Background
In mammals, age-related deterioration of the sensory nervous system is common, while repairing or replacing the injured or dead neuronal cells is difficult due, in part, to the lack of cell proliferation and differentiation after development. Thus, the functionality of sensory systems is dependent on the long-term viability of terminally differentiated cells. Mice in captivity can live for three years, however, by about two years of age, the retina has become significantly thinner and less responsive in the electroretinogram (ERG) than in mature, but much younger, adult mice (e.g. 4 months) [
1]. Additionally, bipolar cells in aged retinas extend dendrites into the outer nuclear layer (ONL) to form ectopic synapses with retracted rod axons [
2,
3]. These changes are recapitulated in mice with deletions of either LKB1 or AMPK, key metabolic pathway controlling proteins, in subsets of retinal cells at just two months of age [
4]. These findings indicate that disturbance in retinal metabolism is likely behind some age-related deterioration in retinal neurons. However, the exact mechanisms remain elusive.
Recently, several reports demonstrate a role for endoplasmic reticulum (ER) stress in aging and neurodegenerative diseases, though the mechanisms remain elusive due to seemingly contradictory results across species and tissues [
5,
6]. ER stress occurs in a cell when there is an abundance of unfolded or misfolded proteins in the ER. To ameliorate ER stress, three pathways of the Unfolded Protein Response (UPR), namely IRE1, ATF6, and PERK, are activated. These pathways work in parallel to reestablish the ER’s homeostasis through increasing the folding capacity of the ER, slowing protein translation, and increasing protein degradation [
7]. However, in scenarios where ER stress is too severe or endures for too long that homeostasis cannot be restored, the UPR can shift to a pro-apoptotic mode promoting cell death, often through increased CHOP expression [
6]. X-Box binding protein 1 (XBP1) is the main effector of the IRE1 pathway, playing a crucial role in cell adaptation to ER stress. XBP1 is activated via a unique RNA splicing event initiated by IRE1. Spliced XBP1 (XBP1s) encodes an active transcription factor that regulates a diverse set of genes including ER chaperones and proteins involved in ER-associated degradation (ERAD) that facilitate the restoration of the homeostasis of the ER [
7]. Long-term or extreme stress, however, can result in differential oligomerization of IRE1, resulting in a distinct array of RNA splicing targets that tilt the cell towards apoptosis and away from XBP1 splicing [
8]. Systemic ablation of XBP1 causes embryonic lethality [
9]. Conditional depletion of XBP1 in the nervous system results in impairment in learning and memory formation, suggesting that XBP1 may play a role in controlling neuronal function [
10]. In addition, the UPR and XBP1 in particular have been shown to regulate longevity. For example, activation of the UPR, including XBP1, or neuronal expression of XBP1s extends lifespan in
C. elegans [
11,
12]. Long-term intranasal injection of exogenous Hsp70 (an ER chaperone) also delays age-related deterioration and improves learning and memory in aging mice [
13]. These studies suggest that successful adaptation to cellular stress through the UPR, and potentially XBP1, is critical for maintaining neuronal viability as well as the integrity of sensory systems as the organism ages. Though the literature is inconsistent across species and tissues, there is strong evidence that the ability to activate the UPR declines with advanced age suggesting a role for ameliorating cellular stresses in longevity and potentially implicating long-term ER stress as a causative aging factor.
In the present study, we identified that aged retinas have a compromised ER stress response with reduced levels of XBP1s under both basal and stressed conditions. To elucidate the role of XBP1 in retinal neuronal maintenance especially during aging, we generated a retina-specific conditional knockout (cKO) mouse line of XBP1 and characterized the retinal structure and function of the mice from an early age to over one year old. Our data suggest that age-related neurodegeneration occurs at an accelerated pace in the retina lacking XBP1 thus indicating a critical role of XBP1 in regulation of age-related retinal degeneration.
Methods
Animals and genotyping
Conditional knockout of XBP1 in the retina was achieved by crossing mice with LoxP sites flanking exon 2 of XBP1 [
14] with a retina-specific Chx10-Cre line [
15]. The two independent lines have been interbred and maintained as XBP1 fl/fl and XBP1 fl/fl; Chx10-cre for multiple generations over several years. Genotyping was performed by PCR with allele-specific primers for XBP1 WT, floxed, and exon 2-deleted alleles. PCR using the primers 3’lox-S - 5’-ACT TGC ACC AAC ACT TGC CAT TTC-3′ and 3’lox-A - 5′- CAT TAC AGG CAG TGA ACC ACC TTG-3′ result in a 140 bp band for WT and a 180 bp band for XBP1-fl. The XBP1-fl allele after Cre-mediated recombination is detected with Int1-S: 5′ - CTT TGT GGT CGT AGG GTA GGA ACC - 3′ and 3’lox-A, resulting in a 352 bp band after recombination, and no detectable band in WT or non-recombined XBP1-fl. The presence of the Chx10-Cre allele was determined by PCR with Cre-F 5′- GCA TTA CCG GTC GAT GCA ACG AGT GAT G-3′ and Cre-R 5′- GAG TGA ACG AAC CTG GTC GAA ATC AGT G - 3′, resulting in a 408 bp band.
Ex-vivo retinal explant cultures and reverse transcription quantitative PCR (qPCR)
Retinas were dissected and one half cultured for 6 h in Neurobasal-A medium supplemented with 25uM glutamic acid, 2 mM glutamine, and B-27 (all ThermoFisher) and treated with DMSO as control and the other half retina cultured in identical media with 5 μm thapsigargin (ThermoFisher). Retinas were washed in PBS and RNA isolated with Trizol per manufacturer instructions. Complementary DNA (cDNA) was made using Bio-Rad cDNA kit by manufacturer instruction and 500 ng of total RNA. Reverse transcription-qPCR was performed using Bio-Rad iQ Sybr Green Supermix by manufacturer instruction with 10 ng of cDNA, and 300 nM of each of the following primers and normalized to 18 s.: XBP1sF 5’-CCA TCA CAT TGC CTA GAG GAT A-3′ and XBP1R 5’-AGC TGA GTG TCA AAC GAC AAT A-3′; ATF4F 5’-CCC CCT TCG ACC AGT CGG GT-3′ and ATF4R 5’-CCG CCT TGT CGC TGG AGA ACC-3′; ATF6F 5’-CAG ACT CGT GTT CTT CAA C-3′ and ATF6R 5’-GGC TTC TCT TCC TTC AGT-3′; p58ipkF 5’-TCC TGG TGG ACC TGC AG TACG-3′ and p58ipkR 5’-CTG CGA GTA ATT TCT TCC CC-3’ CHOPF 5’-GTC CCT AGC TTG GCT GAC AGA-3′ and CHOPR 5′-TGG AGA GCG AGG GCT TTG-3′; 18 sF 5′-GTA ACC CGT TGA ACC CCA TT-3′ and 18 sR 5’-CCA TCC AAT CGG TAG TAG CG-3′.
Electroretinography (ERG)
Visual function was assessed by dark- and light-adapted electroretinogram (ERG) using a Diagnosys Espion ColorDome system and software (Diagnosys LLC, Lowell, MA). For photopic transient ERG the manufacturer installed Light plus OPs protocol was used. Mice were anesthetized with intraperitoneal injection of 120 mg/kg ketamine and 5 mg/kg xylazine, and had pupils dilated with 1% atropine (Falcon Pharmaceuticals) followed by 2.5% Phenylephrine Hydrochloride (Bausch & Lomb). Mice were placed in the ColorDome apparatus, an electrode was inserted into the tail as ground, and a reference electrode placed subcutaneously between the eyes. Gonak (Akorn) ophthalmic gel was applied liberally to each cornea, and wire electrodes placed in contact. Animals were light adapted for a minimum of 15 min prior to stimulation and subjected to five flashes of 4 ms duration at 1 Hz at 10 cd.s/m2 with a background of 5 cd.s/m2. The a-wave amplitude was recorded as the lowest point of the initial response compared to baseline, and the b-wave amplitude was the high point, as calculated from the a-wave amplitude. The average value for both eyes for all flashes is reported.
For the dark-adapted step ERGs, a custom procedure was designed within the Diagnosys software. Briefly, mice were dark adapted overnight (14–16 h) in their home cage. Under dim red light, mice were anesthetized, dilated, and set in the ColorDome as described above. A protocol consisting of ten series of three flashes of light of 4 ms duration was applied. Light intensity increases with each subsequent series. There was a delay between each flash of 15–60 s, with the delay increasing with intensity. Light intensities were (luminance in log(cd.s/m2)): − 3.6, − 3.0, − 2.4, − 1.8, − 1.2, − 0.6, 0.0, 0.6, 1.4, 2.1. Amplitudes for all traces from both eyes were averaged for each light intensity and animal. The reported N value represents independent animals. The peak for the a-wave was the lowest point of the initial response to the flash, and the b-wave peak was determined to be the peak after the oscillatory potentials and within 140 ms of the a-wave, and its magnitude measured from the a-wave peak. The average value for both eyes for all flashes is reported. For ERG on 10 week old animals the average age for WT was 10.8+/− 1 weeks and 10.4+/− 1.2 weeks for XBP1 cKO mice. For the 6–8 month cases WT average age was 32.6 +/− 3 weeks and 29.7+/− 4 weeks for XBP1 cKO mice. For 12–14 month ERGs WT average age was 62.3+/− 2 weeks old and XBP1 cKO average age was 59.1+/− 5 weeks old.
Immunohistochemistry and retinal staining
Retinal morphology was examined with immunohistochemical markers in WT and XBP1 fl/fl; Chx10-Cre in retinal sections in adolescent and 12–15 month old mice (average ages for morphological analyses was 59.2+/− 5 weeks for WT and 56.1+/− 5 weeks for XBP1 cKO mice). Eyes were immersion fixed in 4% w/v paraformaldehyde in phosphate buffered saline (PBS), pH 7.4 for one hour, washed, and equilibrated in 30% sucrose w/v in PBS at 4C. Eyes were embedded in OCT and frozen in molds partially submerged in absolute ethanol chilled on crushed dry ice. Blocks were cryosectioned immediately or stored at -80C. Cryosections were cut at -20C at 20 μm, mounted on Superfrost Plus slides (Statlab), and dried overnight.
Sections were blocked with PBS plus 1% triton X-100 plus 1% BSA fraction V (Calbiochem) for 1 h and incubated with primary antibody overnight at 4C in a light-protected, humidified chamber. Primary antibodies were removed, sections washed with PBS plus 1% Triton X-100, and incubated with the appropriate secondary antibody in block solution for 1 h in a light-protected, humidified chamber. Secondary antibody was washed and sections mounted with Vectashield mounting medium with DAPI (Vector, H-1200) for examination. All antibodies used are listed in Table
1.
Table 1
Antibodies used in immunofluorescence (IF) and western blot analyses (WB)
anti-Ribeye, B-domain | 1:800 (IF) | 192,003 | Synaptic Systems |
anti-Pkc-α | 1:400 (IF) | sc-8393 | Santa Cruz Biotechnology |
anti-Calretinin, clone 6B8.2 | 1:800 (IF) | Mab1568 | Millipore |
anti-Iba1 | 1:800 (IF) | 019–19,741 | Wako |
anti-XBP1 | 1:1000 (WB) | sc-7160 | Santa Cruz Biotechnology |
anti-Glutamine synthetase | 1:800 (IF) | Mab302 | Millipore |
anti-Brn3a | 1:800 (IF) | sc-31,984 | Santa Cruz Biotechnology |
anti-Pax6 | 1:40 (IF) | Pax6-s | DSHB |
anti-β actin | 1:10,000 (WB) | ab8226 | Abcam |
Peroxidase | 1:10,000 (WB) | PI-2000 | Vector |
Peroxidase | 1:10,000 (WB) | PI-1000 | Vector |
Texas Red | 1:800 (IF) | T6391 | Molecular Probes |
Alexa Fluor-488 | 1:800 (IF) | A11001 | Molecular Probes |
Alexa Fluor-594 | 1:800 (IF) | A11005 | Molecular Probes |
Alexa Fluor-594 | 1:800 (IF) | A11080 | Molecular Probes |
anti-Calbindin D-28 K, CB-955-955 | 1:100 (IF) | Sab4200543 | Sigma |
Cryosections of retina were incubated in 1-Step NBT/BCIP alkaline phosphatase labeling reagent (Thermo Scientific, #34070) until an appropriate chromogenic signal was observed.
Photographs were taken with an Olympus DP80 digital camera mounted on an upright Olympus BX53 microscope using 10-40× objectives, processed, montaged, and analyzed with Photoshop (Adobe) and NIH ImageJ software.
Western blot analysis
Retinas were dissected and flash frozen in liquid nitrogen. Tissue was thawed on ice in chilled fresh radio immune precipitation assay (RIPA) buffer with protease inhibitor mixture, PMSF, and sodium orthovanadate and sonicated. Protein samples (approximately 25 μg), MagicMark XP Western Protein Standard (Life Technologies, LC5602), and BenchMark Prestained Protein Ladder (Life Technologies, 10,748–010) were electrophoresed in 10% acrylamide resolving gels and transferred to nitrocellulose blotting membranes using standard techniques. Membranes were incubated with anti-XBP1 (Santa Cruz Biotechnology) at 1:1000 in Tris-buffered saline plus tween overnight at 4C. followed by incubation with peroxidase conjugated secondary antibody (Vector), at 1:10000 for 1 h room temperature before development with SuperSignal West Dura Extended Duration Substrate (Thermo Scientific, #34076) per manufacturer instructions. Blots were imaged and analyzed with a Bio-Rad ChemiDoc MP imaging system using ImageLab software. Bands were re-blotted with anti-beta-actin (Abcam 1:10000, overnight at 4C) and then peroxidase conjugated secondary antibody as above, as loading control.
Retina from 12 to 14 month WT or XBP1 fl/fl; Chx10-Cre mice was dissected into 15–20 explants of nearly equal size and plated in a prepared 24 well assay dish, GCL side down. A single 24 well assay dish was used per experiment and contained ten explants each from a WT retina and an age-matched XBP1 cKO retina. Five explants of each genotype were subjected to glycolytic testing and five explants of each genotype were used for mitochondrial testing in a Seahorse XFe24 Extracellular Flux Analyzer using manufacturer supplied kits. The glycolysis stress test used final well concentrations of 10 mM glucose, 1 μM oligomycin, and 50 mM 2-D-glucose. Compounds were injected just after measurement steps 3, 6, and 9, respectively. For the mitochondrial stress test, FCCP was injected just after measurement 3 to a well concentration of 1 μm and rotenone/antimycin A was injected after measurement 9 to a well concentration 0.5 μM. Readings for ECAR and OCR were taken at each measurement interval. Immediately after the final measurements each explant was removed from the dish, centrifuged briefly, assay media was removed, and RIPA lysis buffer added. Tissue was then sonicated and subjected to BCA protein assay according to the manufacturer protocol (Thermo Scientific). Measurements from each well were normalized to total protein content of the corresponding explant. Average age of WT mice used for metabolic measurements was 58.2+/− 5 weeks and the average age of XBP1 cKO mice used was 59.1+/− 5 weeks old.
Data analysis
RGC cell counts were performed automatically in NIH Image J using the Analyze Particles macro modified to accurately count cells in single-channel images of Brn3a immunohistochemistry using secondary antibodies conjugated to Texas Red and of single channel images of DAPI labeling. Counting was restricted to the GCL layer of cryosections from central retina containing the optic disc, or adjacent, and containing at least the full extent of the GCL from central retina to the periphery. Areas where the plane of section was tangential were excluded. Between 3 and 5 sections per case were counted blind to genotype and averaged. Automated counting paradigms were verified by comparisons to a subset of manually counted images by a second operator. Linear distances were measured through the GCL layer.
Calretinin-labeled synaptic lamina data was assessed by two independent investigators blind to genotype and the presence or absence of an Iba1-positive cell. Areas analyzed were chosen without knowledge of calretinin staining by random selection of Iba1-positive cells A score of ‘continuous’ or ‘discontinuous’ was applied to each case from a photograph of the IPL with calretinin staining. Assessments were made using agreed-upon criteria, and repeated at a later date with re-blinded cases. Assessments were approximately 90% identical between investigators and between assessments.
Retinal layers were measured in photographs of DAPI-stained 20 μm cryosections by at least two independent investigators blind to genotype. Measurements were taken in NIH ImageJ from retinal sections from the central 15% of retina encompassing the optic disc, cut orthogonal to the pupillary plane, and containing at least the full extent of all retinal layers from central retina to the periphery. Layers were measured at two locations approximately 300 μm from the optic disc (or linear center of sections adjacent to the optic disk) in each direction.
Data were compiled and analyzed in Microsoft Excel. Figures and photos used for data analyses were montaged and assembled in Adobe Photoshop, NIH Image J, and/or Microsoft PowerPoint.
Statistical analysis
Statistical analyses were performed in Microsoft Excel. Statistical analyses were unpaired Student’s t-test when comparing two groups. Two-way ANOVA with Bonferroni post-hoc test was used for the light adapted and dark adapted ERG analyses, layer thickness measurements, and glycolysis test ECAR measurements. Statistical differences were considered significant at a P value less than 0.05. N values represent independent animals except where noted and all experiments were performed independently at least twice. Error bars indicate standard deviation except where noted.
Discussion
In the present study, we for the first time investigate the long-term effects of XBP1 deficiency on retinal neuronal development and maintenance. We demonstrate that in aged mouse retinas, the ER stress response, as measured by the activation of XBP1, is less effective as in the younger ones. Furthermore, very old WT retinas (20–24 months) have a lower level of XBP1s than young adult (4 months) WT retina, but a higher level of CHOP. These data suggest an age-related deficit in the IRE1/XBP1 signaling pathway in the retina. Thus, we examined the consequences of a lack of XBP1 in retinal cells using a XBP1 cKO mouse line. We find that the retinas of XBP1 cKO mice at 12–14 months of age demonstrate structural degeneration that closely resembles WT mice aged 20 months or older, and display functional deficits that are not yet present in age-matched WT mice. Interestingly, no significant defects in retinal structure or function were identified in the cKO mice through the age of eight months, suggesting that the lack of XBP1 in a subset of retinal cells does not affect retinal development. However, the deficit in XBP1 may compromise the retinal responses to natural and chronic stresses over a long time period (i.e. 12 months), which may subtly but cumulatively affect cellular metabolism and synaptic transmission [
5,
6,
16], consequently resulting in accelerated functional decline, metabolic deficits, and structural deterioration.
Our data indicate that the deletion of XBP1 in retinal progenitor cells, driven by Chx10 expression, does not affect retinal neuronal development. This finding does seem surprising, given that previous studies have reported an important role of XBP1 in brain-derived neurotrophic factor (BDNF)-mediated neurite outgrowth [
17,
18] as well as in axon regeneration in
C. elegans [
19] and in Drosophila [
20]. Yet currently there is no direct evidence to support that lack of XBP1 results in significant structural defects in the development of central nervous system. It is perhaps logical to speculate that in addition to XBP1, other UPR molecules, e.g. ATF6, that orchestrate the adaptive response to physiological stresses, may play an equally important, and compensatory, role in regulation of neuronal development and function. This is in part supported by several recent reports identifying mutations of human ATF6 that interrupt the proper activation of this UPR molecule during ER stress contributing to the development of achromatopsia, an autosomal recessive retinal disease characterized by cone dysfunction [
21,
22]. Interestingly, we only observed a modest and insignificant increase in ATF6 level in the retina of XBP1 cKO mice. Whether this change, or the changes in other molecules, could exert a compensatory effect in regulation of the adaptive UPR during retinal development in the XBP1 cKO mice remains to be elucidated.
Despite the negative finding in retinal development, our data strongly suggest an essential and beneficial role of XBP1 in protecting retinal neurons against age-related degeneration. At 12–14 months of age, XBP1 cKO mice demonstrate retinal thinning, indicative of cell loss or shrinkage in the IPL and ONL, and ganglion cell loss in the GCL. At this point we cannot differentiate between an appropriate expansion of the IPL after P15, followed by an accelerated reduction in the XBP1 cKO, versus a simple lack of the IPL expansion observed in WT from P15 to 12–14 months. Importantly, we find that bipolar cells in the year-old XBP1 cKO have more than twice as many dendritic extensions into the ONL and more than twice as many ectopic synapses in the ONL, compared to age-matched WT. The increase in bipolar cell dendritic extensions into ONL is strikingly similar to what occurs in very old (i.e. 24 month) mice [
2,
3]. A previous study [
4] shows that bipolar extensions into ONL are following retracted photoreceptor axons. We speculate that in XBP1 cKO mice, bipolar cell dysfunction, as indicated by reduction in ERG b wave, could be the underlying mechanism by which bipolar-photoreceptor synapses break down, leading to axon retraction. Interestingly, we did not observe any significant difference in the number of ONL neurons nor in PKC-α labeled bipolar cells in the retina of XBP1 cKO mice. This is consistent with an earlier report that aged mice do not have significantly fewer bipolar cells or rods than young mice [
2]. Thus, chronic dysfunction of bipolar cells may result in aberrant bipolar extension and ectopic synapsis in aged XBP1 cKO mice.
Another interesting finding from our study supports a role of microglial activation in age-related neuronal degeneration. We demonstrate a significant increase in disrupted calretinin-positive synaptic lamina colocalized with Iba1-positive microglia in the IPL of XBP1 cKO retina. Furthermore, the total number of activated Iba1-positive microglia in the IPL was also increased in the cKO mice. One important function of microglia is to monitor and eliminate unhealthy synapses [
23]. We propose that synaptic dysfunction in the XBP1 cKO retina may trigger microglial activation to eliminate damaged synapsis resulting in discontinuities in the trilaminar calretinin labeling, which was observed only in mice at about 1 year of age or older. Another possibility is that changes in Müller cells, which can be potentially affected by Chx10-driven XBP1 deletion, may influence microglia activation. A close interaction between microglial activation and Müller cell abnormality has been demonstrated previously [
24]. Although our preliminary study shows no significant change in the morphology of Müller cells, whether changes in the metabolism and function of these cells contribute to microglial activation in the cKO retina will be pursued in future studies. Nevertheless, the potential influence of XBP1 in retinal cells on microglial activation is of great interest, as the increased number of activated microglia has also been observed in aged human retinas [
25]. Uncontrolled microglial activation, through producing pro-inflammatory factors and interacting with Müller glia and retinal neurons, may exacerbate neuronal injury eventually resulting in degeneration of the retina with aging.
Finally, our study identifies a potential role of XBP1 in regulation of retinal metabolism, which could in part contribute to the premature decline in retinal function and structural deterioration in XBP1 cKO mice. Specifically, our data imply a more prominent role for XBP1 in the regulation and process of glycolysis. This finding is in line with a recent report demonstrating that silencing XBP1 in glioma cells inhibits glycolysis resulting in reduced ATP production and decreased cancer cell survival [
26]. Long-term activation of the IRE1 signaling has also been shown to reduce glucose metabolism and mitochondrial function; however, it is unclear whether this effect is mediated by XBP1 [
27]. Intriguingly, deficiency of XBP1 has been shown to induce an overactivation of IRE1, contributing to enhanced inflammation and apoptosis [
28]. Whether XBP1 deficiency influences IRE1 activation, which in turn reduces glycolysis in the cKO retina, is yet to be determined. In addition, although we observe no significant difference in mitochondrial function between the XBP1 cKO retina and the WT, a more detailed analysis of mitochondrial function in non-photoreceptor cells is warranted. In line with our findings, a study reports that mice carrying mutations in AMPK or LKB1, an upstream activator of AMPK, have a similar bipolar cell phenotype, as observed in our XBP1 cKO mice, attributed to retraction of rod axons [
4]. Furthermore, Lkb1 fl/fl; Chx10 Cre mice also demonstrate a large decline in the ERG response at 8–9 months of age [
4]. These observations collectively suggest a critical role of metabolic disturbance in the development of age-related retinal neurodegeneration. In addition, our data show that aged WT retina has an increased expression of CHOP, further supporting a potential role of long-term chronic ER stress in retinal aging. Whether increased CHOP expression contributes to age-related retinal structural and functional deterioration should also be evaluated in the future.
Acknowledgements
The authors thank Jonna Sakowski (University at Buffalo) for technical assistance and Zhang lab members for helpful discussion. We thank Dr. Laurie Glimcher (Harvard Medical School) for XBP1 floxed mice, Dr. Ivy Samuels, Dr. Neal Peachey, and Dr. Minzhong Yu (Cleveland Clinic) for detailed and very helpful advice and assistance with the ERG protocols and analyses, Emily Marra, Narayan Dhimal (University at Buffalo) for their contributions to various analyses.