β Subunits of Voltage-gated Sodium Channels Are Novel Substrates of β-Site Amyloid Precursor Protein-cleaving Enzyme (BACE1) and γ-Secretase*

Sequential processing of amyloid precursor protein (APP) by membrane-bound proteases, BACE1 and γ-secretase, plays a crucial role in the pathogenesis of Alzheimer disease. Much has been discovered on the properties of these proteases; however, regulatory mechanisms of enzyme-substrate interaction in neurons and their involvement in pathological changes are still not fully understood. It is mainly because of the membrane-associated cleavage of these proteases and the lack of information on new substrates processed in a similar way to APP. Here, using RNA interference-mediated BACE1 knockdown, mouse embryonic fibroblasts that are deficient in either BACE1 or presenilins, and BACE1-deficient mouse brain, we show clear evidence that β subunits of voltage-gated sodium channels are sequentially processed by BACE1 and γ-secretase. These results may provide new insights into the underlying pathology of Alzheimer disease.

Alzheimer disease is a progressive neurodegenerative disorder and the most common form of age-dependent dementia. The major pathological features of Alzheimer disease are senile plaques and neurofibrillary tangles, which are the deposits of amyloid ␤ peptide (A␤) 1 and hyperphosphorylated tau, respectively. It is widely accepted that the sequential processing of APP, a type I membrane protein, by ␤and ␥-secretases in the brain is crucial for the accumulation of A␤ and disease pathogenesis (1,2). Although ␤-site APP-cleaving enzyme (BACE1) has been identified to be the ␤-secretase (3-6), a growing body of evidence favors presenilins-1 and -2 as the catalytic core of ␥-secretase (7). Although the properties of both proteases as APP processing enzymes are relatively well established, the regulatory mechanisms of sequential cleavage by both proteases in neurons are not completely clear. This is partly because of the fact that APP and its family proteins are still the only substrates identified for both ␤and ␥-secretases, although a number of integral membrane proteins have been reported to be processed either by BACE1 (8,9) or ␥-secretase (10). Identifying new substrates for both ␤and ␥-secretases in neurons would therefore be useful to further explore the precise mechanism by which BACE1 and ␥-secretase function in cohort.
Recently, our laboratory has been focusing on examining the role of voltage-gated sodium channel (VGSC) ␤ in the pathogenesis of Huntington disease and the regulation of APP processing in lipid rafts. 2,3 VGSC is a large, multimeric complex that consists of an ␣ subunit and one or more ␤ subunits. To date, nine functional ␣ subunits and four ␤ subunits have been identified (11,12). Although VGSC␤ subunits are not essential to the basic operation of sodium channels, they are considered to be important auxiliary subunits, because co-expression of ␤ subunits are required to reconstitute full properties of the native sodium channel and to modify channel properties and intracellular localization (11,13). In the course of analyzing the VGSC␤, we found that these subunits are preferentially associated with Lubrol-WX-resistant membranes in the mouse brain and possess putative BACE1 cleavage sites juxtaposed to the transmembrane region (14). In this study, we explored the possibility that VGSC␤ subunits could be processed sequentially by BACE1 as well as ␥-secretase. Here we show direct evidence that these subunits are indeed novel substrates of both proteases.
Generation of BACE1 Knock-out Mice-A BACE1 knock-out mouse line was generated by inserting a neomycin expression cassette from pMC1neopA (Thomas and Capecchi, 1987) (Stratagene) into exon 1 (glutamate 19 of the BACE1 cDNA). The insertion of the neo-cassette introduces a premature translational stop codon into the open reading frame of the BACE1 gene. Successful knock down of BACE was confirmed by Southern, Northern, and Western blot analysis.
Production of End-specific VGSC␤4 Antibodies-A cleavage site-directed antibody was generated as described, with minor modifications (15). Briefly, we synthesized a peptide corresponding to the predicted N-terminal region of C-terminal fragment (CTF) ␤4 from in vitro cleavage assay (NH 2 -QVVDKLEKV) with the C-terminal Cys. The peptide was conjugated with keyhole limpet hemocyanin and injected into rabbits. For antibody purification, a shorter peptide, NH 2 -QVVDKLC, and acetyl-QVVDKLC were synthesized and coupled with SulfoLink coupling gel (Pierce). The antiserum was purified by affinity chromatography on the immobilized NH 2 -QVVDKLC. To obtain N-terminal endspecific antibody, the affinity-purified antibody was further adsorbed against the immobilized acetyl-QVVDKLC. We designated these new antibodies as QV6 antibody.
Preparation of Detergent-soluble Brain Membrane Extracts-Various regions of the mouse brain were dissected from BDF1 (wild type) or C57B6 ϫ 129SV (BACE1-deficient) mice, and the tissues were homogenized in 10 volumes of dissection buffer (50 mM Tris, pH 7.5, 1 mM EDTA, 1 mM EGTA supplemented with 1ϫ Complete TM ). The homogenate was centrifuged at 30,000 ϫ g for 10 min, and the supernatant was discarded. The pellet was rehomogenized in 10 volumes of the same buffer and recentrifuged as above. The final pellet was resuspended in 10 volumes of dissection buffer supplemented with 100 mM NaCl and 1% Triton X-100 and incubated at 4°C for 30 min with constant rocking. Detergent-soluble protein extract was collected as the supernatant after spinning at 40,000 ϫ g for 20 min.
Preparation of Detergent-resistant Membranes (DRMs) from Primary Cultures-Primary cortical neuron cultures were lysed with 1% Lubrol WX at 4°C in MES-buffered saline (pH 6.5) supplemented with Complete TM protease inhibitor mixture (Roche Applied Science). After lysis, the detergent-resistant membranes were prepared essentially as previously described (16).
RNA Interference-BACE1 small interfering RNA (siGENOME TM SMARTpool, Dharmacon) was transfected into MEFs or H4 cells with Lipofectamine reagent (Invitrogen) according to the manufacturer's instructions. The cells were collected for further analysis after 2-3 days.
Phosphatase Assay-Full-length VGSC␤ subunit cDNAs were cloned into pAPtag-4 (GenHunter Corporation) to express VGSC␤ subunits fused with soluble human placental alkaline phosphatase (Alp) at their N termini (AP-␤). Each AP-␤ and BACE1 were transfected into HEK cells with ␤-galactosidase to normalize the expression levels. Alkaline phosphatase activity in the medium was measured after 24 h in essentially the same way as described previously (9).
In Vitro Cleavage Assay and Matrix-assisted Laser Desorption Ionization Time-of-Flight (MALDI-TOF) Mass Spectrometry Analysis-Synthetic peptides (50 M) in 0.1 M sodium acetate, pH 4.5, were incubated with 25 g/ml recombinant human BACE1 (R & D Systems) at 30°C for 12 h. Monoisotopic mass values of the peptides were measured by a Biflex III MALDI-TOF mass spectrometer (Bruker Daltonics) operated in the reflector mode. The peptide solution was desalted with ZipTip C18 (Millipore) and co-crystallized with equal volumes of 10 mg/ml ␣-cyano-4-hydroxycinnamic acid (Fluka) matrix in 50% acetonitrile/ 0.1% trifluoroacetic acid. All MALDI spectra were calibrated externally using a peptide standard. Cleavage sites were searched by calculating monoisotopic masses of possible peptides with PAWS software (Genomic Solutions, 65.219.84.5/paws.html) based on the peptide sequence used as a substrate with Ϯ0.1% mass tolerance.
Immunofluorescence Microscopy-Cells grown on coverslips or growth chambers were fixed with 2% paraformaldehyde in PBS at room temperature for ϳ3 min. The cells were then blocked with 5% nonfat milk for 1 h in PBS, washed extensively, and incubated with the appropriate primary antibody at 4°C overnight. After primary antibody incubation, the cells were washed three times with PBS and subsequently incubated with fluorescence-conjugated secondary antibodies at room temperature for 1 h in the dark. The cells were finally washed three times with PBS and mounted in VECTASHIELD® (Vector) and then analyzed by confocal microscopy (the spectral confocal scanning system TCS SP2 from Leica).
Western Blotting Analysis-Cells that were ready to be analyzed by Western blotting were rinsed with cold PBS and lysed in an appropriate volume of lysis buffer (8.6% sucrose, 1 mM EDTA, 1 mM sodium orthovanadate (Na 3 VO 4 ), 10 mM sodium fluoride (NaF), 50 mM Tris, 0.038% EGTA, 1% Triton X-100 and Complete TM ). Protein concentration was measured with the bicinchoninic acid Protein Assay Kit (Sigma) according to the manufacturer's instructions. Appropriate volumes of protein lysates were then mixed with 4ϫ SDS gel-loading buffer (200 mM Tris-HCl, pH 6.8, 400 mM dithiothreitol, 8% SDS, 40% glycerol, and an appropriate amount of bromphenol blue), boiled at 100°C for 5 min, and resolved in 7.5-15% polyacrylamide gels (Atto). Proteins were subsequently transferred onto polyvinylidene difluoride membrane (0.22 m, Schleicher & Schuell), and the membrane was blocked with 5% nonfat milk in TBS-T (10 mM Tris, pH 7.5, 50 mM NaCl, and 0.1% Tween 20, Sigma) for 1 h. The blots were then incubated with appropriately diluted primary and horseradish peroxidase-conjugated secondary antibodies, each for 1 h at room temperature, with extensive washing with TBS-T after each incubation. Finally, the blots were developed with ECL solutions and exposed onto Hyperfilm TM MP (Amersham Biosciences).
Others-Band intensities on immunoblots were either quantified by Im-ageJ (version 1.32j, NCBI) after scanning with an EPSON ES-8000 scanner or captured by the FUJIFILM LAS-1000ϩ system and analyzed using image analyzing software. The value obtained from one data point was divided by the value obtained for its own internal control (␤-tubulin) and set to 100%. Ratios from other data points were compared accordingly and expressed as relative ratio of change. All significance between two sets of data was determined by unpaired Student's t test.

VGSC␤ Subunits Are Preferentially Associated with Lipid
Rafts-DRMs prepared with Lubrol-WX (Lubrol rafts) have been shown to be especially useful for the analysis of APP processing, because BACE1 and ␥-secretase can be efficiently recovered (16,17). To confirm the association of VGSC␤ subunits with Lubrol rafts in neuronal cells, we fractionated mature postnatal primary cortical neuron cultures (4 weeks) with Lubrol-WX and examined their expression patterns by Western blotting (WB). As shown in Fig. 1A, each of the VGSC␤ subunits was substantially and prominently recovered in fraction 3 of the preparation that was also enriched in prion protein, a marker of lipid raft domains, as well as BACE1, CTFs of presenilin-1/2, and the mature form of APP. These results indicate that VGSC␤ subunits are preferentially associated with lipid rafts in neurons. Indeed, based on the preferential amino acid residues flanking the BACE1 cleavage site (14), we found sev-eral putative BACE1 cleavage sites juxtaposed to the transmembrane region on each VGSC␤ subunit (Fig. 1B, arrows  (2)). The recovery of these proteins in the prion protein-enriched fraction and the presence of putative BACE1 cleavage sites led us to examine whether VGSC␤ subunits are physiological substrates for both proteases. Primary neurons were lysed with Lubrol-WX and fractionated on a sucrose gradient (see "Experimental Procedures"). Equal volumes from each fraction were resolved by 15% SDS-PAGE, and blots were probed with the antibodies indicated in the diagram. GM130, a cis-medial Golgi SNARE protein, serves as a marker for non-raft fractions. The blots were also stained with 0.25% Coomassie Brilliant Blue to show protein loading. To represent the relative distribution of each protein graphically, total protein expression was set at 100%, and the expression of each protein in any individual fraction was compared with the total accordingly. B, potential cleavage sites of BACE1 on each VGSC␤ subunit. Mouse VGSC␤ subunits were aligned using ClustalW (EMBL-EBI). Arrows indicate the potential proteolytic sites by BACE1, and boxes represent the transmembrane domain. C-terminal peptides used to generate antibodies are underlined. VGSC␤ Subunits Are Novel Substrates of BACE1-First we introduced FL VGSC␤ subunits into human embryonic kidney (HEK) cells in the presence or absence of human BACE1 to examine the processing pattern of each VGSC␤ subunit by WB with C-terminal-specific VGSC␤ antibodies generated in our laboratory ( Fig. 2A). For VGSC␤1, -2, and -4, introduction of FL-VGSC␤ subunits revealed strong expression at ϳ37 kDa, and co-expression of BACE1 dramatically reduced the expression of FL-VGSC␤ with a concomitant production of smaller C-terminal fragments (CTF␤s). In contrast, strong expression of CTF␤3 was detected without BACE1 transfection, and coexpression of BACE1 reduced the expression of FL-VGSC␤3 and CTF␤3. This implies that perhaps CTF␤3 could be processed further by other protease(s) in HEK cells. Experiments were repeated with VGSC␤ constructs tagged with V5 at the C terminus, and essentially the same pattern was observed ( Fig.  2A, V5). Because we were unable to detect smaller CTF␤3 by WB upon the addition of BACE1, VGSC␤3 was not analyzed in some of the later experiments.
Next, the specificity of this cleavage was confirmed by knockdown (Fig. 2B) and knock-out experiments (Fig. 2C). As shown in Fig. 2B, using a cell line that expresses high levels of endogenous BACE1 (HTB-148/H4), small interfering RNA-mediated BACE1 suppression significantly reduced (about 80%) the production of CTF␤1, -2, and -4 with an accumulation of the FL proteins. Similar to Fig. 2B, the lack of BACE1 in MEFs led to a reduction of CTF␤ expression (Fig. 2C). In summary, these data suggest that the proteolysis of FL-VGSC␤ subunits and production of CTF␤s are correlated to the expression and/or activity of BACE1.
We have now confirmed that BACE1 is able to cleave FL-VGSC␤ subunits, but whether the extracellular domain of VGSC␤ subunit is shed is unknown. To this end, we generated N-terminal truncated VGSC␤ constructs with (␤1-␤4 -1) or without (␤1-␤4 -2) putative BACE1 cleavage sites (Fig. 3A). We introduced these constructs into HEK cells and examined the expression of VGSC␤, first, with anti-FLAG and anti-c-Myc antibodies by WB. Of note, only results for the VGSC␤4 subunit are shown here, because similar results were obtained for VGSC␤1, -2, and -3. As shown in Fig. 3B, the introduction of ␤4 -1 or -2 alone gave strong FLAG and c-Myc immunoreactivities, and the co-expression of BACE1 with ␤4 -1 led to the complete disappearance of FLAG expression. These results suggest that the FLAG-tagged extracellular domain was processed and released (more evidence to be shown in Fig. 3D). As a result of this cleavage, the expression of full-length ␤4 -1 (Fig.  3B, the upper c-Myc band, middle panel) was largely reduced with a concomitant increase in truncated ␤4 -1 (Fig. 3B, the lower c-Myc band, middle panel). In contrast, co-expression of BACE1 with ␤4 -2 did not lead to any obvious change in FLAG or c-Myc immunoreactivity. Second, to visualize the expression of these proteins with respect to their cell surface expression, we performed immunocytochemistry on non-permeabilized HEK cells with 1) anti-FLAG, 2) N-terminal end-specific CTF␤4 antibodies (the QV6 antibody) that only detect CTF␤4 FIG. 2. VGSC␤ subunits are substrates of BACE1. A, processing of full-length VGSC␤ subunits by BACE1. Cell lysates from HEK cells transiently transfected with GFP or each VGSC␤ subunit with or without BACE1 were collected, resolved on a 15% SDS-PAGE gel, and blots were probed with antibodies as indicated. The generation of CTF␤s was confirmed by anti-V5. ␤-tubulin serves as a loading control in this experiment. B, inhibition of endogenous BACE1 expression reduces the production of CTF␤s of VGSC␤ subunits. HTB-148/H4 cells were first transiently transfected with each VGSC␤ subunit and allowed to express the VGSC␤ subunit for 24 h. One hundred nM BACE1 small interfering RNA was subsequently applied to the cells, and the expression of CTF␤s was analyzed after 48 h by immunoblots. Expression of CTF␤s and BACE1 was quantified, normalized to that of ␤-tubulin, and presented graphically in the lower panel. Inhibition of endogenous BACE1 enhanced the accumulation of full-length protein. Mean Ϯ S.E., n ϭ 3 from three independent experiments is shown. C, reduction of expression of CTF␤s in BACE1-deficient MEFs. GFP serves as a transfection control. Expression of CTF␤s was quantified, normalized to that of GFP, and presented graphically in the right panel. Mean Ϯ S.E., n ϭ 3 from three independent experiments. *, p Ͻ 0.05, (Student's t test).
but not FL-␤4, and 3) anti-BACE1 antibodies that recognize the N-terminal extracellular region of BACE1. As shown in Fig.  3C [aЈ], there were virtually no FLAG immunoreactivities detected on the cell surface membrane when ␤4 -1 and BACE1 co-expressed at the same time, indicating that amino acid residues 143-161 contained a putative BACE1 cleavage site. To show the processing in an alternative way, we made use of the newly generated QV6 antibody that detected only the CTF␤4 produced by BACE1 (see Fig. 3C, right lower panel, for specificity). Using this cleavage site-directed antibody, we found a co-localization between cleaved ␤4 and BACE1 (Fig. 3C,  [b-bЉ]). In contrast, FLAG-labeled ␤4 -2 proteins showed in- tense co-localization with BACE1 on the cell surface, (Fig. 3C, [c-cЉ]), and there was virtually no immunoreactivities detected by the QV6 antibody (Fig. 3C, [d]). Furthermore, to show the N termini of VGSC␤s were actually released from the cell surface, we fused an Alp tag to the N terminus of each FL-␤ subunit and measured the Alp activity in the medium 24 h after transfection. As shown in Fig. 3D, there was a significant increase of Alp activity in the conditioned medium upon co-expression of BACE1 and Alp-tagged FL VGSC␤, compared with the VGSC␤ alone-transfected condition. In summary, these data clearly indicate that the sequence adjacent to the transmembrane domain on each VGSC␤ subunit contains a putative BACE1 cleavage site(s), and the N-terminal part of VGSC␤ is shed and released similar to that of APP.
Confirmation of the Predicted Putative BACE1 Cleavage Sites on VGSC␤ Subunits-To demonstrate that BACE1 is able to cleave the VGSC␤ subunit directly and to confirm the putative cleavage sites that we predicted in Fig. 1B on each VGSC␤ subunit, we employed an in vitro cleavage assay using the recombinant BACE1 catalytic domain and peptides corresponding to the extracellular domain of each VGSC␤ subunit containing putative BACE1 cleavage sites (␤1, amino acids 122-160; ␤2, amino acids 128 -157; ␤3, amino acids 121-159; ␤4, amino acids 132-161) (Figs. 3A and 4). Peptides were synthesized, purified by high pressure liquid chromatography, and incubated with or without recombinant BACE1. The APP peptide (662-679, KTEEISEVKMDAEFRHDS) was also included in this assay to confirm the activity of recombinant BACE1 (data not shown). The resulting peptides were analyzed by MALDI-TOF mass spectrometry. It is worth noting that only the masses that can be assigned to the sequence of each peptide were labeled in the MALDI spectra (Fig. 4, A-D), and we cannot exclude the possibility that other relatively minor processing sites exist. Furthermore, some of the peptides may not be properly recovered in this assay (for example, the TSVVSE peptide for VGSC␤3). As shown in Fig. 4, peptides incubated without BACE1 (Fig. 4, A-D, Con) 4E). Co-incubation of BACE1 with ex-VGSC␤ peptides (Fig. 4,  A-D, ϩBACE1) led to the production of new peaks, indicating that ex-VGSC␤ peptides were processed by BACE1. Fig. 4E summarizes all of the peptide fragments and cleavage sites on each ex-VGSC␤ peptide that are preferentially processed by BACE1. Importantly, all of the processing sites revealed by this in vitro cleavage assay are included in the putative BACE1 cleavage sites that we predicted and indicated in Fig. 1B.
CTF␤s of VGSC␤ Subunits Are Further Processed by ␥-Secretase-Next we tested whether CTF␤s could be further processed by ␥-secretase similar to that of the APP family (1,18). As a first step to examining whether CTF␤s are substrates of ␥-secretase, we incubated HEK cells transfected with VGSC␤ subunits and BACE1 with a ␥-secretase inhibitor, DAPT. As shown in Fig. 5, A and B, co-incubation with DAPT led to a significant accumulation of CTF␤1, -2, and -3. In contrast, a small but consistent increase was observed for CTF␤4. Similar results were also obtained by the addition of another ␥-secretase inhibitor, L-685, 458 (data not shown). Overexpression of BACE1 also led to the production of In each MALDI spectrum, numbers in brackets denote the corresponding amino acid number of peptides derived from each VGSC␤ subunit, whereas figures below the brackets represent the monoisotopic mass of peptide ions in m/z. Newly generated peaks were labeled with arrows. ϩϩ represents doubly protonated peptide ions. E, schematic diagrams summarizing the peptide bonds that are preferentially processed by BACE1 in vitro on each ex-VGSC␤ peptide. Numbers with two decimal places denote the calculated mass of the peptides. small immunoreactive bands for VGSC␤1, -2, and -4, which were DAPT-inhibitable (Fig. 5C). Overexpression indicates that these smaller immunoreactive bands were possibly the intracellular domains and that their production depended on the ␥-secretase activity. To further confirm the role of ␥-secretase in processing these CTF␤s, we transfected BACE1 and VGSC␤ subunits into the MEFs that are either deficient in presenilin-1 (PS1 knockout), presenilin-2 (PS2 knock-out) or both presenilins (double knock-out) (19). Compared with that of wild type MEFs, the lack of PS1 increased the expression of all CTF␤s being examined by 2-fold, whereas the absence of PS2 resulted in a 1.5-2-fold increase in the expression of CTF␤1 and CTF␤3 and a 3-4-fold increase in CTF␤2 and CTF␤4. Interestingly, the lack of both PS1 and PS2 (double knock-out) produced a synergistic accumulation of all CTF␤s, because a Ͼ6-fold increase was observed (Fig.  5D, and quantified in E). This observation is supported by pub-  's t test). C, inhibition of putative intracellular domain expression by ␥-secretase inhibitor. Mouse Neuro2a cells were transfected with various VGSC␤-V5 subunits and BACE1 as indicated in the diagram. Lactacystin (Lac) at 10 M and DAPT at 1 M were added to the cells 4 h and 36 h after transfection, respectively. Cell lysates were collected after 48 h, and expression of VGSC␤ subunits was analyzed using anti-V5 with Western blotting. D, presenilins are required to further process CTF␤s. Mouse embryonic fibroblasts (MEFs) deficient in PS1, PS2, or both were transiently transfected with BACE1 and each VGSC␤ subunit. The expression levels of CTF␤s were analyzed by immunoblots with the corresponding anti-VGSC␤. Blots were reprobed with anti-PS1 and -PS2 to confirm the lack of presenilins. wt, wild type; KO, knock-out; DKO, double knock-out. Asterisks indicate nonspecific bands. E, quantification of the expression of CTF␤s in D. Mean Ϯ S.E., n ϭ 3 from three independent experiments. F, accumulation of CTF␤2 and -4 in mature primary neurons in the presence of ␥-secretase inhibitor. Pure primary neuron cultures (6 weeks old) were incubated with 5 M DAPT for 10 h before collection of protein lysates. Expression levels of CTF␤ were subsequently analyzed with the corresponding anti-VGSC␤ by immunoblots and quantified, normalized, and presented graphically on the right panel. Mean Ϯ S.E., n ϭ 3 from three independent experiments. *, p Ͻ 0.05 (Student's t test). lished data demonstrating that PS1 and PS2 are functionally redundant and the complete deficiency in both presenilins results in a more severe phenotype (20). Furthermore, as shown in Fig.  5F, the addition of DAPT also led to a significant accumulation of CTF␤2 and CTF␤4 (2-2.5-fold increase, right panel) in the mature primary cortical neuron cultures, indicating that these Cterminal fragments were also processed in a similar manner in neurons (of note, there were no CTF␤1 and CTF␤3 detected in primary neurons). These results confirm that CTF␤2 and CTF␤4 are present in neuronal cells and suggest that the inability of ␥-secretase inhibitors to produce a significant accumulation of CTF␤4 in Fig. 5A is likely due to the inhibitor-specific effect on VGSC␤4 in the overexpression system.
In Vivo Cleavage of VGSC␤ by BACE1-The lack of CTF␤1 and CTF␤3 expression in the primary neurons led us to investigate the processing of VGSC␤ subunits in the mouse brain. First, we examined the expression pattern of VGSC␤ subunits in protein lysates prepared from wild type mouse brain, because it has high BACE1 activity. Essentially the same as in the primary neuron system, CTF␤2 and CTF␤4 (but not CTF␤1 and CTF␤3) were readily detected in protein extracted from the cerebral cortex (CX), cerebellum (CB), spinal cord (SP), and subcortical area (SC) of adult mice (Fig. 6A, left panels). Next, to confirm that BACE1 is the protease that cleaves VGSC␤2 and VGSC␤4 in vivo, we examined the expression levels of CTF␤2 and CTF␤4 in the cerebrum (Fig. 6A, CERE) and striatum (STR) of BACE1-deficient mice. Consistent with results obtained in our in vitro overexpression system, the absence of BACE1 in the mouse brain led to reduced expression of both CTF␤2 and CTF␤4 (Fig. 6A, right panels, and quantified in B). We also detected a reduction of CTF␤4 expression in the cerebrum and striatum in the absence of BACE1 using the QV6 antibody. These results indicate that BACE1 is indeed the protease that cleaves VGSC␤2 and VGSC␤4 in the nervous system. Other processing enzyme(s) may exist, as the expression of CTF␤2 and CTF␤4 is not completely absent in the BACE1-deficient mouse brain. Furthermore, the reduction of CTF␤4 expression in the striatum also led to a consistent and significant increase (20 -25%) of FL striatal VGSC␤4 in the BACE1 knock-out mice (Fig. 6C). The lack of significant increase for that of cerebral ␤2/4 and striatal ␤2 might be explained by the fact that CTF␤ expresses at relatively low levels in the mammalian nervous system (compare with the overexpression system); therefore a 40% reduction in expression levels may not be able to be reflected via a significant increase of the full-length protein. It may also be that the processing of VGSC␤ subunits is limited to specific types of neuronal populations in the cerebrum, thus an overall significant increase was not easily observed. In summary, all of the results indicate that VGSC␤ subunits are indeed the physiological substrate of BACE1 in the mammalian brain. DISCUSSION In this article, we report the finding of VGSC␤ subunits as a second group of substrates that can be sequentially processed by both BACE1 and ␥-secretase. First, we found that VGSC␤ subunits, similar to BACE1 and PS1/2, associated with DRMs prepared with Lubrol-WX and that each VGSC␤ subunit contained putative BACE1 cleavage sites (Fig. 1). We subsequently employed an overexpression system to show that the production of CTF␤s, resulting from overexpression of BACE1, could be specifically recognized by antibodies raised against the C terminus of each VGSC␤ subunit. The reduction or lack of BACE1 expression led to the reduction of CTF␤ expression (Fig. 2). We further showed that N termini of VGSC␤ subunits were released from the cell surface upon BACE1 expression (Fig. 3) and determined the putative cleavage sites that are preferentially recognized by BACE1 through an in vitro cleavage assay (Fig. 4), confirming the putative cleavage sites that we predicted in Fig. 1. Interestingly, in addition to being the substrates of BACE1, VGSC-CTF␤s were further processed by ␥-secretase activity, as demonstrated in the overexpression system, presenilin knock-out MEFs, and primary neurons (Fig.  5). More importantly, we confirmed the processing of VGSC␤2 and VGSC␤4 by BACE1 in the mammalian nervous system by FIG. 6. In vivo cleavage of VGSC␤. A, the expression pattern of VGSC␤ in the wild type and BACE1-deficient mouse brain illustrated by anti-VGSC␤ or QV6 antibody. CERE, cerebrum; CX, cerebral cortex; SC, subcortical area; CB, cerebellum; SP, spinal cord; STR, striatum; wt, wild type; ko, knock-out. Tubulin served as a loading control. B, quantification of the expression of CTF␤s in A. Mean Ϯ S.E., n ϭ 3 from three independent BACE1 knock-out mice. C, lack of BACE1 expression in the mouse brain leads to an accumulation of full-length VGSC␤4 in the striatum. Expression of striatal VGSC␤4 was quantified, normalized to that of tubulin, and presented graphically in the lower panel. Note that cerebral ␤2/4 and striatal ␤2 did not produce a significant increase, thus their quantifications were not shown. SE, short exposure. Mean Ϯ S.E., n ϭ 3 from three BACE1 knock-out mouse brains. *, p Ͻ 0.05 (Student's t test). employing newly generated BACE1 knock-out mice and the in vivo cleavage site detected by the QV6 antibody (Fig. 6).
Up to this moment, no protein is found to be processed sequentially by both ␤ and ␥-secretases in a manner similar to that of APP family proteins (1,2,18,(21)(22)(23). In keeping with the current belief that BACE1 recognizes substrate sequences based on residue preference instead of stringent consensus sequence, we initially predicted and subsequently confirmed by an in vitro cleavage assay that each VGSC␤ subunit contains putative BACE1 cleavage sites juxtaposed to the transmembrane region. Our findings correlate well with previous studies, demonstrating the residue preference for subsites of BACE1 using combinatorial inhibitor libraries (14,24); that is, BACE1 preferentially recognizes bulky hydrophobic residues, such as leucine (for VGSC␤1, -2, and -4) or phenylalanine (for VGSC␤3) at the P 1 site. Indeed, as an additional group of BACE1 substrates, the processing site of each VGSC␤ subunit is also comparable with that of rat ␤-galactoside ␣2,6-sialyltransferase (ST6GalI) and P-selectin glycoprotein ligand-1, the two known BACE1 substrates that are also cleaved by BACE1 after leucine residue (8,9,25,26). Furthermore, it is also worth noting that other protease(s), such as BACE2 (27,28), may also process VGSC␤ at the same (or very close) site, because the lack of BACE1 in MEFs and mouse brain only led to the reduction of CTF␤ production (Figs. 2C and 6, A and B). Whether BACE2 is another processing enzyme for VGSC␤ subunits requires further investigation.
We think the processing of voltage-gated sodium channels by BACE1 is important, because VGSCs are one of the most fundamental and abundant types of ion channels that are responsible for the initiation and propagation of action potentials in neurons. Recently, it has been proposed that neural activity is able to regulate the production of A␤ through ␤and ␥-secretase and that the A␤ depresses synaptic transmission and, hence, suppresses neuronal activity (29). Given that CTF␤2 and CTF␤4 are expressed in various areas of the nervous system (Fig. 6A), the turnover of membrane-localized functional sodium channels by sequential processing by BACE1 and ␥-secretase in wild type neurons may be involved in such a feedback mechanism as to modulate the neuronal activity and endogenous A␤ production. It would therefore be interesting to further investigate the involvement and significance of this sequential proteolytic system in activity-dependent processing of APP.
On the other hand, studies from BACE1 knock-out mice demonstrated that complete loss of BACE1 activity in the brain does not result in a grossly abnormal phenotype (30,31), therefore inhibition of BACE1 activity is generally accepted to be an attractive way to control amyloid formation. The finding that VGSC␤ subunits are readily processed by BACE1 may pose a potential threat to this approach as a genuine therapy that is free from mechanism-based toxicity. Although the exact physiological role of this cleavage is still largely unknown, we speculate that the increase of BACE1 activity in the Alzheimer diseased brain (32, 33) may affect the normal physiology of brain sodium channels and result in the neurochemical deficits and behavioral abnormalities that have been reported in a BACE1 transgenic mouse model (34). Furthermore, it is also possible that long term suppression of BACE1 activity may cause other more subtle cognitive and behavioral changes. As it is likely that other physiological substrates for both BACE1 and ␥-secretase exist, it is therefore important to develop therapeutic strategies that specifically target APP processing and A␤ production to prevent the possible adverse side effects that might have arisen in Alzheimer disease patients.