Endoplasmic Reticulum Stress-inducible Protein, Herp, Enhances Presenilin-mediated Generation of Amyloid β-Protein* 210

Presenilin (PS) is essential for the γ-cleavage required for the generation of the C terminus of amyloid β-protein (Aβ). However, the mechanism underlying PS-mediated γ-cleavage remains unclear. We have identified Herp cDNA by our newly developed screening method for the isolation of cDNAs that increase the degree of γ-cleavage. Herp was originally identified as a homocysteine-responsive protein, and its expression is up-regulated by endoplasmic reticulum stress. Herp is an endoplasmic reticulum-localized membrane protein that has a ubiquitin-like domain. Here, we report that a high expression of Herp in cells increases the level of Aβ generation, although not in PS-deficient cells. We found that Herp interacts with both PS1 and PS2. Thus, Herp regulates PS-mediated Aβ generation, possibly through its binding to PS. Immunohistochemical analysis of a normal human brain section with an anti-Herp antibody revealed the exclusive staining of neurons and vascular smooth muscle cells. Moreover, the antibody strongly stained activated microglia in senile plaques in the brain of patients with Alzheimer disease. Taken together, Herp could be involved in Aβ accumulation, including the formation of senile plaques and vascular Aβ deposits.

Mutations in the presenilin genes, PS1 and PS2, cause early onset familial Alzheimer disease (AD) 1 (1)(2)(3) and lead to an increase in the ratio of A␤42/A␤40 generation from ␤-amyloid precursor protein (APP) (reviewed in Ref. 4). PS is required for intramembranous cleavage of APP (termed ␥-cleavage) and Notch (5)(6)(7)(8). Recently, ErbB4 has been found to be another natural substrate of the PS-mediated intramembranous proteolysis (9). Interestingly, although the ␥-cleavage of APP is a critical step toward the production of A␤, the major intramembranous cleavage site of APP was found to be distinct from the ␥-cleavage site (10,11). The mechanism underlying PS-mediated intramembranous proteolysis including ␥-cleavage remains to be clarified. To date, a number of PS-interacting proteins have been identified, but no natural interactors with PS have been found to modulate A␤ generation, although some mutant forms of nicastrin, which was identified as a component of the PS complex, increase A␤ production (12). To elucidate the mechanism underlying the PS-mediated ␥-cleavage, we have recently developed a new functional screening method for identifying cDNAs that increase the degree of ␥-cleavage using a combination of ␥-cleavage-dependent puromycin-resistant assay and A␤ quantitation (see "Experimental Procedures"; the details of this screening method will be described elsewhere). Using this method, we have identified Herp, which was originally identified as a homocysteine-induced protein (13). Interestingly, elevated levels of homocysteine are correlated with multiple neurological disorders (14 -16). Plasma homocysteine levels have been reported to be elevated in some cases of AD (16) and could be an early marker of cognitive impairment in the elderly (17). Herp mRNA is constitutively expressed in various tissues, but its expression is up-regulated by homocysteine or the inducers of ER stress such as 2-mercaptoethanol, A23187, and thapsigargin (13,18). Herp is a membrane protein localized in the ER, and it has an N-terminal ubiquitin-like domain; the function of this protein, however, is not known (13). In contrast to ER stress-induced molecular chaperones including BiP and GRP94 in the ER lumen, a major portion of Herp molecules are present in the cytoplasmic side of the ER membrane (13). Herp may have a role distinct from that of molecular chaperones under ER stress. Here, we show that Herp enhances PS-mediated A␤ generation and that it interacts with PS.

EXPERIMENTAL PROCEDURES
Isolation of Herp cDNA-We designed a functional screening system for isolating cDNAs that increase the degree of ␥-cleavage as follows.
(The details of the screening system will be described elsewhere.) 2 We established a cell line (designated as A5-9), BaF/3 cells (19) stably transfected with pCxN-C53NICD (see below) and pHES1-pac (see below) in which an increase in the degree of ␥-cleavage confers on the cells an increase in puromycin resistance. BaF/3 cells do not express endogenous APP. 3 To initiate screening, a human hippocampus-derived cDNA library in a retroviral vector, pMX, was infected into A5-9 cells as reported previously (19,20). The infected A5-9 cells were treated with a lethal dose of puromycin. The level of A␤40 secreted from surviving clones was measured using a highly sensitive immunoblotting method. The cDNAs from surviving clones expressing a higher level of A␤40 as compared with that of parental cells were extracted and amplified by PCR using vector primers. One of these encoded Herp lacking Gln 76 -Lys 100 . Since we did not know whether this Herp variant is a natural form or a PCR artifact product, we recloned a previously reported Herp cDNA (13) into pcDNA3.1 or pMX. The effect of the Herp cDNA on A␤ generation from the full-length APP was then tested using HEK293 cells stably expressing APP.
Plasmids and Retrovirus-mediated Infection-By the PCR method, we generated a chimeral cDNA encoding C53NICD that consists of the first Met plus the Asp 596 -Leu 648 fragment of APP695 (C100 (N-terminal truncated APP starting at the ␤-secretase site) lacking the intracellular domain) and the C-terminal intracellular domain (from Ser 1745 to the C terminus) of mouse Notch-1. This chimeral cDNA was inserted into pCxN (21) and designated as pCxN-C53NICD. A plasmid named pHES1-pac, which contains a puromycin-resistant gene (pac) driven by the HES1 promoter (a DNA element responsible for Notch-dependent gene expression) (22), was constructed as follows: (i) The HES1 promoter isolated by the PCR method from the mouse genomic library was inserted into a vector, PGV-B (TOYO B-Net Co., Ltd., Tokyo, Japan). This plasmid was designated as PGV-B-HES-1. (ii) pac of pPUR (CLONTECH) was inserted into PGV-B-HES-1. The final plasmid was designated as pHES-pac. We ensured that the cells expressing C53NICD, as well as the cells expressing APP, generate A␤40 and A␤42. We also confirmed that the expression of C53NICD in the cells induced the expression of the luciferase gene driven by the HES1 promoter. The retrovirus-mediated infection was carried out as reported previously (20). The gene encoding Herp tagged with the N-terminal Myc and C-terminal FLAG was prepared as described previously (13).
Cell Lines-HEK293 cells stably transfected with APP695 were generated as reported previously (23). HEK293 cells stably transfected with APP695 and human PS1 were generated from HEK293 cells stably expressing APP695 by the transfection of the cells with the pcDNA 3.1-Hygro vector (Invitrogen) carrying human PS1.
PS-deficient Fibroblasts-Tissue from PS-deficient embryo (24) was digested with collagenase and cultured in Dulbecco's modified Eagle's medium-F12 containing 10% fetal calf serum. Outgrowing cells were immortalized with large T antigen and split twice a week.
Antibodies, Immunoprecipitation, Immunoblotting, and Sandwich ELISA-An affinity-purified rabbit anti-Herp antibody was prepared as described previously (13). A rat anti-PS1 antibody (for the N-terminal fragment of PS1) and a rabbit anti-PS2 antibody (for the N-terminal fragment) were purchased from Chemicon International, Inc. (Temecula, CA) and Santa Cruz Biotechnology, Inc. (Santa Cruz, CA), respectively. A monoclonal antibody 6E10 specific to human A␤1-17 was purchased from Senetek (St. Louis, MO). BA27 specific for the A␤40 terminal site, BC05 specific for the A␤42 terminal site, BAN50 raised against A␤1-6, and BNT77 raised against human A␤11-28 have all been characterized previously (25). Affinity-purified rabbit antibody 369 was raised against the C-terminal residues of APP695 (26). Affinity-purified rabbit antibody B12/4 was raised against 20 C-terminal amino acid residues of APP695 (27). Cultured cells were lysed in RIPA buffer (150 mM NaCl, 10 mM Tris/HCl pH 7.5, 1% Nonidet P-40, 0.1% SDS, and 0.2% sodium deoxycholate) containing a protease inhibitor mixture. The solubilized proteins were subjected to immunoprecipitation as described previously (28). The precipitated proteins were resolved by SDS-PAGE on 4 -20% gel for the detection of PS and Herp and on 7.5% gel for the detection of intracellular APP. Immunoblotting was performed as reported previously (28). The secreted A␤ was immunoprecipitated and detected using a highly sensitive immunoblotting tech-  (14) harboring Herp cDNA or the vector alone. Cells stably expressing Herp were selected in the presence of puromycin. A␤40 secreted from cells (4 ϫ 10 5 ) during 12 h was immunoprecipitated with 6E10 and detected with BA27. Immunoprecipitated A␤ from cells grown on one culture dish was loaded in one lane. The lysates (10 g) were immunoblotted with anti-Herp antibody. WB, Western blots. b, ELISA for A␤ secreted from the cells stably expressing Herp after a 12-h culture (for A␤40) or a 48-h culture (for A␤42). HEK293 cells stably expressing APP were retrovirally infected with the pMX-puro vector harboring Herp cDNA or vector alone. Cells stably expressing Herp were selected in the presence of puromycin. The relative amounts of A␤ were determined by calculating the ratio of the amount of A␤ secreted by cells expressing Herp to the mean amount of A␤ secreted by a mock transfectant (normalized to 1). Values are the means Ϯ S.D. of two independent experiments of two independent Herp or mock-stable transfectants (n ϭ 4). *, p Ͻ 0.02; **, p Ͼ 0.05 (Mann-Whitney's U test). c, PS1/PS2 double-deficient fibroblasts or wild-type fibroblasts (5 ϫ 10 5 ) plated on a 100-mm tissue culture dish and retrovirally infected with retroviral vector, pMX (mock), pMX-Herp, or pMX-PS1, together with pMX-APP695. A␤40 secreted during a 24-h incubation was detected. The level of intracellular APP (holo-APP) was detected with the anti-C-terminal APP antibody, 369. mr(K), molecular size in kDa. d, HEK293 cells stably expressing APP were retrovirally infected with the pMX-puro vector harboring Herp cDNA or vector alone. The upper panel indicates the level of intracellular C99. The intracellular C99 in RIPA-solubilized lysates (1.5 mg) was immunoprecipitated with BAN50 and immunoblotted with B12/4. The lysates (10 g) prepared from HEK293 cells transiently transfected with pcDNA or pcDNA-C100 (41) were also immunoblotted with B12/4 (right two lanes). The lower panel indicates the level of ␣-secretase-cleaved soluble APP (␣-APPs). The media used for a 24-h culture were immunoblotted with 6E10. IP, immunoprecipitates.
nique as described previously (28). Intracellular C99 was immunoprecipitated with BAN50 and detected using a highly sensitive immunoblotting technique with B12/4. ELISA for A␤ was performed as described previously (25). The capture antibody used was BNT77. Detector antibodies were horseradish peroxidase-conjugated BA27 (for A␤40) and horseradish peroxidase-conjugated BC05 (for A␤42). ELISA data were statistically analyzed by analysis of variance (ANOVA) using StatView-J.4.11.
Histochemical Analysis-Acetone-fixed cryostat sections or Kryofixfixed, paraffin-embedded serial sections were prepared from five nondemented and five AD brains. Cryostat sections were used only for the staining of neurons with the anti-Herp antibody since the staining signals were weak in ordinary formalin-paraffin sections. Tissue sections were blocked with 5% normal goat (for anti-Herp antibody) or horse (for anti-HLA-DP, DQ, and DR antibody) serum and then reacted with the anti-Herp antibody (5 g/ml) or the anti-HLA-DP, DQ, and DR (major histocompatibility complex class II antigen, 3 g/ml; DAKO, CR3/43) antibody followed by biotinylated secondary antibodies. Positive signals were visualized by incubating the sections in a diaminobenzidine-H 2 O 2 solution. For the double immunofluorescent study, after pretreatment with 0.1% Sudan black B in 70% ethanol for 7 min to mask autofluorescence, Kryofix paraffin sections were reacted with a primary antibody mixture (anti-Herp, 5 g/ml and anti-A␤, 4G8, monoclonal, 1:1,000) and then reacted with a mixture of fluorescein isothiocyanate-conjugated goat anti-rabbit IgG (40 g/ml, Jackson Immuno-Research Laboratories) and Texas red-conjugated goat anti-mouse IgG (20 g/ml, Jackson). For the control study, anti-Herp antibody was preabsorbed with 10-fold by weight of recombinant Herp.

RESULTS
We first investigated whether transfection of Herp cDNA increases the extent of A␤ generation from full-length APP using HEK293 cells stably expressing APP. We found that a stably high expression level of Herp increased the extent of A␤40 generation (Fig. 1a). Statistical analysis of ELISA results using the cells stably expressing Herp showed that the increase in the A␤40 level caused by the high expression of Herp was ϳ1.8-fold (Fig. 1b). In contrast, the effect of Herp on the A␤42 level was not significant (Fig. 1b). We next tested whether an increase in the level of A␤ generation by Herp expression can occur in PS-deficient cells since it is not conclusive yet whether PS itself is a ␥-secretase. As shown in Fig. 1c, no A␤ was detected in PS-deficient fibroblasts expressing Herp, whereas Herp expression in wild-type fibroblasts increased the A␤ level. We also confirmed that the intracellular APP level was not changed by the high expression level of Herp (Fig. 1c). These results indicate that Herp enhances PS-mediated A␤40 generation. In addition, the level of intracellular ␤-secretase-cleaved C-terminal APP fragment (C99) was significantly reduced by Herp transfection, whereas the level of soluble ␣-secretasecleaved N-terminal APP fragment (␣-APPs) was not changed (Fig. 1d). These results strongly suggest that Herp preferentially increases a degree of ␥-secretase cleavage.
Since both PS and Herp reside in the ER, it is likely that Herp interacts with PS, resulting in an increase in the extent of A␤ generation. Therefore, we next investigated the interaction of PS with Herp by coimmunoprecipitation studies using HEK293 cells stably expressing PS1 plus APP. An anti-PS1 antibody coimmunoprecipitated with Herp (Fig. 2a). The level of the coimmunoprecipitated Herp was approximately equivalent to that of Herp contained in the 10% starting lysate used for immunoprecipitation (Fig. 2a). Considering that the efficiency of immunoprecipitation by the anti-PS1 antibody is about 10% (data not shown), it is likely that almost the entire amount of Herp interacts with PS1. PS undergoes endoproteolysis, forming N-and C-terminal fragments (29). Interestingly, an anti-Herp antibody immunoprecipitated the fulllength PS1, but it only immunoprecipitated a small amount of the N-terminal fragment of PS1 (Fig. 2b). These results indicate that Herp mainly interacts with the full-length PS1. The study of PS2 and Herp interaction was also performed using RIPA-solubilized lysates (100 g) were immunoprecipitated with anti-PS1 N-terminal fragment (NTF) antibody and immunoblotted with anti-Herp antibody. The lysates (10 g) were also immunoblotted with the anti-Herp antibody (right two lanes). WB, Western blots. b, HEK293 cells stably expressing APP and PS1 were transiently transfected with Herp cDNA. RIPA-solubilized lysates (100 g) were immunoprecipitated with anti-Herp and immunoblotted with anti-PS1 N-terminal fragment antibody. The lysates (10 g) were also immunoblotted with anti-PS1 N-terminal fragment antibody (right two lanes). The same result was also obtained when the lysate was prepared using a different detergent such as digitonin, which extracts the complex of Nicastrin and the endoproteolytic product of PS (12). FL, full-length. c, HEK293 cells transiently transfected with C-terminal FLAG-tagged (and also N-terminal Myc-tagged) Herp (Herp-F) and PS2. RIPA-solubilized lysates (100 g) were immunoprecipitated with the anti-PS2 N-terminal fragment antibody and immunoblotted with anti-FLAG antibody. The lysates (10 g) were also immunoblotted with the anti-FLAG antibody (right two lanes). d, RIPA-solubilized lysates (100 g) of HEK293 cells transiently transfected with PS2 and FLAG-tagged Herp were immunoprecipitated with anti-FLAG antibody and immunoblotted with anti-PS2 N-terminal fragment antibody. The lysates (10 g) were immunoblotted with the anti-PS2 antibody (right two lanes).
HEK293 cells that transiently expressed PS2 and FLAGepitope-tagged Herp. As shown in Fig. 2c, full-length PS2 coprecipitated with FLAG-epitope-tagged Herp. Thus, Herp interacts with both PS1 and PS2, and the major PS molecule that interacts with Herp was the full-length PS. The expression of Herp did not alter the steady-state levels of full-length PS and N-terminal fragment (Fig. 2, b and d), suggesting that Herp does not enhance the endoproteolysis of PS. APP and Herp interaction was also studied. Herp only coprecipitated APP at less than 1% of the total APP in the cells (data not shown). Therefore, in all probability, APP does not specifically interact with Herp. Thus, it is likely that the direct interaction of the full-length PS with Herp causes an increase in the degree of PS-mediated ␥-cleavage of APP. Alternatively, at present, we cannot exclude the possibility that Herp indirectly affects the A␤ level, for example, by regulating the intracellular calcium level, since it was suggested that the intracellular calcium level modifies the A␤ level (30). Herp expression is up-regulated by calcium ionophores (13), and Herp may potentially regulate the intracellular calcium level.
Immunohistochemical analysis of Herp in a human brain section was performed to identify the cell types expressing a high level of Herp in the brain; these cells are considered to generate a high level of A␤. As shown in Fig. 3a, only neurons were positive for Herp in the parenchyma cerebral cortex of a nondemented human brain. Interestingly, microglia in senile plaques in the brain of AD patients were strongly stained (Fig.  3b). The staining of HLA-DP, DQ, and DR (markers of activated microglia (31)) and Herp in serial sections (Fig. 3c) indicates that some of the activated microglia in the center of senile plaques are highly immunoreactive with the anti-Herp antibody. Double immunostaining with anti-Herp and anti-HLA-DP, DQ, and DR antibodies also confirmed that activated microglia are strongly positive for Herp (see supplemental data). In addition, smooth muscle cells in meningeal arteries and arterioles were strongly stained (Fig. 3d). DISCUSSION One hallmark of AD is the accumulation of A␤, although the cause of AD may be multifactorial. Here, we showed that ER stress-inducible Herp enhances A␤40 generation. Since a significant increase in the level of A␤42 was not observed, the high expression level of Herp appears to mainly enhance 40-␥-secretase cleavage. A recent report showed that some nonsteroidal anti-inflammatory drugs preferentially decreased the generation of A␤42 (32). Taken together with our result, it appears that the regulation of the generation of A␤42 is distinct from FIG. 3. Immunohistochemical analysis of human brain sections for Herp. a, a cryostat section prepared from the frontal cortex from a nondemented human brain stained with the anti-Herp antibody. Only neurons are positive for Herp (brown stain). Staining signals were not observed when the antibody was preabsorbed with recombinant Herp (see supplemental data). b, a Kryofix paraffin-fixed section from the AD brain double-immunostained with the anti-Herp antibody (green) and the anti-A␤ antibody, 4G8 (red). The staining pattern was visualized by fluorescence microscopy. Dendritic cells, possibly microglia (white arrows), in the center of a 4G8-positive senile plaque were strongly stained with the anti-Herp antibody. c, serial sections from an AD brain immunostained with the anti-HLA-DP, DQ, DR (left), the anti-Herp (center), and the preabsorbed anti-Herp (right) antibodies. Activated microglia, which were positive for HLA-DP, DQ, and DR (left), were intensely labeled for Herp (center), whereas they were negative when anti-Herp antibody was preabsorbed with recombinant Herp (right). The black arrow indicates A␤ deposits. The asterisks indicate an identical capillary. d, smooth muscle cells in the meningeal arterial wall of nondemented brain consistently immunostained with the anti-Herp antibody (brown color). The antibody preabsorbed with recombinant Herp abolished the staining (see supplemental data). Scale bar, 50 m. that of A␤40, even though the generation of both A␤ forms requires PS.
Herp was originally discovered as a homocysteine-induced protein. It is noteworthy that a high plasma homocysteine level was observed in some cases of AD (14). At present, we cannot determine whether Herp directly or indirectly activates ␥-secretase. We have shown that Herp interacts with the fulllength PS rather than with its endoproteolytic products. A transient overexpression of Herp did not enhance the endoproteolytic cleavage of PS nor change the PS level. Since endoproteolytic cleavage of PS is not essential for the biological function of PS or A␤ generation (34), one interpretation for the enhancement of A␤ generation caused by Herp expression is that Herp directly activates the full-length PS-mediated ␥-activity by binding to the full-length PS, possibly through a change in the conformation or the trafficking of PS. The physical interaction of Herp with PS also raises the possibility that the function of PS is modulated by ER stress through Herp expressed at high levels. It was reported that cells exposed to inducers of ER stress, such as calcium ionophore, A23187, or thapsigargin, have increased the extracellular A␤ level, suggesting that the intracellular calcium level regulates A␤ generation (30). Taken together with our results, it is most likely that these inducers cause a Herp-mediated increase in the A␤ level since these inducers strongly up-regulate Herp expression.
Our present study showed that, in the normal brain, Herp was expressed in neurons and highly expressed in vascular cells, suggesting that these cells are the major source of A␤ in the normal brain. Interestingly, Herp expression was up-regulated in activated microglia in senile plaques in the AD brain. Activated microglia are thought to play a role in the clearance of A␤. However, some activated microglia, which strongly express Herp, may generate a high level of A␤ and function in the formation of A␤ deposits. Vascular A␤ deposits known as amyloid angiopathy are also one of the major signs of AD pathology (4,33). The high expression level of Herp in vascular cells may be related to amyloid angiopathy. The cells expressing Herp at high levels could contribute to A␤ accumulation, including the formation of senile plaques or vascular A␤ deposits, in patients with AD.
In particular, it is interesting that Herp has an N-terminal ubiquitin-like domain. Several proteins were reported to have a ubiquitin-like domain in eukaryotic cells, but the biological significance of this domain still remains to be clarified. In some cases, such as ubiquitin itself, a ubiquitin-related domain is involved in targeting proteins for protein degradation (35), but other functions such as chaperon function or protein-protein interaction were also reported (36). Presenilin undergoes degradation via the ubiquitin/proteosome pathway (29). However, the steady-state level of PS in cells expressing Herp was not changed as compared with that in nontransfected cells, strongly suggesting that the high expression level of Herp is not involved in the degradation of PS. Mutations in the PAR-KIN gene were discovered in families with an autosomal recessive form of Parkinson disease (37). Parkin also has the ubiquitin-like domain in the N-terminal site, although the function of the domain is not known, and it is up-regulated by ER stress (38,39). Recent reports indicate that parkin is involved in the ubiquitination of Pael receptor and ␣-synuclein (39,40). Since a strong ubiquitin immunoreactivity is associated with the pathology of both AD and Parkinson disease, the complex of presenilin and Herp with a ubiquitin-like domain may also function in the elimination of misfolded proteins via the ubiquitin/proteasome pathway as parkin does. Further exploration of the biological significance of the interaction of Herp with PS will not only help determine the pathways by which ␥-secretase activity is regulated by PS but may also further clarify the role of PS under ER stress or in AD pathogenesis.