The Cell Adhesion Protein P-selectin Glycoprotein Ligand-1 Is a Substrate for the Aspartyl Protease BACE1*

The aspartyl protease BACE1 cleaves the amyloid precursor protein and the sialyltransferase ST6Gal I and is important in the pathogenesis of Alzheimer’s disease. The normal function of BACE1 and additional physio-logical substrates have not been identified. Here we show that BACE1 acts on the P-selectin glycoprotein ligand 1 (PSGL-1), which mediates leukocyte adhesion in inflammatory reactions. In human monocytic U937 and human embryonic kidney 293 cells expressing endogenous or transfected BACE1, PSGL-1 was cleaved by BACE1 to generate a soluble ectodomain and a C-termi-nal transmembrane fragment. No evidence of the cleavage fragment was seen in primary cells derived from mice deficient in BACE1. By using deletion constructs and enzymatic deglycosylation of the C-terminal PSGL-1 fragments, the cleavage site

The amyloid hypothesis of Alzheimer's disease attributes the pathogenesis of the disease to accumulation of amyloid ␤-peptide (A␤), 1 a fragment of the amyloid precursor protein (APP) (1). APP is one of a large number of membrane proteins that are proteolytically converted to their soluble counterparts. This process is referred to as ectodomain shedding and is an important way of regulating the biological activity of membrane proteins (2,3). Ectodomain shedding has been described in many multicellular organisms, such as Caenorhabditis elegans, Drosophila melanogaster, mice, and humans and is important in embryonic development, the inflammatory response, and other biological processes (3)(4)(5).
The shedding of APP may occur through two different protease activities termed ␣and ␤-secretase, which cleave APP within its ectodomain close to its transmembrane domain (1). ␣-Secretase is a member of the ADAM family of proteases (A Disintegrin And Metalloprotease), which typically carry out the ectodomain shedding of numerous proteins (2,3). Because the ADAM proteases cleave within the A␤-sequence, they preclude the generation of the A␤-peptide. In contrast, the ␤-secretase activity cleaves at the N terminus of the A␤-peptide domain, thereby catalyzing the first step in A␤-peptide generation. The ␤-secretase has recently been identified and is a novel aspartyl protease called BACE1 (␤-site APP-cleaving enzyme) (6 -10). Once APP is cleaved by BACE1, the remaining C-terminal APP fragment may be cleaved by the so-called ␥-secretase within its transmembrane domain at the C terminus of A␤, leading to the secretion of the A␤-peptide (11).
Because the ectodomain shedding of APP by BACE1, but not by ADAM proteases, is the first step of A␤-peptide generation, inhibition of BACE1 activity is considered to be a highly promising approach to treat Alzheimer's disease (1,12,13). However, it remains unclear whether BACE1 predominantly cleaves APP and a recently identified sialyltransferase (14) or also additional proteins. In the latter case, like the ADAM metalloproteases, BACE1 could be a basic cellular mediator of the ectodomain shedding of membrane proteins.
Because APP is mainly cleaved by a metalloprotease of the ADAM family and only to a smaller extent by BACE1 (1), we reasoned that additional membrane proteins known to undergo ectodomain shedding by a metalloprotease might also be cleaved to a smaller extent by BACE1. By testing selected candidate proteins, we found that the P-selectin glycoprotein ligand-1 (PSGL-1) is a novel substrate for BACE1. PSGL-1, which is expressed as a homodimer, mediates leukocyte adhesion to endothelial cells and is critically involved in the inflammatory response both in brain and in peripheral tissues (15). Similar to APP, PSGL-1 is a type I membrane protein. It consists of a signal peptide, a prodomain, which is presumably removed by furin or one of its homologues upon maturation of the protein, a receptor binding domain, 15 to 16 repeats of 10 amino acids (decamer repeats), which are highly O-glycosylated, a juxtamembrane domain, a transmembrane, and a cytoplasmic domain (15).
Cell Culture, Transfections, Retroviral Transductions, and Flow Cytometric Analysis-293-EBNA cells were cultured in Dulbecco's modified Eagle's medium (Invitrogen) containing 10% fetal bovine serum (Hyclone). Human monocytic U937 cells and the human T cell clone Jurkat 10104 were cultured in Iscove's modified Dulbecco's medium (Invitrogen) containing 10% lipopolysaccharide-free fetal bovine serum (Hyclone) and 50 M ␤-mercaptoethanol. 293 cells stably expressing AP-TNFR2 or AP-PSGL-1 were selected in 0.5 g/ml puromycin. Clonal 293 cells expressing AP-APP and Bcl-X L /CrmA were selected in 0.3 g/ml puromycin and 40 g/ml hygromycin. Transfections were carried out using LipofectAMINE 2000 (Invitrogen). In the transient transfections, the medium was replaced with fresh medium 1 day after transfection. After another overnight incubation, the conditioned medium and the cell lysate were collected. For the alkaline phosphatase measurements, aliquots of the conditioned medium were treated for 30 min at 65°C to heat-inactivate the endogenous alkaline phosphatase activity. In transient transfections, where multiple plasmids were cotransfected, luciferase or alkaline phosphatase was included to normalize the expression levels for transfection efficiencies.
To produce retroviral supernatants, plasmids encoding VSV-G, gagpol, and either P12/MMP-GFP or P12/MMP-BACE were transfected into 293 cells. The retroviral transductions were carried out using Polybrene (Sigma). Flow cytometric analysis was carried out on a BD Biosciences FACSCalibur using the indicated antibody.
To analyze the effect of PMA on shedding, the cells were incubated in fresh medium for 3 h with the addition of 1 M PMA in ethanol or with ethanol alone. For treatment with the metalloprotease inhibitor TAPI (25 M), the cells were pretreated for 45 min with inhibitor. Next the medium was replaced with fresh medium for 3 h containing the inhibitor dissolved in Me 2 SO or with Me 2 SO alone. To detect secreted and cellular PSGL-1, aliquots of lysate or conditioned medium were directly loaded onto an electrophoresis gel. The concentration of ␤-mercaptoethanol in the sample buffer was 10%. At lower concentrations PSGL-1 and its C-terminal fragments were detected not only at the monomeric but also at the dimeric apparent molecular weight. Western blot detection was carried out by using the indicated antibodies.
To study the effect of ␥-secretase inhibition on the amount of Cterminal fragments, 293 cells were transiently transfected with PSGL-1 or as a control with APP. Two days after transfection the cells were preincubated for 45 min in the presence of the inhibitor and then incubated for additional 8 h with fresh medium containing the inhibitor. The specific ␥-secretase inhibitors DAPT (1 M; kindly provided by Dr. Boris Schmidt, Darmstadt, Germany) or L-685,458 (5 M; Bachem) were dissolved in Me 2 SO. Control cells were treated with Me 2 SO alone. Aliquots of the cell lysate were analyzed by blot analysis.
Plasmid Construction-All cDNAs (PSGL-1, BACE1, BACE2, ADAM10, APP, luciferase, L-selectin, TNFR2, and mutants and fusion proteins thereof) were cloned into the expression vector peak 12. The cDNA of ADAM10 was amplified by PCR from an activated T cell library; the cDNAs of secretory alkaline phosphatase was kindly provided by Michael Brown, and the cDNA of BACE2 was kindly provided by Hyeryun Choe and Mike Farzan. The identity of all constructs obtained by PCR was confirmed by DNA sequencing. The plasmid encoding Bcl-X L and CrmA was described previously (17).
Phosphatase Assay-For alkaline phosphatase activity measurements, 200 l of reaction solution (0.1 M glycine, pH 10.4, 1 mM MgCl 2 , 1 mM ZnCl 2 containing 1 mg/ml 4-nitrophenyl phosphate disodium salt hexahydrate, Sigma) were added to 20 l of the conditioned medium. The absorbance was read at 405 nm.
Infection of Primary Neurons with Semliki Forest Virus-Cortical neurons were prepared from E14 mice embryos from BACE1 heterozygote crosses as described (18). Embryo tails were used for the genotyping.
The BACE knock out was verified by Northern and Western blot detection and by functional analysis demonstrating that APP cleavage by BACE was virtually eliminated in the neurons. 2 Preparation of recombinant Semliki Forest virus stocks has been described previously (19). Virus was diluted 1:100 in conditioned culture medium and added to 4-day-old neurons. Three hours post-infection, cells were labeled with 100 Ci/ml [ 35 S]methionine for 4 h and lysed in immunoprecipitation buffer (1% Triton X-100, 1% sodium deoxycholate, 0.1% SDS in TBS buffer). PSGL-1 full-length and CTFs were immunoprecipitated using anti-FLAG antibody. Immunoprecipitated material was separated by SDS-PAGE, and dried gels were exposed to PhosphorImager (Amersham Biosciences).
In Vitro BACE1 Cleavage Assay and Mass Spectrometry-Cell lysates from PSGL-1-expressing 293 cells were incubated overnight at 37°C with or without BACE1-containing membrane preparations in 50 mM sodium acetate, pH 4.4. Membranes from BACE1-transfected and non-transfected 293 cells were extracted according to a previously published protocol (20) using 1% Triton X-100 and STE buffer instead of DDM lysis buffer. The following protease inhibitors were included where indicated. The BACE1 inhibitor (GL189) H-EVNstatineVAEF-NH 2 was synthesized by K. Maskos and W. Bode and used at 2 M. Pepstatin A was used at 2 g/ml. and the complete protease inhibitor mixture (Roche Applied Science) was used according to the manufacturer's instructions. 20 g of the synthetic peptide AASNLSVNYPVGAPDHISVKQ-CONH2 were incubated for 1 h at 37°C with or without the purified, soluble BACE1 ectodomain in 50 mM sodium acetate, pH 4.4. The reaction mixture was purified using ZipTips (Millipore) according to the manufacturer's protocol and was directly eluted from the ZipTips with a saturated solution of ␣-cinnamic acid in 50% acetonitrile, 0.3% trifluoroacetic acid onto a stainless steel matrix-assisted laser desorption ionization target plate. Mass spectra were recorded on a Voyager DESTR matrix-assisted laser desorption ionization-mass spectrometer.

Alkaline Phosphatase Fusion Protein Assay Identifies PSGL-1 as a New BACE1 Substrate-To identify new BACE1
substrates, alkaline phosphatase (AP) fusion proteins of Lselectin, TNF-receptor 2, and P-selectin glycoprotein ligand-1 (PSGL-1) were generated. Like APP, the three proteins are known to undergo ectodomain shedding in a metalloproteasedependent manner. As a control an AP fusion protein of APP was included. All four proteins were stably expressed in human embryonic kidney 293 cells, which are widely used for studying APP processing and BACE activity. The AP activity was measured in the conditioned medium. The AP fusion proteins were shed in the same manner as the corresponding wild-type proteins (5,21,22); shedding was stimulated by the phorbol ester PMA and inhibited by the metalloprotease inhibitor TAPI (data not shown), as expected for the cleavage by a metalloprotease of the ADAM family.
Next, the cells were transfected with vectors encoding either BACE1, a catalytically inactive BACE1 mutant (BACE1 D93A) (23), the BACE1 homologue BACE2, a catalytically inactive BACE2 mutant (BACE2 D110N) (24), the metalloprotease ADAM10, which is able to cleave APP (25), or with the empty vector alone (Con, Fig. 1). Transfection of BACE1 strongly induced the shedding of the AP fusion proteins of APP and PSGL-1 but not of L-selectin and TNF receptor 2 ( Fig. 1). This result suggests that PSGL-1, but not L-selectin and TNFR2, could be a novel substrate for the protease BACE1. A catalytically inactive mutant of BACE1 (BACE D93A) (23) had no effect on the shedding of AP-APP and AP-PSGL-1, showing that the proteolytic activity of BACE1 is required for the shedding of both proteins. Interestingly, ADAM10 and BACE2 showed the same cleavage pattern as BACE1 by stimulating the shedding of APP and PSGL-1 but not of L-selectin and TNFR2 (Fig. 1).
Proteolytic Processing of PSGL-1 by BACE1-To analyze in more detail the cleavage of PSGL-1 by BACE1, we used blot analysis to study the proteolytic processing of PSGL-1 in the human monocytic cell line U937, which expresses endogenous PSGL-1, as well as in the human embryonic kidney cell line 293, which has been used for several studies of BACE1 activity. To facilitate detection of the proteolytic fragments, PSGL-1 was tagged with three small epitope tags at the N terminus of the immature protein (HA), in the ectodomain of the mature protein (AU1), and in the cytoplasmic domain (FLAG; Fig. 2A). PSGL-1 was transfected into 293 cells, where it could be detected on the cell surface by flow cytometry using an antibody against PSGL-1 (PL1) or against the AU1 tag (data not shown). In the cell lysate, the immature form and the mature, fully O-glycosylated form of PSGL-1 showed the expected molecular masses of ϳ85 and ϳ120 kDa, respectively (Fig. 2B, lane 2, FLAG blot), and were not visible in control cells not transfected with PSGL-1 (lane 1). Two types of C-terminal fragments (CTFs) of PSGL-1 were observed in the lysate (Fig. 2B, lane 2 Flag blots): two fragments with a molecular mass of ϳ20 kDa and three fragments with a molecular mass of around ϳ30 kDa. The fragments of ϳ20 kDa were dramatically enriched in a dose-dependent manner when BACE1 was overexpressed (Fig. 2B, lanes 3 and 4), suggesting that they arise through cleavage of PSGL-1 by the endogenous BACE1 of the 293 cells. Upon BACE1 overexpression a third C-terminal fragment of a slightly lower molecular weight was detected. This fragment was not detected in cells expressing endogenous BACE1, presumably because its amount was below the detection limit. The apparent molecular weight of these C-terminal fragments is nearly identical to the molecular weight of a PSGL-1 deletion mutant (PSGL-1 noecto) in which the ectodomain was replaced by the short HA epitope tag (Fig. 2C), suggesting that BACE1 cleavage occurs close to the transmembrane domain. Interestingly, similar to the PSGL-1 C-terminal fragments, PSGL-1 noecto was present as three different bands. At present it is unknown whether the three bands represent three conformations with differing electrophoretic mobility or might differ by post-translational modifications. However, the three bands have the same N terminus, since they could be detected with an antibody against their N-terminal HA tag (Fig. 2C).
The C-terminal PSGL-1 fragments of ϳ30 kDa were enriched when the metalloprotease ADAM10 was overexpressed (Fig.  2B, lane 5), and therefore most likely represent the CTFs generated by ADAM protease cleavage of PSGL-1 in the 293 cells. Thus, like those of APP, PSGL-1 CTFs of different length seem to be generated by BACE1 or by metalloproteases of the ADAM family.
Overexpression of BACE1 not only led to the increased generation of PSGL-1 CTFs but also to a reduction in quantity of mature, full-length PSGL-1. This was particularly apparent when more BACE1 plasmid was used for transfection (Fig. 2B, FLAG blot; high BACE and low BACE), suggesting that BACE1 very efficiently cleaves the mature PSGL-1. The reduction of mature, full-length PSGL-1 was accompanied by a reduction of PSGL-1 detected on the surface of 293 cells using flow cytometry (data not shown).
In the conditioned medium, the secreted ectodomain of PSGL-1 was detected with a molecular mass of ϳ120 kDa (Fig.  2B, lane 2, top panel). Additionally, smaller fragments of 80 -100 kDa were observed (Fig. 2B, top panel, PSGL-1 short), suggesting that the soluble form of PSGL-1 is subject to further proteolytic cleavage. This was particularly visible upon BACE1 transfection, where the full-length ectodomain could barely be detected (Fig. 2B, top panel, lanes 3 and 4, secreted PSGL-1), but instead an increased amount of the apparent degradation products was visible.
After the initial ectodomain cleavage the C-terminal fragments of APP and other type I membrane proteins may undergo regulated intramembrane proteolysis by ␥-secretase. For example, treatment of APP-expressing 293 cells with the two specific ␥-secretase inhibitors DAPT (26) and L-685,458 (27) strongly increases the amount of APP C-terminal fragments in the cell lysate (Fig. 2D), resulting from the inhibited turn over by ␥-secretase. In contrast, no accumulation of PSGL-1 Cterminal fragments was observed in PSGL-1-expressing 293 cells (Fig. 2D), suggesting that PSGL-1 is not a substrate for ␥-secretase.
Proteolytic Processing of PSGL-1 in U937 Cells-The monocytic U937 cell line showed essentially the same proteolytic processing of PSGL-1 (Fig. 3) as the 293 cells (Fig. 2B), including the strong reduction of full-length PSGL-1 from the cell surface of U937 cells upon transduction of BACE1 (Fig. 4). The U937 cells were retrovirally transduced with PSGL-1 and additionally with a retrovirus encoding BACE1 or GFP as a control. In a control experiment, the U937 cells were stimulated with the phorbol ester PMA, which induced the secretion of PSGL-1 (not shown), as reported previously (21) for human neutrophils.
In contrast to the findings from 293 cells, only one C-terminal fragment was visible in the lysate of U937 cells overexpressing BACE1 (Fig. 3, lane 3, Flag blot). In control cells transduced with GFP instead of BACE1, the C-terminal fragment was detected after longer exposure of the film (see Fig. 3, bottom panel), similar to the results from 293 cells (Fig. 2B, lane 2, Flag blot), suggesting that this C-terminal fragment resulted from cleavage induced by the endogenous BACE1 of the U937 cells.
Similar results were observed in the human Jurkat T cell line (data not shown), which also expresses endogenous PSGL-1. The nature and the position of the C-terminal epitope tag of PSGL-1 did not influence its proteolytic processing (data not shown).
C-terminal Fragments of PSGL-1 Are Not Generated in BACE1-deficient Cells-Next we analyzed whether the PSGL-1 CTFs were absent in BACE1-deficient cells. To this aim, we made use of BACE1 knock-out mice. 2 Although BACE1 is expressed in most tissues and cell lines including leukocytes (8,10), its highest expression is in neurons, where its detection is relatively easy. Thus, primary neuronal cultures were used for the subsequent experiments. The neurons were infected with Semliki Forest virus encoding the epitope-tagged PSGL-1. Fulllength PSGL-1 and its CTFs were detected by [ 35 S]methionine labeling and immunoprecipitation with an anti-FLAG tag antibody. In neurons of wild-type mice, PSGL-1 was processed to the same C-terminal fragment as detected in U937 and 293 cells (Fig. 5, lane 5). Coinfection of the neurons with virus encoding BACE1 led to an increased generation of the PSGL-1 FIG. 1. Secretion of AP fusion proteins of L-selectin, TNFR2, APP, and PSGL-1. 293 cells stably expressing AP-L-selectin, AP-TNFR2, AP-APP, or AP-PSGL-1 were transfected with control vector (Con) or with a plasmid encoding BACE1, a catalytically inactive BACE1 mutant (BACE1 D93A), BACE2, a catalytically inactive BACE2 mutant (BACE2 D110N) or the ADAM10 protease. Shown is the alkaline phosphatase activity in the conditioned medium. Given are the means and S.D. of at least two independent experiments. Each experiment was carried out in duplicate. CTF (Fig. 5, lane 6). In contrast, in neurons of BACE1Ϫ/Ϫ mice, no CTF of PSGL-1 could be detected (Fig. 5, lane 2). Virally induced expression of BACE1 restored the generation of the C-terminal fragment in the BACE1Ϫ/Ϫ neurons (Fig. 5,  lane 3), showing that BACE1 is required for the generation of the PSGL-1 C-terminal fragment of ϳ20 kDa.
In Vitro, BACE1 Induces the Cleavage of PSGL-1-To analyze whether PSGL-1 is directly cleaved by BACE1, the following in vitro assay was used. Cell lysates of 293 cells expressing full-length PSGL-1 were incubated with membrane extracts from control 293 cells (Fig. 6, lane 3) or from 293 cells overexpressing BACE1 (Fig. 6, lanes 4 -7). In the presence (Fig. 6,  lane 4) but not the absence (Fig. 6, lane 3) of the BACE1 extract, the same PSGL-1 CTFs were generated as in vivo (Fig.  6, lane 2). The addition of a specific BACE1 inhibitor (GL189) (28) suppressed the generation of the PSGL-1 CTFs (Fig. 6, lane 5), whereas the aspartyl protease inhibitor pepstatin A (Fig. 6, lane 7) as well as a general protease inhibitor mix against different classes of proteases had no effect (Fig. 6, lane  6). This experiment suggests that BACE1 does not act in a proteolytic cascade but directly cleaves PSGL-1.
The Ectodomain Cleavage Sites Map to the Juxtamembrane Domain-To define which domains of PSGL-1 ( Fig. 2A) are essential for the BACE1-induced cleavage, a domain deletion analysis of PSGL-1 was carried out. Although most of the deletions of the individual domains reduced the constitutive secretion of AP-PSGL-1 (Fig. 7A), all proteins were correctly transported to the cell surface as measured by flow cytometric analysis (not shown) by using an antibody against an HA epitope tag at the N terminus of the protein (29). The soluble form of AP-PSGL-1 was efficiently secreted and not detected on the cell surface. As in the experiment in Fig. 1, transfection of BACE1 induced the secretion of wild-type AP-PSGL-1 (PSGL fl), albeit to a lower extent. The difference is likely due to the stronger expression level of PSGL-1 in the transient transfection in Fig. 7A compared with the lower expression level in the stably transfected cells in Fig. 1 (data not shown). Transfection of BACE1 also strongly induced the secretion of most mutants (Fig. 7A), showing that most of the PSGL-1 ectodomain (⌬rep) and the cytoplasmic domain (⌬cyto) were dispensable for the shedding. In contrast, a fusion protein lacking the whole ectodomain including the juxtamembrane domain (⌬ecto) was not secreted even when BACE1 was overexpressed. However, the protein could be detected on the cell surface by using flow cytometry and by enzymatic activity in the cell lysate (data not shown). This deletion analysis suggests that the 52-residuelong juxtamembrane domain of PSGL-1 (Fig. 7B) is required both for its constitutive secretion and for its BACE1-stimulated secretion, which is consistent with the data available for many different membrane proteins that undergo ectodomain shedding in their juxtamembrane domain (2).
The juxtamembrane region of PSGL-1 contains an N-glycosylation site (Fig. 7B), such that the PSGL-1 CTFs may or may FIG. 2. Western blot analysis of the proteolytic processing of PSGL-1 in 293 cells. A, schematic drawing of the domain structure of the type I membrane protein PSGL-1, which is present on the cell surface as a homodimer. SP, N-terminal signal peptide; Pro, prodomain, which is cleaved upon maturation of PSGL-1, presumably by furin or a furin-like protease; Rb, receptor binding domain; decamer repeats, 15-16 stretches of 10 amino acids each, which are rich in serine, threonine, and proline and which are highly O-glycosylated (vertical black bars); JM, juxtamembrane region, where the cleavages by BACE1 and by an ADAM protease occur (determined by mass spectrometry and the glycosylation experiment in Fig. 7 and indicated by arrows); TMD, transmembrane domain; cyto, cytoplasmic domain. Also indicated are the sites where three short epitope tags (HA, AU1, and FLAG) were introduced into the PSGL-1 sequence. The lollipops indicate N-glycosylation sites. B, control vector (Mock, lane 1) or a plasmid encoding epitope-tagged PSGL-1 (but lacking alkaline phosphatase) was transfected into 293 cells (lanes 2-5). The transient transfections were carried out as cotransfections with control vector (Con) or BACE1 (BACE) or ADAM10 (AD10) and luciferase or alkaline phosphatase to normalize for transfection efficiencies. Low BACE/high BACE, 10 ng of BACE1 plasmid plus 90 ng of control vector (low BACE) or 100 ng (high BACE) of BACE1 were used for the transfection. Aliquots of the conditioned medium (top panel) and of the cell lysate (lower two panels) were directly loaded onto electrophoresis gels. After Western blotting, PSGL-1 and its CTFs were detected by using the indicated antibodies against the epitope tags AU1 and FLAG. Shown are representative blots of three independent experiments. The bottom panel shows a longer exposure of the anti-FLAG blot to visualize the PSGL-1-CTFs induced by the endogenous BACE1. CTFs (B1), CTFs arising through BACE1 cleavage; CTFs (AD): CTFs arising through ADAM cleavage; PSGL-1 short: secreted forms of PSGL-1 with a lower appar-ent molecular weight than the full-length ectodomain. C, 293 cells were transfected with the PSGL-1 noecto deletion mutant (NE) or cotransfected with PSGL-1 and BACE1 (P ϩ B). PSGL-1 noecto carries a C-terminal FLAG tag and lacks the ectodomain of PSGL-1, which is replaced by an HA epitope tag. The three bands of PSGL-1 noecto have the same apparent molecular weight as the PSGL-1 C-terminal fragments and can be detected with an antibody against their N-terminal HA tag. Mature PSGL-1 full length and its CTFs lack the HA tag and are thus not detected on the HA blot. D, 293 cells were transfected with PSGL-1, APP, or control vector (mock) and incubated for 8 h in the presence or absence of the specific ␥-secretase inhibitors L-685,458 (L, 5 M) or DAPT (1 M). Aliquots of the cell lysate were analyzed by blot analysis and probed with antibodies against the C terminus of APP (6687) or against the C-terminal FLAG tag of PSGL-1. not be glycosylated depending on whether the BACE1-or ADAM10-induced cleavage occurred N-terminally or C-terminally to the glycosylation site. To answer this question, the corresponding CTFs were treated with N-glycosidase F (Fig.  7C). Only the ADAM10-induced CTFs showed a shift in their molecular weight upon deglycosylation, revealing that their N terminus is N-terminal to the N-glycosylation site, whereas the BACE1-induced CTFs must have been generated through proteolysis between the N-glycosylation site and the transmembrane domain (indicated in the scheme in Fig. 7B). This was confirmed by cleaving a synthetic peptide spanning this region of the PSGL-1 sequence with purified BACE1. Mass spectrometric analysis identified the full-length peptide (2167 Da) and one of the cleavage products (1710 Da), which corresponds to the fragment arising through cleavage between the leucine and the serine residue (indicated in Fig. 7D). The second cleavage product of 457 Da comprising the N-terminal 5 amino acids was not detected in the mass spectra, presumably because it was lost during the purification procedure using the ZipTips. The large cleavage product of 1710 Da was not seen in control experiments in the absence of BACE1 (Fig. 7D) or when BACE1 was inhibited by the BACE1 inhibitor GL189 (not shown), clearly showing that it was generated by BACE1 cleavage. DISCUSSION BACE1 cleavage of APP catalyzes the first step in the generation of the amyloid ␤-peptide, which is deposited in the brain of Alzheimer patients. Thus, BACE1 is currently consid-ered a prime drug target for Alzheimer's disease (12). However, the biological function and additional substrates besides APP and the sialyltransferase ST6Gal I (14,30) have not been studied extensively.
Here we used an alkaline phosphatase fusion protein assay to identify additional BACE1 substrates besides APP. We tested three proteins known to undergo ectodomain shedding in a manner similar to APP (5,21,22): PSGL-1, TNF receptor 2, and L-selectin. Among these proteins we found an additional substrate for BACE1, the leukocyte adhesion protein PSGL-1. It is possible that additional BACE1 substrates could be identified among those membrane proteins, which have been reported to undergo ectodomain shedding (2,3). Like the ADAM metalloproteases, BACE1 may therefore have a broader role in the ectodomain shedding of membrane proteins, although BACE1 may have a more restricted set of substrate proteins than the ADAM proteases, because it does not cleave all proteins, which are known to undergo ectodomain shedding by ADAM proteases (e.g. TNF receptor 2 and L-selectin, as analyzed in this study). Using a PSGL-1 shedding assay in addition to an APP shedding assay may help in the identification of substances that specifically inhibit the BACE1 cleavage of APP but not of PSGL-1, ST6Gal I, or other potential BACE1 substrates.  3). Aliquots of the conditioned medium and the lysate were loaded onto electrophoresis gels, and PSGL-1 and its C-terminal fragment were detected by Western blotting using antibodies against the AU1 or the FLAG tag. The lower FLAG panel (longer exposure) is derived from a different experiment and was exposed longer than the upper FLAG panel in order to visualize the endogenously formed CTF. PSGL-1 is expressed constitutively on the surface of most leukocytes and mediates the adhesion of leukocytes to endothelial cells in a process called leukocyte recruitment. This process occurs during inflammation and tissue injury and allows leukocytes to transmigrate from the blood to the underlying tissue (15). The binding of the N-terminal ectodomain of PSGL-1 to its receptor P-selectin has been well studied and is the target for drugs aimed at reducing the inflammatory response in various disease conditions (31,32). In contrast, neither the proteolytic processing of PSGL-1 nor the role of this processing in leukocyte adhesion have been well characterized. A previous study showed that the phorbol ester PMA, which induces the ectodomain shedding of various proteins, may also induce the shedding and secretion of PSGL-1 from human neutrophils (21), suggesting the involvement of a metalloprotease. This is consistent with our finding that PMA and the overexpression of ADAM10 stimulate the shedding of PSGL-1. In addition, we find that the shedding of PSGL-1 is also mediated by BACE1. Interestingly, the BACE1 cleavage site is closer to the transmembrane domain than the putative ADAM protease cleavage site (Fig. 7B), and thus is in opposite order compared with the proteolytic processing of APP (33). However, this finding agrees well with the known properties of BACE1 and ADAM proteases. The ADAM protease cleavage of APP occurs without a strict sequence specificity but at a fixed distance from the membrane (34), whereas the BACE1 cleavage seems to be more sequence-specific but not strongly dependent on the distance of the cleavage site from the membrane (35). We found that the BACE1 cleavage site in PSGL-1 is located after the hydrophobic leucine in the amino acid sequence Asn-Leu-Ser (NL-S), which agrees well with previous in vitro studies showing that BACE1 does not have a very high sequence specificity but prefers hydrophobic residues at the P1 position (36,37). Moreover, the cleavage site in PSGL-1 (NL-S) is very similar to the corresponding cleavage site for APP carrying the Swedish mutation (NL-D) (6,10), which is linked to a familial form of Alzheimer's disease. Interestingly, both BACE1 substrates besides APP identified so far (PSGL-1 and ST6Gal I) have an important function in the immune system. Although BACE1Ϫ/Ϫ mice don't seem to have an overt phenotype (38 -40), these mice have not yet been immunologically challenged extensively, and an immunological phenotype may not have been detected. Moreover, BACE1 is only found in higher organisms and not in C. elegans FIG. 7. Mapping of the cleavage site to the juxtamembrane region of PSGL-1. A, domain deletion analysis of the shedding of PSGL-1. Domain deletion constructs of PSGL-1 tagged with an HA epitope tag and AP were transiently transfected into 293 cells. The AP activity of the secreted AP-PSGL-1 fusion proteins was measured in the conditioned medium. The transient transfections were carried out as cotransfections with control vector (ϩVector) or BACE1 (ϩBACE) and luciferase to normalize for transfection efficiencies. Given are the means and the S.D. of at least two independent experiments. Each experiment was carried out in duplicate. PSGL fl, PSGL-1 full-length; ⌬rep, deletion of the receptor binding domain and the decamer repeats; ⌬glyc, deletion of the Nglycosylation site within the juxtamembrane domain; ⌬ecto, deletion of the whole ectodomain; sol, soluble form of PSGL-1 lacking the transmembrane and the cytoplasmic domain; ⌬cyto, deletion of the cytoplasmic domain. B, amino acid sequence of the 52-residue-long juxtamembrane region (JM) of PSGL-1. The glycosylated asparagine is marked with an asterisk. Cleavage sites for BACE1 and ADAM10 are indicated (as determined by the N-glycosylation experiment in C and the mass spectrometric analysis in D). C, lysates from 293 cells expressing PSGL-1 and either BACE1 or ADAM10 were treated with N-glycosidase F (PNGaseF) to remove possible Nglycosylations from the CTFs of PSGL-1. Subsequently, aliquots of the treated lysates were directly loaded on electrophoresis gels, and the CTFs were visualized with an anti-FLAG tag antibody. D, the synthetic peptide AANLSVNYPVGAPDHIS-VKQC was incubated for 1 h at 37°C in the absence (ϪBACE1) or the presence of purified BACE1 (ϩBACE1). Mass spectrometric analysis identified the full-length peptide (2167 Da). The cleavage product (1710 Da) was only detected in the presence of BACE1. The mass difference between 2167 and 1710 Da corresponds exactly to the mass of the undetected N-terminal cleavage fragment comprising five residues. and D. melanogaster, suggesting that BACE1 may have special functions in cellular processes developed later in evolution such as the elaborate immune system of vertebrates.