Background
Alzheimer's disease (AD) is a neurodegenerative disorder characterized clinically by progressive loss of memory along with other cognitive skills and pathologically by accumulation of amyloid plaques and neurofibrillary tangles [
1]. While most cases occur sporadically some are inherited in an autosomal dominant fashion. Familial cases (FAD) exhibit similar clinical and pathological features as the sporadic disease but have a generally earlier age of onset [
2].
The presenilin-1 gene (
Psen1) was discovered because mutations in it and a homologous gene
presenilin-2 (
Psen2) cause FAD [
2]. To date more than 170 different mutations in
Psen1 have been associated with FAD [
3]. Mutations in
Psen1 are the most commonly recognized cause of early onset FAD accounting for probably 50% of all autosomal dominant cases [
4]. Mutations in
Psen2 are a less common cause of FAD [
2].
The pathophysiological cascade in AD includes deposition of β-amyloid (Aβ) in compact and diffuse plaques as well as production of oligomeric forms of Aβ that are currently thought to be the most toxic species [
5]. The factors that trigger the amyloid cascade remain incompletely understood. Recently there has been interest in the role that vascular disease might play in this process [
6]. Indeed an array of epidemiologic evidence supports an association between vascular disease and its risk factors with cognitive impairment and AD [
6] and postmortem studies have shown that AD is complicated by vascular pathology in about one-third of cases [
7].
These observations could be explained by viewing vascular disease as a co-morbidity that acts in an additive fashion with AD pathology to exacerbate ongoing cognitive decline. Alternatively vascular disease might lead to hypoxia/ischemia that might drive AD pathology itself. Supporting the latter hypothesis both hypoxia and ischemia increase amyloid precursor protein (APP) expression as well as shunt APP processing towards Aβ production [
8‐
13]. Hypoxia also decreases expression of the α-secretase, ADAM-10 [
11] while increasing expression of the principal β-secretase, BACE1 [
14,
15].
Thus aberrant or inefficient hypoxic responses in brain might contribute to AD initiation or progression. The brain's response to hypoxia involves a multiplicity of pathways that as in other tissues includes induction of a set of genes regulated by the transcription factor hypoxia inducible factor-1 (HIF-1)[
16,
17]. HIF-1 is a dimer that consists of α and β subunits. Levels of HIF-1α are regulated by oxygen availability while HIF-1β subunits are constitutively expressed with the level of HIF-1 transcriptional activity regulated primarily by levels of HIF-1α. HIF-1 is a key transcriptional regulator of a broad range of cellular genes that mediate responses to hypoxia. Although originally discovered as an hypoxia induced factor, HIF-1α also increases in response to stimulation by a number of metabolic/growth factor signaling pathways including insulin and insulin like growth factor-1 (IGF-1)[
18‐
23]. Interestingly both Psen1 and Psen2 are induced by hypoxia [
24‐
26] and Psen1 has been suggested to regulate signaling through a number of growth factor related pathways [
27‐
29]. In addition HIF-1α mRNA was found reduced in a hypomorphic
Psen1 mouse mutant [
30,
31]. Here we investigated whether Psen1 might regulate hypoxic responses and show that Psen1 influences both hypoxic and metabolic/growth factor induction of HIF-1α in cells that lack Psen1 although the effects vary in a cell type dependent manner. We also show that a
Psen1 FAD mutation alters metabolic induction of HIF-1α.
Discussion
Psen1 is a polytopic transmembrane protein that was first discovered because of its association with FAD [
2,
48]. Psen1 influences multiple molecular pathways being best known for its role as a component of the γ-secretase complex [
48]. HIF-1 is a transcription factor that was first recognized as a DNA-binding protein that mediates hypoxia-inducible expression of the erythropoietin (EPO) 3' enhancer. It is now known that HIF-1 is a key transcriptional regulator of a wide variety of cellular genes that are induced by and regulate cellular responses to hypoxia [
16,
17,
49,
50].
Here we show that Psen1 regulates induction of HIF-1α. Specifically we demonstrate that there is impaired induction of HIF-1α in fibroblasts lacking Psen1 following stimulation with the hypoxia mimetic cobalt chloride as well as treatment with insulin and calcium chelation. Associated with impaired induction there was reduced and delayed activation of the HIF-1α target genes Vegf and Glut-1. We further show that decreased basal levels of HIF-1α best account for this impairment in fibroblasts lacking Psen1.
Depressed levels of basal HIF-1α in Psen1-/- fibroblasts may in part be the result of lowered levels of HIF-1α RNA. However the ≈ 50% reduction in HIF-1α RNA levels did not appear sufficient to explain the more than 5-fold reduction in basal levels of HIF-1α protein in Psen1-/- fibroblasts. Rather these results suggested that in addition either constitutive synthesis of HIF-1α was decreased at a post-transcriptional level or that constitutive degradation was increased. In further studies we found that increased degradation best accounts for these effects showing that when proteasomal degradation of HIF-1α was blocked, HIF-1α levels accumulated as rapidly in Psen1-/- fibroblasts as in Psen1+/+ cells. By contrast when translation was blocked with cycloheximide levels of HIF-1α decreased more rapidly in Psen1-/- fibroblasts than Psen1+/+ with an approximate 50% reduction in HIF-1α half-life.
Mechanistically we show that Psen1's γ-secretase activity is not needed for HIF-1α induction by insulin nor cobalt chloride and that in the case of insulin induction of HIF-1α, impaired signaling through the PI3K/Akt pathway cannot account for the effect. Rather we show that Psen1 and HIF-1α physically interact suggesting that Psen1 may stabilize HIF-1α and protect it from degradation.
Both mouse and human Psen1 are initially produced as an ≈ 45-48 kDa holoprotein. However, in most cells and tissues nearly all of the holoprotein is cleaved into an ≈ 30 kDa NTF and an ≈18 kDa CTF, that are produced by endoproteolytic cleavage within a large cytoplasmic loop region [
48]. NTFs and CTFs associate with each other as heterodimeric components of a larger multimeric protein complex. After incorporation into these high molecular weight complexes, NTFs and CTFs remain associated and attain relatively long half-lives.
Our co-IP studies suggest that HIF-1α associates with both the Psen1 NTF and CTF and likely forms a part of these complexes. Psen1 holoprotein that remains monomeric is rapidly degraded by a proteasome-dependent mechanism [
51,
52] and complex incorporation is likely necessary for the biological activity of presenilins [
48]. Psen1 and 2 proteolysis is also regulated by the proteasome interacting proteins ubiquilin-1 and -2 [
53‐
55] and Psen1 influences proteasomal degradation of another component of the γ-secretase complex Pen-2 [
43]. Indeed the effects of Psen1 on Pen-2 appear relatively similar to the effects of Psen1 on HIF-1α with in both cases the absence of Psen1 leading to accelerated protein degradation through the proteasome. Interestingly, overexpression of Psen1 has also been reported to increase levels of HIF-1β in pancreatic islet cells [
56] pointing to a possibly larger role of Psen1 in HIF-1 signaling although the mechanistic basis for Psen1's effect on HIF-1β is unknown.
In contrast to the more global effects that the absence of Psen1 has in fibroblasts, hypoxic signaling proceeded normally in
Psen1-/- neurons and in fibroblasts harboring the M146V FAD mutation while in these same cells metabolic signaling through insulin and IGF-1 was impaired. HIF-1α levels were also normal in
Psen1-/- embryos arguing that activation in neuronal cells during development which is probably mostly driven by hypoxia [
47] can proceed normally without Psen1.
While hypoxia regulates HIF-1α levels mostly through effects on protein degradation by the proteasome, metabolic/growth factor stimulation principally increases HIF-1α protein synthesis by stimulating transcription/translation while degradation proceeds at a constant rate [
33]. Of these two pathways, hypoxic induction of HIF-1α results in more dramatic elevations of HIF-1α than metabolic/growth factor stimulation likely because degradation is the key rate limiting step in setting basal HIF-1α levels while factors such as insulin and IGF-1 induce HIF-1α by stimulating synthesis on a background of continued degradation.
Interestingly, unlike in Psen1-/- fibroblasts, basal levels of HIF-1α were not reduced in fibroblasts harboring the M146V FAD mutation or in Psen1-/- primary neurons. While the basis for these differences in HIF-1α basal levels remains to be explained, they provide a potential explanation for why HIF-1α responsiveness differs between cell types in that hypoxic induction of HIF-1α depends on a basal level of HIF-1α being present. Metabolic/growth factor induction being weaker requires a higher basal level of HIF-1α to be present in order to produce detectible induction. Since Psen1-/- fibroblasts have dramatically reduced levels of basal HIF-1α neither hypoxic nor metabolic/growth factor induction is possible. By contrast in FAD mutant fibroblasts or Psen1-/- primary neurons, a basal level of HIF-1α is present that is sufficient to support hypoxic induction while metabolic/growth factor induction is impaired.
Of the metabolic pathways implicated in HIF-1α regulation the PI3K/Akt and MAP kinase are the best-documented [
33] and many of insulin's as well as IGF-1's cellular effects are mediated by the PI3K/Akt pathway [
40]. In at least some cells, insulin's induction of HIF-1α has been shown to occur through a PI3K/Akt dependent mechanism [
19] and induction of HIF-1α by IGF-1 has also been shown to be dependent on PI3K/Akt activation [
21,
34]. Psen1 is known to influence signaling through the PI3K/Akt pathway, an effect that is γ-secretase independent [
27,
29].
Interestingly we did not find effects on PI3K/Akt induction in Psen1-/- fibroblasts or neurons although we found that in fibroblasts harboring the Psen1 M146V FAD mutant insulin receptor activation as well as activation of PI3K/Akt and Akt's downstream target GSK-3β were impaired. Thus while a direct effect of Psen1 on insulin/IGF-1 signaling might explain the failed activation of HIF-1α in FAD mutant fibroblasts, it cannot account for the failed activation in Psen1-/- neurons. Future studies will be needed to sort out the cell type specific differences as well as the differences between the effects of the absence of Psen1 and the presence of Psen1 FAD mutants.
The implications of these observations for the pathophysiological effects of
Psen1 FAD mutations remains speculative. Insulin receptors are expressed widely in brain including areas involved in AD such as the hippocampus [
57]. The widespread expression of insulin receptors in brain suggests that insulin signaling plays a significant functional role in brain and indeed signaling through the insulin receptor is known to influence a wide variety of cellular functions related to cellular homeostasis and growth factor signaling including affecting synaptic transmission and neurotransmitter levels as well as exerting effects on learning and memory [
58,
59]. IGF-1 signaling is also of interest since it has many links to brain function as well as potential pathophysiological connections to neurodegenerative diseases including AD [
60‐
63]. There has also been much interest in the potential role of insulin resistance in AD with most epidemiologic studies suggesting diabetes to be a risk factor for AD [
64]. Diabetes and anti-diabetic medication have also been shown to influence AD related neuropathology [
65,
66].
Both insulin and IGF-1 signaling have connections to Aβ production [
67‐
74] and tau phosphorylation [
75‐
80] as well as have roles in modulating neural inflammatory reactions and neuronal apoptosis [
81,
82]. Thus an effect of Psen1 FAD mutations on insulin and IGF-1 related signaling could impact AD-related pathology including but not limited to effects on HIF-1α induction.
Materials and methods
Genetically modified mice
The
Psen1-/- mice utilized were those generated by Shen et al. [
84]. These animals were provided on a mixed genetic background and have been maintained by breeding to C57BL/6 wild type mice. Genotypes were determined by PCR utilizing primers from the
Psen1 intron 1 (5'ACCTCAGCTGTTTGTCCCGG3'), the neo gene (5'GCACGAGACTAGTGAGACGTG3') and
Psen1 exon 3 (5'TCTGGAAGTAGGACAAAGGTG3') as described in Shen et al. [
84]. Heterozygous mice were mated to produce
Psen1-/- embryos with the day a vaginal plug was detected designated as E0.5. Pregnant female mice were euthanized with carbon dioxide and
Psen1-/- embryos were presumptively identified based on their gross dysmorphic appearance. A portion of the body was saved and used to isolate DNA and confirm genotypes. Due to the mixed genetic background,
Psen1+/+ and
Psen1+/- littermates were used as controls. Since
Psen1+/- mice have never been observed to exhibit any developmental abnormalities, they were treated as wild type controls.
Transgenic mice expressing wild type human Psen1 from a P1 bacteriophage artificial chromosome (PAC) were produced using the clone RP1-54D12 (204 kb; accession number AC 006342) containing the entire human Psen1 transcription unit (> 75 kb). Founders were generated by injecting C57BL/6 × C3 H F
1 oocytes. In order to generate PAC transgenic mice expressing the M146V FAD mutation, the 54D12 clone was retrofitted using the Rec A-mediated homologous recombination system described by Ali Imam et al. [
85] and transgenic founders were generated as described above. Further details concerning production and characterization of the
Psen1 wild type and FAD mutant PAC transgenic mice can be found elsewhere [
46]. All protocols were approved by the Institutional Animal Care and Use Committees of the James J. Peters Department of Veterans Affairs Medical Center (Bronx, NY, USA) and the Mount Sinai School of Medicine (New York, NY, USA) and were conducted in conformance with Public Health Service (PHS) policy on the humane care and use of laboratory animals and the NIH Guide for the Care and Use of Laboratory Animals.
Cell culture
Generation of immortalized
Psen1+/+ and
Psen1-/- mouse embryonic fibroblasts has been described previously [
27]. Immortalized fibroblast lines were also developed from the Psen1 wild type and M146V PAC transgenic mice described above. In order to eliminate effects of the endogenous mouse
Psen1, both PAC transgenic lines were bred onto the mouse
Psen1-/- background and fibroblast cell lines were established from E16.5 embryos. After removing the head and liver, embryos were finely minced and treated with 0.25% trypsin (Invitrogen, Carlsbad, CA, USA), 0.1% EDTA. Trypsinized tissues were plated in Dulbecco's Modified Eagle's Medium (DMEM) supplemented with 20% heat-inactivated fetal calf serum (Mediatech, Manassas, VA, USA), penicillin, streptomycin, fungizone, and glutamine (Invitrogen) at 37°C in 5% CO
2. Explanted cells were continuously passed in the same medium until lines were established. Fibroblast cell lines lacking both presenilins (
Psen-/-) were a gift from Dr. Nikolaos Robakis (Mount Sinai School of Medicine, New York, NY, USA).
Immortalized fibroblasts were maintained in DMEM medium supplemented with 20% FCS at 37°C and 5% CO2. To study HIF-1α induction under hypoxic conditions confluent cells were treated with the hypoxia mimetic CoCl2 (100 μM, Sigma Aldrich, St. Louis, MO, USA) for 4 hrs. For insulin or IGF-1 treatment, cells were serum starved overnight and treated with 2.4 μg/ml of insulin (porcine pancreas, Sigma Aldrich) or 100 ng/ml IGF-1 (Sigma Aldrich) for 4-6 hours. Cells were washed with phosphate buffered saline (PBS) and lysed in 10 mM Na phosphate buffer, pH 7.4, 150 mM NaCl, 2 mM EDTA, 1%Triton X-100, 0.25% Na deoxycholate, and 0.5% SDS, supplemented with HALT protease inhibitor cocktail (Pierce, Rockford, IL, USA) and phosphatase inhibitor cocktails I and II (Sigma Aldrich). The lysates were briefly sonicated and centrifuged at 14,000 rpm for 15 min. The cleared supernatant was then collected. Protein concentrations were determined with the BCA reagent (Pierce).
In some experiments Psen1 +/+ and Psen1-/- fibroblasts at ≈70% confluence were treated with 5 μM 1,2-bis(o-aminophenoxy)ethane-N,N,N',N'-tetraacetic acid tetrakis (acetoxymethylester) (BAPTA-AM, Sigma Aldrich), 10 μm MG132 (Calbiochem, San Diego, CA, USA), or 100 μg/ml cycloheximide (Sigma-Aldrich). For γ-secretase inhibition experiments, Psen1 +/+ cultures were treated overnight with 1 μM inhibitor XXI (also known as compound E, Calbiochem) or 1 μM L-685,458 (Calbiochem).
Primary neuronal cultures were prepared from the cerebral cortex isolated from E15.5-16 embryos as previously described [
86,
87] and maintained in Neurobasal medium/B27 supplement (Invitrogen) for 5-7 days
in vitro (DIV). For treatment with insulin or IGF-1, neurons were maintained overnight in neurobasal medium devoid of B27 supplement and then treated as described above.
Tissue harvesting
Brains were homogenized in 10 volumes of RIPA buffer (50 mM Tris HCl, pH 7.6, 0.15 M NaCl, 1 mM EDTA, 1% Triton X100, 1% sodium deoxycholate, 0.1% SDS) containing HALT protease and phosphatase inhibitor cocktails (Pierce Biotechnologies). The samples were centrifuged at 15,000 rpm for 20 minutes and the supernatant collected. Protein concentrations were determined using the BCA reagent assay (Pierce).
Western Blotting
Protein samples were separated by SDS-PAGE and blotted onto polyvinylidene difluoride (PVDF) membranes (Millipore Corporation, Billerica, MA, USA). Blots were blocked with 50 mM Tris HCl, pH 7.6, 0.15 M NaCl, 0.1% Tween-20 (TBST), 5% non fat dry milk and probed overnight with the relevant primary antibody diluted in blocking solution or in 5% BSA/TBST in the case of phospho-specific antibodies. Blots were then incubated for 1.5 hours with the appropriate horseradish peroxidase (HRP) conjugated secondary antibody (GE Healthcare Bio-Sciences Corporation, Piscataway, NJ, USA) diluted in blocking solution (1:5000-1:10,000) and bands visualized by ECL+ (GE Biosciences) after exposure to CL-Xposure film (Pierce) and/or imaged on a Kodak Image Station 4000R (Carestream Molecular Imaging, New Haven, CT, USA). Quantification was performed using Image Quant TL software (GE Biosciences). For reprobing, the membranes were treated with Re-Blot Plus strong stripping solution (Millipore) according to the manufacturer's instructions.
The primary antibodies utilized were rabbit polyclonal (1:1500 dilution; Novus Biologicals, Littleton, CO, USA) or monoclonal (Mab5382, 1:1000; Millipore) antibodies to HIF-1α, a monoclonal antibody to the cytoplasmic domain of N-cadherin (1:1000; BD Biosciences, San Jose, CA, USA), rabbit monoclonal antibodies to phospho-Akt (Ser473), total Akt, phospho-GSK-3β (Ser9), total GSK-3β, and the phospho-insulin receptor (all at 1:1500 dilutions; Cell Signaling, Beverly, MA, USA), as well as a mouse monoclonal antibody to the insulin receptor β (MAB 1139, 1:500; Millipore). Mouse monoclonal anti-actin antibody AC15 (1:500; Sigma Aldrich) or rabbit polyclonal anti β-tubulin (1:1500; Abcam, Cambridge, UK) antibodies were used as loading controls.
Lentivirus infection
A lentiviral construct (2 μg) containing human Psen1 cloned in vector pReceiver-LV31 (Z 0049, Genecopeia, Rockville, MD, USA) was transfected with 10 μg of the packaging plasmids pLV-PK-FIV and pLV-PK-VSG (Genecopeia) into 293Ta cells using the Lipofectamine reagent (Invitrogen). Forty-eight hours post-tranfection the culture supernatant containing the recombinant lentiviruses was centrifuged for 10 minutes at 1500 rpm to clear debris and filtered through a 0.45 μm PVDF filter. To infect fibroblasts 1 ml of the lentiviral supernatant containing 8 μg/ml of polybrene was added (multiplicity of infection = 10). After overnight incubation, the viral supernatant was replaced with DMEM containing 20% fetal calf serum and cultured for 48 h. The cells were then treated with cobalt chloride and analyzed for HIF-1α as described above.
Co-immunoprecipitations
An expression ready plasmid for mouse HIF-1α (clone ID 4019056) was obtained from Open Biosystems (Huntsville, AL, USA) and for human Psen1 (clone sc125532) from Origene (Rockville, MD, USA). Psen1 and HIF-1α expression plasmids (3 μg each) or empty vector were transfected singly or together into HEK293 cells using the Fugene 6 reagent as recommended by the manufacturer (Roche, Indianapolis, IN, USA). After 48 hours cells were treated for 1 hour with 100 μM MG132 and harvested. The cells were lysed in an IP buffer consisting of 50 mM Tris HCl pH 7.4, 150 mM NaCl, 1 mM EDTA, 0.5% NP-40 supplemented with a protease inhibitor cocktail (Halt, Pierce) and phosphatase inhibitor cocktails I and II (Sigma Aldrich). Lysates (500-750 μg protein) were pre-cleared with protein A/G (Santa Cruz Biotechnology, Santa Cruz, CA, USA) for 2 hours and then incubated overnight at 4°C with one of the following antibodies: rabbit anti-Psen1 NTF (1 μg; Santa Cruz), mouse monoclonal antibody NT.1 which recognizes the NTF of human Psen1 (3 μg, gift of Dr. Paul Mathews, Nathan Kline Institute, Orangeburg, NY, USA), a mouse monoclonal anti-Psen1 loop antibody (2 μg, MAB5232 Millipore), a mouse monoclonal anti-HIF-1α antibody (3 μg, clone ESEE 122, Novus) or rabbit IgG (Santa Cruz). 20 μl of protein A/G was then added for 2 hours at 4°C and the samples were washed 4 times (20 min each) with IP buffer. SDS-PAGE buffer was added and samples were boiled or for Psen1 blots heated at 50°C for 10 minutes and analyzed by Western blot using the antibodies described as well as the mouse monoclonal antibody 33B10 (1:3,000, gift of Dr. Nikolaos Robakis, Mount Sinai School of Medicine, New York, NY, USA) which recognizes both the human and mouse Psen1 CTF. Clean-Blot IP HRP detection reagent (Pierce) was used to minimize reactivity with denatured IgG in co-immunoprecipitations when the blots were probed with rabbit polyclonal antibodies. For co-immunoprecipitation of endogenous HIF-1α in Psen1+/+ fibroblasts, cells were treated with 100 μM CoCl2 and 10 μM MG132 for 4 hours and then immunoprecipitations were performed with anti-Psen1 antibodies as described above.
Quantitation of mRNA expression by real time quantitative PCR
Total RNA was isolated from cultured cells using the Ribopure kit (Ambion, Austin, TX, USA). Residual genomic DNA was removed using the DNA-free kit (Ambion) and cDNA was synthesized from 0.8 μg of total RNA using the High Capacity cDNA Reverse Transcription Kit (Applied Biosystems, Foster City, CA, USA). Real-time quantitative polymerase chain reaction (qPCR) was performed using an ABI Prism 7500 or 7700 Sequence Detector and TaqMan FAM/MGB gene-specific fluorogenic assays mostly as previously described [
88]. Pre-designed TaqMan Gene Expression assays (probe and primer mix) for all target and control genes were purchased from Applied Biosystems (Mm00437304_m1 VegfA, Mm00441480 Glut-1, Mm00468875_m1 HIF-1a, Mm00446953_m1 Gusb and Mm02342429_m1 Ppia). Each 20 μl reaction contained 5 μl of the relevant cDNA (diluted 25 times in H
2O), 1 μl of a specific TaqMan assay, and 10 μl of the 2×PCR Universal Master Mix (Applied Biosystems). The thermal cycling program consisted of 2 min at 50°C, 10 min at 95°C, followed by 40 cycles of 15 sec at 95°C and 1 min at 60°C. Only one cDNA was amplified in each qPCR (monoplex). The reactions were run in triplicate for each sample. Relative expression values of target genes were calculated using the 2
DDCt method [
89], with the amount of target normalized to the geometric mean of the expression values for the endogenous control genes peptidylprolyl isomerase A (Ppia) and Glucuronidase beta (Gusb) relative to a calibrator (generated by pooling all the samples).
Statistical procedures
All data are presented as mean ± the standard error of the mean (S.E.M.). Statistical comparisons were made using unpaired t-tests (Student's t if the variances did not differ significantly, p > 0.05, by Levene's test; otherwise the Welch correction for unequal variances) or one-way analysis of variance (ANOVA) with Dunnett's post-test. Statistical tests were performed using the program GraphPad Prism 5.0 (GraphPad Software, San Diego, CA, USA) or SPSS 16.0 (SPSS, Chicago, IL, USA).
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
RDG participated in the design, execution and data analysis of all the experiments as well as participated in the manuscript writing; MAGS participated in the experimental design as well as generation and maintenance of the fibroblast cell lines and manuscript writing; SP provided expert advice and assistance for the qPCR experiments including data analysis; GAE participated in the experimental design, data analysis and manuscript writing. All authors read and approved the final manuscript.