Amyloid Precursor Proteins Inhibit Heme Oxygenase Activity and Augment Neurotoxicity in Alzheimer's Disease

Neuron, Vol. 28, 461–473, November, 2000, Copyright 2000 by Cell Press

Amyloid Precursor Proteins Inhibit Heme Oxygenase Activity and Augment Neurotoxicity in Alzheimer’s Disease Masaaki Takahashi,1,10 Sylvain Dore´,1,2,10 Christopher D. Ferris,1,3,11 Taisuke Tomita,7 Akira Sawa,1,6 Herman Wolosker,1 David R. Borchelt,1,4 Takeshi Iwatsubo,7 Seong-Hun Kim,8 Gopal Thinakaran,8 Sangram S. Sisodia,8 and Solomon H. Snyder1,5,6,9 1 Department of Neuroscience 2 Department of Anesthesiology/Critical Care Medicine 3 Department of Medicine, Division of Gastroenterology 4 Department of Pathology 5 Department of Pharmacology and Molecular Sciences 6 Department of Psychiatry The Johns Hopkins University School of Medicine 725 North Wolfe Street Baltimore, Maryland 21205 7 Department of Neuropathology and Neuroscience Faculty of Pharmaceutical Science University of Tokyo Tokyo 113 Japan 8 Department of Neurobiology Pharmacology and Physiology The University of Chicago Chicago, Illinois 60637

Summary Amyloid precursor protein (APP) generates the ␤-amyloid peptide, postulated to participate in the neurotoxicity of Alzheimer’s disease. We report that APP and APLP bind to heme oxygenase (HO), an enzyme whose product, bilirubin, is antioxidant and neuroprotective. The binding of APP inhibits HO activity, and APP with mutations linked to the familial Alzheimer’s disease (FAD) provides substantially greater inhibition of HO activity than wild-type APP. Cortical cultures from transgenic mice expressing Swedish mutant APP have greatly reduced bilirubin levels, establishing that mutant APP inhibits HO activity in vivo. Oxidative neurotoxicity is markedly greater in cerebral cortical cultures from APP Swedish mutant transgenic mice than wild-type cultures. These findings indicate that augmented neurotoxicity caused by APP–HO interactions may contribute to neuronal cell death in Alzheimer’s disease.

9

To whom correspondence should be addressed (e-mail: ssnyder@ jhmi.edu). 10 These authors contributed equally to this work. 11 Present address: Department of Medicine, Division of Gastroenterology, C-2104 Medical Center North, Vanderbilt University Medical Center, Nashville, Tennessee 37232.

Introduction Alzheimer’s disease (AD) is characterized by the presence of amyloid plaques and tangles in the brains of affected individuals (Price and Sisodia, 1998). The principal components of plaques are derived by the proteolytic processing of amyloid precursor proteins (APP) by ␤ and ␥ secretases, leading to the formation of amyloid-␤ (A␤) peptides (Selkoe, 1996). Of the two major forms, the 42 amino acid peptide (A␤42) is more closely linked to AD pathophysiology than the 40 amino acid form A␤40 (Price and Sisodia, 1998). Familial Alzheimer’s disease (FAD), reflecting mutations in APP and presenilins, is associated with augmented the formation of the neurotoxic A␤42 (Jarrett and Lansbury, 1993; Yankner, 1996), which can directly elicit brain damage (Hardy, 1997a, 1997b; Younkin, 1995). Physiological functions of APP have not been fully defined. Though some groups have postulated that APP may have a neuroprotective action (Mucke et al., 1996), this hypothesis is still controversial (White et al., 1998). APP is a member of a multigene family that contains at least two other members known as amyloid precursorlike protein (APLP)-1 and -2 (Wasco et al., 1992, 1993; Sprecher et al., 1993; Slunt et al., 1994). APLP isoforms contain most of the domains and motifs of APP but lack the A␤ region. Structural similarities between APP and APLP and the viability of mice with ablated APP (Zheng et al., 1995) and APLP2 (von Koch et al., 1997) have suggested that APLP functionally compensate for some physiologic functions of APP (Crain et al., 1996; McNamara et al., 1998). Heme oxygenase (HO) was first characterized as an enzyme that degrades heme from aging red blood cells (Maines, 1997). This isoform of HO, HO1, is induced by a large number of stimuli and is also designated heat shock protein-32 (HSP-32) (Shibahara et al., 1985). A second isoform, HO2, is not inducible and displays highest concentrations in the brain (Rotenberg and Maines, 1990). In catabolizing heme, HO gives rise to three products, carbon monoxide (CO), iron, and biliverdin. Biliverdin is immediately reduced by biliverdin reductase to bilirubin, while CO functions as a neurotransmitter in the autonomic nervous system (Zakhary et al., 1997) and possibly in the brain (Verma et al., 1993; Maines, 1997). HO activity protects against oxidative stress (Abraham et al., 1995; Lee et al., 1996; Dennery et al., 1997; Poss and Tonegawa, 1997b). Recent studies indicate that HO1 is cytoprotective by increasing iron transport out of cells (Ferris et al., 1999), accounting for iron accumulation in tissues of HO1 and HO2-deficient mice (Poss and Tonegawa, 1997a; Dennery et al., 1998). Bilirubin has antioxidant activity (Stocker et al., 1987), and recently we have found that bilirubin appears to be a physiologic neuroprotectant (Dore´ et al., 1999a, 1999b). In spite of the recognized importance of HO in oxidative stress, mechanisms regulating its activity have not been well characterized. To find molecules that specifically interact with HOs, we conducted a yeast twohybrid analysis. We identified HO interactions with APP

Neuron 462

Figure 1. Interaction of HO and APP/APLP (A) Determination of the binding domain of APLP1 to HO2 by yeast two-hybrid analysis. The expression vector encoding rat HO2 C-terminal domain (169–296) fused to the Gal 4 DNA binding domain was cotransformed into yeast Y190 cells with the indicated deletion mutants of APLP1 fused to the Gal 4 activation domain expression vector. Plus signs represent activation of HIS3 and ␤-gal reporters (⬍1 hr). Plus-minus signs represent slow growth on a plate lacking histidine and weak ␤-gal activities (⬍8 hr). (B) In vivo interaction of APP with HO1 and HO2. HEK293 cells were transiently transfected with human HO constructs and (His)6tagged APLP1 (APLP1 His) or full-length APP/ APLP2 constructs (APP695, APP770, and APLP2). After 40 hr, extracts were prepared and immunoprecipitated with anti HO antibody. Coprecipitated APP was detected by immunoblot analysis using an anti-(His)6 or anti-APP/APLP2 monoclonal antibody (22C11). The arrow indicates APLP1-His. (C) Specificity of the interaction of APP with HO1. HEK293 cells were transfected with HO1 expression vector and various APP constructs. APP695 (1–477) and APP770 (1–552) are soluble deletion mutants that have the signal peptide on the N terminus and are localized into the lumenal side of the ER. APP695 (46–477) and APP770 (46–552) lack the signal peptide on the N terminus and are localized in the cytosol. After 40 hr, extracts were prepared and immunoprecipitated with anti-HO1 antibody. Coprecipitated APP was detected by immunoblot analysis using the anti-APP monoclonal antibody. (D) Subcellular fractionation of mouse neuroblastoma (N2A cells) was accomplished using 0%–26% Iodixanol-based linear density gradient as indicated in Experimental Procedures. Eighteen fractions, from top to bottom, were collected. The density of each fraction was estimated from refractory index and was linear throughout the gradient. Aliquots of each fraction were run on 8%–15% gradient SDS–PAGE and analyzed by immunoblotting with anti-HO2, anti-APP (CT15), and antiKDEL antibodies. Anti-KDEL recognizes the ER luminal markers, the Grp94, and BiP.

family members leading to inhibition of HO activity. Mutant APP associated with FAD inhibit HO activity significantly more than wild-type APP. In cortical cultures from APP Swedish mutant mice, bilirubin levels are substantially diminished, while oxidative stress-induced neurotoxicity is markedly increased. These findings suggest that reduced neuroprotective activity of HO, resulting from APP-HO interaction, may participate in the neurotoxicity of AD. Results HO Interacts with and Is Colocalized with APP/APLP We conducted a yeast two-hybrid analysis employing a portion of HO2 (amino acids 169–296) as “bait.” After screening five million colonies from a cDNA library de-

rived from rat cerebral cortex, we identified five different positive clones and subsequently focused on APLP-1 (Figure 1). To eliminate the possibility that only HO or APLP was necessary to activate transcription of GAL4 in yeast, we monitored growth on histidine-minus plates and ␤-galactosidase (␤-gal) activity following transformation of yeast with HO or APLP alone or in combination with another irrelevant construct (see Experimental Procedures). In these experiments, we were able to detect interaction in yeast only following transformation of both HO and APLP (data not shown). To examine the precise portions of HO2 and APLP-1 required for the interaction, we conducted yeast two-hybrid studies with constructs of different lengths. Truncation of APLP-1 reveals that interaction with HO2 is restricted to an acidic domain, encoding amino acids 237–271, near the zinc binding site. APP and APLP-1 share 36% amino acid identity

Amyloid Precursor Proteins Inhibit Heme Oxygenase 463

and 41% amino acid similarity over this region. Amino acids 169–296 of HO2, which are closely associated with its catalytic domain, are required for interaction with APLP, while a construct containing the N-terminal portion (amino acids 2–143) is inactive (data not shown). Using immunoprecipitation, we compared interactions of HO1 and HO2 with two physiologic isoforms of APP, APP695, and APP770 (Figures 1B and 1C). In these experiments, HEK293 cells were transiently transfected with cDNA encoding the various forms of APP, and some of these were cotransfected with HO1 or HO2 cDNA. Immunoprecipitation was conducted with antisera to HO1 or HO2, and the presence of APP in the immunoprecipitates was evaluated by immunoblotting. In all cases, APP are precipitated only in cells transfected with HO1 or HO2. These immunoprecipitation findings confirm the protein–protein interactions first identified in yeast twohybrid analysis. In order for HO-APP interactions to be physiologically relevant, the two proteins should be in the same subcellular compartment. APP is cotranslationally translocated into the endoplasmic reticulum via its signal peptide and then undergoes maturation during passage through the Golgi. A minor percentage of mature APP is transported to the plasma membrane via secretory vesicles. Thus, the majority of APP remains in the Golgi (Selkoe, 1996) and endoplasmic reticulum (ER), where the acidic domain of APP is within the lumen (Selkoe, 1996). HOs are ER proteins with an N-terminal domain anchored to the ER by a single transmembrane domain at the extreme C terminus (Maines, 1997). However, the transmembrane topology of HO has not been definitively established (Hino et al., 1979; Yoshida and Sato, 1989). To determine if HO and APP are located in the same subcellular compartment, we conducted immunoprecipitation experiments with native or truncated forms of APP695 or APP770 (Figure 1C). C-terminal truncation transforms APP from a particulate ER protein to a soluble lumenal protein, while truncation of the signal peptide results in a cytoplasmic localization of APP. We find that coimmunoprecipitation of APP/HO persists following truncation of the transmembrane portion of APP, while truncation of the signal peptide abolishes interaction of APP with HO (Figure 1C). To further define the subcellular localization of APP and HO, we monitored the distribution of APP and HO following fractionation of membranes from mouse neuroblastoma (N2a) cells on Iodoxanol-based linear density gradients (Kim et al., 2000). HO2 and APP comigrate in fractions (10–13), coinciding with ER lumenal markers GRP94 and BiP (Figure 1D). In other experiments, we monitored HO immunofluorescence in 293/cytochrome P450 reductase (CPR)-HO2 cells following selective permeabilization of the plasma membrane with digitonin or complete membrane permeabilization with saponin (Otto and Smith, 1994) (Figures 2A–2F). For these experiments, we chose to monitor HO2 localization since under physiologic conditions HO2 is the only HO isoform expressed in the brain. To confirm permeabilization of the plasma membrane, we monitored immunostaining for the C terminus of calnexin, which is located on the cytosolic side of the ER (Ou et al., 1993). Calnexin staining appears similar following either digitonin or saponin permeabilization (Fig-

ures 2A and 2B). To confirm permeabilization of the ER membranes with saponin, we stained cells for Grp78, which is localized to the lumen of the ER (Doan et al., 1996). Grp78 staining is nearly absent in digitonintreated cells, while staining is evident in saponin-treated cells (Figures 2C and 2D). Immunofluorescence for HO2 is strikingly similar to that seen with Grp78 antibodies, with no staining evident in digitonin-treated preparations but clear staining in the saponin-treated cells (Figures 2E and 2F). To compare the intracellular localization of APP and HO2 in brain tissue without transfection, we conducted immunohistochemical staining with primary cultures of rat cortical neurons. HO2 and APP staining in neurons reveals overlapping subcellular distributions consistent with colocalization of the two proteins (Figures 2G–2I). To ascertain whether HO2 and APP exist as a complex in brain under native conditions, we immunoprecipitated HO2 from rat brain extracts and monitored coprecipitation of APP by Western blot. As shown, APP coprecipitates upon immunoprecipitation of HO2 but is not detected in immunoprecipitates generated using an anti-calnexin antibody (Figure 2J). We also detect coprecipitation of HO2 in immunoprecipitates of APP (data not shown). In addition, we immunoprecipitated HO2 from mouse brain extracts prepared from wild-type and HO2⫺/⫺ mice. In these experiments, we detect coprecipitation of APP in extracts derived from wild-type but not HO2⫺/⫺ brains (data not shown). Thus, coprecipitation studies, subcellular fractionation, and immunofluorescence studies all indicate that the N-terminal region of HO2 is located within the ER lumen, where it interacts with the acidic portion of APP. HO1 and HO2 Expression Do Not Affect A␤ Peptide Production To ascertain whether HO1 or HO2 influence the processing of APP into the A␤ peptides, we employed HEK293 cell lines stably transfected with CPR (293/ CPR), which provides electron donation for HO activity (Maines, 1997), and cells stably transfected with HO1 (293/CPR-HO1) or HO2 (293/CPR-HO2). These cell lines manifest 6- to 10-fold augmentation of HO1 or HO2 by both Western blot and enzyme activity. These levels are similar to endogenous levels in tissues in which HO1 or HO2 are enriched (Maines, 1997). Each line was transiently transfected with cDNA encoding APP695 or APP770, and levels of A␤ in the conditioned medium were quantified using a two-site ELISA (Tomita et al., 1997) (Table 1). Expression of HO1 or HO2 does not significantly alter levels of secreted A␤40 or A␤42. Moreover, the ratio of A␤42 to total A␤ is essentially the same in 293/CPR-HO1 or 293/CPR-HO2 cells as it is in 293/ CPR cells. Thus, there is no significant effect of HO1 or HO2 on the conversion of APP to A␤ peptides in these cells. APP/APLP Diminish HO Activity and Accentuate Hemin Toxicity To ascertain the influence of APP interaction upon HO activity, we utilized 293/CPR-HO1 and 293/CPR-HO2 cells transiently transfected with cDNA encoding APP or green fluorescent protein (GFP) as a control. The data

Neuron 464

Figure 2. Immunocytochemical Analysis of HO and APP (A–F) Immunocytofluorescent staining of selectively permeabilized 293/CPR-HO2 cells. Cells were fixed and permeabilized with either digitonin or saponin. Cells were then stained with either an anti-calnexin C terminus polyclonal antibody (anti-calnexin CT) or costained with an anti-KDEL monoclonal antibody and an anti-HO2 polyclonal antibody (SPA897). The bar represents 10 ␮m. (G–I) Primary cultured rat cortical neurons were stained by an anti-HO2 polyclonal antibody or an anti-APP/APLP2 monoclonal antibody (22C11). Colocalization was conducted by confocal microscopy. The bars represent 50 ␮m. (J) APP/APLP2 and HO2 complex are detectable in rat brain extracts. An anti-HO2 antibody (OSA200) specifically coprecipitates APP/APLP2, while comparable amounts of an anti-calnexin CT antibody fail to coprecipitate APP.

shown are derived from measurements of HO activity following a 5 min incubation because, in other time course experiments, we determined that the 5 min time point was reliably in the linear range of enzyme activity under our experimental conditions. Mock transfection and transfections with GFP cDNA have no influence on HO1 or HO2 activity in cell lysates, while APP695 and APP770 cDNA transfections reduce HO1 and HO2 activity by 25%–35% (Figure 3A). Expression of HO1 and HO2 is not changed by the transfections (Figure 3A, inset). Since the efficiency of transfection, as estimated by cell counting using GFP, is about 60%, the extent of enzyme inhibition in the cells that are successfully transfected is likely to be substantially greater than observed values. In these experiments, cells are lysed in glycerol containing buffer to preserve protein–protein interaction. In other experiments, where cells are lysed in buffer without glycerol, we often fail to detect significant inhibition of HO activity, suggesting that the HO–APP interaction may be labile. Because the lysis of cells, required to conduct HO enzymatic assays, may disrupt endogenous protein–protein interactions, we sought to assess HO activity in intact cells using hemin toxicity.

In an effort to assess in vitro the influences of purified APP on HO activity, we incubated purified His-tagged APP695 or APLP with Triton X-100 solubilized extracts of cells stably transfected with HO1 or HO2 and cytochrome P450 reductase. Concentration of APP, APLP, HO1, HO2 were approximately stoichiomatric. We do not detect inhibition of HO1 or HO2 catalytic activity under these conditions (data not shown). There are several possible reasons for failure to demonstrate inhibition. Preparation of APP and APLP utilized denaturing conditions followed by refolding, which may not have reconstituted two optimal conformations for HO inhibition. Interactions of APP/APLP and HO may require physiologic juxtaposition of two proteins on intracellular membranes. Such interaction may not be readily reconstituted in vitro. Cells expressing HO are highly resistant to hemin toxicity (Abraham et al., 1995), while fibroblasts from HO1deficient mice are sensitive to hemin (Poss and Tonegawa, 1997b). Since hemin can induce HO1 mRNA and protein synthesis in a variety of cell types (Maines, 1997), we examined HO1 and HO2 protein content by Western blot in hemin-treated cells. No increase of HO1 protein

Amyloid Precursor Proteins Inhibit Heme Oxygenase 465

Table 1. Amyloid-␤ Production in the 293/CPR, 293/CPR-HO1, and 293/CPR-HO2 Cells

APP695

APP770

A␤ Production (pmol/mg cell protein)

Ratio

cDNA Cell Line

A␤40

(A␤42/A␤40 ⫹ A␤42)

293/CPR 293/CPR-HO1 293/CPR-HO2 293/CPR 293/CPR-HO1 293/CPR-HO2

15.2 14.9 15.7 12.6 13.7 12.8

⫾ ⫾ ⫾ ⫾ ⫾ ⫾

A␤42 0.8 1.2 0.4 0.8 0.5 0.1

expression is observed with hemin administration in any of the three cell lines (Figure 3D). To determine whether APP/APLP mediated inhibition of HO activity affects toxicity in intact cells, we monitored cell death following treatment of cells with hemin by Trypan blue exclusion. In 293/CPR cells, hemin is highly toxic with 40% killing of cells evident at 50 ␮M (Figure 3C). In contrast, 293/ CPR-HO1 and 293/CPR-HO2 cells are markedly protected with only 10%–20% cell death elicited at 200 ␮M hemin, a concentration that kills over 80% of 293/CPR

1.55 2.03 1.68 1.40 1.70 1.38

⫾ ⫾ ⫾ ⫾ ⫾ ⫾

0.09 0.04 0.02 0.02 0.14 0.11

9.7% 12.0% 9.8% 10.0% 11.4% 9.8%

cells (Figure 3C). These findings establish that hemin toxicity reflects HO activity. Next, we evaluated the influence of transient transfection of APP695, APP770, APLP1, and APLP2 in 293/CPR, 293/CPR-HO1, or 293/CPR-HO2 cells. Transfection with GFP has no influence on hemin toxicity in 293/CPR-HO1 and 293/CPR-HO2 cells (Figures 3E and 3F). By contrast, APP695, APP770, APLP1, or APLP2 transfections all markedly increases hemin toxicity (Figures 3E and 3F). In wild-type cells or CPR cells without HO overexpresFigure 3. APP and APLP Inhibit HO Catalytic Activity (A) HEK293/CPR-HO1 or (B) 293/CPR-HO2 cells were transfected with APP expression vectors by electroporation. After 20 hr, cells were harvested and disrupted by freezethaw. Then, HO activities in 12 ␮g (293/CPRHO1) or 20 ␮g (293/CPR-HO2) of cell extracts were determined using [55Fe]hemin as a substrate. Insets show the expression of HO1 or HO2 in the cells transfected with GFP (lane 1), APP695 (lane 2), APP770 (lane 3), APLP1 (lane 4), or APLP2 (lane 5) cDNA. (C) Survival of cells overexpressing HO1 or HO2 exposed to hemin. HEK293/CPR, 293/ CPR-HO1, or 293/CPR-HO2 cells were exposed for 20 hr to several concentrations of hemin. Cell survival was determined by Trypan blue exclusion. (D) HO expression after 20 hr exposure of the cells to hemin. HO1 and HO2 protein levels in 10 ␮g of cell extracts were determined by immunoblot analysis. The numbers indicate concentrations of hemin (0, 50, or 100 ␮M). Induction of endogenous HO1 was not detected in these cell lines. (E and F) Survival of 293/CPR-HO1 or 293/ CPR-HO2 cells transfected with APP695, APP770, APLP1, or APLP2 expression vectors. Cells were cultured for 5 hr after transfection and then exposed for 20 hr to hemin (0, 50, 100, 200 ␮M). Results are mean of duplicate determinations and the experiments were repeated three times with similar results.

Neuron 466

Figure 4. APP and APLP Inhibit HO in Intact Cells To monitor HO activity in intact cells, cell viability of 293/CPR-HO2 cells transfected with various APP/APLP-1 deletion constructs was monitored in response to hemin. Schematic representation of APP/APLP-1 deletion mutants shows signal peptide (Signal), the zinc binding site (Zn), the acidic domain (Acidic), the Kunitz protease inhibitor domain (KPI), the transmembrane domain (TM), and the A␤ fragment (A␤). The ␣ and ␤ cleavage sites are indicated by arrowheads. The cells were cotransfected with a LacZ expression vector (1.5 ␮g) and membranebound APP constructs (3 ␮g), APP (3 ␮g), or the other soluble APP deletion constructs (20 ␮g) by electroporation. After transfection, cells were cultured for 5 hr, and hemin was added into the culture medium at a final concentration of 100 ␮M. After 12 hr of incubation with hemin, cell viability was determined by a ␤-gal assay. The values represent ␤-gal activity relative to the same cells without hemin treatment (percent of control). Results are mean of triplicate determinations and the experiments were repeated three times with similar results.

sion, we do not detect in increased hemin toxicity following APP transfection establishing that APP transfection mediates the increased toxicity through inhibition of HO. Inhibition of HO Activity in Intact Cells by APP/APLP Requires the Acidic Domain but Not the Transmembrane Region We employed several truncations of APP/APLP to ascertain the sequence required for regulation of HO activity in intact cells (Figure 4). In these experiments, 293/CPRHO2 cells were cotransfected with APP and LacZ constructs. After exposure to hemin, cell viability was monitored by ␤-gal activity in cell lysates. In this way, we can selectively monitor transfected cells and easily quantify cell death. Since APP695 and APP770 show similar activities, the two alternatively spliced exons in APP770 are not required for activity. Deletion of the C-terminal portion of APP (constructs ␣APP695s and ␤APP695s) does not significantly affect the increase in hemin toxicity. Thus, the A␤ peptide is not required for augmented hemin toxicity (Figure 4). While APP695⌬305–590, APP695(1–477), and APP695(1–305) retain significant activity, APP695⌬198–590 and APP695(1–197) lack the acidic domain and have markedly reduced activity. Of note, APP695⌬305–590 does contain the transmembrane domain, indicating that the transmembrane domain is not sufficient to interact with HO. Deletion of the

N-terminal signal peptide in MycAPP(46–477) disrupts the subcellular localization of APP and markedly attenuates the increased hemin toxicity. While three different APLP1 constructs provide substantial increases in hemin toxicity, APLP1(1–223) lacks the acidic domain and again has substantially attenuated activity (Figure 4). Familial Alzheimer’s Disease Mutants Exhibit Increased HO Inhibition FAD has been associated with distinct mutations of APP (Hardy, 1997b). We wondered whether mutant forms of APP influence HO activity differently than the wild type. We examined the Swedish mutation (K670N/M671L), the Dutch mutation (E692Q), and the three London mutations (V717G, V717F, V717I). We transiently transfected the mutants or wild-type APP695 into 293/CPR-HO1 or 293/CPR-HO2 cells. Expression of each of the five FAD mutant APP provides 45%–50% inhibition of HO1 and HO2 activity, about twice that elicited by wild-type APP (Figures 5A and 5B). To ensure that the total expression of APP/APLP was similar following all transfections, we monitor APP/APLP expression by Western blot. Total APP/APLP expression is similar following transfection of the five FAD mutant forms and wild-type APP695 (Figure 5C). To assess the effect of the FAD mutations on hemin toxicity in intact cells, we cotransfected a LacZ con-

Amyloid Precursor Proteins Inhibit Heme Oxygenase 467

Figure 5. APP Mutations Associated with FAD Augment HO Inhibition (A and B) Inhibition of HO activity by APP. 293/CPR-HO1 or 293/CPR-HO2 cells were transfected with the wild-type APP695 or FAD mutant expression vectors by electroporation. After 20 hr, HO activities in 12 ␮g (293/ CPR-HO1) or 20 ␮g (293/CPR-HO2) of cell extracts were determined using [55Fe]hemin as a substrate. A single asterisk indicates p ⬍ 0.05, while a double asterisk indicates p ⬍ 0.01. (C) Expression of APP695 in the transfected 293/CPR-HO2 cells were determined in 10 ␮g of cell extracts by immunoblot analysis using an anti-APP monoclonal antibody (22C11). Of note, mock-transfected cells do not display any immunoreactivity under these conditions, though longer exposures of the blots do reveal modest baseline APP/APLP immunoreactivity. (D and E) Survival of 293/CPR-HO1 and CPRHO2 cells transfected with wild-type or FAD mutant APP. HEK293/CPR-HO1 cells were cotransfected with 1.5 ␮g of a LacZ expression vector and 3 ␮g of APP695 constructs. 293/CPR-HO2 cells were transfected 0.5 ␮g of the LacZ construct and 1 ␮g of APP constructs. After treatment with hemin, cell viability was determined by a ␤-gal assay. Results are mean of duplicate determinations and the experiments were repeated three times with similar results.

struct with the various APP constructs and monitored cell death. The Swedish, Dutch, and the London (V717G) mutations all are associated with greater hemin-elicited cell death than the wild-type APP695 or APP770 (Figures 5D and 5E). These data indicate that FAD-associated mutations result in greater inhibition of HO activity and augmented hemin toxicity. HO Inhibition and Oxidative Neurotoxicity Are Increased in Cortical Cultures from APP (Swedish Mutant) Transgenic Mice We wondered whether the increased HO inhibitory activity elicited by FAD mutant APP would be evident in intact organisms with such mutations. Recently transgenic mice have been developed expressing APP with the Swedish FAD mutation (Mo/Hu APP695swe; Borchelt et al., 1996). We reasoned that if the mutant APP diminishes HO2 activity in the intact transgenic organism, this should be evident as decreased bilirubin formation. Using antibodies to detect bilirubin formation in mouse neuronal cultures, we previously found that stimulation of protein kinase C activity by phorbol ester markedly augments HO2 activity and results in increased bilirubin staining (Dore´ et al., 1999a). We utilized cortical cultures made from E19 embryos of APP Swedish mutant or nontransgenic mice and monitored bilirubin accumulation in response to treatment with 0.1 ␮M phorbol myris-

tyl acetate (PMA), a concentration that reproducibly stimulates bilirubin accumulation in such cultures (Dore´ et al., 1999a). PMA treatment markedly augments bilirubin staining in the nontransgenic littermate cultures but has no effect in cultures from transgenic littermates (Figure 6). Since PMA may have effects other than activation of HO, we induced neurotoxicity with H2O2 in wild-type and HO2⫺/⫺ cortical cultures. Pretreatment of the cultures with PMA (0.1 ␮M) could offer complete neuroprotection for wild-type neurons, whereas no significant protection occurred in the HO2⫺/⫺ cells (data not shown) similar to our previous observations (Dore´ et al., 1999a). No protection occurs with the inactive isomer of PMA (4␣-PMA, 0.1 ␮M). Moreover, the protective effect of PMA in wild-type cultures is blocked by the HO inhibitor (Tin protoporphyrin-IX [SnPPIX], 5 ␮M). The combination of diminished bilirubin levels and decreased HO activity in mutant-derived cultures implies a causal link. However, altered expression of biliverdin reductase might influence bilirubin levels. Western blot analysis indicates no difference between biliverdin reductase expression between mutant and wild-type brains (data not shown). We also examined hemin toxicity in primary cortical cultures derived from nontransgenic and Mo/Hu APP695swe transgenic mice. At all concentrations of hemin examined, cultures prepared from Mo/Hu APP695swe are more susceptible to toxicity than wild-type cultures (Fig-

Neuron 468

Figure 6. Phorbol Myristyl Acetate (PMA) Augmentation of Bilirubin (BR) Is Abolished in Familial Alzheimer’s Disease Swedish Mutation (Mo/Hu APP695swe) Mouse Cortical Cultures Cultured cortical neurons from nontransgenic (non-TG) and Mo/Hu APP695swe transgenic mice (TG) were incubated with the antibody against BR to assess BR accumulation with and without PMA treatment. (A) Photomicrograph of staining performed without (Ctl) or with 0.1 ␮M PMA (⫹PMA). (B) Graph representing the relative intensity of bilirubin accumulation in non-TG and TG mice without or with PMA (plus sign) treatment. Results are mean of triplicate determinations and the experiments were repeated three times with similar results.

ures 7A and 7B). Conceivably, the greater toxicity in Mo/Hu APP695swe cultures might arise from some effect unrelated to inhibition of HO. Accordingly, we treated Mo/Hu APP695swe and nontransgenic cultures with 5 ␮M SnPPIX, a selective HO inhibitor when employed in mammalian tissues at 1–10 ␮M (Drummond and Kappas, 1981; Yoshinaga et al., 1982; Zakhary et al., 1996). SnPPIX treatment increases the hemin-mediated toxicity in nontransgenic neuronal cultures but has no effect in Mo/Hu APP695swe cultures. In the presence of SnPPIX, the concentration-response relationship for hemin is similar in Swedish mutant and wild-type cultures (Figure 7B). These data strongly suggest that the greater susceptibility of the Mo/Hu APP695swe cultures to hemin toxicity reflects decreased HO activity. In other experiments, we examined the effect of Mo/Hu APP695swe mutation on toxicity in response to H2O2. Cultures from Mo/Hu APP695swe transgenic mice are more susceptible to H2O2-mediated toxicity than nontransgenic cultures (Figure 7C). Previously, we demonstrated that H2O2 toxicity in hippocampal and cortical cultures is reduced

by increased HO activity following treatment with PMA (Dore´ et al., 1999a). While PMA treatment decreases H2O2 toxicity in the nontransgenic cultures, PMA has no effect on toxicity in the Mo/Hu APP695swe cultures (Figure 7D). The HO inhibitor SnPPIX blocks the effect of PMA in the nontransgenic cultures (Figure 7D). We monitored expression of HO2, HO1 and APP by Western blot (Figure 7E). Expression of APP is about 3-fold higher in the cultures from transgenic mice. HO2 and APP levels are not influenced by hemin or H2O2 treatment, while these treatments do augment HO1 levels. Discussion In the present study, we have identified an interaction between APP/APLP and HO. The protein–protein interactions have been established both by yeast two-hybrid analysis and coimmunoprecipitation. The site of interaction between these proteins appears to be in the lumen of the endoplasmic reticulum (ER), as mutations in APP that disrupt the lumenal ER localization prevent interac-

Amyloid Precursor Proteins Inhibit Heme Oxygenase 469

Figure 7. Primary Cortical Neurons from Mo/ Hu APP695swe Transgenic Mice Display HO Inhibition and Manifest Increased Cell Death in Response to Stress (A and B) After 6 days in culture, cortical neuronal cultures from nontransgenic and Mo/ Hu APP695swe transgenic mice were preincubated with or without SnPPIX (5 ␮M) for 2 hr and then exposed for 24 hr to different concentrations of hemin. (C and D) After 5 days culture, the cortical cultures were pretreated with or without 0.1 ␮M of PMA for 24 hr and then exposed to H2O2 for 24 hr. Cell survival was assayed by MTT. Asterisk, p ⬍ 0.05; double asterisk, p ⬍ 0.01 versus nontransgenic mice. (E) Expression of HO1, HO2, and APP were determined in 10 ␮g of cell extracts by immunoblot analysis using anti-HO1 (SPA895), HO2 (SPA897), and APP (CT15) antibodies after treatment with hemin and H2O2 for 6 hr. Results are mean of triplicate determinations, and the experiments were repeated three times with similar results.

tions with HO, and HO2 itself is localized to the lumen of the ER. HO has physiologic antioxidant properties (Poss and Tonegawa, 1997b; Dennery et al., 1998; Ferris et al., 1999), while generation by HO of intracellular free iron might increase oxidative stress (Van Lenten et al., 1995). However, since the catalytic site of HO is facing the lumenal side of the ER, free iron produced by HO should be rapidly released outside of cells. In cells cotransfected with APP and HO isoforms, we failed to find any effect of HO1 or HO2 on the processing of APP695 and APP770 into A␤ peptides. However, we have found a striking inhibitory effect of APP on HO activity in lysates and intact cells. FAD mutant forms of APP elicit substantially greater inhibition of HO and augmentation of hemin toxicity than wild-type APP. How might HO and APP be juxtaposed to facilitate HO inhibition? HO binds APP near the carboxyl terminus of HO and adjacent to the catalytic domain, allowing APP binding to inhibit enzyme activity. HO binds to the acidic domain of APP located about one-third of the distance from the N-terminal to the C-terminal portion. Since the C-terminal portion of the protein is attached to the membrane, this acidic domain would be distant from the membrane if it were fully extended. A folded

conformation of APP may bring the acidic domain close to the cell membrane facilitating interactions with HO. This hypothesis is supported by the finding that APP695⌬305–590 has about the same HO inhibitory activity as full-length APP695. Such a localization for the acidic domain would bring it into juxtaposition with the transmembrane, carboxyl portion of APP, which contains FAD mutations in the APP ectodomain. In this way, the Swedish and Dutch mutations might influence the ability of the acidic domain to bind HO. The Londontype mutation involves the transmembrane domain of APP. Since both HO and APP are membrane-bound proteins, transmembrane interactions between the two proteins might be stabilized due to the mutant APP sequences. Though aberrations in APP processing certainly are associated with Alzheimer’s disease, mechanisms whereby neurotoxic insults originate have not been definitively established. Because A␤ peptide is a component of the amyloid plaques that accumulate in the disease, it is reasonable to assume that A␤ is a neurotoxic culprit. Indeed, A␤42 peptide is neurotoxic to cultured neurons (Yankner, 1996). However, most lines of APP transgenic mice with high levels of amyloid plaques do not experi-

Neuron 470

ence major neural degeneration (Games et al., 1995; Hsiao et al., 1996), though recently some lines have displayed such alterations (Calhoun et al., 1998). To ascertain whether the inhibition of HO activity by mutant APP is physiologically relevant, we stained cortical neuronal cultures from APP695swe transgenic mice for bilirubin, which reflects HO activity. The phorbol ester stimulation of bilirubin accumulation in nontransgenic litter cultures is abolished in cultures from APP965swe transgenic littermates. These findings parallel the lack of neuroprotection by PMA in cultures from the mutant mice. Neuroprotection by PMA is due to activation of HO2, as the neuroprotection is lost in cultures from mice with deletion of HO2 as observed previously (Dore´ et al., 1999a) and replicated here. Moreover, nanomolar bilirubin concentrations applied to neuronal cultures produce neuroprotection similar to the influences of PMA. The concentrations of bilirubin that are neuroprotectant correspond to physiologic brain levels in rodents (Hansen et al., 1999). Together these findings indicate that the failure of PMA to provide neuroprotection to mutant cultures is due to the inhibition of the generation of bilirubin by HO2 in the mutant cultures. Basal bilirubin staining in transgenic mice is close to background and appears to be the same as in nontransgenic mice. This suggests that the very low basal HO2 activity is not involved in neuroprotection but presumably in the turnover of heme, with HO2 participating in neuroprotection only following its activation by protein kinase C or other stimuli. Our findings indicate that HO inhibition may be neurotoxic by diminishing the formation of bilirubin and/or leading to less iron efflux from cells. HO catalytic activity gives rise to bilirubin, a known antioxidant (Stocker et al., 1987). H2O2 and hemin toxicities are markedly augmented in cultures from HO1- and HO2-deficient mice and low nanomolar concentrations of bilirubin reverse this toxicity. Reversal of the toxicity by phorbol esters appears to involve stimulation of PKC to activate HO2 (Dore´ et al., 1999a). Recent evidence also indicates a role for HO in facilitating the transport of iron out of cells. Thus, several organs of HO-deficient mice are overloaded with iron, while serum iron levels in these mice are reduced (Poss and Tonegawa, 1997a; Dennery et al., 1998). Transfection of HO1 or HO2 into cells augments iron efflux, which is diminished in cells from HO1deficient animals (Ferris et al., 1999). Evidence in favor of the notion that HO inhibition is relevant to the neurotoxicity of Alzheimer’s disease includes the greater neurotoxicity elicited by the FAD mutant forms of APP. Also, in primary cultures from embryonic cerebral cortex of mice with transgenic expression of the Swedish FAD mutant APP, both hemin and H2O2 toxicity is greater than in nontransgenic cultures. APP transgenic mice display neural oxidative stress as well as a striking induction of HO1 activity in the brain (Pappolla et al., 1998; Smith et al., 1998). These animals evince higher susceptibility to ischemic brain damage (Zhang et al., 1997). Similarly, neural damage following middle cerebral artery occlusion is markedly accentuated in HO2-deficient mice (Dore´ et al., 1999b). Increased oxidative stress was also reported in cultured fetal Down’s syndrome cortical neurons (Busciglio and Yankner, 1995; Sawa et al., 1997).

Sporadic AD may reflect multiple etiologies and pathogenic mechanisms. Free radical induced oxidative stress has been implicated in the pathophysiology of sporadic AD (Markesbery, 1997). Increased oxidative damage (Smith et al., 1996; Gabbita et al., 1998), iron accumulation (Good et al., 1992; Loeffler et al., 1995; Lovell et al., 1998), and HO1 induction (Smith et al., 1994; Schipper et al., 1995) have been reported in sporadic AD brain. These phenotypes resemble those from APP transgenic mice and HO-deficient mice. Thus, HO may modulate oxidative damage in sporadic AD. Conceivably, decreased neuroprotection associated with inhibited HO activity facilitates neurotoxicity elicited by A␤ peptides. If HO plays a role in neurotoxicity associated with various APP forms, then drugs that block interactions of APP and HO might be therapeutic. Additionally, such agents would facilitate our understanding of physiologic and pathophysiologic implications of HO-APP associations. Experimental Procedures Reagents and Cell Lines Rabbit anti-HO1 (SPA896), anti-HO2 (SPA897, OSA200), anti-Calnexin (SPA870) antiserum, and an anti-KDEL (Grp78) monoclonal antibody (SPA827) were purchased from StressGen (Victoria, Canada). Anti-(His)6 and anti-APP/APLP (22C11) monoclonal antibodies were purchased from Invitrogen (Carlsbad, CA) and Boehringer Mannheim (Indianapolis, IN), respectively. Other anti-APP antibodies (CT15) have been described previously (Thinakaran et al., 1995). HEK293 cell lines stably expressing human CPR and human HO1 (HEK 293/CPR-HO1) or human HO2 (293/CPR-HO2) have been described previously (Dore´ et al., 1999a). These cell lines were cultured in DMEM supplemented with 10% FCS, 500 ␮g/ml of G418, and 500 ␮g/ml of Zeocin (Sigma, St. Louis, MO). (Gly)5(His)6 tag was introduced on the C terminus of APLP1 by PCR. Yeast Two-Hybrid Methods Two-hybrid screening and the construction of the parent vectors pPC97 and pPC86 were as described (Chevray and Nathans, 1992). Plasmid pPC97-HO2C was prepared by the insertion of an HO2 PCR product corresponding to amino acids 169–296 of rat HO2 (Rotenberg and Maines, 1990) into the SalI and BglII sites of pPC97, resulting in an open reading frame encoding a GAL4 BD-HO2 fusion protein. The HO2 fragment was constructed by PCR using the following primers: 5⬘-GGATCCGTCGACCGTGGCTCAGCGGGCA CTA-3⬘ (coding strand) and 5⬘-GGATCCAGATCTCACTGCAGGC TAGGCTTCC-3⬘ (noncoding strand). A rat cortex/hippocampal cDNA library in pPC86 was amplified once in DH10B (GIBCO–BRL, Gaithersburg, MD) and transformed into yeast Y190 containing the pPC97-HO2C (Walensky et al., 1998). pPC86-C105 was identified as a 1.3 kb clone that activated LacZ transcription and conferred histidine prototrophy in the presence of pPC97-HO2C. Truncated APLP1 fragments comprising amino acids 68–436 and 68–385 were generated by restriction of pPC86-C105 with NcoI–NotI and PflMI– NotI, respectively, followed by klenow filling of the ends and selfligation. Other truncated APLP1 fragments were prepared by PCR. Specificity of the interaction was confirmed for HO2 and APLP-1, and no activation was observed with HO2 and the vector only and the vector only with APLP-1. In addition, we did not detect activation with HO2 and PIN or with GAPDH and APLP-1. Immunoprecipitations Human HO1, human HO2, human APP695, human APP770, mouse APLP1, and mouse APLP2-763 cDNA were subcloned into pRK5 under the transcriptional control of the cytomegalovirus (CMV) immediate-early promoter-enhancer (Schall et al., 1990). HEK293 cells on 10 cm dishes were transfected with 10 ␮g of the APP constructs alone or in combination with 5 ␮g of the HO1 or HO2 constructs by the calcium phosphate precipitation method (Ausubel et al., 1994).

Amyloid Precursor Proteins Inhibit Heme Oxygenase 471

After 40 hr, cells were solubilized with 500 ␮l of lysis buffer (50 mM Tris-HCl [pH 7.4], 150 mM NaCl, 0.5% Triton X-100) on ice for 30 min and then cleared by centrifugation. For immunoprecipitation, 20 ␮l aliquots of the cleared lysates were diluted to 100 ␮l with dilution buffer (50 mM Tris-HCl [pH 7.4], 150 mM NaCl) and incubated with 1 ␮l of the anti-HO1 antibody (SPA896) or the anti-HO2 antibody (OSA200). The mixture was incubated at 4⬚C for 1 hr, mixed with 20 ␮l of a 1:1 slurry of protein A–agarose (Calbiochem, La Jolla, CA), and incubated for another 30 min at 4⬚C. The beads were washed twice with wash buffer A (50 mM Tris-HCl [pH 7.4], 300 mM NaCl, 0.1% Triton X-100), twice with wash buffer B (50 mM Tris-HCl [pH 7.4], 150 mM NaCl, 0.1% Triton X-100), and then proteins were eluted in 50 ␮l of SDS–PAGE sample buffer by boiling. The immunoprecipitates were fractionated on 4%–15% SDS–PAGE gels, transferred to nitrocellulose membranes, and immunoblotted with an anti-APP/APLP (22C11) or anti-His antibodies. Immunoblots were visualized using an enhanced chemiluminescence system (ECL, Amersham, Arlington Heights, IL). For immunoprecipitation from rat brain, one brain was homogenized in 3 ml of homogenization buffer (Tris-HCl [pH 7.4], 150 mM NaCl, and 10% glycerol) and centrifuged at 100,000 ⫻ g for 30 min. The pellets were solubilized in 3 ml of homogenization buffer containing 0.5% Triton X-100 and recentrifuged at 100,000 ⫻ g for 30 min. The supernatant (100 ␮l) was diluted 2-fold with homogenization buffer and incubated with 30 ␮l of protein G–agarose that was cross-linked to 20 ␮l of an anti-HO2 antibody (OSA200) or the anti-Calnexin antibody (SPA 870) for 60 min at 4⬚C (Harrow and Lane, 1988). The immunoprecipitates were then washed five times with wash buffer B and eluted in 50 ␮l of SDS–PAGE sample buffer. Subcellular Fractionation Subcellular fractionation of the mouse neuroblastoma N2a cell lines (maintained in 50% DMEM and 50% OptiMEM [Life Technologies, Rockville, MD] supplemented with 5% FBS) was performed as described before (Kim et al., 2000). Immunofluorescence HEK293/CPR-HO2 cells grown on lab chambers were washed with PBS and fixed with 2% paraformaldehyde-PBS for 20 min. The lab chambers were washed with PBS and incubated with permeabilization buffer (10 mM PIPES-NaOH [pH 6.8], 0.3 M sucrose, 0.1 M KCl, 2.5 mM MgCl2, and 1 mM EDTA) with or without 5 ␮g/ml digitonin for 15 min on ice (Otto and Smith, 1994). The coverslips were then washed with PBS and blocked in blocking solution (PBS, 1% BSA, and 1% normal goat serum). Subsequent washings and primary antibody (22C11 and SPA897) incubations were conducted in blocking buffer with or without 0.2% saponin for 5 hr at 4⬚C. Then, cells were stained with a FITC-labeled goat anti-rabbit IgG antibody or a Texas red–labeled goat anti-mouse IgG antibody (Jackson Lab). Rat primary cortical neuron cultures were prepared as described (Dore´ et al., 1997). Cells were fixed with 4% paraformaldehyde and permeabilized with 0.1% Triton X-100. Immunofluorescent staining was conducted using an anti-HO2 antiserum (SPA897) or anti-APP/ APLP2 antibody (22C11) as described above. Quantification of A␤ by Two-Site ELISA HEK293/CPR, 293/CPR-HO1, or 293/CPR-HO2 cells were plated on 6-well dishes at a density of 5 ⫻ 105 cells/well. Cells were transfected the next day with 3 ␮g of the pRK-APP constructs by Lipofectamine plus (GIBCO–BRL). After 24 hr the medium was changed, and cells were cultured for an additional 24 hr. The conditioned media were collected and A␤ concentrations were determined by a two-site ELISA as described (Tomita et al., 1997).

Boston, MA) as described (Zakhary et al., 1997) with slight modifications. Briefly, the assay was performed in 250 ␮l of 50 mM HEPESNaOH [pH 7.4], 1 mM EDTA, 10% glycerol, 0.2 mM NADPH, and 0.6 ␮M [55Fe]hemin (30,000 cpm). After 5 min at 37⬚C, the reaction was stopped with addition of 10 ␮M of SnPPIX. The reaction mixture was added to Dowex AG1X-8 (Bio-Rad Lab, Hercules, CA), and free iron produced was eluted with 50 mM HEPES-NaOH [pH 7.4], 1 mM EDTA, and 1 M NaCl. In routine assays, total activity corresponds to about 4000 cpm, while background is less than 700 cpm. Immunocytochemistry for Bilirubin After primary neuronal cultures were incubated for 24 hr with and without PMA (0.1 ␮M), cells were rinsed with ice-cold PBS and fixed with 6% EDC (Pierce) at 4⬚C for 2 hr as described before (Dore´ et al., 1999a). Cells were washed with PBS, incubated with 0.1% Triton X-100 for 15 min at room temperature, and washed with PBS. Cells were then preincubated for 1 hr with 4% NGS in PBS followed by incubation with anti-bilirubin (BR) antibody in 2% NGS for 2 hr at room temperature. Cells were then washed and incubated with the secondary antibody and developed. Hemin Toxicity Assay HEK293/CPR-HO1 or 293/CPR-HO2 cells were transfected with various concentrations of APP constructs by electroporation and then plated onto 6-well dishes. Total plasmid concentration was normalized with the pRK5 parent vector to 2 ␮g. After 5 hr incubation, hemin was added into the culture medium. After another 24 hr, cells were collected and cytotoxicity was determined by Trypan blue exclusion using a hematocytometer. In some experiments, cells were cotransfected with APP constructs and a LacZ-pRK5 construct so that ␤-gal enzymatic activity could be used for a quantitative and rapid determination of hemin toxicity. After 12 hr of incubation with hemin, cells were harvested and disrupted in 0.2 M Tris-HCl [pH 8.0]. Then, ␤-gal activity was determined by a ␤-gal assay kit using o-Nitrophenyl-␤-D-galactopyranoside as a substrate (Invitrogen). Primary Cultures of Neuronal Cells Mo/Hu APP695swe transgenic mice have been described previously (Borchelt et al., 1996). Cortical neuronal cells were isolated from litters of heterozygous Mo/Hu APP695swe transgenic or nontransgenic pups at embryonic day 18 as described (Dore´ et al., 1999a). For hemin toxicity assay, neurons were plated onto poly-D-lysineand laminin-coated 24-well plates at a density of 0.75 ⫻ 106 cells/ well in Neurobasal medium supplemented with B27. After 6 days in culture, cells were incubated in fresh medium with or without SnPPIX (5 ␮M) for 2 hr, and then hemin was added into the culture medium. Cell survival was determined 24 hr later using 3-(4,5-dimethylthiazol2yl)2,5-diphenyl tetrazolium bromide (MTT) assay (Dore´ et al., 1997) or a LDH cytotoxicity detection kit (Boehringer Mannheim). For H2O2 toxicity assay, neurons were placed in the N2 supplement HEPESbuffered high-glucose Neurobasal medium at day 5. Cells were treated with or without PMA for 24 hr and then exposed to H2O2 for 24 hr. Acknowledgments This work was supported by USPHS grants DA-00266 and Research Scientist Award DA 00074 (to S. H. S.). S. D. has a Grant-in-Aid from the American Heart Association. C. D. F. has a Howard Hughes Medical Institute Fellowship for Physicians. This article is dedicated to the memory of Robert Dore´. Received February 2, 2000; revised September 11, 2000.

HO Enzymatic Assay For enzyme assay, 6 ⫻ 106 cells of 293/CPR-HO1 or 293/CPR-HO2 were transiently transfected with 2 ␮g of pRK-GFP or pRK-APP by electroporation (260 V, 1000 ␮F in the culture medium) (Ausubel et al., 1994) and then spread on poly-D-lysine-coated dishes. Twenty hours after transfection, cells were harvested, resuspended in 50 mM HEPES-NaOH [pH 7.4], and 20% glycerol, and then disrupted by freezing and thawing. The cell extracts were used within 2 hr. HO activity was determined using [55Fe]hemin (NEN Life Sciences,

References Abraham, N.G., Lavrovsky, Y., Schwartzman, M.L., Stoltz, R.A., Levere, R.D., Gerritsen, M.E., Shibahara, S., and Kappas, A. (1995). Transfection of the human heme oxygenase gene into rabbit coronary microvessel endothelial cells: protective effect against heme and hemoglobin toxicity. Proc. Natl. Acad. Sci. USA 92, 6798–6802. Ausubel, F.M., Brent, R., Kingston, R.E., Moore, D.D., Seidman, J.G.,

Neuron 472

Smith, J.A., and Struhl, K. (1994). Transfection of DNA into eukaryotic cells. Curr. Prot. Mol. Biol. 1, 9.1.1–9.1.3.

Hardy, J. (1997a). The Alzheimer family of diseases: many etiologies, one pathogenesis? Proc. Natl. Acad. Sci. USA 94, 2095–2097.

Borchelt, D.R., Thinakaran, G., Eckman, C.B., Lee, M.K., Davenport, F., Ratovitsky, T., Prada, C.M., Kim, G., Seekins, S., Yager, D., et al. (1996). Familial Alzheimer’s disease-linked presenilin 1 variants elevate Abeta1-42/1-40 ratio in vitro and in vivo. Neuron 17, 1005– 1013.

Hardy, J. (1997b). Amyloid, the presenilins and Alzheimer’s disease. Trends Neurosci. 20, 154–159.

Busciglio, J., and Yankner, B.A. (1995). Apoptosis and increased generation of reactive oxygen species in Down’s syndrome neurons in vitro. Nature 378, 776–779.

Hino, Y., Asagami, H., and Minakami, S. (1979). Topological arrangement in microsomal membranes of hepatic haem oxygenase induced by cobalt chloride. Biochem. J. 178, 331–337.

Calhoun, M.E., Wiederhold, K.H., Abramowski, D., Phinney, A.L., Probst, A., Sturchler-Pierrat, C., Staufenbiel, M., Sommer, B., and Jucker, M. (1998). Neuron loss in APP transgenic mice. Nature 395, 755–756.

Hsiao, K., Chapman, P., Nilsen, S., Eckman, C., Harigaya, Y., Younkin, S., Yang, F., and Cole, G. (1996). Correlative memory deficits, Abeta elevation, and amyloid plaques in transgenic mice. Science 274, 99–102.

Chevray, P.M., and Nathans, D. (1992). Protein interaction cloning in yeast: identification of mammalian proteins that react with the leucine zipper of Jun. Proc. Natl. Acad. Sci. USA 89, 5789–5793.

Jarrett, J.T., and Lansbury, P.T., Jr. (1993). Seeding “one-dimensional crystallization” of amyloid: a pathogenic mechanism in Alzheimer’s disease and scrapie? Cell 73, 1055–1058.

Crain, B.J., Hu, W., Sze, C.I., Slunt, H.H., Koo, E.H., Price, D.L., Thinakaran, G., and Sisodia, S.S. (1996). Expression and distribution of amyloid precursor protein-like protein-2 in Alzheimer’s disease and in normal brain. Am. J. Pathol. 149, 1087–1095.

Kim, S.H., Lah, J.J., Thinakaran, G., Levey, A., and Sisodia, S.S. (2000). Subcellular localization of presenilins: association with a unique membrane pool in cultured cells. Neurobiol. Dis. 7, 99–117.

Dennery, P.A., Sridhar, K.J., Lee, C.S., Wong, H.E., Shokoohi, V., Rodgers, P.A., and Spitz, D.R. (1997). Heme oxygenase-mediated resistance to oxygen toxicity in hamster fibroblasts. J. Biol. Chem. 272, 14937–14942. Dennery, P.A., Spitz, D.R., Yang, G., Tatarov, A., Lee, C.S., Shegog, M.L., and Poss, K.D. (1998). Oxygen toxicity and iron accumulation in the lungs of mice lacking heme oxygenase-2. J. Clin. Invest. 101, 1001–1011. Doan, A., Thinakaran, G., Borchelt, D.R., Slunt, H.H., Ratovitsky, T., Podlisny, M., Selkoe, D.J., Seeger, M., Gandy, S.E., Price, D.L., and Sisodia, S.S. (1996). Protein topology of presenilin 1. Neuron 17, 1023–1030. Dore´, S., Kar, S., and Quirion, R. (1997). Insulin-like growth factor I protects and rescues hippocampal neurons against beta-amyloidand human amylin-induced toxicity. Proc. Natl. Acad. Sci. USA 94, 4772–4777. Dore´, S., Takahashi, M., Ferris, C.D., Hester, L.D., Guastella, D., and Snyder, S.H. (1999a). Bilirubin, formed by activation of heme oxygenase-2, protects neurons against oxidative stress injury. Proc. Natl. Acad. Sci. USA 96, 2445–2450. Dore´, S., Sampei, K., Goto, S., Alkayed, N.J., Guastella, D., Blackshaw, S., Gallagher, M., Traystman, R.J., Hurn, P.D., Koehler, R.C., and Snyder, S.H. (1999b). Heme oxygenase-2 is neuroprotective in cerebral ischemia. Mol. Med. 5, 656–663. Drummond, G.S., and Kappas, A. (1981). Prevention of neonatal hyperbilirubinemia by tin protoporphyrin IX, a potent competitive inhibitor of heme oxidation. Proc. Natl. Acad. Sci. USA 78, 6466– 6470. Ferris, C.D., Jaffrey, S.R., Sawa, A., Takahashi, M., Brady, S.D., Barrow, R.K., Tysoe, S.A., Wolosker, H., Baran˜ano, D.E., Dore´, S., et al. (1999). Haem oxygenase-1 prevents cell death by regulating cellular iron. Nat. Cell Biol. 1, 152–157. Gabbita, S.P., Lovell, M.A., and Markesbery, W.R. (1998). Increased nuclear DNA oxidation in the brain in Alzheimer’s disease. J. Neurochem. 71, 2034–2040. Games, D., Adams, D., Alessandrini, R., Barbour, R., Berthelette, P., Blackwell, C., Carr, T., Clemens, J., Donaldson, T., Gillespie, F., et al. (1995). Alzheimer-type neuropathology in transgenic mice overexpressing V717F beta-amyloid precursor protein. Nature 373, 523–527. Good, P.F., Perl, D.P., Bierer, L.M., and Schmeidler, J. (1992). Selective accumulation of aluminum and iron in the neurofibrillary tangles of Alzheimer’s disease: a laser microprobe (LAMMA) study. Ann. Neurol. 31, 286–292. Hansen, T.W., Allen, J.W., and Tommarello, S. (1999). Oxidation of bilirubin in the brain-further characterization of a potentially protective mechanism. Mol. Genet. Metab. 68, 404–409.

Harrow, E., and Lane, D. (1988). Preparing protein A bead-antibody affinity columns-direct coupling. Antibodies: A Laboratory Manual 1, 522–523.

Lee, P.J., Alam, J., Wiegand, G.W., and Choi, A.M. (1996). Overexpression of heme oxygenase-1 in human pulmonary epithelial cells results in cell growth arrest and increased resistance to hyperoxia. Proc. Natl. Acad. Sci. USA 93, 10393–10398. Loeffler, D.A., Connor, J.R., Juneau, P.L., Snyder, B.S., Kanaley, L., DeMaggio, A.J., Nguyen, H., Brickman, C.M., and LeWitt, P.A. (1995). Transferrin and iron in normal, Alzheimer’s disease, and Parkinson’s disease brain regions. J. Neurochem. 65, 710–724. Lovell, M.A., Robertson, J.D., Teesdale, W.J., Campbell, J.L., and Markesbery, W.R. (1998). Copper, iron and zinc in Alzheimer’s disease senile plaques. J. Neurol. Sci. 158, 47–52. Maines, M.D. (1997). The heme oxygenase system: a regulator of second messenger gases. Annu. Rev. Pharmacol. Toxicol. 37, 517–554. Markesbery, W.R. (1997). Oxidative stress hypothesis in Alzheimer’s disease. Free Radic. Biol. Med. 23, 134–147. McNamara, M.J., Ruff, C.T., Wasco, W., Tanzi, R.E., Thinakaran, G., and Hyman, B.T. (1998). Immunohistochemical and in situ analysis of amyloid precursor-like protein-1 and amyloid precursor-like protein-2 expression in Alzheimer disease and aged control brains. Brain Res. 804, 45–51. Mucke, L., Abraham, C.R., and Masliah, E. (1996). Neurotrophic and neuroprotective effects of hAPP in transgenic mice. Ann. NY Acad. Sci. 777, 82–88. Otto, J.C., and Smith, W.L. (1994). The orientation of prostaglandin endoperoxide synthases-1 and -2 in the endoplasmic reticulum. J. Biol. Chem. 269, 19868–19875. Ou, W.J., Cameron, P.H., Thomas, D.Y., and Bergeron, J.J. (1993). Association of folding intermediates of glycoproteins with calnexin during protein maturation. Nature 364, 771–776. Pappolla, M.A., Chyan, Y.J., Omar, R.A., Hsiao, K., Perry, G., Smith, M.A., and Bozner, P. (1998). Evidence of oxidative stress and in vivo neurotoxicity of beta-amyloid in a transgenic mouse model of Alzheimer’s disease: a chronic oxidative paradigm for testing antioxidant therapies in vivo. Am. J. Pathol. 152, 871–877. Poss, K.D., and Tonegawa, S. (1997a). Heme oxygenase 1 is required for mammalian iron reutilization. Proc. Natl. Acad. Sci. USA 94, 10919–10924. Poss, K.D., and Tonegawa, S. (1997b). Reduced stress defense in heme oxygenase 1-deficient cells. Proc. Natl. Acad. Sci. USA 94, 10925–10930. Price, D.L., and Sisodia, S.S. (1998). Mutant genes in familial Alzheimer’s disease and transgenic models. Annu. Rev. Neurosci. 21, 479–505. Rotenberg, M.O., and Maines, M.D. (1990). Isolation, characterization, and expression in Escherichia coli of a cDNA encoding rat heme oxygenase-2. J. Biol. Chem. 265, 7501–7506. Sawa, A., Oyama, F., Cairns, N.J., Amano, N., and Matsushita, M.

Amyloid Precursor Proteins Inhibit Heme Oxygenase 473

(1997). Aberrant expression of bcl-2 gene family in Down’s syndrome brains. Mol. Brain Res. 48, 53–59.

mer’s associated amyloid beta protein precursor. Nat. Genet. 5, 95–100.

Schall, T.J., Lewis, M., Koller, K.J., Lee, A., Rice, G.C., Wong, G.H., Gatanaga, T., Granger, G.A., Lentz, R., Raab, H., et al. (1990). Molecular cloning and expression of a receptor for human tumor necrosis factor. Cell 61, 361–370.

White, A.R., Zheng, H., Galatis, D., Maher, F., Hesse, L., Multhaup, G., Beyreuther, K., Masters, C.L., and Cappai, R. (1998). Survival of cultured neurons from amyloid precursor protein knock-out mice against Alzheimer’s amyloid-beta toxicity and oxidative stress. J. Neurosci. 18, 6207–6217.

Schipper, H.M., Cisse, S., and Stopa, E.G. (1995). Expression of heme oxygenase-1 in the senescent and Alzheimer-diseased brain. Ann. Neurol. 37, 758–768. Selkoe, D.J. (1996). Amyloid beta-protein and the genetics of Alzheimer’s disease. J. Biol. Chem. 271, 18295–18298. Shibahara, S., Muller, R., Taguchi, H., and Yoshida, T. (1985). Cloning and expression of cDNA for rat heme oxygenase. Proc. Natl. Acad. Sci. USA 82, 7865–7869. Slunt, H.H., Thinakaran, G., Von Koch, C., Lo, A.C., Tanzi, R.E., and Sisodia, S.S. (1994). Expression of a ubiquitous, cross-reactive homologue of the mouse beta-amyloid precursor protein (APP). J. Biol. Chem. 269, 2637–2644.

Yankner, B.A. (1996). Mechanisms of neuronal degeneration in Alzheimer’s disease. Neuron 16, 921–932. Yoshida, T., and Sato, M. (1989). Posttranslational and direct integration of heme oxygenase into microsomes. Biochem. Biophys. Res. Commun. 163, 1086–1092. Yoshinaga, T., Sassa, S., and Kappas, A. (1982). Purification and properties of bovine spleen heme oxygenase. Amino acid composition and sites of action of inhibitors of heme oxidation. J. Biol. Chem. 257, 7778–7785. Younkin, S.G. (1995). Evidence that A beta 42 is the real culprit in Alzheimer’s disease. Ann. Neurol. 37, 287–288.

Smith, M.A., Kutty, R.K., Richey, P.L., Yan, S.D., Stern, D., Chader, G.J., Wiggert, B., Petersen, R.B., and Perry, G. (1994). Heme oxygenase-1 is associated with the neurofibrillary pathology of Alzheimer’s disease. Am. J. Pathol. 145, 42–47.

Zakhary, R., Gaine, S.P., Dinerman, J.L., Ruat, M., Flavahan, N.A., and Snyder, S.H. (1996). Heme oxygenase 2: endothelial and neuronal localization and role in endothelium-dependent relaxation. Proc. Natl. Acad. Sci. USA 93, 795–798.

Smith, M.A., Perry, G., Richey, P.L., Sayre, L.M., Anderson, V.E., Beal, M.F., and Kowall, N. (1996). Oxidative damage in Alzheimer’s. Nature 382, 120–121.

Zakhary, R., Poss, K.D., Jaffrey, S.R., Ferris, C.D., Tonegawa, S., and Snyder, S.H. (1997). Targeted gene deletion of heme oxygenase 2 reveals neural role for carbon monoxide. Proc. Natl. Acad. Sci. USA 94, 14848–14853.

Smith, M.A., Hirai, K., Hsiao, K., Pappolla, M.A., Harris, P.L., Siedlak, S.L., Tabaton, M., and Perry, G. (1998). Amyloid-beta deposition in Alzheimer transgenic mice is associated with oxidative stress. J. Neurochem. 70, 2212–2215. Sprecher, C.A., Grant, F.J., Grimm, G., O’Hara, P.J., Norris, F., Norris, K., and Foster, D.C. (1993). Molecular cloning of the cDNA for a human amyloid precursor protein homolog: evidence for a multigene family. Biochemistry 32, 4481–4486. Stocker, R., Yamamoto, Y., McDonagh, A.F., Glazer, A.N., and Ames, B.N. (1987). Bilirubin is an antioxidant of possible physiological importance. Science 235, 1043–1046. Thinakaran, G., Kitt, C.A., Roskams, A.J., Slunt, H.H., Masliah, E., von Koch, C., Ginsberg, S.D., Ronnett, G.V., Reed, R.R., Price, D.L., et al. (1995). Distribution of an APP homolog, APLP2, in the mouse olfactory system: a potential role for APLP2 in axogenesis. J. Neurosci. 15, 6314–6326. Tomita, T., Maruyama, K., Saido, T.C., Kume, H., Shinozaki, K., Tokuhiro, S., Capell, A., Walter, J., Grunberg, J., Haass, C., et al. (1997). The presenilin 2 mutation (N141I) linked to familial Alzheimer disease (Volga German families) increases the secretion of amyloid beta protein ending at the 42nd (or 43rd) residue. Proc. Natl. Acad. Sci. USA 94, 2025–2030. Van Lenten, B.J., Prieve, J., Navab, M., Hama, S., Lusis, A.J., and Fogelman, A.M. (1995). Lipid-induced changes in intracellular iron homeostasis in vitro and in vivo. J. Clin. Invest. 95, 2104–2110. Verma, A., Hirsch, D.J., Glatt, C.E., Ronnett, G.V., and Snyder, S.H. (1993). Carbon monoxide: a putative neural messenger. Science 259, 381–384. von Koch, C.S., Zheng, H., Chen, H., Trumbauer, M., Thinakaran, G., van der Ploeg, L.H., Price, D.L., and Sisodia, S.S. (1997). Generation of APLP2 KO mice and early postnatal lethality in APLP2/APP double KO mice. Neurobiol. Aging 18, 661–669. Walensky, L.D., Gascard, P., Fields, M.E., Blackshaw, S., Conboy, J.G., Mohandas, N., and Snyder, S.H. (1998). The 13-kD FK506 binding protein, FKBP13, interacts with a novel homologue of the erythrocyte membrane cytoskeletal protein 4.1. J. Cell Biol. 141, 143–153. Wasco, W., Bupp, K., Magendantz, M., Gusella, J.F., Tanzi, R.E., and Solomon, F. (1992). Identification of a mouse brain cDNA that encodes a protein related to the Alzheimer disease-associated amyloid beta protein precursor. Proc. Natl. Acad. Sci. USA 89, 10758– 10762. Wasco, W., Gurubhagavatula, S., Paradis, M.D., Romano, D.M., Sisodia, S.S., Hyman, B.T., Neve, R.L., and Tanzi, R.E. (1993). Isolation and characterization of APLP2 encoding a homologue of the Alzhei-

Zhang, F., Eckman, C., Younkin, S., Hsiao, K.K., and Iadecola, C. (1997). Increased susceptibility to ischemic brain damage in transgenic mice overexpressing the amyloid precursor protein. J. Neurosci. 17, 7655–7661. Zheng, H., Jiang, M., Trumbauer, M.E., Sirinathsinghji, D.J., Hopkins, R., Smith, D.W., Heavens, R.P., Dawson, G.R., Boyce, S., Conner, M.W., et al. (1995). beta-Amyloid precursor protein-deficient mice show reactive gliosis and decreased locomotor activity. Cell 81, 525–531.