Discovery of colorectal cancer biomarker candidates by membrane proteomic analysis and subsequent verification using selected reaction monitoring (SRM) and tissue microarray (TMA) analysis.

Recent advances in quantitative proteomic technology have enabled the large-scale validation of biomarkers. We here performed a quantitative proteomic analysis of membrane fractions from colorectal cancer tissue to discover biomarker candidates, and then extensively validated the candidate proteins identified. A total of 5566 proteins were identified in six tissue samples, each of which was obtained from polyps and cancer with and without metastasis. GO cellular component analysis predicted that 3087 of these proteins were membrane proteins, whereas TMHMM algorithm predicted that 1567 proteins had a transmembrane domain. Differences were observed in the expression of 159 membrane proteins and 55 extracellular proteins between polyps and cancer without metastasis, while the expression of 32 membrane proteins and 17 extracellular proteins differed between cancer with and without metastasis. A total of 105 of these biomarker candidates were quantitated using selected (or multiple) reaction monitoring (SRM/MRM) with stable synthetic isotope-labeled peptides as an internal control. The results obtained revealed differences in the expression of 69 of these proteins, and this was subsequently verified in an independent set of patient samples (polyps (n = 10), cancer without metastasis (n = 10), cancer with metastasis (n = 10)). Significant differences were observed in the expression of 44 of these proteins, including ITGA5, GPRC5A, PDGFRB, and TFRC, which have already been shown to be overexpressed in colorectal cancer, as well as proteins with unknown function, such as C8orf55. The expression of C8orf55 was also shown to be high not only in colorectal cancer, but also in several cancer tissues using a multicancer tissue microarray, which included 1150 cores from 14 cancer tissues. This is the largest verification study of biomarker candidate membrane proteins to date; our methods for biomarker discovery and subsequent validation using SRM/MRM will contribute to the identification of useful biomarker candidates for various cancers. Data are available via ProteomeXchange with identifier PXD000851.

A-to-I editing in the miRNA seed region regulates target mRNA selection and silencing efficiency.

Hydrolytic deamination of adenosine to inosine (A-to-I) by adenosine deaminases acting on RNA (ADARs) is a post-transcriptional modification which results in a discrepancy between genomic DNA and the transcribed RNA sequence, thus contributing to the diversity of the transcriptome. Inosine preferentially base pairs with cytidine, meaning that A-to-I modifications in the mRNA sequences may be observed as A-to-G substitutions by the protein-coding machinery. Genome-wide studies have revealed that the majority of editing events occur in non-coding RNA sequences, but little is known about their functional meaning. MiRNAs are small non-coding RNAs that regulate the expression of target mRNAs with complementarities to their seed region. Here, we confirm that A-to-I editing in the miRNA seed duplex globally reassigns their target mRNAs in vivo, and reveal that miRNA containing inosine in the seed region exhibits a different degree of silencing efficiency compared to the corresponding miRNA with guanosine at the same position. The difference in base-pairing stability, deduced by melting temperature measurements, between seed-target duplexes containing either C:G or I:C pairs may account for the observed silencing efficiency. These findings unequivocally show that C:G and I:C pairs are biologically different in terms of gene expression regulation by miRNAs.

A novel mechanism of keratin cytoskeleton organization through casein kinase Iα and FAM83H in colorectal cancer.

Keratin filaments form cytoskeletal networks in epithelial cells. Dynamic rearrangement of keratin filament networks is required for epithelial cells to perform cellular processes such as cell migration and polarization; however, the mechanism governing keratin filament rearrangement remains unclear. Here, we describe a novel mechanism of keratin cytoskeleton organization mediated by casein kinase Iα (CK-1α) and a newly identified keratin-associated protein, FAM83H. Knockdown of FAM83H induces keratin filament bundling, whereas overexpression of FAM83H disassembles keratin filaments, suggesting that FAM83H regulates the filamentous state of keratins. Intriguingly, keratin filament bundling is concomitant with the dissociation of CK-1α from keratin filaments, whereas aberrant speckle-like localization of CK-1α is observed concomitantly with keratin filament disassembly. Furthermore, CK-1α inhibition, similar to FAM83H knockdown, causes keratin filament bundling and reverses keratin filament disassembly induced by FAM83H overexpression, suggesting that CK-1α mediates FAM83H-dependent reorganization of keratin filaments. Because the N-terminal region of FAM83H interacts with CK-1α and the C-terminal region interacts with keratins, FAM83H might tether CK-1α to keratins. Colorectal cancer tissue also shows keratin filament disassembly accompanied with FAM83H overexpression and aberrant CK-1α localization, and FAM83H-overexpressing cancer cells exhibit loss or alteration of epithelial cell polarity. Importantly, knockdown of FAM83H inhibits cell migration accompanied by keratin cytoskeleton rearrangement in colorectal cancer cells. These results suggest that keratin cytoskeleton organization is regulated by FAM83H-mediated recruitment of CK-1α to keratins, and that keratin filament disassembly caused by overexpression of FAM83H and aberrant localization of CK-1α could contribute to the progression of colorectal cancer.

[Biomarker Discovery of Colorectal Cancer Using Membrane Proteins Extracted from Cancer Tissues].

Colorectal cancer is the leading cause of cancer-related death in women and third leading cause of death in men. The prognosis worsens when cancer metastasizes to other organs. Thus, there is an urgent need to develop biomarkers for the early diagnosis of metastasis as well as cancer development. Mass spectrometry-based technologies have been applied to the discovery of protein biomarkers, especially in the field of cancer. These results have identified numerous candidate protein biomarkers. Unfortunately, only a few are currently being applied in clinical diagnostics. Recent advances in proteomic technology such as selected/multiple reaction monitoring (SRM/MRM) facilitated the detection and quantitation of specific proteins in complex samples without the need for antibodies. We performed a quantitative proteomic analysis of membrane proteins extracted from colorectal cancer tissues using the iTRAQ shotgun method to discover biomarker candidates, and then extensively validated the biomarker candidate proteins by SRM/MRM. A total of 5,566 proteins were identified in tissue samples obtained from adenoma and cancer with and without metastasis. Differences were observed in the expression of about 400 proteins. Among them, 105 biomarker candidates which were predicted to be membrane proteins and extracellular proteins by gene ontology analysis were quantitated using SRM/MRM. As a result, we could confirm differences in the expressions of 69 of these proteins using the same set of patient samples as discovery experiments, and this was subsequently verified in an independent set of samples. Significant differences were observed in the expression of 44 of these proteins. Moreover, some of the biomarker candidates were detected and quantitated in the serum of colorectal cancer patients. These biomarker candidates are promising diagnostic tools on investigating the development and progression of colorectal cancer.

In-depth membrane proteomic study of breast cancer tissues for the generation of a chromosome-based protein list.

The Chromosome-centric Human Proteome Project (C-HPP) aims to define all proteins encoded in each chromosome and especially to identify proteins that currently lack evidence by mass spectrometry. The C-HPP also prioritizes particular protein subsets such as membrane proteins, post-translational modifications, and low-abundance proteins. In this study, we aimed to generate deep profiling of the membrane proteins of human breast cancer tissues on a chromosome-by-chromosome basis using shotgun proteomics. We identified 7092 unique proteins using membrane fractions isolated from pooled breast cancer tissues with high confidence. A total of 3282 proteins were annotated as membrane proteins by Gene Ontology analysis, which covered 45% of the membrane proteins predicted in 20,859 protein-coding genes. Furthermore, we were able to identify 851 membrane proteins that currently lack evidence by mass spectrometry in neXtProt. Our results will contribute to the accomplishment of the primary goal of the C-HPP in identifying so-called "missing proteins" and generating a whole protein catalog for each chromosome.

A strategy for large-scale phosphoproteomics and SRM-based validation of human breast cancer tissue samples.

Protein phosphorylation is a key mechanism of cellular signaling pathways and aberrant phosphorylation has been implicated in a number of human diseases. Thus, approaches in phosphoproteomics can contribute to the identification of key biomarkers to assess disease pathogenesis and drug targets. Moreover, careful validation of large-scale phosphoproteome analysis, which is lacking in the current protein-based biomarker discovery, significantly increases the value of identified biomarkers. Here, we performed large-scale differential phosphoproteome analysis using IMAC coupled with the isobaric tag for relative quantification (iTRAQ) technique and subsequent validation by selected/multiple reaction monitoring (SRM/MRM) of human breast cancer tissues in high- and low-risk recurrence groups. We identified 8309 phosphorylation sites on 3401 proteins, of which 3766 phosphopeptides (1927 phosphoproteins) were able to be quantified and 133 phosphopeptides (117 phosphoproteins) were differentially expressed between the two groups. Among them, 19 phosphopeptides were selected for further verification and 15 were successfully quantified by SRM using stable isotope peptides as a reference. The ratio of phosphopeptides between high- and low-risk groups quantified by SRM was well correlated with iTRAQ-based quantification with a few exceptions. These results suggest that large-scale phosphoproteome quantification coupled with SRM-based validation is a powerful tool for biomarker discovery using clinical samples.

Strategy for SRM-based verification of biomarker candidates discovered by iTRAQ method in limited breast cancer tissue samples.

Since LC-MS-based quantitative proteomics has become increasingly applied to a wide range of biological applications over the past decade, numerous studies have performed relative and/or absolute abundance determinations across large sets of proteins. In this study, we discovered prognostic biomarker candidates from limited breast cancer tissue samples using discovery-through-verification strategy combining iTRAQ method followed by selected reaction monitoring/multiple reaction monitoring analysis (SRM/MRM). We identified and quantified 5122 proteins with high confidence in 18 patient tissue samples (pooled high-risk (n=9) or low-risk (n=9)). A total of 2480 proteins (48.4%) of them were annotated as membrane proteins, 16.1% were plasma membrane and 6.6% were extracellular space proteins by Gene Ontology analysis. Forty-nine proteins with >2-fold differences in two groups were chosen for further analysis and verified in 16 individual tissue samples (high-risk (n=9) or low-risk (n=7)) using SRM/MRM. Twenty-three proteins were differentially expressed among two groups of which MFAP4 and GP2 were further confirmed by Western blotting in 17 tissue samples (high-risk (n=9) or low-risk (n=8)) and Immunohistochemistry (IHC) in 24 tissue samples (high-risk (n=12) or low-risk (n=12)). These results indicate that the combination of iTRAQ and SRM/MRM proteomics will be a powerful tool for identification and verification of candidate protein biomarkers.

Tri-view two-photon microscopic image registration and deblurring with convolutional neural networks.

Two-photon fluorescence microscopy has enabled the three-dimensional (3D) neural imaging of deep cortical regions. While it can capture the detailed neural structures in the x-y image space, the image quality along the depth direction is lower because of lens blur, which often makes it difficult to identify the neural connectivity. To address this problem, we propose a novel approach for restoring the isotropic image volume by estimating and fusing the intersection regions of the images captured from three orthogonal viewpoints using convolutional neural networks (CNNs). Because convolution on 3D images is computationally complex, the proposed method takes the form of cascaded CNN models consisting of rigid transformation, dense registration, and deblurring networks for more efficient processing. In addition, to enable self-supervised learning, we trained the CNN models with simulated synthetic images by considering the distortions of the microscopic imaging process. Through extensive experiments, the proposed method achieved substantial image quality improvements.

[Clinical features of idiopathic restless legs syndrome in Japanese patients].

BACKGROUND: Little is known about the diagnosis and management of restless legs syndrome (RLS) in Japanese neurology clinics. OBJECTIVE: To validate the diagnostic criteria of the International RLS Study Group (IRLSSG) and the treatment algorithm of the Mayo Clinic in a Japanese neurology clinic setting and to clarify the features of Japanese patients with idiopathic RLS. METHODS: Patients with RLS symptoms were examined by a neurologist and the assessment included neurological examination, tests for periodic limb movements (PLM) and dopaminergic response, and the clinical diagnosis was made according to IRLSSG diagnostic criteria. Patients diagnosed with idiopathic RLS were treated with dopaminergic agents and the efficacy was evaluated. RESULTS: The study subjects were 151 Japanese patients who presented with RLS symptoms. Idiopathic RLS was diagnosed in 113 patients, secondary RLS in 16 and RLS mimics in 22. The cause of RLS mimics was either myelopathy, radiculopathy or neuropathy in 11 patients. The mean age of patients with idiopathic RLS was 50.1 (SD 20.0) years, 63% were woman, 97% had daily RLS, 31% had family history (40% of the early-onset subgroup), 86% reported unpleasant sensations in the lower legs, 43% had PLM in the daytime suggested immobilization test, 81% suffered from insomnia, 49% had limitations of work and activities, 71% reported impaired mood, 27% had consulted physicians about their symptoms, 4% had been diagnosed with RLS, 73% improved after dopaminergic treatments, and 33% experienced complete remission. CONCLUSIONS: The clinical features of Japanese patients with idiopathic RLS were identical to those reported in western countries, which suggests that IRLSSG diagnostic criteria and Mayo Clinic treatment algorism are valid in Japanese neurology clinics. Both patients and physicians were not fully aware of RLS in this country. Neurological examination was important in excluding RLS mimics and making a diagnosis of RLS.

Generative and discriminative model-based approaches to microscopic image restoration and segmentation.

Image processing is one of the most important applications of recent machine learning (ML) technologies. Convolutional neural networks (CNNs), a popular deep learning-based ML architecture, have been developed for image processing applications. However, the application of ML to microscopic images is limited as microscopic images are often 3D/4D, that is, the image sizes can be very large, and the images may suffer from serious noise generated due to optics. In this review, three types of feature reconstruction applications to microscopic images are discussed, which fully utilize the recent advancements in ML technologies. First, multi-frame super-resolution is introduced, based on the formulation of statistical generative model-based techniques such as Bayesian inference. Second, data-driven image restoration is introduced, based on supervised discriminative model-based ML technique. In this application, CNNs are demonstrated to exhibit preferable restoration performance. Third, image segmentation based on data-driven CNNs is introduced. Image segmentation has become immensely popular in object segmentation based on electron microscopy (EM); therefore, we focus on EM image processing.

IGF-1 promotes beta-amyloid production by a secretase-independent mechanism.

Beta-amyloid peptide (Abeta) is generated via the sequential proteolysis of beta-amyloid precursor protein (APP) by beta- and gamma-secretases, and plays a crucial role in the pathogenesis of Alzheimer's disease (AD). Here, we sought to clarify the role of insulin-like growth factor-1 (IGF-1), implicated in the AD pathomechanism, in the generation of Abeta. Treatment of neuroblastoma SH-SY5Y cells expressing AD-associated Swedish mutant APP with IGF-1 did not alter cellular levels of APP, but significantly increased those of beta-C-terminal fragment (beta-CTF) and secreted Abeta. IGF-1 also enhanced APP phosphorylation at Thr668. Treatment of beta-CTF-expressing cells with IGF-1 increased the levels of beta-CTF and secreted Abeta. The IGF-1-induced augmentation of beta-CTF was observed in the presence of gamma-secretase inhibitors, but not in cells expressing beta-CTF with a Thr668 to alanine substitution. These results suggest that IGF-1 promotes Abeta production through a secretase-independent mechanism involving APP phosphorylation.

The two-hydrophobic domain tertiary structure of reticulon proteins is critical for modulation of beta-secretase BACE1.

beta-Site amyloid precursor protein (APP) cleaving enzyme 1 (BACE1) is a membrane-bound protease that is essential for the production of beta-amyloid protein (Abeta). Given the crucial role of Abeta accumulation in Alzheimer's disease (AD), inhibition of BACE1 activity may represent a feasible therapeutic strategy in the treatment of AD. Recently, we and others identified reticulon 3 (RTN3) and reticulon 4-B/C (RTN4-B/C or Nogo-B/C) as membrane proteins that interact with BACE1 and inhibit its ability to produce Abeta. In this study, we employed various mutants of RTN3 and RTN4-C and C. elegans RTN to investigate the molecular mechanisms by which RTNs regulate BACE1. We found that RTN3 mutants lacking the N-terminal or C-terminal or loop domain as well as a RTN4-C mutant lacking the C-terminal domain bound to BACE1 comparably to wild-type RTN3 and RTN4-C. Furthermore, overexpression of wild-type RTN3, RTN4-C, and these RTN mutants similarly reduced Abeta40 and Abeta42 secretion by cells expressing Swedish mutant APP. C. elegans RTN, which has low homology to human RTNs, also interacted with BACE1 and inhibited Abeta secretion. In contrast, two RTN3 mutants containing deletions of the first or second potential transmembrane domains and an RTN3 swap mutant of the second transmembrane domain bound BACE1 but failed to inhibit Abeta secretion. Collectively, these results suggest that the two-transmembrane-domain tertiary structure of RTN proteins is critical for the ability of RTNs to modulate BACE1 activity, whereas N-terminal, C-terminal and loop regions are not essential for this function.

Quantitation of putative colorectal cancer biomarker candidates in serum extracellular vesicles by targeted proteomics.

At the moment, there is no sensitive clinical test for detecting early-stage colorectal cancer (CRC). Target proteomics has enabled high-throughput verification of hundreds of biomarker candidate proteins. Using this technology, we verified 725 previously reported CRC biomarker candidate proteins that are functionally correlated with CRC in extracellular vesicles (EVs) from patients. Of these, 356 proteins were quantified, and 34 peptides (22 proteins) showed significant differences in the serum EVs between healthy controls and CRC patients of two independent cohorts (n = 77 and 84). These peptides were evaluated as single or multiple markers, and four single peptides in annexin family proteins and eight combinations of peptides showed area under the curve > 0.9 for discriminating between healthy controls and CRC patients. The sensitivities of annexins A3, A4, and A11 peptides for detecting early-stage CRC greatly exceed those of carcinoembryonic antigen. These peptides are promising biomarkers for early detection of CRC.

Proteomic analysis of urinary extracellular vesicles from high Gleason score prostate cancer.

Extracellular vesicles (EVs) are microvesicles secreted from various cell types. We aimed to discover a new biomarker for high Gleason score (GS) prostate cancer (PCa) in urinary EVs via quantitative proteomics. EVs were isolated from urine after massage from 18 men (negative biopsy [n = 6], GS 6 PCa [n = 6], or GS 8-9 PCa [n = 6]). EV proteins were labeled with iTRAQ and analyzed by LC-MS/MS. We identified 4710 proteins and quantified 3528 proteins in the urinary EVs. Eleven proteins increased in patients with PCa compared to those with negative biopsy (ratio >1.5, p-value < 0.05). Eleven proteins were chosen for further analysis and verified in 29 independent urine samples (negative [n = 11], PCa [n = 18]) using selected reaction monitoring/multiple reaction monitoring. Among these candidate markers, fatty acid binding protein 5 (FABP5) was higher in the cancer group than in the negative group (p-value = 0.009) and was significantly associated with GS (p-value for trend = 0.011). Granulin, AMBP, CHMP4A, and CHMP4C were also higher in men with high GS prostate cancer (p-value < 0.05). FABP5 in urinary EVs could be a potential biomarker of high GS PCa.

Disturbance of cerebellar synaptic maturation in mutant mice lacking BSRPs, a novel brain-specific receptor-like protein family.

By DNA cloning, we have identified the BSRP (brain-specific receptor-like proteins) family of three members in mammalian genomes. BSRPs were predominantly expressed in the soma and dendrites of neurons and localized in the endoplasmic reticulum (ER). Expression levels of BSRPs seemed to fluctuate greatly during postnatal cerebellar maturation. Triple-knockout mice lacking BSRP members exhibited motor discoordination, and Purkinje cells (PCs) were often innervated by multiple climbing fibers with different neuronal origins in the mutant cerebellum. Moreover, the phosphorylation levels of protein kinase Calpha (PKCalpha) were significantly downregulated in the mutant cerebellum. Because cerebellar maturation and plasticity require metabotropic glutamate receptor signaling and resulting PKC activation, BSRPs are likely involved in ER functions supporting PKCalpha activation in PCs.

Casein kinase 1 is recruited to nuclear speckles by FAM83H and SON.

In some fibroblasts, casein kinase 1α (CK1α) is localized to nuclear speckles, which are sub-nuclear compartments supplying splicing factors, whereas it is recruited on keratin filaments in colorectal cancer cells such as DLD1 cells. In order to obtain a deeper understanding of why CK1α is localized to these different subcellular sites, we herein elucidated the mechanisms underlying its localization to nuclear speckles. CK1α and FAM83H were localized to nuclear speckles in RKO and WiDr colorectal cancer cells, which do not express simple epithelial keratins, and in DLD1 cells transfected with siRNAs for type I keratins. The localization of FAM83H to nuclear speckles was also detected in colorectal cancer cells with a poorly organized keratin cytoskeleton in colorectal cancer tissues. Using an interactome analysis of FAM83H, we identified SON, a protein present in nuclear speckles, as a scaffold protein to which FAM83H recruits CK1α. This result was supported by the knockdown of FAM83H or SON delocalizing CK1α from nuclear speckles. We also found that CK1δ and ε are localized to nuclear speckles in a FAM83H-dependent manner. These results suggest that CK1 is recruited to nuclear speckles by FAM83H and SON in the absence of an intact keratin cytoskeleton.

Developmental genetic bases behind the independent origin of the tympanic membrane in mammals and diapsids.

The amniote middle ear is a classical example of the evolutionary novelty. Although paleontological evidence supports the view that mammals and diapsids (modern reptiles and birds) independently acquired the middle ear after divergence from their common ancestor, the developmental bases of these transformations remain unknown. Here we show that lower-to-upper jaw transformation induced by inactivation of the Endothelin1-Dlx5/6 cascade involving Goosecoid results in loss of the tympanic membrane in mouse, but causes duplication of the tympanic membrane in chicken. Detailed anatomical analysis indicates that the relative positions of the primary jaw joint and first pharyngeal pouch led to the coupling of tympanic membrane formation with the lower jaw in mammals, but with the upper jaw in diapsids. We propose that differences in connection and release by various pharyngeal skeletal elements resulted in structural diversity, leading to the acquisition of the tympanic membrane in two distinct manners during amniote evolution.

Mutant presenilin (A260V) affects Rab8 in PC12D cell.

Most familial early-onset Alzheimer's disease (FAD) is caused by mutations in the presenilin-1 (PS1) gene. Abeta is derived from amyloid precursor protein (APP) and an increased concentration of Abeta 42 is widely believed to be a pathological hallmark of abnormal PS function. Therefore, the interaction between PS1 and APP is a central theme in attempts to clarify the molecular mechanism of AD. To examine the effect of PS1 mutations on APP metabolism, we made PC12D cell lines that express human PS1 or mutant PS1 (A260V). In PC12D cells expressing the PS1A260V mutant, we found that Rab8, a GTPase involved in transport from the trans-Golgi network (TGN) to the plasma membrane (PM), was significantly reduced in PC12D cells expressing the A260V mutant and that APP C-terminal fragment (CTF), the direct precursor of Abeta, accumulated in the heavy membrane fraction including membrane vesicles involved in TGN-to-PM transport. Furthermore, the total intracellular Abeta production was reduced in these cells. Combined together, we have observed that PS1 mutation disturbs membrane vesicle transport, resulting in prolonged residence of APP CTF during TGN-to-PM transport pathway. Therefore, it is highly likely that reduction of Abeta is closely related to the retention of APP CTF during TGN-to-PM transport.

Intracellular domain generation of amyloid precursor protein by epsilon-cleavage depends on C-terminal fragment by alpha-secretase cleavage.

A significant accumulation of Abeta is major pathological feature in the brain in patients with Alzheimer's disease (AD). Amyloid precursor protein (APP) is cleaved by alpha- and beta-secretase, resulting in secretion of extracellular domain. Then the remaining membrane-anchored C-terminal fragments (CTFalpha or CTFbeta) are cleaved by gamma-secretase. Therefore, Abeta is derived from APP by sequential proteolytic processing involving 2 proteases, called beta- and gamma-secretase. Gamma-secretase cleavage results in the generation of the APP intracellular domain (AICD), as well as Abeta. Recently, AICD is generated by epsilon-cleavage site near the cytoplasmic membrane boundary of APP, not by gamma-cleavage site in the middle of transmembrane domain. We show that elevated beta-secretase levels induced the increase in CTFbeta and Abeta generation and reduction of CTFalpha and AICD generation in vitro. Furthermore, elevated alpha-secretase levels induced an increase of AICD. The results suggest that generation of AICD by epsilon-cleavage depends on CTFalpha and that gamma- and epsilon-cleavage are differentially regulated.

gamma-Secretase can cleave amyloid precursor protein fragments independent of alpha- and beta-secretase pre-cutting.

Abeta is the major component of amyloid in the brain in Alzheimer's disease and is derived from an amyloid precursor protein (APP) by the sequential proteolytic processing of two putative proteases, called beta- and gamma-secretase. To clarify the mechanism of gamma-secretase processing, we created constructs contained the C-terminal domain of APP and analyzed the processing in COS-1 cells. We found that C-terminal fragments (CTFs) containing a short extra N-terminal region before the beta-secretase cleavage site were directly processed at gamma-secretase cleavage site. This suggests that gamma-secretase cleavage occurs independently and is not dependent on alpha- and beta-secretase cleavage.