Hypoxia-inducible Factor 1 ( HIF-1 )-mediated Hypoxia Increases BACE 1 Expression and-Amyloid Generation *

Xian Zhang, Kun Zhou, Ruishan Wang, Jiankun Cui, Stuart A. Lipton, Francesca-Fang Liao, Huaxi Xu, and Yun-wu Zhang From the Institute for Biomedical Research and School of Life Sciences, Xiamen University, Xiamen 361005, China, the Institute of Health Sciences, Shanghai Institutes for Biological Sciences, Chinese Academy of Sciences and Shanghai Jiao Tong University School of Medicine, Shanghai 200025, China, and the Burnham Institute for Medical Research, La Jolla, California 92037

these transcription factors have been found altered in AD brain regions containing A␤ plaques, suggesting that increased BACE1 expression resulting from altered levels of its transcriptional activators may contribute in part to AD pathogenesis (17)(18)(19)(20).
Hypoxia-inducible factor 1 (HIF-1) is a major transcription factor that responds to cellular oxygen reduction. HIF-1 has two subunits, HIF-1␣ and HIF-1␤. HIF-1␣ levels are low under physiological conditions but increase dramatically under hypoxia. HIF-1␣ binds to HIF-1␤ to form functional HIF-1 heterodimers, consequently regulating a series of transcriptional events (21). HIF-1␣ is essential for adaptation to low oxygen levels, and a complete deficiency in HIF-1␣ leads to embryonic lethality (22). Cerebral ischemia results from an insufficient oxygen supply to the brain and causes neuronal damage in vulnerable brain areas (23). Accumulating evidence indicates that cerebral ischemia/stroke significantly increases AD risk (24 -26). Moreover, APP expression is elevated in post-ischemic brain, and cleavage of APP leading to amyloidogenic A␤ peptides may hence be increased by ischemia (27)(28)(29). Although it is likely that cellular stresses such as ischemia or hypoxia facilitate or worsen the pathogenesis of AD, molecular links between these conditions have yet to be established. We recently characterized the promoter of the APH-1a gene, which encodes a key component of the PS1/␥-secretase com-plex, and identified a functional HIF-1 binding element. We further demonstrated that activation of HIF-1 by short-term hypoxia increases expression of APH-1a mRNA and protein, leading to increased ␥-cleavage of APP and the Notch receptor (30). Sequence analysis of the mouse BACE1 promoter revealed a putative HIF-1 binding element, although there are no reports of regulation of BACE1 mRNA expression by HIF-1 or hypoxia. Given that BACE1 cleavage of APP is a prerequisite for ␥-cleavage and thus widely viewed as the key regulatory step in inhibiting A␤ production, we investigated the effects of acute hypoxia (up to 8 h) and HIF-1 on BACE1 expression.
A␤ ELISA Assay-Conditioned media from hypoxia-treated and untreated cells were collected. The concentrations of A␤40 and A␤42 were quantified using commercial ELISA kits FIGURE 1. Acute hypoxia increases protein levels and activity of BACE1 and secretion of A␤. N2a-APP cells were treated with 1% O 2 for 2, 4, or 8 h. A, equal amounts of protein from lysates were analyzed and immunoblotted with antibodies against HIF-1␣, BACE1, PS1-NTF, TACE, APP, and ␣-tubulin (as a loading control). B, lysates of hypoxia-treated and untreated N2a-APP cells were assayed for ␤-secretase activity using BACE1 activity assay kit. C, lysates of N2a-APP cells were immunoblotted with the 6E10 antibody, which detects human APP ␤CTF (upper panel). Conditioned media were assayed for A␤ by immunoprecipitation with the 4G8 antibody, followed by Western blot analysis using 6E10 (lower panel). D, conditioned media were collected and secreted A␤40 and A␤42 were quantified using commercial ELISA kits. All data represent means Ϯ S.E. of levels relative to that of control from three independent experiments. *, p Ͻ 0.05, hypoxia-treated versus untreated samples at the indicated time. C, control; H, hypoxia.
In Vitro ␤-Secretase Activity Assay-Lysates of hypoxiatreated or untreated cells were assayed for ␤-secretase activity using a kit from Calbiochem (San Diego, CA), following the manufacturer's protocol.
Pulse-Chase Experiments-To assay the kinetics of BACE1 metabolism, 4-h hypoxia-treated and untreated N2a-APP cells were labeled with [ 35 S]methionine (500 Ci/ml) for 5 or 10 min at 37°C and collected for analysis. In some experiments, N2a-APP cells were pretreated for 4 h with hypoxia and then subjected to 15 min of pulse-labeling followed by chase for 15, 30, 60, and 120 min under hypoxic conditions. Cell lysates were immunoprecipitated with anti-BACE1 antibody B690, followed by SDS-PAGE analysis and autoradiography. Data were quantified using Scion Image (Scion Corp., Frederick, MD). The level of labeled BACE1 after 5 min of labeling of untreated cells was defined as one arbitrary unit.
Construction of Luciferase Reporter Plasmids, Site-directed Mutagenesis, and Electrophoretic Mobility Shift Assay (EMSA)-PCR was performed to amplify a 5Ј-flanking region of the BACE1 gene by using genomic DNA from mouse N2a cells as template. Primers used were BACE primer5, 5Ј-GGCTGGCA-TGCATGACAGGGTGCGCACGGGGGTGTG-3Ј and BACE primer3, 5Ј-CAGCACCTAGGCAGGCTGGGGAGGCGGA-AAGGCTTG-3Ј. The mutBACE primer5, 5Ј-GGCTGGCAT-GCATGACAGGGTGCGTCACGGGGTGTG-3Ј, was paired with BACE primer3 for PCR amplification to introduce mutations (boldface and underlined) into the potential HIF-1 binding site. After amplification, PCR products were inserted into the pCR2.1-TOPO vector (Invitrogen) for sequencing. After cleavage of the vectors with KpnI and XhoI, released fragments were introduced into the pGL3-enhancer vector containing the firefly luciferase gene (Promega, Madison, WI). Firefly luciferase vectors were co-transfected with phRL-SV40 containing the Renilla luciferase gene (Promega) into N2a cells, and the cells were treated with or without hypoxia for 4 h. Firefly luciferase activities were assayed and normalized to those of Renilla luciferase.
Tissue Isolation from HIF-1␣ Conditional Knock-out Mice-The hippocampus and cortex were dissected from HIF-1␣ conditional knock-out mice (in which Cre is driven by the conditions. Equal amounts of protein in cell lysates were immunoprecipitated with an anti-BACE1 antibody, followed by SDS-PAGE analysis and autoradiography. BACE1 levels were quantified (by densitometry of autoradiographic bands), and the level of labeled BACE1 after 5 min labeling of untreated cells was defined as one arbitrary unit. B, total RNA was extracted from hypoxiatreated or untreated cells and reverse-transcribed for quantitative real time-PCR. BACE1 mRNA levels were normalized to ␤-actin and compared with untreated controls (defined as one arbitrary unit). Data represent means Ϯ S.E. from three independent experiments. *, p Ͻ 0.05, hypoxia-treated versus untreated samples at indicated time point.
calcium/calmodulin-dependent kinase CaMKII␣ promoter) and littermates at 2 months of age (39). Samples were either lysed in RIPA for immunoblotting or used for RNA extraction and quantitative RT-PCR.

RESULTS
Acute Hypoxia Increases the Protein Level and Activity of BACE1 by Up-regulating Its Transcription-To analyze effects of acute hypoxia on key proteins involved in A␤ generation, we treated mouse neuroblastoma N2a cells stably expressing human APP695 with hypoxic conditions (1% O 2 ) for 2, 4, and 8 h. As shown in Fig. 1A, hypoxia significantly increased the protein level of BACE1, accompanied by a concomitant increase in HIF-1␣ protein levels, as expected. The levels of TACE, the N terminus of PS1 (PS1-NTF) and full-length APP were not affected by hypoxic treatment. A modest but significant increase in total BACE1 activity was observed, reflecting increased cellular protein levels (Fig. 1B). Consistent with our previous observation (30), hypoxic treatment led to a marked increase in overall levels of secreted A␤ (both A␤40 and A␤42) (Fig. 1, C and D). While we previously observed a modest decrease in the total APP CTFs in HeLa cells by chemical hypoxia, the current study showed a slight increase in APP ␤CTF, an immediate cleavage product of BACE1, by low oxygen hypoxic treatments in N2a cells (Fig. 1C).
The increased steady state level of BACE1 may result from increased biogenesis, decreased degradation, or both. To determine at which step(s) hypoxia affects BACE1 expression, we undertook pulsechase experiments. Cells pretreated with hypoxia were pulse-labeled with [ 35 S]methionine for 5 or 10 min without chase, or 15 min followed by chase periods up to 120 min under hypoxia. Biosynthesis of nascent BACE1 was significantly higher in cells treated with hypoxia than in untreated cells, whereas turnover curves for both hypoxiatreated and untreated cells were similar, suggesting that hypoxia mainly affects biosynthesis rather than degradation of BACE1 protein ( Fig. 2A). Quantitative real-time PCR analysis showed 1.5-2-fold increases in BACE1 mRNA following hypoxic treatment, confirming the effect of hypoxia on BACE1 transcription (Fig. 2B). Identification of a Functional HIF-1␣ Binding Element in the BACE1 Promoter-Given that HIF-1 is a critical transcription factor activated in response to hypoxic stresses, we looked for a potential HIF-1 binding element in the BACE1 promoter region and asked whether binding by HIF-1 could be responsible for up-regulation of BACE1 transcription under hypoxic conditions. Indeed, sequence analysis of the 5Ј-flanking sequence of the mouse BACE1 gene revealed a potential HIF-1 binding site on the minus strand at Ϫ835 to Ϫ821 (acgcGTGCccccaca) upstream of the BACE1 start ATG (defined as ϩ1) (Fig. 3A). To determine whether this site was important for promoter activity, we constructed luciferase reporter vectors containing this region or a mutant form that lacks HIF-1 binding ability and transfected both constructs into N2a-APP cells. As expected, the promoter activity of this BACE1 promoter region was significantly higher in cells treated under hypoxia than in untreated cells (Fig. 3B). Moreover, mutation of the putative HIF-1 binding site resulted in a significant reduction in promoter activity in untreated cells and hypoxia did not significantly potentiate the promoter activity of this mutant (Fig.  3B). These results indicate the functionality of this HIF-1 site. To further investigate the interaction between this HIF-1 binding site and HIF-1 protein, we performed gel shift analysis. Such an analysis revealed a DNA-protein complex forming after incubation of this probe with nuclear extracts of N2a-APP cells treated under hypoxia for 4 h (Fig. 3C, lane  2). This complex migrated at the same position as a HIF-1 consensus probe-protein complex (positive control, Fig. 3C,  lane 1). Formation of the BACE1 HIF-1 probe-protein complex was abolished by competitive binding of unlabeled HIF-1 consensus probe and BACE1 HIF-1 probe (Fig. 3C,  lanes 3 and 4) but was not affected in the presence of unlabeled BACE1 HIF-1 mutant probe (Fig. 3C, lane 5) or SP1 probe (Fig. 3C, lane 6). These data support the idea that the identified BACE1 HIF-1 binding element can form a complex with nuclear HIF-1 protein.
HIF-1␣ Regulates the Level of BACE1-To determine whether HIF-1␣ overexpression mediates effects on BACE1 expression similar to those of hypoxia, we transfected N2a-APP cells with various doses (0, 0.5, 1, 2, 5, and 10 g; only 0, 2, and 5 g were shown here) of HIF-1␣ cDNA. Following HIF-1␣ overexpression we observed a maximum of an ϳ2.5-fold increase in BACE1 protein level at 2 g of HIF-1␣ DNA. We also observed a 2-fold increase in BACE1 mRNA level, along with minimal changes in levels of PS1 NTF or TACE (Fig. 4, A and B). The observation that a higher dose (5 g) of HIF-1␣ expression failed to further up-regulate BACE1 expression suggests that the regulatory effect of HIF-1 on BACE1 expression is saturable.
On the other hand, we tested whether down-regulating HIF-1␣ could reduce the level of BACE1. As shown in Fig. 4C, when the level of HIF-1␣ was down-regulated (by ϳ80%) by siRNA treatment, the level of BACE1 was also reduced in both hypoxia-treated and untreated cells. However, hypoxic treatment still stimulated, but to a much lesser extent, The Hypoxic Effect Is Mediated Mainly by HIF-1-To determine whether effects of HIF-1␣ overexpression and hypoxia on BACE1 expression are additive, we treated HIF-1␣ overexpressing N2a cells with hypoxic conditions for various time periods. As shown in Fig. 5, while cellular HIF-1␣ levels were further elevated by hypoxia, the levels of BACE1 mRNA did not increase. Similarly, combining hypoxic treatment with overexpression of HIF-1␣ did not further increase A␤ levels (Fig. 5A). Together with the siRNA knockdown data shown in Fig. 4C, these results strongly suggest that the hypoxic effect of BACE1 expression and A␤ generation are largely mediated by HIF-1.
BACE1 Protein Levels Are Reduced in Brains of HIF-1␣ Conditional Knock-out Mice-Because BACE1 activity is higher in neuronal than in non-neuronal tissues, and neuronal A␤ in particular is thought to be the major contributor to AD pathogenesis, we analyzed the patho-/physiological relevance of HIF-mediated BACE1 expression using a mouse model of neuron-specific HIF-1␣ deficiency (39). A total deficiency of HIF-1␣ mRNA in isolated tissues was confirmed by RT-PCR (data not shown). We then examined and compared BACE1 protein levels in total brain lysates of conditional knock-out mice with those from lysates from littermate controls and observed no significant differences in BACE1 total protein (data not shown). We then dissected the hippocampus and the cortex, two brain regions most affected in AD, from conditional knock-out and littermate control mice and found that HIF-1␣ deficiency led to a marked reduction in BACE1 protein levels in these two brain regions, suggesting an important role for HIF-1 in regulating BACE1 expression (Fig. 6).

DISCUSSION
BACE1 provides the predominant neuronal ␤-secretase activity and is the major protease catalyzing ␤-cleavage of APP (7)(8)(9)(10)(11)(12). Several studies indicate that protein levels and BACE1 activity are elevated in brain regions affected by AD, suggesting that abnormal BACE1 activity contributes significantly to AD pathogenesis (19,20). Ischemia or stroke gives rise to hypoxic conditions known to greatly increase the incidence of AD (24 -26). These observations indicate a potential link between hypoxia-activated signaling pathways triggering HIF-1 induction and activation of the A␤ generation machinery, which requires up-regulation of BACE1 level/activity.
Although several studies have addressed the effects of ischemia or hypoxia on enzymes involved in APP processing/A␤ production, the results have been contradictory. Wen et al. (40) found that both activity and expression of BACE1 were significantly increased in rats under transient cerebral ischemia. Another study, however, using AD transgenic mice overexpressing mutant APP showed that neither expression nor activity of BACE was significantly affected by ischemic insult; instead ADAM10 was markedly increased in the early stage of ischemic insult and then down-regulated at later stages (41). Other studies report a drastic decrease in ADAM10 and TACE protein levels with unchanged BACE1 levels in human neuroblastoma SH-SY5Y cells subjected to chronic hypoxic treatments (42)(43)(44). While it is hard to reconcile these results, differing experimental procedures, such as cells/tissues examined, culture conditions, the percentage of O 2 utilized, or duration of hypoxic treatment, could all contribute to discrepant results. In fact, we observed that effects of hypoxia and/or elevated levels of HIF-1␣ on BACE1 expression and A␤ generation were gradually reduced when cells were exposed to prolonged periods of hypoxic stress or when high levels of HIF-1␣ expression were sustained (data not shown); both outcomes could be due to the cellular ability to adapt to deleterious conditions. Cellular adaptation to a non-physiological condition such as HIF-1␣ deficiency is also manifested by the fact that non-neuronal cells and/or cells of brain regions other than the forebrain can compensate for decreases in BACE1 expression. 5 Our study provides the first evidence for a link between ischemia/hypoxia and APP processing/A␤ production at the molecular level, which is defined mechanistically by transcriptional regulation of BACE1 by HIF-1. In addition, our observation of reduced BACE1 expression in the hippocampus and cortex of HIF-1␣-deficient mice further emphasizes the patho-physiological relevance of FIGURE 5. Acute hypoxia does not potentiate BACE1 expression in HIF-1␣-overexpressing cells. N2a-APP cells were transfected with a HIF-1␣ expression vector for 24 h and then divided into 6 plates and treated with or without hypoxia for 2, 4, or 8 h. A, equal amounts of cell lysate proteins were immunoblotted with antibodies against HIF-1␣, BACE1, and ␣-tubulin as a control. A␤ in conditioned media was assayed by immunoprecipitation using the 4G8 antibody followed by Western blot analysis with the 6E10. B, total RNA was extracted and reverse-transcribed for RT-PCR. BACE1 mRNA levels were normalized to ␤-actin. Data represent means Ϯ S.E. from three independent experiments. C, control; H, hypoxia. FIGURE 6. Decreased BACE1 expression in the hippocampus and cortex of HIF-1␣ conditional knock-out mice. Hippocampus and cortex were dissected from 6 HIF-1␣ conditional knock-out mice (HIF-1 KO) and 5 littermates (control) at 2 months of age, and samples were lysed in RIPA buffer. Equal amounts of lysate proteins were analyzed and immunoblotted with antibodies against BACE1, PS1-NTF, TACE, APP/APP CTFs, and ␣-tubulin. Data represent means Ϯ S.E. p Ͻ 0.05 compared with littermate controls. regulation of BACE1 expression and activity by hypoxia/HIF-1 pathways.
Although a recent study showed that when BACE1 was overexpressed at very high levels in transgenic mice, the generation of A␤ was reduced because of an altered BACE1 subcellular localization and hence a shift on APP cleavage, modest increase of BACE1 level indeed promoted A␤ deposition (45). In addition, more and more reports have shown that increased endogenous BACE1 levels in cell and animal models and more significantly in AD brains result in an increased amyloidogenic cleavage of APP and may contribute directly to (sporadic) AD pathogenesis (19,46,47). Consistent with this, our study also showed that A␤ generation (both A␤40 and A␤42) was increased accompanied by an increased endogenous BACE1 level upon hypoxia treatment (Fig. 1). The greater increase of A␤ detected by ELISA (2-fold) than that by the IP-Western method (1.5-fold) reflects the more sensitive and quantitative feature of the ELISA technique which detects more A␤ variants than the IP-Western method. These results suggest that the generation of A␤ is positively correlated with an increase of BACE1 within the patho/physiological range.
Inhibiting BACE1 activity or reducing levels of BACE1 in vivo has been shown to decrease production of A␤ without severe detrimental phenotypes (13)(14)(15), suggesting that BACE1 is a valuable candidate for therapeutic targeting. However, given the difficulty of developing specific small-molecule inhibitors, it is critical to explore alternative approaches such as down-regulating BACE1 transcription, particularly in cases of ischemia and stroke in which elevated BACE1 expression/activity may contribute to a higher incidence of AD.