Cystatin C Shifts APP Processing from Amyloid-β Production towards Non-Amyloidgenic Pathway in Brain Endothelial Cells

Amyloid-β (Aβ), the major component of neuritic plaques in Alzheimer’s disease (AD), is derived from sequential proteolytic cleavage of amyloid protein precursor (APP) by secretases. In this study, we found that cystatin C (CysC), a natural cysteine protease inhibitor, is able to reduce Aβ40 secretion in human brain microvascular endothelial cells (HBMEC). The CysC-induced Aβ40 reduction was caused by degradation of β-secretase BACE1 through the ubiquitin/proteasome pathway. In contrast, we found that CysC promoted secretion of soluble APPα indicating the activated non-amyloidogenic processing of APP in HBMEC. Further results revealed that α-secretase ADAM10, which was transcriptionally upregulated in response to CysC, was required for the CysC-induced sAPPα secretion. Knockdown of SIRT1 abolished CysC-triggered ADAM10 upregulation and sAPPα production. Taken together, our results demonstrated that exogenously applied CysC can direct amyloidogenic APP processing to non-amyloidgenic pathway in brain endothelial cells, mediated by proteasomal degradation of BACE1 and SIRT1-mediated ADAM10 upregulation. Our study unveils previously unrecognized protective role of CysC in APP processing.


Introduction
Alzheimer's disease (AD) is the most common neurodegenerative disorder among the elderly population. Progressive accumulation of amyloid-β peptide (Aβ) in the brain parenchyma, caused by imbalance between Aβ production and clearance, is the primary mechanism driving AD pathogenesis [1]. In more than 80% of AD individuals, Aβ is deposited within cerebral vessel wall, termed as cerebral amyloid angiopathy (CAA) [2,3]. CAA was previously interpreted as the result of insufficient clearance of neuronal Aβ from brain parenchyma in AD because cerebrovascular system is the major pathway mediating brain Aβ elimination [4,5]. Recently, several groups reported endogenous Aβ generation in brain microvascular endothelial cells [6,7], suggesting an alternative endothelial-dependent pathway in Aβ deposition in CAA.
Aβ is generated from amyloid protein precursor (APP) through sequential proteolytic cleavage. There are two mutually exclusive pathways of APP processing, amyloidogenic and nonamyloidogenic pathway [8,9]. In Aβ-forming amyloidogenic route, APP is cleaved at its N-terminus by β-secretase to generate a membrane-bound soluble C-terminal fragment, and subsequent cleavage of this C-terminal fragment by γ-secretase produces Aβ peptides predominantly including Aβ40 and Aβ42 [8,9]. In the non-amyloidogenic pathway, APP is cleaved within the Aβ peptide sequence by α-secretase, producing a soluble N-terminal fragment named as soluble amyloid protein precursor α (sAPPα) [8,9]. The balance between amyloidogenic and non-amyloidogenic APP processing is critical to pathogenesis of AD. Proteolysis through the amyloidogenic pathway is associated with accumulation of the neurotoxic Aβ peptide [10], while the non-amyloidogenic pathway not only prevents the Aβ production, but also generates sAPPα that exhibits neuroprotective properties [11,12].
Cystatin C (CysC), also known as γ-trace, is a 13-kDa secreted cysteine protease inhibitor ubiquitously expressed in all nucleated cells and presented in all body fluids [13]. CysC plays various roles in many pathological processes, including tumor metastasis, atherosclerosis, inflammatory responses and immunomodulation [13]. CysC is highly abundant in brain tissue and the alteration of CysC levels in the cerebrospinal fluid (CSF) of neurodegenerative diseases have been reported. Recently, the protective role of CysC in Aβ deposition in AD is emerging [14]. In clinically diagnosed AD patients, the levels of CysC in the CSF are reduced compared to the non-dementia controls [15]. CysC could interact with Aβ [16,17] and this interaction results in a concentration-dependent inhibition of Aβ fibril formation [17] as well as inhibition of Aβ oligomerization [18,19].
Interestingly, a novel role of CysC in intracellular APP processing was revealed in this study. We found CysC is able to shift the amyloidogenic APP processing to non-amyloidogenic pathway, causing reduced Aβ40 and increased sAPPα secretion in brain endothelial cells. Furthermore, the inhibition of Aβ40 production is mediated by CysC-induced degradation of βsecretase BACE1 (β-site APP cleaving enzyme 1) through ubiquitin/proteasome pathway. The increased sAPPα secretion is caused by upregulation of α-secretase ADAM10 (a disintegrin and metalloproteinase 10) by CysC via SIRT1 (silent information regulator 1) in brain endothelial cells.
containing protease inhibitor cocktail (Roche, Mannheim, Germany). Cells were harvested by scraping and lysed on ice for 30 min. The lysates were centrifuged for 15 min at 12000×g at 4°C. The supernatant was collected and protein concentration of each sample was quantified using BCA protein assay kit (Thermo Scientific, CergyPontoise, France). Equal amounts of samples were separated by SDS-PAGE and transferred to PVDF membranes (Millipore, Billerica, MA). The PVDF membranes were blocked with 5% nonfat milk and incubated with the primary antibody at 4°C overnight. Then the blots were incubated with a horseradish peroxidase (HRP)-conjugated secondary antibody (Santa Cruz Biotech, Santa Cruz, CA) for 1 h at room temperature. Immunoreactive bands were visualized by Super Signal West Pico Chemiluminescent Substrate (Pierce, Rockford, IL) using LAS-3000mini imaging system (Fuji Film, Tokyo, Japan). The antibodies recognizing BACE1, BACE2, NICASTRIN, PS1, PS2, APH-1, PEN-2 and ADAM10 were obtained from Abcam (Cambridge, MA). Anti-SIRT1 was from Millipore and anti-ubiquitin was from Cell Signaling Technology (Danvers, MA). For quantitative analysis, the mean density of each band was measured by Image J software, and the band density of the interested protein was divided by the density of the corresponding loading control band to obtain the normalized values. Data are plotted as percentages of the control.

Enzyme-Linked Immunosorbent Assay (ELISA)
The concentrations of Aβ40 and sAPPα in the culture medium of HBMEC were determined with ELISA kits (IBL, Gunma, Japan) according to the manufacturer's instructions.

Real-Time Reverse Transcription (RT)-PCR
The total RNA isolated with TRIzol reagent (Sigma-Aldrich, St. Louis, MO) was reverse transcribed using Moloney murine leukemia virus (M-MLV) reverse transcriptase (Promega, Madison, WI). Real-time PCR was performed on an ABI 7500 real-time PCR system (Applied BioSystems) with a SYBR premix Ex Taq kit (Takara Biotechnology, Dalian, China), according to the manufacturer's instructions. The primer sequences for BACE1 were GGCGGGAGTGG TATTATGA (forward) and TTTCTTGGGCAAACGAAGGT (reverse); primer sequences for ADAM10 were ATGGGAGGTCAGTATGGGAATC (forward) and ACTGCTCTTTTGGCA CGCT (reverse). Primers for GAPDH were GAAGGTGAAGGTCGGAGTC (forward) and GAAGATGGTGATGGGATTTC (reverse). The comparative cycle threshold (CT) method was used to calculate the relative gene expression level, with GAPDH as the internal control. Real-time PCR products were analyzed on agarose gel electrophoresis and verified by DNA sequencing.
L-glycerophosphate, 1 mM phenylmethylsulfonyl fluoride) containing protease inhibitor cocktail. The cell lysates were centrifuged and the supernatant was collected. The protein content was determined by the Bradford method. A total of 1 mg of protein was incubated with anti-BACE1 antibody (Proteintech, WuHan, China) overnight at 4°C and incubated for 2 h with protein A/G-agarose (Santa Cruz Biotech). The proteins from immune complexes were eluted in SDS sample buffer for western blot analysis.

Statistical Analysis
All values are presented as mean ± SEM of at least three independent experiments. Statistical significance between two groups was analyzed by Student's t test. One-way analysis of variance (ANOVA) or two-way ANOVA was used to compare multiple groups. A P value of <0.05 was considered significant.

CysC Affects the Releases of Aβ40 and sAPPα from Brain Endothelial Cells
To evaluate the effect of CysC on APP processing in HBMEC, the concentrations of Aβ40 and sAPPα in the culture medium (supernatant) of HBMEC was measured by ELISA. As the physiological concentrations of CysC in the CSF are 0.135-0.693 μM [21], HBMEC were treated with 0.4 μM CysC for indicated times. The results showed that CysC reduced the levels of Aβ40 in the culture medium of HBMEC in a time-dependent manner, with the decrease reaching statistical difference at 8 hr and 12 hr after CysC application ( Fig 1A). Meanwhile, the concentration of secreted sAPPα was significantly increased in HBMEC treated with CysC, reaching the peak at 8 hr ( Fig 1B). In contrast, secretion of Aβ40 and sAPPα in HBMEC in the absence of CysC showed slightly increase without statistical significance (S1 Fig). The protein expression level of APP in HBMEC was not changed upon CysC treatment (S2 Fig). Then the effect of CysC on HBMEC was examined with different concentrations of CysC. As shown in Fig 1C and 1D, the levels of Aβ40 reduced whereas sAPPα increased in HBMEC treated with increasing concentrations of CysC, both of them reached the peak at 0.4 μM CysC. These results suggested that CysC inhibited endogenous secretion of Aβ40 and promoted endogenous sAPPα secretion in brain endothelial cells.
It has been shown that oxidative stress enhanced Aβ production in HEK293 cells transfected with Swedish mutant form of APP [22,23]. To investigate whether CysC regulates APP processing in HBMEC under oxidative stress condition, HBMEC were treated with H 2 O 2 (50 μM), which did not affect cell viability (S3 Fig), to mimic the oxidative stress-induced responses, then the concentrations of Aβ40 and sAPPα in the culture medium of HBMEC were measured by ELISA. The secreted Aβ40 increased in a time-dependent manner after H 2 O 2 treatment, which was effectively abolished by pre-treatment with CysC ( Fig 1E). However, the secreted sAPPα in HBMEC was not changed by H 2 O 2 stimulation (Fig 1F), suggesting H 2 O 2 -induced oxidative stress specifically promoted Aβ40 secretion without any effect on sAPPα. In addition, similar to the findings in Fig 1B, we found the sAPPα secretion were enhanced in a timedependent manner upon CysC treatment in the presence of H 2 O 2 ( Fig 1F). These results indicated that CysC is able to regulate intracellular APP processing in brain endothelial cells.

CysC Promotes Proteasomal Degradation of BACE1 in Brain Endothelial Cells
To clarify the mechanism of CysC-triggered BACE1 reduction in H 2 O 2 -induced HBMEC, real-time RT-PCR was performed to analyze the mRNA level of BACE1. The results showed that CysC treatment did not affect BACE1 mRNA expression in H 2 O 2 -treated HBMEC ( Fig  3A), suggesting the decrease of BACE1 protein induced by CysC was caused by degradation of intracellular BACE1 protein. It has been revealed that BACE1 can be degraded via the ubiquitin-proteasome pathway [27] as well as the lysosomal pathway [28]. Thus, to determine the involved pathway for the CysC-induced BACE1 reduction in H 2 O 2 -treated HBMEC, the cells were pre-incubated with proteasome inhibitor MG132 and lysosomal inhibitors including chloroquine and NH 4 Cl, lysed and subjected to western blot to measure BACE1 levels. We found that MG132 treatment significantly attenuated the CysC-induced BACE1 reduction in H 2 O 2 -treated HBMEC, whereas chloroquine and NH 4 Cl had no such effect (Fig 3B), suggesting CysC-induced BACE1 reduction was caused by ubiquitin-proteasome pathway. Then, H 2 O 2 -treated HBMEC in the absence or presence of CysC were subjected to immunoprecipitation assay with BACE1 antibody, and the precipitates were examined by western blot using ubiquitin antibody (Fig 3C). We found a significant high level of ubiquitinated BACE1 in H 2 O 2treated HBMEC pre-incubated with CysC compared to HBMEC exposed to H 2 O 2 alone. These results demonstrated that CysC promotes BACE1 degradation through the ubiquitin/proteasome pathway in brain endothelial cells under oxidative stress conditions.

CysC-Induced sAPPα Secretion Is Associated with α-Secretase ADAM10 in Brain Endothelial Cells
It is known that sAPPα is the non-amyloidogenic product of APP cleaved by α-secretases. ADAM10, a transmembrane metalloprotease, has been demonstrated as the major α-secretase producing sAPPα [29,30]. To determine the mechanism of increased sAPPα secretion induced by CysC (Fig 1B, 1D and 1F), the expression of ADAM10 in HBMEC treated with CysC was assessed by western blot. We found the protein levels of ADAM10 were significantly elevated in HBMEC upon CysC treatment, reaching the peak at 8 hr after treatment (Fig 4A). Then siRNA-mediated RNA interference was used to knockdown ADAM10 in HBMEC (Fig 4B and  4C). The ADAM10 siRNA were synthesized and transiently transfected into HBMEC, and the knockdown effect was evaluated by western blot. The results showed that ADAM10 in HBMEC was reduced by two different siRNA recognizing ADAM10 (Fig 4B and 4C) compared to the non-silencing siRNA control. Also, we found the CysC-induced ADAM10 upregulation in HBMEC was effectively abolished by ADAM10 knockdown (Fig 4B and 4C). Then, HBMEC with ADAM10 knockdown were incubated with or without CysC followed by measurement of sAPPα secretion in the culture medium. As shown in Fig 4D, ADAM10 knockdown in HBMEC significantly prevented the CysC-induced sAPPα secretion compared to the control. These results indicated that ADAM10 is essential for the CysC-promoted sAPPα secretion in brain endothelial cells.

CysC Upregulates ADAM10 mRNA via SIRT1 to Promote sAPPα Secretion in Brain Endothelial Cells
To further dissect the mechanism of increased ADAM10 protein expression induced by CysC, the mRNA levels of ADAM10 were analyzed by real-time RT-PCR. The results showed that ADAM10 mRNA were significantly increased in HBMEC incubated with CysC ( Fig 5A). The ADAM10 mRNA increased to reach statistical significance at 2 hr time point after CysC  treatment, which is earlier than the 8 hr time point in ADAM10 protein changes (Fig 4A). Also, the peak of ADAM10 mRNA increase, which was at 4 hr after CysC treatment, is earlier than the peak time (8 hr) of ADAM10 protein changes (Fig 4A). Thus, the increased ADAM10 mRNA occurred earlier than the changes of ADAM10 protein levels in HBMEC upon CysC treatment.
In neuroblastoma N2a cells expressing human APP Swedish mutant, overexpression of the SIRT1 gene increased ADAM10 protein expression [31]. These prompt us to test whether the CysC-induced ADAM10 upregulation is caused by SIRT1. Western Blot results showed that CysC significantly promoted protein levels of SIRT1 in HBMEC in a time-dependent manner (Fig 5B). To verify that CysC-induced ADAM10 upregulation is mediated by SIRT1, siRNA targeting to SIRT1 were transfected into HBMEC to reduce SIRT1 protein levels in HBMEC. The subsequent results showed that the mRNA and protein levels of ADAM10 were significantly attenuated in HBMEC with silenced SIRT1 in response to CysC treatment (Fig 5C and  5D). In other words, SIRT1 knockdown effectively prevented the CysC-induced ADAM10 upregulation in HBMEC. Moreover, CysC failed to promote the secretion of sAPPα in HBMEC with silenced SIRT1 compared to the non-silencing siRNA control (Fig 5E). These results demonstrated that CysC upregulates ADAM10 at transcriptional level, mediated by SIRT1 signaling, to facilitate sAPPα secretion in brain endothelial cells.

Discussion
Aβ is a proteolytic product of sequential cleavage of APP protein by secretases. In AD, pathological Aβ deposition in the brain forms senile plaques and Aβ accumulation in cerebral vessel wall produces CAA, both of which are the characteristic lesions of AD [1,3]. Aβ40 and Aβ42 are the predominant Aβ species with quite similar sequences, and the only difference between them is an extra isoleucine and analanine at the C-terminus of Aβ42. Aβ42 is more amyloidogenic than Aβ40, and is deposited earlier than Aβ40 in the brain parenchyma in AD patients. Aβ42 is the major isoform in the amyloid plaque in the brain of AD, whereas Aβ40 aggregates are predominantly found in the vascular wall in CAA [1,3,9]. Modulating the processing of APP has important implications for intervention strategies to prevent Aβ deposition in AD. In this study, we found CysC reduced Aβ40 secretion via proteasomal degradation of β-secretase BACE1 in brain endothelial cells. Meanwhile, CysC promoted sAPPα release by transcriptional upregulation of α-secretase ADAM10. Thus CysC is able to shift the balance of APP processing from the amyloidogenic β-cleavage to non-amyloidogenic α-cleavage, producing less Aβ40 and more sAPPα in brain endothelial cells.
Several compounds (synthetic or natural) have been shown to switch APP processing to nonamyloidogenic route. L-3-n-butylphthalide (L-NBP), an extract from seeds of Apium graveolens Linn (Chinese celery), promotes sAPPα release and reduces Aβ generation in neuroblastoma cells after 24-hr treatment [32]. Neuroprotectin D1 (NPD1), a stereoselective mediator derived from the omega-3 essential fatty acid docosahexaenoic acid (DHA), suppresses Aβ42 peptide shedding and upregulates intracellular sAPPα expression in neuronal-glial co-cultured cells over-expressing APP sw (Swedish double mutation with K595N and M596L) after 48-hr treatment [33]. Carbachol, a muscarinic receptor agonist, caused an increase of sAPPα secretion in teratocarcinoma-derived neurons, as well as a decrease in Aβ production in the medium [34]. To our knowledge, CysC is the first naturally occurring protein described to direct APP metabolism from the amyloidogenic pathway towards non-amyloidogenic pathway. In the brain, CysC was found to be present in neurons and microglial cells [35,36] (but not in astrocytes [36]), whereas the expression of CysC in brain endothelial cells was undetectable (S4 Fig), these support the scenario that neuronal cell-derived extracellular CysC acts directly on brain endothelial cells via paracrine mechanism to affect endothelial processing of APP.
The association of CysC with brain disorders has been reported. Hereditary CysC amyloid angiopathy (HCCAA), also called hereditary cerebral hemorrhage with amyloidosis of Icelandic type, is an autosomal dominant form of CAA. The amyloid deposition in the vessel walls caused fatal brain hemorrhage in normotensive young adults because of a Leu68Gln mutation in CysC [37]. Also, variant B of CysC, containing a single mutation A25T, is associated with age-related macular degeneration (AMD) and AD [38]. A most recent study indicated that variant B of CysC is inefficiently secreted which impairs its protective effect against Aβ aggregation [39]. CysC was identified to interact with APP protein within the Aβ region [17]. The influence of CysC on Aβ aggregation was studied and the results showed CysC could inhibit formation of Aβ fibrils [17,40,41] and Aβ oligomers [18,19]. These studies suggested CysC may exert protective effect against Aβ deposition in AD. In this study, we found endogenous Aβ production was reduced upon CysC administration due to proteasomal degradation of β-secretase BACE1 in brain endothelial cells. Our findings thus unveil a previously unrecognized effect of CysC to reduce Aβ secretion. In addition, CysC stimulates release of sAPPα in brain endothelial cells. It has been demonstrated that sAPPα has protective properties against glucose deprivation, glutamate neurotoxicity [11] and Aβ-induced oxidative injury [42] in cultured neurons as well as ischemic injury of rat hippocampus in vivo [43]. Thus, the ability of CysC to reduce Aβ secretion and promote sAPPα release indicated its protective function, which is in line with the neuroprotective effect of CysC in AD [14].
When the concentration reached to 0.4 μM, CysC significantly inhibited Aβ40 secretion ( Fig  1C) and promoted sAPPα release (Fig 1D) in HBMEC. Interestingly, with the increase of concentration to 0.8 and 2.0 μM, the effect of CysC on Aβ and sAPPα secretion was less prominent than at 0.4 μM (Fig 1C and 1D). These indicated the effect of CysC on APP processing is strictly associated with its concentrations and is saturated at 0.4 μM. Similar to our findings, Martinez-Vargas et al. found that lower-dose (3.5 fmoles) injection of CysC into the rat brain with traumatic injury reduced bleeding and mortality, whereas high doses (35 and 175 fmoles) had little effect on bleeding and mortality [44]. Based on these results, we recommend to be more cautious regarding the concentration of CysC used in the evaluation of the effect of exogenously applied CysC.
In Pawlik et al.'s pioneer study on CysC, transgenic mice expressing either wild-type or the Leu68Gln variant CysC genes were generated [45]. They found that the CysC transgenic mice are fertile and their appearance are indistinguishable from littermate controls. Those mice showed no obvious behavioral defects, without any gross pathological or histopathological abnormalities up to 6 month of age. Similar levels of Aβ40 and Aβ42 were found in the brain homogenates of CysC transgenic mice compared to littermate controls [45], which appeared inconsistent with our study. This discrepancy may reflect the different manipulations in Pawlik M et al.'s and our study. The acute application of recombinant CysC to treat brain endothelial cells in our study revealed that CysC caused a rapid reduction of Aβ40 secretion within a short time window, from 4 hr to 12 hr after application of protein CysC. In contrast, the in vivo overexpression of CysC in 3-8 month transgenic mice has little effect on brain Aβ level [45] are likely due to developmental compensation that could mask the acute effect of CysC during the 3-8 month development. In addition, we used brain endothelial cells to analyze its Aβ secretion in response to CysC treatment, which is different from Pawlik M et al.'s study in which they measured the Aβ level in the whole brain homogenates [45].
So far the effects of CysC on brain Aβ levels were more complicated than expected. It has been reported that overexpression of CysC reduced plaque loads without affecting soluble brain Aβ levels in mice [40,41]. Surprisingly, Sun et al. found that both the soluble Aβ levels and plaque load were reduced in CysC knockout mice due to cathepsin B-induced Aβ degradation [46]. In this study, our results showed that application of recombinant CysC protein decreased Aβ40 secretion in brain endothelial cells. It is difficult to reconcile these puzzling findings of CysC with current understandings of CysC. Thus further study is necessary to clarify the effect of CysC on Aβ metabolism as well as the underlying mechanism.
BACE1 is the major β-secretase enzyme for the production of Aβ from proteolytic processing of APP [8,9]. We found the increased BACE1 in brain endothelial cells upon H 2 O 2 stimulation was significantly attenuated by CysC (Fig 2C). In contrast, the H 2 O 2 -induced increase of γ-secretases (including NICASTRIN, PS1, PS2, APH-1 and PEN2) remained unchanged after CysC application (Fig 2C). These suggested that CysC specifically down-regulates BACE1 expression in brain endothelial cells. Moreover, we found CysC could effectively reduce H 2 O 2induced Aβ secretion (Fig 1E) though the γ-secretases remained increased. Thus in this context, we concluded that β-secretase BACE1 is the critical enzyme in the production of Aβ from APP in brain endothelial cells. This is compatible with previous findings that BACE1 processing is the key step for Aβ generation in the brain [47,48].
The expression of BACE1 was modulated by transcriptional and post-transcriptional controls [49]. Here we found the protein (Fig 2A), but not mRNA (Fig 3A), levels of BACE1 were elevated in brain endothelial cells upon H 2 O 2 stimulation, which was significantly attenuated by CysC (Fig 2C), suggesting CysC affects BACE1 expression by post-transcriptional regulation. Our further results revealed that CysC promotes proteasomal degradation of BACE1 (Fig 3B and  3C) which points out post-translational modification of BACE1 initiated by CysC. Previous studies have documented the transcriptional control of BACE1, whereas the post-transcriptional regulation of BACE1 was reported until recently. It was shown that peroxisome proliferatoractivated receptor-γ coactivator 1 (PGC)-1α and E3-ligase CHIP promotes BACE1 degradation via proteasomal pathway [50,51]. Regarding the CysC-induced BACE1 proteasomal degradation in brain endothelial cells, whether it is dependent on PGC-1α or CHIP, or an alternative unrecognized signaling pathway, remains to be determined in future studies.
In summary, our study demonstrated that CysC reduces Aβ40 secretion and facilitates sAPPα secretion in brain endothelial cells. The inhibition of Aβ40 secretion is caused by the CysC-induced degradation of BACE1 through the ubiquitin/proteasome pathway, whereas the enhanced sAPPα secretion is due to increased expression of ADAM10 mediated by SIRT1. Our findings point out the novel role of CysC in APP processing which suggests a potential therapeutic application in AD.