|Home | About | Journals | Submit | Contact Us | Français|
The mechanisms leading to neuronal death in neurodegenerative disease are poorly understood. Many of these disorders, including Alzheimer’s (AD), Parkinson’s (PD) and prion diseases, are associated with the accumulation of misfolded disease-specific proteins. The unfolded protein response (UPR) is a protective cellular mechanism triggered by rising levels of misfolded proteins. One arm of this pathway results in the transient shutdown of protein translation, through phosphorylation of the alpha subunit of eukaryotic translation initiation factor, eIF2α. UPR activation and/or increased eIF2α–P levels are seen in patients with AD, PD and prion disease 1-4, but how this links to neurodegeneration is unknown. Here we show that accumulation of prion protein (PrP) during prion replication causes persistent translational repression of global protein synthesis by eIF2α–P, associated with synaptic failure and neuronal loss in prion-diseased mice. Further, we show that promoting translational recovery in hippocampi of prion-infected mice is neuroprotective. Over-expression of GADD34, a specific eIF2α–P phosphatase, as well as reduction of PrP levels by lentivirally-mediated RNAi, reduced eIF2α–P levels. As a result, both approaches restored vital translation rates during prion disease, rescuing synaptic deficits and neuronal loss, and thereby significantly increasing survival. In contrast, salubrinal, an inhibitor of eIF2α-P dephosphorylation5 increased eIF2α-P levels, exacerbating neurotoxicity and significantly reducing survival in prion diseased mice. Given the prevalence of protein misfolding and UPR activation in several neurodegenerative diseases, our results suggest that manipulation of common pathways such as translational control, rather than disease-specific approaches, may lead to new therapies preventing synaptic failure and neuronal loss across the spectrum of these disorders.
Neurodegenerative diseases pose an ever-increasing challenge for society and health care systems worldwide, but their molecular pathogenesis is still largely unknown and no curative treatments exist. Alzheimer’s (AD), Parkinson’s (PD) and prion diseases are separate clinical and pathological conditions, but it is likely they share common mechanisms leading to neuronal death. Mice with prion disease show misfolded prion protein (PrP) accumulation and develop extensive neurodegeneration (with profound neurological deficits), in contrast to mouse models of AD or PD, in which neuronal loss is rare. Uniquely therefore, prion-infected mice allow access to mechanisms linking protein misfolding with neuronal death. Prion replication involves the conversion of cellular PrP, PrPC, to its misfolded, aggregating conformer, PrPSc, a process leading ultimately to neurodegeneration6. We have previously shown rescue of neuronal loss and reversal of early cognitive and morphological changes in prion-infected mice by depleting PrP in neurons, preventing prion replication and abrogating neurotoxicity7-9. However, the molecular mechanisms underlying both the progression of disease, and those underlying recovery in PrP-depleted animals, were unknown.
In order to understand these processes better, we now analysed the evolution of neurodegeneration in prion diseased mice. We examined hippocampi from prion-infected tg37 mice used in our previous experiments7-10, in which the time course of impairment and recovery are clearly defined. Hemizygous tg37 mice express mouse PrP at ~3x wild type levels and succumb to RML (Rocky Mountain Laboratory) prion infection within 12 weeks post infection (wpi)10. They first develop behavioural signs with decreased burrowing activity at ~9wpi, following reduction in hippocampal synaptic transmission and first neuropathological changes7,8. This is the window of reversibility when diseased neurons can still be rescued: PrP depletion up to 9wpi, but not later, rescues neurotoxicity, as by 10wpi neuronal loss is established7-9. We measured PrP levels, synapse number, levels of synaptic proteins and synaptic transmission in prion-infected mice weekly from 5wpi, and burrowing behaviour from 6wpi. We examined brains histologically and counted CA1 neurons. (Cohorts of at least 30 animals were used per group; biochemical and histological analyses were done on 3 mice per time point, burrowing behaviour on 12, n for other analyses is indicated in figure legends). We found an early decline in synapse number in asymptomatic animals at 7wpi to ~55% of control levels (Fig. 1a), despite unchanged levels of several pre- and post-synaptic marker proteins (Fig. 1b). Reduced synapse number with normal synaptic protein levels is likely to reflect impaired structural plasticity of synapses at this early stage of disease. At 9wpi, however, there was a sudden decline in synaptic protein levels to ~50% of control levels for several pre- (SNAP-25 and VAMP-2) and post-synaptic (PSD-95 and NMDAR1) proteins (Fig. 1b and Supplementary Fig. 1b). This was associated with further decline in synapse number, and the critical reduction in synaptic transmission, both in amplitude of evoked excitatory post-synaptic currents (EPSCs) and in the number of spontaneous EPSCs (mEPSCs) in CA1 neurons (Fig.1c and Supplementary Fig. 1e). This was coincident with behavioural change (Fig. 1d) and first spongiform pathology (Supplementary Fig. 1d), and was rapidly followed by the onset of neurodegeneration, resulting in 50% reduction in hippocampal pyramidal neurons at 10wpi (Fig. 1e). All animals developed overt motor signs and were terminally sick by 12wpi.
The abrupt loss of synaptic proteins at 9wpi appeared to be a critical factor in the evolution of disease, occurring when synapse number and transmission were already declining. This could result from increased degradation, or decreased synthesis. Prion infection in mice is known to impair the ubiquitin proteasome system, causing reduction - not increase - in protein degradation11. We therefore asked if protein synthesis was reduced through translational control mechanisms. Given that total PrP levels rise during disease (Fig. 2a), and that PrP is synthesized in the endoplasmic reticulum (ER), we examined the translational repression pathway of the UPR. Rising levels of unfolded proteins detected by BiP/Grp78 (BiP) in the ER membrane cause auto-phosphorylation of protein kinase-like ER kinase (PERK). PERK-P phosphorylates eIF2α, which blocks the initiation step of translation, reducing new protein synthesis12. eIF2α–P then induces ATF4 and CHOP expression, ultimately leading to caspase 12 cleavage, and expression of GADD34, the stress-induced eIF2α-P specific phosphatase and key effector of a negative feedback loop that terminates UPR signaling, allowing translational recovery (Fig. 2g).
Upregulation of various steps in the pathway are seen in human prion cases13,14 and in prion-infected mice13; and increased phosphorylation of eIF2α occurs in AD and PD1-4. We characterized this pathway in prion diseased mice. We found that PERK-P and eIF2α-P increased throughout the course of disease (Fig. 2b,c and Supplementary Fig. 2b-d), in parallel with rising levels of total PrP, and the presence of detectable protease-resistant PrPSc (Fig. 2a). GADD34 levels did not change, despite rising eIF2α–P levels, suggesting insufficient GADD34 for dephosphorylation of increased eIF2α–P (Fig. 2d, Supplementary Fig. 2e). Caspase 12 cleavage occurred at 10wpi, following rising CHOP expression (Fig. 2e,f), coincident with onset of neuronal loss (Fig. 1e; see also Hetz et al13). However, the exact effector mechanism of neuronal death is unclear: we found neither apoptosis, nor autophagy, nor necrosis on examination of hippocampal slices (Supplementary Fig. 3); and neither Bax deletion, nor Bcl-2 overexpression15, nor caspase 12 deficiency16 are neuroprotective in prion disease.
We asked what effects the marked rise in eIF2α–P levels at 9wpi had on overall protein synthesis in hippocampi. We found abrupt, significant reduction in global translation rates, with a 50% decline in 35S-methionine incorporation in hippocampal slices from prion-infected mice at 9wpi compared to mice at 8wpi and uninfected controls (Fig. 2h), confirming sudden onset of reduced protein synthesis. We also looked at translation of specific mRNAs. We extracted polysomes from hippocampi of prion-infected mice. In successive polysomal fractions, mRNAs are associated with increased numbers of actively translating ribosomes. The change from a single fraction to the next reflects a large change in translation rate for any specific mRNA. Polysomal profiles at 9wpi showed a reduction in the overall number of actively translating ribosomes, represented by the smaller area under the curve between fractions 6-11 in prion-infected mice (Fig. 2i). Northern blots for specific mRNAs in individual polysomal fractions confirmed changes in actively translated messages consistent with eIF2α–P induction. Thus SNAP-25 and β-actin mRNAs showed a left shift to a lower polysomal fraction (Fig. 2j), representing reduced active translation (Fig. 1b). In contrast, ATF4 mRNA (which escapes eIF2α–P-mediated inhibition of translation, due to the presence of upstream open reading frames (uORFs) in its 5′ UTR17,18), showed increased active translation, represented by a right shift to a higher polysomal fraction (Fig. 2j, Supplementary Fig. 4 and 5). PrP mRNA did not show reduced translation, possibly due to the existence of similar translational control elements within the PrP gene. Indeed, human PrP mRNA has multiple upstream AUG uORFs in its 5′ UTR, which could allow it to escape eIF2α-P translational inhibition in the same way as ATF4 does17,18 (Supplementary Fig. 6).
Overall, these findings confirm that reduction in protein synthesis in prion disease is controlled at the translational, not the transcriptional, level, as it is rates of translation not levels of total mRNA that change (Supplementary Figs. 1c and 2g,h).
We propose that the key trigger to prion neurodegeneration is the continued unchecked activation of the UPR due to rising levels of PrP during disease, with fatal repression of translation rates. Importantly, prion neurotoxicity relates in a dose-dependent manner to PrP expression19-21. We therefore asked if levels of eIF2α-P and onset of neurodegeneration were related to levels of PrP in different strains of mice. We found that in homozygous tg37 mice, which over-express PrP ~6-fold, eIF2α-P was induced at 6wpi, and mice succumbed to prion infection at ~8wpi. In wild type C57/Bl6 mice, which express 1x levels of PrP, eIF2α-P was induced at 16wpi and animals succumbed at ~22wpi. Thus, as for hemizygous tg37 mice where PrP was expressed at 3x wild type levels and eIF2α-P was induced at 9wpi followed by death at 12wpi (Fig 2c), in each case there was a corresponding critical decline in synaptic proteins and synapse number after eIF2α-P induction (Supplementary Fig. 7; Fig 1a,b).
Transient eIF2α phosphorylation is beneficial to cells overloaded with misfolded proteins - reducing protein synthesis and increasing availability of chaperones, promoting refolding22,23. However, persistently high levels of eIF2α-P are detrimental in vitro24. To directly test the role of eIF2α-P in prion neurodegeneration in vivo, we first asked if reduction of eIF2α-P levels in prion disease would be neuroprotective. We used two approaches. We overexpressed GADD34, the eIF2α-P specific phosphatase to directly reduce eIF2α-P levels. In a separate experiment, we used targeted RNAi of PrP to abrogate UPR activation and prevent eIF2α-P formation. We then asked if increased levels of eIF2α-P exacerbate prion neurotoxicity by using salubrinal, a specific small molecule inhibitor of eIF2α-P dephosphorylation5, in infected mice. Salubrinal penetrates the blood brain barrier (Supplementary Fig. 8) and has been used for modulation of eIF2α-P dependent effects in ER stress-mediated processes in the central nervous system in vivo after peripheral administration25-27. Mice were inoculated with prions and received hippocampal injections of lentiviruses expressing GADD34 (LV-GADD34), anti-PrP shRNA (LV-shPrP), or YFP only (LV-control) at 5wpi, allowing 4 weeks for lentiviral expression to occur, before testing the effects of treatment on eIF2α-P levels and neurotoxicity at 9wpi. (All virally expressed constructs were driven by the CAMKII promoter for neuron-specific expression, Supplementary Fig. 9a-c). Another group of prion-infected mice received daily intra-peritoneal injections of salubrinal (1mg/kg), for 1 week, from 8wpi, with controls receiving vehicle alone. Two further control groups received normal brain homogenate, or RML prion inoculation alone (Fig. 3a).
We examined mice from each group at 9wpi when eIF2α–P-mediated translational repression occurs (Fig. 2c,h). LV-GADD34 treatment did not reduce PrP levels (Fig. 3b) and PERK-P levels were equivalent to those in prion only or LV-control treated animals (Fig. 3c and Supplementary Fig. 9d), confirming UPR activation in these mice. Critically, however, eIF2α–P levels were reduced (Fig. 3d and Supplementary Fig. 9e), strongly supporting its dephosphorylation by lentivirally-mediated GADD34 expression. LV-shPrP treatment reduced PrP levels (Fig. 3b) and prevented the PrP-induced rise in PERK-P and eIF2α–P seen in untreated animals (Fig. 3c,d), confirming prevention of UPR activation. Both GADD34 over-expression and PrP knockdown prevented prion-induced eIF2α–P-mediated translational repression, with restoration of global rates at 9wpi (Fig. 3e) and prevention of eIF2α–P-induced changes in translation of specific mRNAs (Fig. 3f and Supplementary Fig. 4). As a result, synaptic protein levels, synaptic transmission and synapse number in prion-diseased mice treated with GADD34 or PrP knockdown were protected and equivalent to levels in uninfected control mice (Fig. 3g,h,j). Burrowing deficits were prevented (Fig. 3i) and there was extensive neuronal protection in the hippocampus, with no neuronal loss and markedly reduced spongiform change (Fig. 3k). Further, targeted expression of LV-GADD34 and focal PrP knockdown had a modest, but highly significant, effect on survival, increasing this to 90±3 days and 92±5 days respectively, compared to 83±2 days for prion only mice (Fig. 4), and to 82±2 days for LV-control injected mice (Supplementary Fig. 10). For both LV-GADD34 and LV-shPrP, treatment was localised to the dorsal hippocampus, a very small area of the brain, so prion infection in the rest of the brain was fatal, but neuroprotection in GADD34 treated animals was seen even when the animals were terminally sick (Fig. 3k). More extensive brain-wide delivery of GADD34, or targeting of this pathway, would be predicted to further increase survival and give more widespread neuroprotection. Critically, treatment with salubrinal had the opposite effect, by preventing dephosphorylation of eIF2α-P. Thus, eIF2α-P levels were markedly higher at 9wpi than in prion-only controls (Fig. 3d and Supplementary Fig. 9e), causing further repression of global translation (Fig. 3e) and reduction of synaptic proteins (Fig. 3g). Salubrinal treatment resulted in earlier severe neuronal loss (Fig. 3k), and significantly accelerated disease, compared to untreated prion-infected mice (Fig. 4).
In conclusion, we have shown that PrP replication causes sustained UPR induction with persistent, deleterious expression of eIF2α-P in prion disease. The resulting chronic blockade of protein synthesis leads to synaptic failure, spongiosis and neuronal loss. Promoting eIF2α-P dephosphorylation rescues vital translation rates and is thereby neuroprotective, while preventing this further reduces translation and enhances neurotoxicity. The data support the development of generic proteostatic approaches22,28 to therapy - fine-tuning protein synthesis - in prion, and perhaps other neurodegenerative disorders involving protein misfolding.
All animal work conformed to UK regulations and institutional guidelines, performed under Home Office guidelines. tg3710 and C57/Bl6N mice (Harlan) were inoculated with 1% brain homogenate of Chandler/RML (Rocky Mountain Laboratories) prions aged 3-4 weeks, as described7. Animals were culled when they developed clinical signs of scrapie. Control mice received 1% normal brain homogenate. Hippocampi were processed for protein, RNA or histological analysis7-10. For all analyses n=3 mice unless otherwise stated.
Lentiviral plasmids were generated using the Invitrogen Gateway cloning system29. The neuron-specific promoter CAMKII was used to drive shPrP expression, C-terminal GADD34 expression, or YFP alone (control virus). Viruses were injected stereotaxically into the CA1 region of the hippocampus as described9.
Mice received daily intraperitoneal injections of 1mg/kg of salubrinal (Calbiochem), or vehicle (diluted DMSO in saline (Sigma))25, for 7 days from 8wpi.
Synapse numbers were counted in EM images of the stratum radiatum of the hippocampal CA1 region30. Synaptic marker proteins, UPR pathway proteins and PrP were analysed by immunoblotting of brain homogenates. PrPSc was detected after PK digestion7. Whole-cell recordings were done in acute hippocampal slices to measure synaptic transmission31. Global translation levels were detected using 35S−Methionine incorporation in acute hippocampal slices and translation of specific transcripts in polysomal fractions from hippocampi were analysed by Northern blotting32. Neuronal counts were determined by quantifying NeuN positive pyramidal CA1 neurons9. All analyses were performed using hippocampi from 3 mice in triplicate unless otherwise stated. Burrowing behaviour was performed as described on groups of 10 or more mice8. Statistical analyses were performed using using Prism v5 software, using Student’s t test for data sets with normal distribution and a single intervention. ANOVA testing was performed using one-way analysis with Tukey’s post-hoc test for multiple comparisons.
Lentiviral plasmids were generated by using the Invitrogen Gateway cloning system as described29. att-flanked cassettes were constructed by PCR amplification using att recombination site sequences and 20-25 template specific sequences. The C-terminus GADD34 cassette was amplified from FLAG-tagged C-term GADD34 using attB5 5′-ggggacaactttgtatacaaaagttggcaccatgcgttcaggagaggcgtccga-3′; and attB4 5′-ggggacaactttgtatagaaaagttgggtgttggtctcagccacgcctcccac3′ primers; WPRE and YFP cassettes were amplified from pLL3.7-shPrP plasmid using attB3 5′- ggggacaactttgtataataaagttgtcaacctctggattacaaaatttgt-3′ and attB2 5′- ggggaccactttgtacaagaaagctgggtatgcggggaggcggcccaaagggaga-3′ primers for WPRE and attB4r 5′ accatggtgagcaagggcga-3′ and attB3r 5′ ttacttgtacagctcgtccatgccg-3′ primers for YFP . The CAMKII promoter was amplified from using attB1 5′- ggggacaagtttgtacaaaaaagcaggctacttgtggactaagtttgttcgc-3′ and attB5r primer 5′- ggggacaacttttgtatacaaagttgtctgcccccagaactaggg-3′primers. Cassettes were amplified and recombined into appropriate pDONR entry vectors. Once the entry plasmids were generated they were recombined with the pLenti6/BLOCK-iT/DEST vector (Invitrogen) to construct lentiviral plasmids. We used lentiviral plasmids containing shPrP and empty vector constructs as described9. Lentiviruses were generated using DNA/Ca2+ phosphate transfection of HEK293 cells33. Additional stocks of virus were generated by GenTarget, Inc. (San Diego, CA) and titre determined using FACS (BD FACS Calibur). Viruses were used with a final titre of 0.6-1.5 × 108 TU.
Mice were anesthetized using isofluorane and injected into CA1 region of the hippocampus as described9.
Semi-thin sections for electron microscopy were obtained by terminal perfusion of mice as described30. Ultrathin sections (70 nm) were examined in a JEOL 100-CXII electron microscope (JEOL (UK) Ltd) equipped with a ‘Megaview III’ digital camera (Olympus Soft Imaging Solutions GmbH, Munster, Germany). Synapses were scored using criteria of structures showing a postsynaptic density, containing synaptic vesicles and a synaptic junction. 32 images from two mice were used for scoring.
Whole-cell recordings were made from identified CA1 neurons and recording performed as described30. In brief, neurons were voltage-clamped using a Multiclamp 700B amplifier and pClamp 10.3 software (Molecular Devices) and EPSCs were evoked by stimulation with bipolar platinum electrode at 37°C. Pipettes (2.5-3.5MΩ) were filled with a solution containing (mM): KCl 110, HEPES 40, EGTA 0.2, MgCl2 1, CaCl2 0.1, pH was adjusted to 7.2 with KOH. Neurons were visualized with 60x objectives on a Nikon FS600 microscope fitted with differential interference contrast (DIC) optics. 4-8 cells were measured per mouse in at least 2 animals per experiment.
Protein samples were isolated from hippocampi using protein lysis buffer (50mM Tris, 150mM NaCl, 2mM EDTA, 1mM MgCl2, 100mM NaF, 10% glycerol, 1% Triton X-100, 1% Na deoxycholate, 0.1% SDS and 125mM sucrose) supplemented with Phos-STOP and protease inhibitors (Roche). Synaptic protein levels were determined by resolving 20mg of protein on SDS-PAGE gels, transferred onto nitrocellulose membrane and incubated with primary antibodies, SNAP25, (1:10000; Abcam), VAMP2 (1:5000; Synaptic Systems), NMDA-R1 (1:1000; Sigma) and PSD95 (1:1000; Millipore). Odyssey IRDye800 seconday antibodies (1:5000; LI-COR) were applied, visualized and quantitated using Odyssey infrared imager (LI-COR; software v3.0). Protein for PrP levels and UPR pathway activation were determined using the primary antibodies, 8H4 for total PrP (1:1000; Abcam), ICSM35 for PrPSc (1:10,000; D-GEN), PERK-P, total PERK, P-eIF2a, eIF2a (1:1000; Cell Signaling), CHOP (1:1000; ThermoScientific), Caspase12 (1:1000; Exalpha), BiP/Grp78 (1:1000; Stressgen) and GADD34 (1:1000; ProteinTech). Horseradish peroxidase (HRP) conjugated secondary antibodies (1:5000; DAKO) were applied and protein visualized using enhanced chemiluminescence (GE Healthcare) and quantitated using ImageJ. Antibodies against GAPDH (1:5000; Santa Cruz) or β-tubulin (1:5000; Millipore) were used to determine loading.
Slices were dissected in an oxygenated cold (2–5°C) sucrose artificial cerebrospinal fluid (ACSF) containing (mm): 26mM NaHCO3, 2.5mM KCl, 4 mM MgCl2, 0.1 mM CaCl2, and 250mM sucrose. Hippocampal slices were prepared using a tissue chopper (McIlwain). Slices were allowed to recover in normal ACSF buffer while being oxygenated at 37°C for 1hr, then incubated with [35S]-Methionine label for one hour, then homogenized. Proteins were TCA precipitated and incorporation of radiolabel was measured by scintillation counting (Winspectal, Wallac Inc.).
Sucrose density gradient centrifugation was used to separate hippocampal homogenates into polysomal and subpolysomal fractions. Polysomal fractions were isolated as described32. Briefly, hippocampi were dissected in ice-cold gradient buffer (0.3 M NaCl, 15 mM MgCl2, 15 mM Tris-HCL (pH7.4), 0.1 mg/ml cyclohexamide, and 1 mg/ml heparin). The hippocampal tissue was homogenised in gradient buffer containing RNase inhibitors and 1.2% TritonX-100 added. Samples were centrifuged and the supernatants layered onto 10-60 % sucrose gradients. The gradients were sedimented at 38,000 rpm for 120 minutes at 4°C. 1 ml fractions were collected from the gradients into 3 ml of 7.7 M guanidine-HCL using a Foxy R1 gradient fractionator (Teledyne ISCO; ISCO peak Trak v1.10 software) with continuous measurement of the absorbance at 254 nm. RNA was then isolated and equal volumes of each fraction were analysed by Northern blot analysis32.
Paraffin embedded brains were either stained with NeuN antibody (1:200; Millipore) for neuronal counts. CA1 pyramidal neuron counts were determined using three serial sections from three separate miceAll images were taken on using Axiovision 4.8 software (Zeiss) and counted using volocity imaging system.
Student’s t tests were applied to all data sets with two tails (two samples; unequal variance). ANOVA testing was performed using one-way analysis with Tukey’s post-hoc test for group effects. Statistical tests were performed using Prism v5. All data in bar charts show mean ± s.e.m.
Supplementary information: Supplementary figures 1-10 and legends, Supplementary methods and references.
We thank David Read for imaging analysis, and Jenny Edwards, Tim Smith, Judy McWilliam, Paul Glynn and Colin Molloy for technical assistance; Professor John Collinge (MRC Prion Unit) for the original RML prion inoculum and Professor Ken Liddle for critical reading of the manuscript. This work was funded by the Medical Research Council, UK.
Author contributions: JAM did most of the experimental work and analysis. NV and MGM performed stereotaxic surgery and prion inoculations. HR, DP, MH and JM performed various experiments, DD performed EM analyses, JRS performed electrophysiological analysis, CAO and DAB performed mass spectrometry analysis, PT and AB worked with JAM in Cambridge, AEW and MB contributed expertise and direction on translational control mechanisms, GRM directed and supervised the project. JAM and GRM wrote the paper. All authors contributed to discussion and analysis of data and final draft of paper.
The authors declare no conflict of interest.