Alcadein Cleavages by Amyloid β-Precursor Protein (APP) α- and γ-Secretases Generate Small Peptides, p3-Alcs, Indicating Alzheimer Disease-related γ-Secretase Dysfunction*

Alcadeins (Alcs) constitute a family of neuronal type I membrane proteins, designated Alcα, Alcβ, and Alcγ. The Alcs express in neurons dominantly and largely colocalize with the Alzheimer amyloid precursor protein (APP) in the brain. Alcs and APP show an identical function as a cargo receptor of kinesin-1. Moreover, proteolytic processing of Alc proteins appears highly similar to that of APP. We found that APP α-secretases ADAM 10 and ADAM 17 primarily cleave Alc proteins and trigger the subsequent secondary intramembranous cleavage of Alc C-terminal fragments by a presenilin-dependent γ-secretase complex, thereby generating “APP p3-like” and non-aggregative Alc peptides (p3-Alcs). We determined the complete amino acid sequence of p3-Alcα, p3-Alcβ, and p3-Alcγ, whose major species comprise 35, 37, and 31 amino acids, respectively, in human cerebrospinal fluid. We demonstrate here that variant p3-Alc C termini are modulated by FAD-linked presenilin 1 mutations increasing minor β-amyloid species Aβ42, and these mutations alter the level of minor p3-Alc species. However, the magnitudes of C-terminal alteration of p3-Alcα, p3-Alcβ, and p3-Alcγ were not equivalent, suggesting that one type of γ-secretase dysfunction does not appear in the phenotype equivalently in the cleavage of type I membrane proteins. Because these C-terminal alterations are detectable in human cerebrospinal fluid, the use of a substrate panel, including Alcs and APP, may be effective to detect γ-secretase dysfunction in the prepathogenic state of Alzheimer disease subjects.

p3-Alc ␥ were not equivalent, suggesting that one type of ␥-secretase dysfunction does not appear in the phenotype equivalently in the cleavage of type I membrane proteins. Because these C-terminal alterations are detectable in human cerebrospinal fluid, the use of a substrate panel, including Alcs and APP, may be effective to detect ␥-secretase dysfunction in the prepathogenic state of Alzheimer disease subjects.
Alcadein (Alc) 5 proteins comprise a family of evolutionarily conserved, type I membrane proteins that are predominantly expressed in neuronal tissues. Alc has been independently identified as a binding protein for the neuron-specific adaptor protein X11L (X11-like) (1) and as a postsynaptic Ca 2ϩ -binding protein, where it is known by the name calsyntenin (2). Alc functions as a cargo-receptor for the kinesin-1 motor that mediates anterograde transport of APP (3,4), and a mutation in a nematode ortholog of the Alc gene is reported to cause a defect in associative learning (5,6). Thus, Alc plays important roles in vesicular transport at the subcellular level and in learning behavior at the organismal level. Alc exists as four isoforms in mammals: Alc ␣1 (971 amino acids in humans), Alc ␣2 (981 amino acids in humans), Alc ␤ (956 amino acids in humans), and Alc ␥ (955 amino acids in humans) (1). Alc ␣ , Alc ␤ , and Alc ␥ are encoded by independent genes, whereas Alc ␣1 and Alc ␣2 are splice variants derived from the Alc ␣ gene.
In neurons, Alc proteins are complexed to X11L molecules, which, in turn, are complexed with the amyloid ␤-precursor protein (APP), a type I transmembrane protein that is believed to play a seminal role in the pathogenesis of familial and sporadic Alzheimer disease (reviewed for AD in Refs. 7-9 and for X11L in Refs. 10 and 11). In the absence of X11L, both Alc and APP proteins are rapidly cleaved in a coordinated manner (12). Levels of the endogenous APP metabolite, amyloid-␤ protein (A␤), are elevated in the brains of X11L-deficient mice, indicating that the APP-X11L interaction is physiologically important in the regulation of APP metabolism in the brain in vivo (13,14). Alc proteins are also cleaved successively by secretases and release soluble Alc ectodomain (sAlc, corresponding to the soluble APP ectodomain (sAPP)) and p3-Alc (corresponding to the APP fragment, p3) (12). Taken together with similarities and/or identities in their structure, cellular distribution, and neural function, the physiological and pathophysiological metabolic fate of Alc would be predicted to parallel that of APP (1,3,12).
In this study, we report that all three members of the Alc family (Alc ␣ , Alc ␤ , and Alc ␥ ) are cleaved by ADAM 10 and ADAM 17, which have been identified as the ␣-secretases for APP (15)(16)(17). Subsequent cleavage of the remaining Alc C-terminal fragments involves the presenilin-1 (PS1)-dependent ␥-secretase, and this reaction liberates into cell-conditioned medium and into cerebrospinal fluid (CSF) a short peptide, p3-Alc, previously designated "␤-Alc." Our other analysis using CSF from three groups of human subjects (n ϭ 158) indicates that p3-Alc ␣ variant ratio (minor p3-Alc ␣ 38/major p3-Alc ␣ 35) correlated with the A␤42/40 ratio in the sporadic AD (clinical dementia rating 0.5 ϩ 1 patients) but not elderly non-demented and other neurological disease controls. 6 Therefore, the detailed biochemical analysis for the cleavages of Alc proteins is significant for understanding the features of p3-Alc peptides in human subjects. We found that various FAD-linked PS1 mutations appeared, at different magnitudes, with the alterations of C termini of p3-Alcs, suggesting that one type of ␥-secretase dysfunction appears in various phenotypes upon cleavage of Alc proteins and APP. In other words, one type of ␥-secretase dysfunction largely alters the cleavage of one Alc species and APP, but the same dysfunction slightly alters the cleavage of another Alc species. When the cleavage phenotypes appear on APP to increase pathogenic A␤42, the subject may experience the onset of AD. Testing the hypothesis that AD-related variant processing of p3-Alc peptide might yield surrogate markers for ␥-secretase dysfunction is important. In this study, we characterized all p3-Alc species generated from Alc ␣ , Alc ␤ , and Alc ␥ in detail.
Cells, Transfection, and Western Blot Assay-Mouse embryonic fibroblasts (MEFs) derived from ADAM 10 homozygous (Ϫ/Ϫ) and heterozygous (ϩ/Ϫ) gene knock-out mice were described previously (20). HEK293, Neuro 2a, and MEFs (0.3-1.0 ϫ 10 6 ) were subjected to gene transfection with the indicated amounts of various combinations of plasmids in Lipofectamine 2000 or Lipofectamine according to the manufacturer's protocol (Invitrogen). After transfection for 24 h, the medium was changed, and the cells were cultured for a further ϳ24 h. In order to analyze secreted proteins, sAlc and sAPP were recovered from the conditioned medium by immunoprecipitation with an anti-FLAG antibody and Protein G-Sepharose. To analyze cellular proteins, the cells were harvested and lysed in Hepes-buffered saline with Triton X-100 (12). The cell lysates and the immunoprecipitates were analyzed by Western blotting with the indicated antibodies, detected by ECL (GE Healthcare), and quantified using the VersaDoc imaging System (Bio-Rad).
Antibodies-Anti-Alc ␣ polyclonal rabbit UT135 antibody was raised against a peptide that was composed of Cys plus the sequence between positions 839 and 851 (NPHPFAVVPSTATϩC) of human Alc ␣1 . The anti-Alc ␣ monoclonal antibody 3B5 was raised against a peptide that was composed of Cys plus the sequence between positions 821 and 826 (CϩFVHPEH). The anti-Alc ␤ polyclonal rabbit UT143 antibody was raised against a GST-fusion protein containing the sequence between 819 and 847 (FLHRGHQPPPEMAGH-SLASSHRNSMIPSA) of human Alc ␤ . The anti-Alc ␥ polyclonal antibody UT166 was raised against a peptide composed of Cys plus the sequence between positions 823 and 834 (CϩIQHSSVVPSIAT) of human Alc ␥ . These Alc-specific antibodies were raised against the extracellular juxtamembrane region of Alc family proteins and were specific for their respective p3-Alc targets with the exception of UT166, which exhibited cross-reactivity to p3-Alc ␣ (data not shown). These antibodies were used to isolate and detect p3-Alc. The regions recognized by the specific antibodies are shown in supplemental Fig. S1. The anti-Alc ␣ and anti-Alc ␤ cytoplasmic domain antibodies UT83 and UT99 were described previously (12). The monoclonal anti-FLAG (M2, Sigma) and anti-HA (12CA5, BD Biosciences) antibodies were purchased from vendors as noted. Anti-mouse and anti-rabbit IgG peroxidase-linked species-specific whole antibodies were purchased from GE Healthcare.
For experiments with human samples (CSF was furnished by Choju Medical Institute and Tottori University), a mixture of human CSF (0.3-1.0 ml) from 5-10 individuals (70 -90-yearold AD patients with clinical dementia rating 0.5, 1, or 2) was subjected to immunoprecipitation with the above antibodies. The beads were washed, and samples were eluted as described above. The dissolved samples were dried on a target plate, and MALDI-TOF/MS analysis was performed as described above.
The quantitative accuracy of mass spectrometric analysis with immunoprecipitation was confirmed by studies with a mixture of synthetic p3-Alc ␣ 35 and p3-Alc ␣ 39 peptides and a mixture of synthetic p3-Alc ␤ 37 and p3-Alc ␤ 40 peptides (supplemental Fig. S2). Molecular masses of p3-Alc species measured with MALDI-TOF/MS were compared with theoretical values to show the accuracy of mass spectrometric analysis (supplemental Table 1).
A␤40 and A␤42 levels were quantified with sandwich enzyme-linked immunosorbent assay systems (Immuno-Biological Laboratories, Takasaki, Japan). Informed consent for the use of human CSF in this study was obtained from the patients and their families.

Identification of the Proteinases That Initiate
Processing of Alc ␣ , Alc ␤ , and Alc ␥ -We have previously shown that the initial cleavage of Alc occurs in the extracellular juxtamembrane region and liberates the Alc ectodomain (sAlc) (12). The cellassociated C-terminal Alc fragment (AlcCTF) is proteolyzed by an intramembrane-cleaving protease, and both Alc cleavages are apparently catalyzed by the identical secretase enzymes that process APP (12). ADAM 10 is the enzyme that usually underlies the constitutive ␣-site cleavage of APP (16,17); therefore, we initially examined whether this ADAM also cleaved Alc.
As expected, we found that the primary cleavage of APP was deficient in homozygous (Ϫ/Ϫ) but not in heterozygous (ϩ/Ϫ) ADAM 10-deficient MEF cells (Fig. 1D); secretion of sAPP by Ϫ/Ϫ cells was 40% below the level of sAPP in ADAM 10 ϩ/Ϫ cells (compare lane 2 with lane 1 in D). Release of sAlc ␣ , sAlc ␤ , and sAlc ␥ was deficient in ADAM 10 Ϫ/Ϫ cells to a similar extent ( Fig. 1, A-C). These observations strongly suggest that Alc ␣ , Alc ␤ , and Alc ␥ are subjected to primary cleavage by ADAM 10 in a fashion similar to the ␣-secretase processing of APP.
The cleavage of Alc family proteins by ADAM 10 was rescued following expression of HA-tagged ADAM 10 in ADAM 10 Ϫ/Ϫ cells ( Fig. 1, E-H). In this experiment, we initially confirmed that sAPP secretion was restored by the expression of an exogenous cDNA for ADAM 10 ( Fig. 1H). Secretion of sAPP from ADAM 10 Ϫ/Ϫ cells that were expressing exogenous ADAM 10 increased ϳ1.7-fold as compared with the level of sAPP that was secreted from ADAM 10 Ϫ/Ϫ cells. As expected, we also found that the level of intracellular mature APP holoprotein (which is the substrate of ADAM 10 that gives rise to sAPP) was diminished in ADAM 10 Ϫ/Ϫ cells following expression of exogenous ADAM 10. The wild type secretion patterns of sAlc ␣ , sAlc ␤ , and sAlc ␥ were restored in ADAM 10 Ϫ/Ϫ cells following expression of exogenous ADAM 10 ( Fig. 1, E-G). In a dose-dependent fashion, these cells exhibited a 1.8 -2.2-fold increase in sAlc in the medium when compared with sAlc secretion from ADAM 10 Ϫ/Ϫ cells ( Fig. 1, E-G). Taken together, these results suggest that ADAM 10 is an important secretase that proteolyzes Alc ␣ , Alc ␤ , and Alc ␥ . APP is also cleaved by another ␣-secretase, ADAM 17 (15,16,21), although ADAM 17 is largely expressed in glial cells rather than neurons (22). Cleavage of the Alc family was examined in Neuro 2a cells overexpressing an HA-tagged ADAM 17. We initially confirmed that ADAM 17 cleaved APP in these cells (supplemental Fig. S3). As expected, increased expression of ADAM 17 enhanced sAPP␣ secretion, indicating that, in our cells, exogenously expressed ADAM17 was active in proteolysis of APP. The secretion of sAlc ␣ , sAlc ␤ , and sAlc ␥ from Neuro 2a cells expressing exogenous ADAM 17 was then assayed and quantified using experimental conditions that were identical to those used for the analysis of APP. We found that the secretion of sAlc (sAlc ␣ , sAlc ␤ , and sAlc ␥ ) increased in a dose-dependent manner in response to the exogenous expression of exogenous ADAM 17. These observations suggest that, like ADAM 10, ADAM 17 also cleaves Alc ␣ , Alc ␤ , and Alc ␥ as well as APP. Taken together, these data show that Alc and APP are metabolized by the same two APP ␣-secretases.
Identification of the Primary and Secondary Cleavage Sites of Alc-ADAM 10 and ADAM 17 have been identified as the ␣-secretases that cleave the peptide bond between Lys 612 and Leu 613 of APP 695 , thereby destroying the A␤ domain (Fig. 2). This cleavage at the juxtamembranous region triggers a sec-ondary intramembranous ␥-cleavage, which results in the secretion of a nonamyloidogenic 3-kDa peptide (p3) composed of 24 -26 amino acids (for a review, see Refs. 7 and 8).
We have previously reported that Alc family proteins, like APP family proteins, are also cleaved within the intramembranous region. In our earlier study, we showed that the small peptide, previously designated "␤-Alc," was secreted by the secondary cleavage of alcadein C-terminal fragments (CTFs) by ␥-secretase, but the exact cleavage sites were not previously identified (12). Now we have demonstrated that the peptide previously designated "␤-Alc" is generated by the successive action of ␣and ␥-secretases. Furthermore, ␤-secretase/␤-APP sitecleaving enzyme (BACE) (23) is not likely to be involved in the primary cleavage of Alc ␣ , Alc ␤ , and Alc ␥ (supplemental Fig. S4). In HEK293 cells overexpressing human BACE1, the generation of sAPP␤ and CTF␤, the products of APP cleaved by BACE, increased (supplemental Fig.  S4A), whereas no remarkable changes were observed in the cleavages of Alc ␣ , Alc ␤ , and Alc ␥ (supplemental Fig. S4B) in quality. The results suggest that Alcs are not a preferential substrate for BACE. Therefore, the ␤-Alc peptide has been renamed "p3-Alc" so as to maintain consistency with an APPbased nomenclature for the peptides produced by the proteolytic processing of Alc.
The recent production of anti-Alc ␣ , anti-Alc ␤ , and anti-Alc ␥ antibodies raised against the respective extracellular juxtamembrane sequences has enabled us to recover p3-Alc ␣ , p3-Alc ␤ , and p3-Alc ␥ secreted into the culture medium by HEK293 cells expressing Alc ␣ , Alc ␤ , and Alc ␥ . The molecular masses of p3-Alc ␣ , p3-Alc ␤ , and p3-Alc ␥ were determined by analyses with MALDI-TOF/MS. The spectrum of p3-Alc ␣ showed two major peaks with molecular masses of 3804.8 and 4007.0 (supplemental Fig.  S5A, indicated by arrows). The spectrum of p3-Alc ␤ showed two   Fig. S5B). The spectrum of p3-Alc ␥ showed one major peak with a molecular mass of 3377.8 Da and a minor peak with a molecular mass of 3689.2 Da (supplemental Fig. S5C).
The amino acid sequences of the respective major peak products were determined by MALDI-MS/MS analysis; the amino acid sequence of p3-Alc ␣ was composed of 35 and 37 amino acids, that of p3-Alc ␤ was composed of 37 and 40 amino acids, and that of p3-Alc ␥ was composed of 31 and 34 amino acids (supplemental Fig. S5, middle and right; the sequences are also double-underlined in Fig. 2). Thus, the major p3-Alc ␣ species, p3-Alc ␣ 35 and p3-Alc ␣ 2Nϩ35, that were secreted from HEK293 cells expressing Alc ␣1 were peptides that include the sequence from Ala 817 to Thr 851 and from Met 815 to Thr 851 of human Alc ␣1 . The major p3-Alc ␤ species, p3-Alc ␤ 37 and p3-Alc ␤ 40, that were secreted from HEK293 cells expressing Alc ␤ were peptides that included the sequences from Val 813 to Thr 849 and from Val 813 to Ile 852 of human Alc ␤ . The major p3-Alc ␥ species, p3-Alc ␥ 31 and p3-Alc ␥ 34, that were secreted from HEK293 cells expressing Alc ␥ were peptides that included the sequences from Leu 804 to Thr 834 and from Leu 804 to Ile 837 of human Alc ␥ . These p3-Alc ␣ 35, p3-Alc ␤ 37, and p3-Alc ␥ 31 species are also the major species in human CSF (see supplemental Fig. S6).
In our previous report, we used a gas phase protein sequencer to identify Ala 816 (numbering for the Alc ␣1 isoform) as the N-terminal amino acid of the Alc ␣ CTF (12). We therefore expected that ADAM 10 and ADAM 17 would cleave Alc ␣ at the peptide bond between Met 815 and Ala 816 . In the present study, we used MALDI-MS/MS to show that the N-terminal amino acid is Ala 817 for p3-Alc ␣ 35 and Met 815 for p3-Alc ␣ 2Nϩ35) (supplemental Fig.  S5A). The N-terminal Met 815 and/or Ala 816 of Alc ␣ CTF generated by primary cleavage may be removed by an N-terminal exopeptidase during p3-Alc ␣ secretion from cells.
We were unable to determine the N-terminal amino acid sequence of the Alc ␤ CTF using a gas phase protein sequencer; however, Val 813 , which was identified by MALDI-MS/MS analysis of p3-Alc ␤ 37 in this study (supplemental Figs. S5B and S6B), is likely to be the N-terminal amino acid of Alc ␤ CTF. We also determined the N-terminal sequence of Alc ␥ CTF (Leu 804 -Ile-Val-Gln-Pro-Pro-Phe-Leu-Gln 812 ) using a gas phase protein sequencer. This result was identical to the result obtained from the MALDI-MS/MS analysis (supplemental Fig.  S5C), indicating that the primary cleavage site of Alc ␥ is the peptide bond between His 803 and Leu 804 . The primary cleavage sites for the Alc family are shown in Fig. 2 (black arrowhead and arrow); these sites are also compared with the primary cleavage sites (␣and ␤sites) of APP 695 .
MALDI-TOF/MS and -MS/MS analyses enabled us to determine the secondary cleavage sites of Alc ␣ , Alc ␤ , and Alc ␥ (supplemental Figs. S5 and S6). The major secondary cleavage site of Alc ␣ is the peptide bond between Thr 851 and Val 852 ; cleavage at this site generates p3-Alc ␣ 35 and p3-Alc ␣ 2Nϩ35. The major secondary cleavage site of Alc ␤ is the peptide bond between Thr 849 and Leu 850 ; cleavage at this site generates p3-Alc ␤ 37. Another secondary cleavage site of Alc ␤ is the peptide bond between Ile 852 and Val 853 , which generates p3-Alc ␤ 40. The major secondary cleavage site of Alc ␥ is the peptide bond between Thr 834 and Val 835 , which generates p3-Alc ␥ 31. Another secondary cleavage site of Alc ␥ includes the peptide bond between Ile 837 and Ile 838 , which generates p3-Alc ␥ 34. Secondary cleavage sites (open arrowheads) of Alc family proteins, determined by MALDI-MS/MS analysis, are also shown in Fig.  2, together with the major secondary ␥-secretase-dependent cleavage sites of APP that generate A␤40 and A␤42.
Modulation of the ␥-Cleavage Sites of Alc by FAD-linked PS1 Mutations-We have previously reported that secondary cleavage of Alc ␣ is performed by a PS1-dependent ␥-secretase (12). PS1 is known to be the catalytic component of the ␥-secretase complex, and several mutations in the PS1 and PS2 genes cause FAD. Several FAD-linked PS1 mutations are known to alter the FIGURE 2. Amino acid sequence of p3-Alc ␣ , p3-Alc ␤ , and p3-Alc ␥ . The amino acid sequences of the major p3-Alc species are shown. p3-Alc ␣ 35, p3-Alc ␤ 37, and p3-Alc ␥ 31, are indicated as double-underlined letters along with the sequences of p3 and A␤40 of APP and the sequence of p3-Alc ␣ 2Nϩ35. The major primary (closed arrowheads) and secondary (open arrowheads) cleavage sites of Alc ␣1 , Alc ␤ , and Alc ␥ are indicated together with those of APP (␣, the cleavage site by ␣-secretase or ADAM 10/17; ␤, the cleavage site by ␤-secretase or BACE). Another primary cleavage site of Alc ␣1 is also indicated (arrow). Numbers on amino acids indicate their positions. Gray letters with a broken underline along with the sequence indicate the putative transmembrane region suggested by the Swiss-Prot protein knowledge base. The N terminus of p3-Alc ␣ is Ala 817 as determined by MALDI-TOF/MS analysis; however, that of p3-Alc ␣ 2Nϩ"X" is Met 815 . The p3-Alc ␣ 35 with N-terminal Ala 817 is a major species in CSF, whereas the p3-Alc ␣ 2Nϩ35 with N-terminal Met 815 is a major species in HEK293 cells (see supplemental Figs. S5A, S6A, and S7A). The N terminus of p3-Alc ␤ is Val 813 , which was determined by MALDI-MS/MS analysis (see supplemental Figs. S5B and S6B). The N terminus of p3-Alc ␥ is Leu 804 (supplemental Fig. S5C); this coincides with the N-terminal sequence of Alc ␥ CTF that was determined using a gas phase peptide sequencer.
cleavage of APP ␤ CTFs and increase the generation of C-terminally altered A␤ species, such as A␤42 (24 -27). Because Alc and APP are cleaved by the identical ␥-secretase, we sought to determine whether Alc family proteins demonstrate displaced secondary cleavage sites in the presence of FAD-linked PS1 mutations. HEK293 cells stably expressing wild type PS1 (WT in Fig. 3) or PS1 carrying either FAD-linked M146L, L166P, A246E, A278T, L286V, or A434C mutations or vector alone (Mock) were co-transfected with Alc ␣ , Alc ␤ , and Alc ␥ or without plasmid (Ϫ).
p3-Alc ␣ , p3-Alc ␤ , and p3-Alc ␥ were recovered from the culture medium by immunoprecipitation with anti-Alc ␣ (UT135), anti-Alc ␤ (UT143), and anti-Alc ␥ (UT166) antibodies and analyzed by MALDI-TOF/MS (Fig. 3). The expression of PS1 was confirmed by Western blotting with anti-PS1 N-and C-terminal antibodies (data not shown), and the effect of the FADlinked mutations on the A␤42/A␤40 ratio was examined in the medium of HEK293 cell lines, each of which expressed PS1 stably plus APP 695 transiently (Fig. 4A, lower right).
As expected, HEK293 cells stably expressing wild type PS1 (WT) or vector alone (Mock) plus Alc ␣ generated p3-Alc ␣ species with C-terminal end of Thr 851 (p3-Alc ␣ 2Nϩ35) as the major species (Fig. 3A). We determined the amino acid sequences of this p3-Alc ␣ species with MALDI-MS/MS analysis and confirmed that this is p3-Alc ␣ 2Nϩ35 and not p3-Alc ␣ 37. This p3-Alc ␣ 2Nϩ35 was a minor p3-Alc ␣ species in human CSF in which p3-Alc ␣ 35 was major (left panel in supplemental Figs. S5A and S6A). In any case, HEK293 cells expressing PS1 (WT) generated largely p3-Alc ␣ species with the C-terminal Thr 851 (right panel in supplemental Figs. S5A and S6A).
Six FAD-linked PS1 mutants (L166P, R278T, A434C, A246E, M146L, and L286V) increased the A␤42/40 ratio at various magnitudes (Fig. 4A, lower right), whereas some of them did not show a similar effect to alter the minor species/major species ratio in the p3-Alc species. To analyze the correlation coefficients of ␥-cleavage alteration between APP and Alc, the A␤42/40 ratio was plotted to certain p3-Alc minor/major ratios (Fig. 4B). We have confirmed that the p3-Alc minor/major ratio of peak area detected by MALDI-TOF/MS analysis correlated well to those of theoretically calculated values in quantity (supplemental Fig. S2). As expected from the Fig. 4A on the basis of visual inspection, p3-Alc ␤ showed a property different from p3-Alc ␣ and p3-Alc ␥ in correlation coefficient to A␤42/40. The p3-Alc ␣ 2Nϩ38/p3-Alc ␣ 2Nϩ35 and p3-Alc ␥ 34/p3-Alc ␥ 31 ratios showed a strong correlation to the A␤42/40 ratio (R 2 Ͼ 0.5), whereas the p3-Alc ␤ 40/p3-Alc ␤ 37 ratio showed a positive but weak correlation to the A␤42/40 ratio. These results suggest that phenotypes of ␥-secretase dysfunction appeared in the altered cleavages of APP and/or Alc family proteins in variety. The magnitudes of C-terminal alteration of p3-Alc ␣ , p3-Alc ␤ , and p3-Alc ␥ along with A␤ were not equivalent, suggesting that one type of ␥-secretase dysfunction does not appear in the phenotype equivalently in the cleavage of type I membrane proteins.
The p3-Alc Species in Human CSF-Secreted A␤ species are detectable in human CSF, and it has been proposed that they might be useful as possible biomarkers or endophenotypes for the diagnosis and classification of AD (28 -31). Above, we discussed how p3-Alc speciation and A␤42/40 were modulated by certain FAD mutant PS1 molecules, which reflect ␥-secretase dysfunction (Figs. 3 and 4). In our separate study, we sought to characterize CSF p3-Alc/A␤ relationships with a special interest in determining whether sporadic AD CSF might display a covariance between p3-Alc and A␤ as a potential indicator of underlying ␥-secretase dysfunction. 6 We herein examined whether the p3-Alc ␣ , p3-Alc ␤ , and p3-Alc ␥ species that were identified in cell study were present in human CSF. The p3-Alc species were recovered from pooled human CSF samples by immunoprecipitation with anti-p3-Alc ␣ 3B5, anti-p3-Alc ␤ UT143, and anti-p3-Alc ␥ UT166 antibodies, followed by analyses with MALDI-TOF/MS ( Fig. 5 and supplemental Fig. S6).
The antibody UT143 recovered peptides of molecular masses 3963.9 and 4303.2 ( Fig. 5B and supplemental Fig. S6B), which were designated p3-Alc ␤ 37 and p3-Alc ␤ 40, respectively, according to MALDI-MS/MS analysis. The antibody UT166 recovered a peptide of molecular mass 3377.6 ( Fig. 5C and supplemental Fig. S6C). The species with a molecular mass of 3377.6 (arrow in supplemental Fig. S6C) could not be analyzed by MALDI-MS/MS for amino acid sequence because the amount of peptide that was recovered by immunoprecipitation was not sufficient to obtain significant signals; however, the molecular mass coincided with that of p3-Alc ␥ 31 (see supplemental Fig. S5C). These results demonstrate that the intramembrane cleavage sites of Alc in humans are identical to those determined by our cell-conditioned medium studies and that p3-Alc ␣ 35, p3-Alc ␣ 37, p3-Alc ␤ 37, p3-Alc ␤ 40, and p3-Alc ␥ 31 are the major p3-Alc peptides recovered from human CSF.
To confirm the major p3-Alc peptide detected in CSF is the exact major product in brain, we compared p3-Alc ␣ species in mouse CSF and brain (supplemental Fig. S8). The major p3-Alc ␣ 35 in CSF was also major product in the brain, along with the similar profile of p3-Alc ␣ minor species between CSF and brain. Furthermore, the p3-Alc ␣ species profile of mouse CSF was identical to those of human CSF (supplemental Fig.  S8), suggesting that p3-Alc ␣ 35 is major product in mouse and human brain.

DISCUSSION
In this study, we show that the Alc family proteins Alc ␣ , Alc ␤ , and Alc ␥ are cleaved by APP ␣-secretases ADAM 10 and ADAM 17. Alc is expressed largely in neurons (1); therefore, ADAM 10 is thought to be the most likely candidate for Alc cleavage in the central nervous system because ADAM 17 is predominantly expressed in glial cells (22). Thus, in neurons, Alc and APP are primarily cleaved by the same enzymes when both proteins are liberated from their individual or coordinated complexes with X11L. These results agree with our previous report that APP and Alc are likely to be metabolized in a coordinated fashion (12). Although APP is also cleaved by BACE, Alc proteins are not likely to be major substrates for this enzyme (supplemental Fig. S4).
In addition to the similar mechanisms regulating the primary cleavage of Alc family proteins and APP, we also found that Alc family proteins are cleaved by the same ␥-secretase complex as is APP (12). In this study, we have demonstrated that p3-Alc species with altered C termini are secreted together with the increased ratio of A␤42 to A␤40 by cells expressing FAD-linked PS1 mutants, although the magnitude of alteration was diversified among Alc family proteins in cells expressing various FADlinked PS1 mutants. p3-Alc species are not prone to aggrega-tion 7 and are detectable in human CSF; this raises the possibility that some p3-Alc changes in the pathological state might serve in clinical situations as markers of variant protein processing by the dysfunction of ␥-secretase.
In the cell culture experiments, we observed that wild type and several pathogenic PS1 mutants modulated p3-Alc ␣ , p-3-Alc ␤ , and p3-Alc ␥ C-terminal speciation at various magnitudes. This indicates that APP is not the only ␥-secretase substrate that undergoes variant processing in association with PS1 mutations and suggests that some of the PS1 mutations caused more alteration in the cleavage of Alc protein(s) than APP and vice versa.
In studies with cells expressing FAD-linked PS1 mutant, a covariance was observed when A␤42/40 and certain variant p3-Alc minor/major species ratios, such as p3-Alc ␣ 2Nϩ38/p3-Alc ␣ 2Nϩ35 or p3-Alc ␥ 34/p3-Alc ␥ 31, were plotted against each other (Fig. 4B). We observed similar phenomena in the CSF of subjects with sporadic AD, in that there existed disease-related covariant signature relationships between p3-Alc ␣ 38/p3-Alc ␣ 35 ratios and A␤42/40 ratios. 6 In the case of p3-Alc ␣ 38/35 ratios, the covariant signature showed a positive slope in AD subjects, whereas the signatures of both the aged non-demented control subjects and the other neurological disease control subjects were similar to each other and negative in slope, suggesting a pathogenic malfunction of the ␥-secretase complex in sporadic AD patients, 6 as observed in a pathogenic state of FAD-linked PS mutations in this study (Fig. 4B).
The biophysical mechanisms by which pathogenic PS1 mutations exert their effects are poorly understood, and it is unclear how p3-Alc ratios from the CSF of humans who do not harbor PS1 mutations are altered, as can be seen in this study with FAD-linked PS1 mutant. It is possible that wild type PS1 is behaving as if a PS1 mutation were present as a result of changes in some co-factors that modulate ␥-secretase function, a result of the conformational changes of PS1, and/or a result of peripheral environmental changes at the place that ␥-secretase is active. Therefore, the present results suggest that qualitative changes of p3-Alc species with altered C termini may be useful in diagnosing dysfunction of ␥-secretase prior to the pathogenesis of sporadic Alzheimer disease. Taking these findings together with those of another study of human subjects, 6 we propose a hypothesis that ␥-secretase dysfunction may be a feature of the pathogenesis in some population of common sporadic AD and that measurement of unusual p3-Alc fragments may provide surrogate markers to detect clinical ␥-secretase dysfunction.