, HIF-1α was identified as a possible regulator of hypoxic preconditioning-induced tolerance to phototoxic injury in the retina. ¹⁰  

We also investigated changes in the retinal expression of heme oxygenase (HO)-1, an HIF-1α target gene, ¹¹ after hypoxic preconditioning, and after ischemia in untreated and hypoxia-preconditioned mice. HO-1 (also known as heat shock protein-32) is the isoform of heme oxygenase that is induced by a variety of cellular stresses, including hypoxia. ^{11 12} HO-1 catalyzes the rate-limiting oxidation of heme to the biologically active molecules biliverdin and bilirubin, releasing iron and carbon monoxide in the process (for reviews, see Refs. ^{13 14} ); these HO-1 enzyme-generated products exhibit free radical scavenging, ¹⁵ vasodilating, ¹⁶ anti-inflammatory, ^{17 18 19} and antiapoptotic ²⁰ actions. Given these protective effects, we hypothesized that RHP-induced elevations in HO-1 activity contribute to the long-lasting retinal protection we observed in our LTT model. 

Materials and Methods

Hypoxic Preconditioning and Acute Retinal Ischemia

All experimental methods and animal care procedures adhered to the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research and were approved by the Animal Studies Committee at Washington University School of Medicine. Studies were conducted in adult (10–16 weeks of age) male Swiss-Webster ND4 mice (25–34 g; Harlan Sprague-Dawley, Indianapolis, IN). The procedures for hypoxic preconditioning and acute retinal ischemia have been described in detail in our previous reports. ⁵ In brief, conscious mice were exposed to either a single or a repetitive hypoxic preconditioning protocol (SHP and RHP, respectively) by being placed in a hypoxic (11% oxygen) chamber. For SHP, animals were subjected to one hypoxic preconditioning stimulus of 2 hours’ duration. For RHP, animals were subjected to six hypoxic preconditioning stimuli, one 2-hour exposure to 11% oxygen per day, every other day, for 12 days. After each preconditioning treatment, the mice were returned to their home cages in the animal facility. 

The mice were then killed by cervical dislocation at various times after SHP or RHP for analysis of retinal protein profiles, or subjected to retinal ischemia 1 week after SHP, or 4 weeks after RHP. For the latter, mice were anesthetized and acute retinal ischemia was induced by elevation of intraocular pressure (IOP) to 90 mm Hg for 30 minutes by placement of a needle (connected to a saline reservoir) into the anterior chamber. ⁵ All animals were randomized to these experimental groups and treatment conditions. 

Histopathology

The animals were euthanatized at 1 week after ischemia, the eyes were perfusion fixed and embedded in paraffin, and 5-μm-thick serial sections were obtained and stained with hematoxylin-eosin. Image analysis software (Image-Pro Plus; Media Cybernetics, Silver Spring, MD) was used to measure the thickness of different retinal layers, including the distance from the outer limiting membrane (OLM) to the inner limiting membrane (ILM; OLM–ILM), the thickness of the inner nuclear layer (INL), and the thickness of the inner plexiform layer (IPL). The total number of morphologically identified, Nissl-positive viable cells was also quantified in the INL and ganglion cell layer (GCL). All counts were performed in the same topographic region of the retina, to avoid the possibility of regional anatomic variation in the results. ²¹ As described in our earlier publication, ⁵ the three morphometric measurements were made in each section over a region 750 to 1000 μm from the edge of the optic nerve head and averaged across triplicate sections obtained at a 50-μm distance from one another, in all four retinal quadrants. Cell counts in the INL and GCL were made over a 75- and 250-μm region of retina. All measures were performed in a masked fashion with regard to prior sham or experimental procedures. 

Electroretinography

Full-field electroretinograms (ERGs) were recorded simultaneously from both eyes (UTAS-E 3000 Visual Electrodiagnostic System; LKC Technologies, Gaithersburg, MD). Simultaneous recordings allow ERGs to be obtained from the treated and untreated eyes under identical states of anesthesia and adaptation. The animals were dark adapted overnight and anesthetized with chloral hydrate and xylazine, as previously described. ⁵ The mice were placed on a heating pad, and core temperature was maintained at 36.5°C to 37.0°C throughout the procedure. The pupils were dilated for recordings with 1% tropicamide and 2.5% phenylephrine, and the corneas were kept moist with application of 1.0% carboxymethylcellulose, as needed. Stimuli were brief white flashes delivered via a Ganzfeld integrating sphere, and signals were recorded with band-pass settings of 0.3 to 500 Hz. After a 10-minute stabilization period in the dark, a scotopic intensity–response series (11 recordings, from −3.60 to 0.88 log cd-s/m²) was recorded that included rod-specific and bright-flash responses. Wave amplitudes were calculated according to standardized methods. ²²  

Immunoblot Analysis

At 0, 2, 24, and 48 hours, and at 1, 2, and 4 weeks after the last SHP or RHP treatment, the mice were killed by cervical dislocation, and the retinas were rapidly isolated and frozen. Whole cell lysates were prepared from retinas pooled from four retinas of two mice, as follows: Retinas were added to 160 μL 1× cell lysis buffer (Cell Signaling Technologies, Beverly, MA), homogenized in a glass tube (40 strokes), and then placed on ice for 30 minutes. After vortexing and centrifuging at 12,000 rpm for 20 minutes at 4°C, supernatants were collected, and total protein concentration was determined by BCA protein assay kit (Pierce Biotechnology, Rockford, IL). For HIF-1α, 120 μg of protein was loaded; for HO-1, HO-2, and β-actin, 30 μg of protein was loaded. The proteins were separated on 4% to 20% of SDS-gradient gels (Bio-Rad Laboratories, Hercules, CA) and transferred to nitrocellulose membranes (Hybond-ECL; GE Healthcare Biosciences, Piscataway, NJ). After the reaction was blocked for 1 hour in 5% nonfat dried milk in TBST, the membranes were incubated at 4°C overnight with 1:200 polyclonal rabbit anti-mouse HIF-1α antibody (Santa Cruz Biotechnology, Inc., Santa Cruz, CA), 1:2,000 polyclonal rabbit anti-mouse HO-1 or rabbit anti-mouse HO-2 antibodies (Stressgen Biotechnologies, Inc., San Diego, CA), or 1:10,000 monoclonal mouse anti-mouse β-actin antibody (Sigma-Aldrich, St. Louis, MO). The membranes were subsequently washed and probed for 1 hour with an HRP-linked antibody (Cell Signaling Technology) to goat anti-rabbit IgG at a dilution of 1:5000, or horse anti-mouse IgG at a dilution of 1:8000. An enhanced chemiluminescence (ECL) Western blot detection system (Cell Signaling Technology) was used to detect the signals according to the manufacturer’s instructions. After scanning, protein in the bands was quantified (Image Pro Plus software; Media Cybernetics) and then normalized to protein levels in naive controls. 

Immunohistochemical Analysis

Paraffin-embedded sections of mouse retina were used to identify the cellular localization of HIF-1α and HO-1 within this tissue by immunohistochemistry. After deparaffinization and rehydration, antigen retrieval was achieved by immersing the slides in boiling antigen unmasking solution (Vector Laboratories, Burlingame, CA) for 10 minutes. Nonspecific binding was blocked with prediluted normal goat serum (Vectastain Elite ABC kit; Vector Laboratories) for 20 minutes at room temperature. The sections then were incubated overnight at 4°C with primary antibody (a chicken anti-mouse HIF-1α antibody [1:50], and a polyclonal rabbit anti-mouse HO-1 antibody [1:200; Stressgen]). To visualize the immunolabels, we incubated the sections with secondary antibodies, which included a rabbit anti-chicken IgY, HRP conjugate (1:500; Promega, Madison, WI) for HIF-1α with a diaminobenzidine-peroxidase conjugate (Vector Laboratories), and a biotinylated goat anti-rabbit IgG for HO-1 with an avidin-biotin-peroxidase conjugate (Vector Laboratories). Hematoxylin was used for nuclear counterstaining. As a control for autofluorescence, negative control slides were run without primary antibody. Slide sections were examined by light microscopy. 

Statistics

Differences in the morphologic endpoints and protein expression by immunoblot analysis were compared by the nonparametric Kruskal-Wallis ANOVA rank sum test. Paired data in the ERG studies was compared by the nonparametric signed rank test. P < 0.05 was accepted as significant. 

Results

Long-Term Tolerance: Histologic Protection

We showed previously that exposing mice to a single episode of mild systemic hypoxia provides ischemic tolerance to the mouse retina when ischemia is induced 24 hours later. ⁵ Related purposes of the present study were to determine how long this neuroprotective phenotype lasts and whether multiple exposures to hypoxia would extend the duration of this protective effect even further. The histology data in Figure 1show that, in eyes without any preconditioning treatment, 30 minutes of retinal ischemia reduced OLM–ILM thickness by 23%, and the thickness of the INL and IPL by 21% and 31%, respectively. Cell loss in the INL and GCL was 24% and 33%, respectively. We found that, even when administered 1 week before ischemia, an SHP exposure completely protected against all these morphologic indices of retinal ischemic injury. The extent of protection was 104% to 111% for retinal thickness, and 75% to 94% for cell counts (relative to the ischemia-induced loss). Representative histologic sections of control, untreated ischemic, and preconditioned ischemic retinas are shown in Figure 2 . However, waiting 2 weeks after SHP to induce ischemia was only partially effective in protecting the retina histologically; specifically, no improvements in ischemia-induced cell loss in the INL and GCL were noted, but ischemia-induced reductions in layer thicknesses were still reversed significantly by SHP (data not shown). 

Figure 1also shows the ability of repetitive hypoxic preconditioning (RHP), consisting of six exposures to systemic hypoxia over 2 weeks, the last exposure of which was 4 weeks before 30 minutes of injurious retinal ischemia, to provide complete histologic protection against the resultant ischemic injury. In these animals, the protection in layer thickness ranged from 89% to 95%, and, in cell counts, from 72% to 85%. Figure 2includes a representative histologic section documenting such protection. If we waited 8 weeks between the last of the RHP stimuli and the induction of retinal ischemia, no histologic protection was observed in terms of ischemia-induced layer thinning, but, cell counts in the INL and GCL were still significantly preserved, although not to the extent observed after waiting only 4 weeks (data not shown). 

Long-Term Tolerance: Functional Protection

Having documented both short-term and long-term tolerance (LTT) to retinal ischemia using several morphologic criteria, we then pursued functional evidence of such protection. Performing scotopic flash ERG studies 1 week after retinal ischemia in animals without prior preconditioning demonstrated that, at maximum stimulus intensities, ischemia reduced a- and b-wave amplitudes by 42% (P = 0.016) and 46% (P = 0.008), respectively (n = 8; data not shown). However, if mice were treated with SHP 1 week before ischemia, postischemic a-wave amplitudes recovered 46% (P = 0.04 vs. ischemia only; n = 7; data not shown), as did b-wave amplitudes (P = 0.029 vs. ischemia only; n = 7; data not shown). 

Full-field bilateral ERG recordings were obtained at baseline and at 1 week after ischemia on animals subjected to ischemia alone or ischemia with RHP 4 weeks earlier. ERGs were also obtained in mice subjected to RHP only and from matched, nonpreconditioned, nonischemic control animals (Fig. 3) . As observed in the group of mice matched for the SHP studies, 30 minutes of retinal ischemia impaired the ERG response, evidenced by fairly equal reductions in the amplitude of the inner retina-dominant b-wave response across low to high stimulus intensities, whereas the photoreceptor-dominant a-wave response was reduced in proportion to increasing stimulus intensity. At maximum stimulus intensities, a 33% (P = 0.008) and 43% (P = 0.008) reduction in a- and b-wave amplitudes, respectively, was evidenced (Fig. 3B) . However, when ischemia was preceded 4 weeks earlier by the RHP stimulus, neither waveform was attenuated, and the functional response of the ischemic retina was indistinguishable from the contralateral nonischemic eye at all recorded stimulus intensities (Fig. 3D) . 

Repetitive Hypoxic Preconditioning Stimulus Is Noninjurious

To demonstrate that the SHP and RHP stimuli we used to establish short- and long-term ischemic tolerance did not cause any deleterious effects, morphologic and functional analyses were undertaken in mice preconditioned by both methods, in the absence of any subsequent ischemia. When measured 2 weeks after SHP (a time point coincident with that used to measure injury in SHP-treated ischemic mice), no significant changes in our morphologic (Figs. 1 2)or functional (n = 3; data not shown) end points (layer thicknesses and cell counts, and scotopic flash ERG, respectively) was noted. Similarly, 5 weeks after the last RHP stimulus, no histologic changes (Figs. 1 2)or functional deficits (Fig. 3F)were noted. 

Molecular Mediators of Retinal Ischemic Tolerance: HIF-1α

To begin to address the mediators involved in establishing short- and long-term tolerance to retinal ischemia in response to a single or repeated preconditioning with hypoxia, we initially examined the temporal basis of SHP- and RHP-induced changes in expression of the hypoxia-responsive transcriptional activator HIF-1α. Figure 4reveals that HIF-1α protein levels are elevated approximately 50% immediately after an SHP stimulus, and remain above baseline for 2 days after the stimulus. Although protection against ischemic injury was observed 1 week after SHP, HIF-1α levels at that time had returned to baseline levels and remained at baseline at 2 weeks after SHP. Figure 4also shows that both the magnitude and duration of the preconditioning-induced elevation in HIF-1α expression were much greater after RHP than after SHP. Specifically, after the last RHP stimulus, HIF-1α expression levels were more than twice as high as baseline, and levels of expression remained elevated for considerably longer (at least 1 week) than that noted with SHP. Although retinas were protected from ischemic injury 4 weeks after the last RHP stimulus, retinal HIF-1α expression was not elevated at this time but had returned to baseline levels 2 weeks earlier. 

We also undertook immunohistochemical analysis of HIF-1α expression after SHP and RHP (Fig. 5) . Under baseline conditions, faint HIF-1α expression was noted in the INL and ONL and in the plexiform layers. However, when examined after SHP, prominent HIF-1α immunostaining was noted in cells of the INL, ONL, and GCL, as well as in the nerve fiber layer and in the inner plexiform layer (Fig. 5C) . After RHP (Fig. 5D) , the cellular distribution pattern was similar, but the intensity of the staining appeared slightly greater in both the cellular and axonal layers relative to that observed after the SHP stimulus. 

Molecular Mediators of Retinal Ischemic Tolerance: HO-1

We then examined whether the elevations in retinal HIF-1α protein expression we confirmed by immunoblotting and immunohistochemistry would lead to corresponding changes in the levels of any HIF-1α gene targets. Immunoblot measures of HO-1, a stress-inducible gene that is also an HIF-1α target, ¹⁴ revealed that both SHP and RHP resulted in robust and long-lasting changes in retinal HO-1 protein expression (Fig. 6) . In response to SHP, HO-1 protein levels almost tripled, and remained elevated for at least 1 week, coincident with the time over which the retina of the SHP-treated mice were protected from ischemic injury. Although the peak expression level of HO-1 was not that much greater after RHP, this preconditioning stimulus resulted in longer-lasting elevations of the protein, with levels three times that of baseline at 2 weeks, and levels still elevated 1 month after the last RHP stimulus, at a time when ischemic neuroprotection was still evidenced in RHP-treated retinas. In contrast, no changes were noted in the levels of the constitutive HO-2 protein from hours to weeks after the RHP stimulus (Fig. 6) . 

Immunohistochemical analyses (Fig. 7)reveal that HO-1 expression, virtually absent under baseline control conditions, was evident in cells of the INL, ONL, and GCL at 1 week after the SHP stimulus (Fig. 7C)and even 4 weeks after the last RHP stimulus (Fig. 7D) , particularly in cell bodies; under both conditions, diffuse staining for HO-1 also appeared in the IPL and nerve fiber layer. In eyes subjected to retinal ischemia and examined 1 week later, we observed HO-1 immunostaining, primarily in the damaged IPL and GCL in nonpreconditioned eyes (Fig. 7E) , consistent with an ischemia-induced triggering of HO-1 expression. However, in SHP-treated (Fig. 7F)and RHP-treated (Fig. 7G)eyes, which, as shown in Figure 2 , were significantly protected morphologically, postischemic HO-1 expression was considerably more robust and widespread throughout the neural retina relative to untreated ischemic eyes. 

Discussion

Results of the present study in mice show for the first time that repeated preconditioning with systemic hypoxia (RHP) activates endogenous adaptive mechanisms in the retina that result in a protracted (>4 weeks) period of protection from retinal ischemia. Immunoblot and immunohistochemistry examinations showed that, relative to a single hypoxic preconditioning stimulus, RHP increased the magnitude and duration of hypoxia-induced increases in panretinal levels of the transcriptional activator HIF-1α as well as HO-1, a HIF-1α gene target with pleiotropic prosurvival effects. The 1- and 4-week periods of sustained HIF-1α and HO-1 expression, respectively, coincident with the time window over which LTT was realized, implicates these molecules as key participants in the endogenous molecular pathways that underlie this unique cytoprotective phenotype. 

In the brain, the duration of the neuroprotective phenotype induced by a single preconditioning stimulus is on the order of one to several days. ² In the retinas of mice ⁵ and rats, ³ protection from ischemia lasts only 24 and 72 hours after preconditioning, respectively. Qualitatively, the transient nature of preconditioning-induced protection fits well with the fundamental physiologic principle of adaptation (i.e., presentation of a stressful stimulus leads to an adaptive response to increase resistance to that stress, but on removal of the stress, the adaptation wanes, and the original set point is reestablished). The long-lasting molecular and cellular adaptations to continuously presented stimuli, notably chronic hypoxia, in the CNS and other tissues ²³ are not germane to the present work, since the phenotype we observed persisted weeks after the inducing stimulus was no longer present. Rather, more relevant to our model are the paradigms of protracted neuronal plasticity to intermittent stimuli, including long-term potentiation underlying learning and memory, ^{24 25} changes in behavioral responses to adverse psychological stimuli, ²⁶ structural and functional alterations in neuronal networks related to drug and alcohol addiction, ²⁷ and long-lasting “cellular memory” associated with immune tolerance. Although these changes are not always thought of as adaptive per se, discontinuous or sporadic stimuli can trigger them, and the phenotypes so induced can be very enduring. We contend that the prolonged changes in phenotype that we documented herein for the retina represent a new kind of endogenous “adaptive plasticity” ²⁸ for this tissue—one that may be relevant to the protection of retinal cells from other types of ischemic retinopathies, both acute and chronic. 

We showed previously that exposing conscious mice to brief systemic hypoxia, with the same intensity–duration parameters as the SHP stimulus used herein, effectively protected against injury when retinal ischemia was induced 24 hours later. ⁵ The present study extends this finding to show that both morphologic and functional indices of protection can actually be documented as long as 1 week after SHP. Brief hypoxic preconditioning also protects mouse photoreceptors from light-induced apoptosis, ¹⁰ corneal stromal cells from UV light-induced apoptosis, ²⁹ and cultured retinal ganglion cells from anoxic cell death. ⁶ However, we are unaware of any study in any tissue that manipulated the hypoxic preconditioning stimulus as we did to extend either the duration or the magnitude of the resulting injury-tolerant phenotype. Our results clearly show that, when administered at an appropriate severity, frequency, and duration, hypoxia can have beneficial effects, not only by triggering ischemic tolerance but by extending tolerance from a few days to many weeks. Moreover, using both histology and electroretinography, we confirmed that this RHP stimulus was not harmful to the retina. 

Mechanistically, we explored the possible involvement of HIF-1α and its downstream gene target HO-1 in affording ischemic protection in our LTT model. HIF-1 is an oxygen-sensitive heterodimeric protein that, when stabilized, binds to core sequences in hypoxia response elements of its target genes to activate transcription. Many studies have provided details as to its regulation. ^{7 8 9} Our documentation of elevations in retinal HIF-1α protein expression after our SHP stimulus of 11% oxygen falls within the range of hypoxia (6%–14% oxygen) shown recently to raise retinal HIF-1α protein levels ¹⁰ ; in both studies, the most robust increase in HIF-1α expression occurred immediately after the hypoxic preconditioning stimulus. Our immunohistochemical analysis revealed that SHP and RHP elevated HIF-1α expression throughout the retina, in both cellular and axonal layers. Most important, we showed that, with RHP, not only did higher levels of retinal HIF-1α protein result after the last of the six hypoxic stimuli were presented, but the duration of elevated HIF-1α expression thereafter was extended from 2 days to 1 week. To our knowledge, such a stimulus–response feature of HIF-1α regulation is unprecedented in any tissue. The conventionally accepted view is that HIF-1α expression increases in proportion to the severity of a single hypoxic stimulus, as well as to its duration (with tachyphylaxis noted after many hours of continuous hypoxia), ³⁰ but declines rapidly after the reestablishment of normoxia. ³¹ However, our finding that RHP changed the extent and duration over which retinal HIF-1α expression was elevated is consistent with the emerging consensus of recent studies of high-frequency systemic hypoxia and reoxygenation that both HIF-1α protein expression and HIF-1α transcriptional activation after intermittent hypoxia differ from that resulting from acute or continuous hypoxia. ^{32 33 34} How HIF-1α protein is stabilized for such protracted periods remains to be clarified, but could involve changes in the level of HIF-1α DNA binding, subcellular localization, phosphorylation, and/or recruitment of transcriptional coactivators. ^{8 35} Ultimately, the effects of these repetitive hypoxia treatments may be mimicked pharmacologically ^{19 36} to extend the duration of HIF-1α expression in specific pathologic conditions where prolonged activation of survival gene transcription is desired. ³⁷  

Our retinal immunoblot and immunohistochemistry results support the general finding in mammalian tissues that the expression of HO-1 message and protein is low under resting conditions, whereas HO-2 is constitutively expressed in specific retinal cells. ^{38 39} Although bright light and oxidative stress can trigger retinal HO-1 expression, aside from the report that hypoxia upregulated HO-1 message in cultured retinal pigment epithelial cells, ⁴⁰ nothing was known about hypoxia-driven HO-1 induction in the retina before our study. Indeed, hypoxia induced retinal HO-1 protein expression in vivo but, as with HIF-1α, the temporal characteristics of this response were uniquely dependent on the nature of the hypoxic preconditioning stimulus. Specifically, SHP caused substantial elevations in retinal HO-1 expression that, surprisingly, persisted for nearly 2 weeks; however, even this robust response was overshadowed by that measured in response to RHP, wherein HO-1 protein levels 2.5 times greater than baseline were present even 4 weeks after the last hypoxic challenge. Both acute and chronic hypoxia are known to trigger HO-1 message and protein expression in other tissues, ^{11 12 41 42} but ours is the first study to document the spatiotemporal nature of the response in the in vivo retina and implicates HO-1 as a uniquely long-lasting cytoprotectant. Pharmacologic and genetic evidence has already implicated HO-1 upregulation as contributing to the traditional, transient period of ischemic tolerance induced by a single preconditioning stimulus in liver, ⁴¹ skeletal muscle, ⁴³ and heart. ⁴⁴ Unlike the robust induction noted for HO-1, our RHP stimulus did not change retinal expression levels of HO-2 protein, either in the initial hours after the stimulus, or out to 4 weeks later. Even retinal ischemia is without effect on retinal HO-2 message and protein levels during the initial 3 to 48 hours of reperfusion, ³⁸ consistent with the notion that this constitutively expressed isoform ^{38 39} performs different functions than does HO-1. 

Although our SHP and RHP stimuli caused temporally parallel, appropriately time shifted increases in both HIF-1α and HO-1 expression, and although HO-1 is a HIF-1α target gene, ¹¹ the extent to which the upregulation in HO-1 protein levels in each model was mediated by the respective period of stabilized HIF-1α preceding it cannot be ascertained directly based on this correlative data set. There are many well-described hypoxia-sensitive transcription factors and hypoxia-driven genes. With respect to the HIF family, we make note of the possibility that SHP- and RHP-induced increases in the activity of the HIF-2α isoform may have contributed to the protection we observed in each model. HIF-2α expression is also oxygen dependent but exhibits unique regulatory features relative to HIF-1α ^{45 46} and may preferentially promote the expression of VEGF and other angiogenesis-related prosurvival genes. ⁴⁷ HO-1 induction by hypoxia ^{11 12} is not solely dependent on HIF-1α but also on NFκB, AP-1, STAT, and other stress-related transcriptional regulators. ^{17 48} Conversely, other HIF-1α-driven survival genes, ^{8 9} in addition to HO-1, are likely to contribute to the protection we observed in our LTT model. For example, erythropoietin, the first identified gene target of HIF-1α, is implicated in hypoxic preconditioning of the retina against phototoxicity, ¹⁰ shows a relatively protracted profile of hypoxia-induced expression in the CNS relative to other tissues, ³⁰ and can be used directly as a preconditioning treatment to reduce retinal cell death in ischemia ⁴⁹ and glaucoma. ⁵⁰  

The findings in many tissues ^{19 51} that injurious ischemia leads to elevations in HO-1 expression, whereas blockade of HO-1 activity during and/or after ischemia exacerbates ischemic injury, indicate that HO-1 induction by ischemia serves as an endogenous, self-protection mechanism. ^{13 14} In the ischemic rat retina, HO-1 protein levels, immunolocalized to Müller cells, are elevated from 6 to 48 hours of reperfusion, and silent RNA-directed knockdown of endogenous HO-1 increases Müller cell death, as well as macrophage infiltration, after retinal ischemia. ³⁸ Moreover, adenoviral transfection and functional expression of human HO-1 in retinal ganglion cells of the rat retina increases their ischemic resistance. ⁵² Our immunohistochemical examinations revealed a panretinal elevation in HO-1 expression still present 1 week after retinal ischemia that was similar in magnitude, duration, and distribution pattern to that triggered for HO-1 one day after a single period of hypoxic preconditioning. Moreover, our data show that prior SHP and RHP cause an even greater increase in this innate, ischemia-induced retinal HO-1 upregulation response. Although it is unlikely that HO-1 is the only example of this, we contend that hypoxic preconditioning not only triggers alterations in retinal phenotype that serve to increase ischemic resistance in advance of the actual ischemic insult, but also results in a unique genomic response to ischemia, a concept directly supported by microarray studies of cerebral ischemic tolerance. ²  

The protective effects of HO-1 collectively derive from the crucial actions of carbon monoxide, bilirubin, and iron, all of which are products the HO-1-mediated enzymatic degradation of heme. ^{13 14} In particular, carbon monoxide serves as a vasodilator, exerts antiaggregatory and antiadhesive effects on platelets and leukocytes, inhibits the expression of proinflammatory cytokines, and prevents endothelial cell apoptosis. Bilirubin scavenges peroxyl radicals, downregulates endothelial adhesion molecule expression, stabilizes mast cells, and is directly neuroprotective. Finally, free iron produced as a result of HO-1 activity not only regulates gene expression, but is sequestered into ferritin, which in turn promotes additional ferritin upregulation, resulting in better iron sequestration capacity. 

In summary, we characterized a new model of retinal preconditioning in which intermittent hypoxia promotes a robust neuroprotective phenotype in the retina characterized most notably by a resistance to ischemic injury that lasts for weeks after presentation of the preconditioning stimulus. The RHP protocol we used to establish this protracted period of ischemic tolerance resulted in uniquely extended temporal expression profiles for the transcriptional activator HIF-1α and the pleiotropic cytoprotectant HO-1, a downstream gene target of HIF-1α, thereby providing initial glimpses into the mechanistic basis for this long-lasting protection. Moreover, we documented by morphologic and functional criteria that the protracted period of ischemic tolerance afforded by the RHP stress was not at the expense of any discernible tissue injury or dysfunction. Our demonstration of LTT in the retina may represent a novel form of retinal plasticity wherein activation of survival mechanisms endogenous to the tissue could offer protection against glaucoma and other chronic retinopathies with ischemia-like features. 

 Supported by National Eye Institute Grant R03 EY014938 (JMG), EY02687 (Department of Ophthalmology and Visual Sciences, Washington University), and The Glaucoma Foundation, New York, NY (JMG).

 Submitted for publication September 1, 2006; revised October 20, 2006; accepted January 11, 2007.

 Disclosure: Y. Zhu, None; Y. Zhang, None; B.A. Ojwang, None; M.A. Brantley, Jr, None; J.M. Gidday, None

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

 

Figure 1.

 

Histologic evidence that SHP or RHP can protect the retina from ischemic injury and that neither treatment is injurious to the retina. Morphologic indices include determinations of the thickness of the OLM–ILM (A), INL (B), and IPL (C), and quantification of viable cells in the INL (D) and GCL (E), in six different groups of mice. These groups include: nonischemic, nonpreconditioned controls (C; gray bars; n = 20), animals treated with SHP or RHP in the absence of ischemia (hatched and cross-hatched bars, n = 3 and 4, respectively), and three groups of mice subject to 30 minutes of retinal ischemia; the latter groups of ischemic mice include those without preconditioning treatment (NT; black bars; n = 10), those treated with SHP 1 week before ischemia (hatched bars; n = 12), and those treated with RHP 4 weeks before ischemia (cross-hatched bars; n = 5). *P < 0.05 vs. control group; #P < 0.05 vs. ischemia only (NT) group.

Figure 2.

 

Representative thin sections of retina from the six animal groups illustrated in Figure 1 , showing no histologic evidence of injury in response to SHP and RHP in the absence of ischemia (B, C, respectively), relative to a control retina (A), as well as preservation of retinal layer thicknesses and cell counts in mice subjected to 30 minutes of retinal ischemia preceded by SHP 1 week earlier (E), or RHP 4 weeks earlier (F), relative to the thinning of retinal layers and the cell loss occurring in untreated ischemic eyes (D). Scale bar, 60 μm.

Figure 3.

 

Bilateral scotopic flash ERG recordings during baseline (A, C, E, left) and after different experimental interventions (B, D, F, right) in three groups of mice. Circles: b-wave data; triangles: a-wave data; open symbols for the left eye and filled symbols for the right. ERG before (A) and 1 week after (B) unilateral retinal ischemia in the right eye (n = 8), showed increasing loss in b- and a-wave amplitudes with progressively higher stimulus intensities. ERG before (C) and 1 week after (D) unilateral retinal ischemia in the right eye, in mice preceded by RHP 4 weeks before ischemia (n = 10), showed that RHP completely prevented the ischemia-induced functional deficits seen in untreated mice (B). ERG before (E) and 5 weeks after (F) RHP only (n = 4), showed the lack of any functional deficits as a result of the RHP stimulus. *P < 0.05 vs. fellow eye, for amplitudes recorded at the highest stimulus intensity.

Figure 4.

 

Temporal changes in HIF-1α protein expression (as measured by retinal immunoblotting; n = 3 individual samples per time point) in response to RHP (measured after the last of six hypoxic preconditioning treatments) and in response to SHP. Note that elevations in retinal HIF-1α expression lasted considerably longer after the last RHP stimulus relative to the SHP stimulus. C, baseline control. *P < 0.05 vs. baseline; #P < 0.05 vs. SHP; 2-hour time point for SHP not shown in the line graph for clarity. Representative immunoblots for each preconditioning protocol are provided, normalized to β-actin (not shown).

Figure 5.

 

HIF-1α immunohistochemistry after SHP and RHP in representative 5-μm sections of mouse retinas obtained under the following conditions: (A) nonpreconditioned control without primary HIF-1α antibody; (B) nonpreconditioned control with primary antibody, showing little HIF-1α expression throughout the neural retina; (C) immediately after SHP, revealing widespread expression of HIF-1α throughout all cellular and axonal layers of the retina; (D) immediately after RHP, with a similarly distributed, but perhaps slightly more intense, staining pattern relative to that noted after SHP. Scale bar, 60 μm.

Figure 6.

 

Temporal changes in HO-1 protein expression (as measured by retinal immunoblot analysis, n = 3–6 individual blots per time point) in response to RHP (measured after the last of six hypoxic preconditioning treatments) and in response to SHP stimuli. Note that both the magnitude and the duration of the increased HO-1 expression was greater in retinas of mice exposed to RHP relative to those exposed to SHP. C, baseline control. *P < 0.05 vs. baseline; #P < 0.05 vs. SHP. Representative immunoblots for each preconditioning protocol are provided. Also, a representative immunoblot for changes in HO-2 protein in response to RHP is shown, wherein no changes in HO-2 expression were observed at any time point. β-Actin is shown as the loading control.

Figure 7.

 

HO-1 immunohistochemistry after SHP and RHP (C, D) and after ischemia with and without SHP and RHP (E–G). (A) Nonpreconditioned control without the primary HO-1 antibody; (B) Nonpreconditioned control with the primary antibody, revealing very low HO-1 expression; (C) 1 week after SHP, with HO-1 expression evident in cell both cellular and axonal layers; (D) 4 weeks after RHP, with a similar distribution pattern, but with stronger expression in the GCL, INL, and ONL relative to the SHP-treated mice in (C); (E) nonpreconditioned ischemic control at 1 week after ischemia, with HO-1 expression primarily limited to damaged inner retina; (F) ischemic retina with SHP 1 week earlier, showing HO-1 expression in both cellular and axonal layers; (G) ischemic retina with RHP 4 weeks earlier, with HO-1 present throughout the neural retina, but expressed even more robustly in the GCL, INL, and ONL, relative to the SHP-treated ischemic mice in (F). Scale bar, 60 μm.

The authors thank Max Gassmann, DVM, PhD, of the Institute of Veterinary Physiology and Zurich Center for Integrative Human Physiology in Zurich, Switzerland, for providing to us the chicken anti-mouse HIF-1α antibody and Emily E. Barr and Peter D. Lukasiewicz, PhD, for expert technical assistance with the ERG studies.