Research ArticleCell Biology

SERP1 is an assembly regulator of γ-secretase in metabolic stress conditions

See allHide authors and affiliations

Science Signaling  17 Mar 2020:
Vol. 13, Issue 623, eaax8949
DOI: 10.1126/scisignal.aax8949

Stress and amyloid

The accumulation of β-amyloid (Aβ) plaques in the brain contributes to the symptoms of Alzheimer’s disease (AD). Aβ is generated by the γ-secretase–mediated cleavage of amyloid precursor protein (APP), and γ-secretase activity is induced by cell stress, such as occurs with obesity and diabetes, which are major risk factors for AD. However, the use of γ-secretase inhibitors is limited by their disruption of other γ-secretase substrates. Using a gain-of-function screen in cells, diabetic AD model mice, and postmortem samples from patients with AD, Jung et al. identified the protein SERP1 as a cell stress–induced, APP-biased activator of γ-secretase. Knocking down SERP1 in cells or the mouse hippocampus decreased Aβ production, suggesting that blocking SERP1 might be a more selective therapeutic to reduce Aβ plaque formation in patients with AD.


The enzyme γ-secretase generates β-amyloid (Aβ) peptides by cleaving amyloid protein precursor (APP); the aggregation of these peptides is associated with Alzheimer’s disease (AD). Despite the development of various γ-secretase regulators, their clinical use is limited by coincident disruption of other γ-secretase–regulated substrates, such as Notch. Using a genome-wide functional screen of γ-secretase activity in cells and a complementary DNA expression library, we found that SERP1 is a previously unknown γ-secretase activator that stimulates Aβ generation in cells experiencing endoplasmic reticulum (ER) stress, such as is seen with diabetes. SERP1 interacted with a subcomplex of γ-secretase (APH1A/NCT) through its carboxyl terminus to enhance the assembly and, consequently, the activity of the γ-secretase holoenzyme complex. In response to ER stress, SERP1 preferentially recruited APP rather than Notch into the γ-secretase complex and enhanced the subcellular localization of the complex into lipid rafts, increasing Aβ production. Moreover, SERP1 abundance, γ-secretase assembly, and Aβ production were increased both in cells exposed to high amounts of glucose and in diabetic AD model mice. Conversely, Aβ production was decreased by knocking down SERP1 in cells or in the hippocampi of mice. Compared to postmortem samples from control individuals, those from patients with AD showed increased SERP1 expression in the hippocampus and parietal lobe. Together, our findings suggest that SERP1 is an APP-biased regulator of γ-secretase function in the context of cell stress, providing a possible molecular explanation for the link between diabetes and sporadic AD.


Alzheimer’s disease (AD) is a progressive neurodegenerative disorder and the most common form of dementia (1). Two major pathological hallmarks of AD are the presence of plaques composed of β-amyloid (Aβ) peptides (2) and neurofibrillary tangles consisting of hyperphosphorylated tau protein (3). The neurotoxic Aβ peptides are derived from Aβ precursor protein (APP) through sequential cleavage by β-site APP-cleaving protein (BACE1) and γ-secretase. γ-Secretase is a rate-limiting enzyme composed of four core subunits: presenilin 1 (PS1) or PS2, nicastrin (NCT), anterior pharynx-defective phenotype 1 (APH1), and PS enhancer 2 (PEN2) (4, 5). Among these, NCT functions to sterically hinder non–ectodomain-shed substrates (6), and APH1 serves as a scaffold to join PS and NCT (7). To date, additional proteins—such as TMP21 that is a type I transmembrane protein of the p24 cargo protein family, heterotrimeric GTP-binding protein–coupled receptor 3 (GPR3), and β-arrestin—have been identified as regulators of γ-secretase (810), but their in vivo role in AD pathogenesis remains to be elucidated. In addition, the presence of several substrates of γ-secretase limits therapeutic development of γ-secretase inhibitors (11). Thus, identifying an APP processing–biased regulator of γ-secretase may reveal a new drug development option for AD.

The activity of γ-secretase is regulated by many physiological and pathological signals, including hypoxia, growth factor signaling, reactive oxygen species, and endoplasmic reticulum (ER) stress (1216). Especially, ER stress caused by aging, environmental factors, or genetic mutations is highly associated with neurodegenerative disorders, including AD. In addition, metabolic diseases, such as obesity and diabetes, are also major risk factors for late-onset AD (1720). Epidemiologic studies have suggested that people with diabetes mellitus have a 1.5- to 2.5-fold greater risk of AD than those without diabetes (21). Aβ production is enhanced by high blood glucose levels (or hyperglycemia), and insulin resistance associated with ER stress (22, 23) and Aβ deposition is increased in AD model mice with either diabetic phenotypes (20) or diet-induced insulin resistance (24). Here, we explored the molecular link between AD pathogenesis and ER stress and diabetes by identifying a factor that regulates γ-secretase activity under these conditions.


Stress-associated ER protein 1 enhances Aβ generation through the regulation of γ-secretase activity

To identify a crucial genetic factor regulating γ-secretase activity under stressful and pathologic conditions, we previously performed a genome-wide functional screen in human embryonic kidney (HEK) 293T cells using a γ-secretase activity–based cell assay with a cDNA expression library encoding thousands of membrane proteins (25). This gain-of-function screen (GOFS) (table S1 and data file S1) identified stress-associated ER protein 1 (SERP1), a single-pass transmembrane protein, as a promoter of γ-secretase activity. Relatively little is known about the function of SERP1 aside from roles in ER stress (26) and protein modification in the liver (27, 28).

Given this link between SERP1 and γ-secretase, we tested whether the expression of SERP1 affects Aβ generation by assessing the amounts of secreted Aβ40 and Aβ42, by enzyme-linked immunosorbent assay (ELISA), in SH-SY5Y cells expressing the Swedish mutant form of APP (APPswe), which causes early-onset familial AD. Consistent with the results of the GOFS, ectopic expression of SERP1 notably increased the amount of Aβ40 and Aβ42 by 30 to ~50% in the conditioned media (Fig. 1A). Conversely, knockdown of SERP1 expression reduced the secretion of Aβ40 and Aβ42 (Fig. 1B). When similar assays were performed in CHO-7PA2 cells, which express human V717F APP mutant (29), Western blotting showed that SERP1 increased total Aβ abundance by 70% in the media, whereas levels of full-length APP and α-secretase–cleavage products of APP [sAPPα and C83 (α-APP-CTF)] were not affected (Fig. 1C). These results suggest that SERP1 stimulates Aβ production in cultured cells.

Fig. 1 SERP1 stimulates γ-secretase activity for Aβ generation.

(A and B) Amount of Aβ40 and Aβ42 in the media of SY5Y-APPswe cells transfected with pSERP1-HA (A) and HEK293T/APPswe cells transfected with SERP1 shRNA (B) and treated with 10 nM Compound E (Comp.E) for 24 hours, measured with an ELISA kit. (C) Conditioned media from pSERP1-HA–transfected cells was subjected to immunoprecipitation (IP) assay using preimmune serum (Pre) or Aβ antibody (4G8), and the immunoprecipitates and whole-cell lysates (WCL) were analyzed with Western blotting (top). The immunoprecipitated Aβ abundance was quantified. Data are means ± SD; n = 3 experiments. *P < 0.05 and ***P < 0.001 by unpaired t test. (D and E) HEK293T cells were transfected with pSERP1-HA (D) or SERP1 shRNA (E) or the respective control, and 1% CHAPS-soluble membrane fractions were subjected to enzyme assay using fluorogenic substrate in the presence or absence of Compound E. Data are means ± SD; n = 5 (D) and n = 3 (E). *P < 0.05, **P < 0.01, and ***P < 0.001 by unpaired t test.

To examine whether SERP1 affected cleavage of C99 (β-APP-CTF) by γ-secretase, in vitro enzyme assay and APP intracellular domain (AICD) generation assays were performed. An in vitro enzyme assay using a fluorogenic substrate in HEK293T cell extracts revealed that SERP1 overexpression increased γ-secretase activity by 50% (Fig. 1D). Examination of enzymatic activity in the microsomal membrane fractions using purified C99 substrate showed that SERP1 overexpression resulted in marked increase of AICD generation, which was inhibited by treatment with γ-secretase inhibitor, Compound E (fig. S1A). Conversely, knockdown of SERP1 expression reduced γ-secretase activity by 30% (Fig. 1E). Consistently, in vitro AICD generation assay revealed reduction of AICD generation by SERP1 knockdown in HEK293 cells overexpressing wild-type (WT) APP695 (fig. S1B). Collectively, these results indicate that SERP1 enhances APP processing through the increase of γ-secretase activity.

SERP1 negatively regulates Notch processing by γ-secretase

Because SERP1 increased γ-secretase activity and consequent Aβ generation, we examined whether SERP1 affected the cleavage of other substrates of γ-secretase. Specifically, we focused on its effect on γ-secretase–mediated Notch cleavage. Cytoplasmic pattern analysis of green fluorescent protein (GFP)–tagged C-terminal 99 amino acids of APP (SC100-GFP) under fluorescence microscopy revealed that, as reported previously (30), AICD-GFP, the cleaved peptide from SC100-GFP by γ-secretase, was observed with a diffused distribution in mouse embryo fibroblasts (MEFs) (Fig. 2A). When MEFs were treated with Compound E, SC100-GFP was observed as a small vesicle pattern because the generation of AICD-GFP was inhibited. In contrast, SC100-GFP was observed with numerous small vesicles in Serp1 knockout (KO) MEFs, the same pattern observed in cells after inhibition of γ-secretase with Compound E (Fig. 2A). Compared to the pattern of SC100-GFP, GFP-tagged Notch intracellular domain (NICD-GFP) generated from the NotchΔE-GFP construct accumulated in the nucleus of WT MEFs, as also observed previously (10), and this pattern was not altered in Serp1 KO MEFs (Fig. 2B), indicating that Notch is processed by γ-secretase even in the absence of SERP1. In contrast, Western blot analysis showed that SERP1 overexpression rather decreased NICD generation by about 30% (Fig. 2C). Furthermore, mRNA levels of NICD downstream genes—such as Hes1, Hey1, and Tcf4—were also reduced by SERP1 overexpression (Fig. 2D). Thus, SERP1 may negatively regulate the processing of Notch by γ-secretase.

Fig. 2 SERP1 reduces NotchΔE processing and NICD target gene expression.

(A and B) WT and SERP1 KO MEFs were transfected with pSC100-GFP (A) or pNotchΔE-GFP (NΔE-GFP) (B) and then treated with Compound E (10 nM) for another 24 hours. Fluorescence signals were observed by confocal microscopy. Scale bars, 10 μm. Images are representative of three experiments. (C) HEK293T cells were transfected with pNΔE-GFP and pSERP1 for 24 hours with or without Compound E (1 μM). Cell extracts were analyzed by Western blotting (left), and the amount of NICD in control and SERP1-overexpressed lanes without Compound E (middle two lanes) was quantitated by densitometry analysis (right). (D) Total RNA from SERP1-HA–overexpressed HEK293T cells was analyzed by RT-PCR (left), and the signals on the blots were quantitated (right). (E and F) HeLa cells were transfected for 48 hours with pSC100-GFP, pAPH1A-RFP, and either pSERP1-HA or SERP1 shRNA (E) or with pNΔE-GFP, pAPH1A-RFP, and either pSERP1-HA or SERP1 shRNA (F). Fluorescence signals were observed under a confocal microscope (top). Scale bars, 10 μm. Pearson’s correlation coefficients (PCCs) of SC100-GFP/APH1A-RFP [(E), bottom] and NΔE-GFP/APH1A-RFP [(F), bottom] images (random) were measured using ImageJ software. Data are means ± SD; n = 4 (C) and n = 3 (D to F). *P < 0.05 and **P < 0.01 by unpaired t test.

These distinct effects of SERP1 on the cleavage of C99 and NotchΔE by γ-secretase enabled us to further examine whether SERP1 could differently affect the docking of γ-secretase with these substrates. From colocalization assays, we determined that the colocalization of SC100-GFP with red fluorescent protein (RFP)–tagged APH1A (APH1A-RFP) was increased by SERP1 overexpression but slightly decreased by SERP1 knockdown (Fig. 2E). In contrast, the effect of SERP1 on the colocalization of NotchΔE-GFP with APH1A-RFP was opposite; the colocalization of NotchΔE-GFP with APH1A-RFP was decreased by SERP1 overexpression but increased by SERP1 knockdown (Fig. 2F). Furthermore, physical interaction of PS1-N-terminal fragment (NTF) and NotchΔE-GFP was decreased in HEK293T cells overexpressing SERP1 (fig. S2, A and B). Together, the results suggest that SERP1 might differently affect docking of γ-secretase with its substrates, which makes it more favorable to cleave APP than Notch.

SERP1 increases γ-secretase complex formation by stabilizing the APH1A/NCT subcomplex

To investigate the molecular mechanism underlying SERP1-mediated stimulation of γ-secretase activity, the expression level of each subunit in the γ-secretase complex was first examined by Western blotting. We found that amounts of APH1A and NCT were increased after SERP1 overexpression (fig. S3A). Next, we performed reverse transcription polymerase chain reaction (RT-PCR) analysis to address the question of whether the increase of γ-secretase subunits by SERP1 resulted from transcriptional regulation. The results revealed that SERP1 expression had no significant effect on mRNA levels of γ-secretase subunits (fig. S3C). Furthermore, determination of the amounts of γ-secretase complex using a blue native (BN)–polyacrylamide gel electrophoresis (PAGE) assay revealed that SERP1 expression notably increased the amount of mature γ-secretase complex with ~440 kDa by about 50% (Fig. 3A). Consistent to these observations, genetic deletion of SERP1 markedly reduced the amounts of mature γ-secretase complex (Fig. 3B) and γ-secretase components (fig. S3D). These results indicate that SERP1 positively regulates the amount of γ-secretase complex.

Fig. 3 SERP1 increases the amounts of APH1A/NCT subcomplex and γ-secretase complex.

(A and B) Amount of γ-secretase holoenzyme in response to SERP1 abundance. Digitonin (1%)–soluble crude membrane fractions prepared from HEK293T cells transfected with pSERP1-HA (A) or psgSERP1 (B) were separated by blue native (BN)-PAGE or SDS-PAGE and analyzed by Western blotting (left), and the signals on the blots were quantitated by densitometry analysis (right). (C and D) Formation of APH1A/NCT subcomplex in response to SERP1 abundance. Digitonin (1%)–soluble membrane fractions prepared from PS1/2 DKO MEF cells transfected with pSERP1-HA (C) or psgSERP1 (PS1/2 DKO/SERP1 KO MEFs) (D) were separated by BN-PAGE and analyzed by Western blotting (left). The signals on the blots were quantitated by densitometry analysis (right). (E to G) Localization of γ-secretase components to lipid rafts in response to SERP1 abundance. HEK293T cells were transfected with pSERP1-HA for 24 hours and solubilized in 1% CHAPS lysis buffer. Equal amount of the soluble lysates was subjected to discontinuous (5, 35, and 45%) sucrose gradient fractionation assay (top, fraction 1; bottom, fraction 12), and each fraction was analyzed by Western blotting (E). DRM fractions (fractions 3 to 5) and non-DRM fractions (fractions 10 to 12) were collected and then analyzed by Western blotting (“D,” DRM; “N,” non-DRM) (F). The relative intensity of PS1-NTF was measured by densitometry analysis (G). Data in all panels are means ± SD; n = 3. *P < 0.05 by unpaired t test.

These results led us to further investigate the effect of SERP1 on the formation of γ-secretase subcomplex in PS1/2 double-KO (PS1/2 DKO) MEFs. The results of the BN-PAGE analysis revealed that the cellular level of APH1A/NCT subcomplex with ~240 kDa was markedly increased by SERP1 overexpression (Fig. 3C) but was decreased by SERP1 knockdown in PS1/2 DKO MEFs (Fig. 3D). Collectively, these results suggest that SERP1 positively regulates the amount of APH1A/NCT subcomplex and, hence, that of the γ-secretase complex.

To investigate how SERP1 could differentially regulate the docking of γ-secretase complex with its substrates, we focused on the subcellular distribution of γ-secretase complex in lipid rafts on the membrane. Previous studies have suggested that Notch processing occurs in nonlipid rafts (31). Cellular fractionation assay revealed that the locations of γ-secretase components in the detergent-resistant membrane (DRM) fraction were increased by SERP1 overexpression in HEK293T cells (Fig. 3, E to G) but was reduced in SERP1-knockdown cells (fig. S3B). These results imply that increase of γ-secretase complex in lipid rafts by SERP1 might differently affect its docking with APP and Notch.

SERP1 interacts with γ-secretase through the APH1A/NCT subcomplex

To investigate how SERP1 regulated the formation of γ-secretase complex, we addressed whether SERP1 physically interacted with γ-secretase. The immunoprecipitation assays revealed that γ-secretase subunits were coimmunoprecipitated together with SERP1 in the cortical extracts of mouse brain (Fig. 4A) and in the transfected HEK293T cells (fig. S4A). In addition, SERP1 comigrated on the BN gel together with γ-secretase holoenzyme complex at ~440 kDa in WT MEFs and with APH1A/NCT subcomplex at ~240 kDa in PS1/2 DKO MEFs (Fig. 4B). Using a glycerol velocity gradient fractionation assay of cell extracts, we also found that SERP1 was detected in the fractions overlapping with APH1A/NCT subcomplex and in the fractions containing γ-secretase complex as well (fig. S4B). Moreover, a protein overlay assay revealed that purified His-tagged SERP1 protein interacted in vitro with γ-secretase components in mouse brain lysates (fig. S4C). These results show that SERP1 physically interacts with γ-secretase through the APH1A/NCT subcomplex.

Fig. 4 SERP1 forms a complex with γ-secretase through interacting with APH1A/NCT.

(A) Mouse cortical tissue lysates were subjected to immunoprecipitation assay with immunoglobulin G (IgG) or antibody to SERP1, followed by Western blotting. L.C., light chain of immunoglobulin. (B) Detection of endogenous SERP1 in γ-secretase subcomplex in membrane fractions prepared from WT and PS1/2 DKO MEFs. Samples were separated by BN-PAGE and analyzed by Western blotting. (C and D) Assessment of SERP1-HA interaction with APH1A and NCT. HEK293T cells were cotransfected with pSERP1-HA and pAPH1A-FLAG (C) or pNCT-V5 (D), and cell lysates (1% Triton X-100) were analyzed by immunoprecipitation assay. H.C., heavy chain of immunoglobulin. (E) PS1-Myc and SERP1-HA–overexpressed HEK293T cells were solubilized in either 1% CHAPS or 1% Triton X-100 lysis buffer and analyzed by immunoprecipitation assay. Blots in (A) to (E) are representative of three experiments. (F) Schematic diagram of SERP1 deletion and chimeric mutants. TMD, transmembrane domain. (G) HEK293T cells were transfected with pNCT-V5 and pSERP1 mutants and then analyzed by immunoprecipitation assay. Blots are representative of three experiments. H.C., heavy chain of immunoglobulin. (H) BiFC assay showing the interactions between SERP1 and NCT in HeLa cells. Scale bars, 10 μm. Images are representative of three experiments. (I) The effect of SERP1 C-terminal region on γ-secretase subcomplex. Membrane fractions prepared from SERP1 WT or SERP1ΔC-overexpressed PS1/2 DKO MEFs were separated by BN-PAGE (top). The amount of APH1A was quantified (bottom). Data are means ± SD; n = 3. *P < 0.05 by unpaired t test.

Next, we analyzed which of the APH1A/NCT subunits interacted with SERP1 by using different detergents. A BN-PAGE assay allowed detection of APH1A/NCT subcomplex in the 1% CHAPS/digitonin-soluble fraction but not in the 1% Triton X-100–soluble fraction (fig. S5A), indicating that APH1A/NCT subcomplex is dissociated by 1% Triton X-100 (fig. S5B). Under this condition using 1% Triton X-100, the results from immunoprecipitation assay revealed that SERP1-HA (human influenza hemagglutinin) interacted with APH1A-FLAG or NCT-V5 epitope tag (Fig. 4, C and D). As expected, SERP1-HA also interacted with PS1-NTF of γ-secretase complex in the 1% CHAPS-soluble membrane fraction (Fig. 4E). However, such interaction was not observed in the 1% Triton X-100–soluble fraction. In addition, we observed colocalization of SERP1-RFP and APH1A-GFP in HeLa cells (fig. S6). Thus, we conclude that SERP1 directly interacts with both APH1A and NCT.

Although the functionally conserved domain of SERP1 is not identified yet, SERP1 consists of the N terminus (residues 1 to 35), transmembrane region (residues 35 to 61), and the C-terminal lumenal domain (residues 62 to 66) (32). Because SERP1 was too small to generate deletion constructs, we generated chimeric constructs replacing each domain with a domain from SEC61G that has similar topology to SERP1 (Fig. 4F) (33). To identify the SERP1 region involved in the interaction with NCT, we performed immunoprecipitation assay using SERP1 and its chimeric constructs. To our surprise, the SERP1ΔC mutant lacking the C-terminal four residues did not interact well with V5-tagged NCT (Fig. 4G). In addition, the SERP1ΔNTM mutant carrying the C-terminal four residues of SERP1 at its C terminus strongly bound to NCT. To confirm this observation, we used bimolecular fluorescence complementation (BiFC) assay, which allowed fluorogenic detection of the protein-protein interaction between N-terminal fragment of Venus-tagged NCT (NCT-VN) and C-terminal fragment of Venus-tagged SERP1 (VC-SERP1) in living cells (34). When both NCT-VN and VC-SERP1 were coexpressed in HeLa cells, fluorescence complementation was observed under fluorescence microscope (Fig. 4H). On the contrary, the fluorescence was not observed in cells expressing both NCT-VN and VC-SERP1ΔC. Furthermore, compared to SERP1 WT, SERP1ΔC mutant had little effect on the formation of γ-secretase subcomplex in PS1/2 DKO MEFs (Fig. 4I) and the activity of γ-secretase in HEK293 cells (fig. S7, A and B). The results suggest that the C terminus of SERP1 is critical for the interaction with NCT and regulation of γ-secretase activity.

SERP1 mediates ER stress–dependent γ-secretase activation and Aβ generation

Because it was previously reported that SERP1 is up-regulated under ER stress (35) and ER stress enhances γ-secretase activity (14), we examined a role of SERP1 in the regulation of γ-secretase activity under ER stress. As reported, monitoring the effect of ER stress on Aβ generation confirmed that the treatment with thapsigargin, an inhibitor of the ER Ca2+ channel, increased SERP1 expression (fig. S8A) and promoted Aβ40 and Aβ42 production in HEK-APP695 cells (Fig. 5, A and B). On the other hand, ER stress did not stimulate Aβ40 and Aβ42 production in SERP1-knockdown cells. In addition, the results of enzyme assays showed similar results as Aβ generation that the treatment with thapsigargin increased the DNA-binding/transactivation domain-tagged C99 (C99-GVP)/upstream activator sequence (UAS)–luciferase reporter activity reflecting γ-secretase activity (Fig. 5C) and AICD generation (Fig. 5D) in control cells but not in SERP1-knockdown HEK-APP695 cells. Thus, SERP1 increases Aβ generation through γ-secretase activation under ER stress.

Fig. 5 ER stress increases γ-secretase activity and Aβ generation via SERP1.

(A and B) HEK293T cells were cotransfected with pAPPswe-FLAG and either pSuper-neo or SERP1-shRNA (shSERP1), incubated in 0.5 μM thapsigargin (Tg) for another 24 hours, and the conditioned medium was analyzed for Aβ40 (A) and Aβ42 (B) using an ELISA kit. (C) HEK293T/shSERP1 cells were transfected with pC99-GVP and pUAS-luciferase, treated with 1 μM thapsigargin for 24 hours, and luciferase reporter assay was performed. (D and E) HEK293-APP695 (D) and HEK293T (E) cells were transfected with pshSERP1 and treated with 0.5 μM thapsigargin or 0.5 μM tunicamycin (Tuni) for 24 hours. Membrane fractions were subjected to AICD generation assay [(D), left] and Western blotting [(E), top]. The signals on the blots were quantitated by densitometry analysis [(D), right; (E), bottom]. (F) Effect of SERP1 on the formation of γ-secretase complex during ER stress. Crude membrane fractions prepared from HeLa/shCtrl and HeLa/shSERP1 cells after treatment with thapsigargin (0.5 μM for 24 hours) were separated by BN-PAGE for Western blot analysis (left), and the signals on the blots were quantitated by densitometry analysis (right). Data are means ± SD; n = 3 experiments (A to C, E, and F) or n = 4 experiments (D). *P < 0.05 and **P < 0.01 by unpaired t test.

Furthermore, Western blotting–based examination of γ-secretase component levels showed that thapsigargin treatment increased the amount of APH1A in control cells but that this increase was impaired in SERP1-knockdown cells (Fig. 5E). Under this condition, the level of binding immunoglobulin protein (BIP), an ER chaperone, was not affected by SERP1 knockdown. Consistent with the increase of APH1A by ER stress, formation of active γ-secretase complex was also concomitantly regulated by ER stress; ER stress increased the amount of γ-secretase holoenzyme in WT cells but not in SERP1-knockdown cells (Fig. 5F). These results show that the increase in γ-secretase activity under ER stress is attributable to the increase in the formation of γ-secretase complex by SERP1.

ER stress response is mediated by three ER stress sensors, namely, PERK, activating transcription factor 6, and inositol-requiring enzyme 1 (IRE1) (36). To investigate which sensor was involved in SERP1 up-regulation under ER stress, we examined the increase of SERP1 in SH-SY5Y cells in the presence of GSK2656157, an inhibitor of PERK, or APY29, an inhibitor of IRE1α. We found that the increase of SERP1 by thapsigargin was inhibited by APY29 (fig. S8, A and B). Furthermore, we confirmed this regulation; the increases of APH1A and SERP1 by thapsigargin were also inhibited in HEK293T cells expressing IRE1α dominant negative mutant (IRE1αΔC) (fig. S8, C to E) (37). These results suggest that the SERP1–γ-secretase axis is regulated through IRE1α under ER stress.

SERP1 increases γ-secretase formation for Aβ generation in diabetic AD model mice

It is well known that ER stress is a pathologic factor in diabetes (38), and diabetes is a critical risk factor for late-onset AD (20, 21). We thus hypothesized that SERP1 abundance might promote a link between diabetes and AD. To test this hypothesis, we assessed the expression of SERP1 in the brain of diabetic model mice. Because hyperglycemia, a hallmark sign of diabetes, increases ER stress (39, 40), we measured glucose level and examined the effect of SERP1 on γ-secretase in a diabetic mouse model. When mice were injected with streptozotocin (STZ), a compound that has a preferential toxicity toward pancreatic β cells (41), for 12 and 19 days, blood glucose levels in STZ-injected mice were increased by about twofold compared to control group (fig. S9A). As expected, the amount of SERP1 was highly elevated in both the cortices and the hippocampi of the STZ-induced diabetic mouse brains (fig. S9, B and C). Moreover, a BN-PAGE analysis showed that the level of the active γ-secretase complex was also increased in the diabetic mouse brains (fig. S9, D and E). In addition, incubation with high glucose increased the amounts of SERP1 and γ-secretase complex in SH-SY5Y cells (fig. S10, A and B).

We then examined the effects of SERP1 on Aβ generation in the STZ-induced diabetic 3x Tg-AD mice, an AD mouse model that expresses APP, PS, and tau transgenes (42). To knock down SERP1 abundance in the mouse brain, the hippocampi of 3x Tg-AD mice were administered with lentiviral vector encoding nontarget short hairpin RNA (shRNA) (shCtrl) or shSERP1 bilaterally. Compared to control group, blood glucose levels in STZ-injected mice were increased by about twofold regardless of SERP1 abundance (Fig. 6A). As reported previously (4345), Aβ40 and Aβ42 levels were substantially increased in the STZ-treated mice that were administered with nontarget shRNA (Fig. 6, B and C). In contrast, Aβ abundance was not increased after STZ administration in the shSERP1-injected mouse group. Accordingly, a BN-PAGE analysis showed that amounts of γ-secretase complex in the STZ-treated mice was also regulated dependently of SERP1 (Fig. 6D). These results suggest that SERP1 contributes to an increased abundance of γ-secretase and, thus, Aβ generation in diabetic AD model mice.

Fig. 6 SERP1 contributes to Aβ generation in STZ-induced diabetic AD mouse model.

(A) Effect of STZ treatment on serum glucose levels in 3x Tg-AD mice. High-titer lentiviral vectors expressing shSERP1 or shCtrl (as a control) were intracranially injected into the hippocampi of 3x Tg-AD mice (2 months old). After 19 days of the daily intraperitoneal injection with vehicle (Veh.) or STZ, serum, glucose levels were measured (n = 5 for control and shSERP1 with vehicle and n = 6 for control and shSERP1 with STZ). (B and C) Total amounts of Aβ40 and Aβ42 were measured by ELISA in the hippocampal extracts of each group (n = 5 for each group). (D) Digitonin (1%)-soluble membrane fractions prepared from control and diabetic AD model mice were subjected to BN-PAGE or SDS-PAGE and then analyzed by immunoblotting (left). The amount of PS1-NTF and SERP1 was quantified by densitometry analysis (right). Data in (A) to (D) are means ± SEM; n = 5. *P < 0.05 and **P < 0.01 by unpaired t test. (E) The relative amount of SERP1 was measured in the hippocampi of patients with AD. Hippocampal extracts from normal (control; n = 6) and AD patient (n = 8) postmortem brain tissues were analyzed by Western blotting and densitometry analysis. Data are means ± SEM. *P < 0.05, ***P < 0.001, or not significant (n.s.), by unpaired t test. (F) Immunohistochemical detection of SERP1 and APH1A in the hippocampi of patients with AD. Scale bars, 20 μm. Images are representative of three experiments.

In addition, we found that compared to postmortem hippocampal brain tissues from age-matched, non-AD individuals, those from patients with AD had increased abundance of SERP1 by ~9.6-fold (Fig. 6E). Similar up-regulation of SERP1 (about 5.5-fold) was also observed in the parietal lobe of patients with Braak stage VI AD (fig. S11, A and B). Immunohistochemical analysis at the cell level revealed that SERP1 was abundant in neurons (fig. S12). In addition, APH1A abundance was greater in the hippocampal tissues from patients with AD than in those of age-matched controls and correlated with a greater amount of SERP1 immunoreactivity (Fig. 6F). These data altogether suggest a pathogenic link between SERP1 abundance and AD pathology.


In this study, we identified SERP1 as a previously unidentified activator of γ-secretase through previously performed three rounds of a GOFS procedure, including C99 cleavage–induced changes of GFP intensity in cells, measurement of γ-secretase activity using a C99-GVP/UAS-luciferase reporter, and detection of Aβ levels using ELISA (25). The cell-based reporter system was not always parallel with γ-secretase activity because some false-positive gene could promote the expression of reporter gene directly. After the top hits were selected, we excluded false-positive hits using γ-secretase enzymatic assay, such as AICD generation assay. In a previous study, we successfully identified dual-specificity phosphatase 26 as a γ-secretase activator with the same approach using cDNA expression libraries that encode phosphatase proteins (46). The cDNA expression libraries that we used in the GOFS included focused libraries of 3000 cDNAs that encode plasma membrane and ER-resident proteins. We used a GOFS approach with a cDNA expression library over a loss-of-function screening with a small interfering RNA library because the objective was to identify a γ-secretase activator and a loss of function may not have always been detected in the cell-based assays using a loss-of-function screen under certain conditions.

SERP1 is a small protein consisting of 64 amino acids and is one of the members of the Sec family. According to our results, SERP1 seems to have several roles for regulating γ-secretase complex. First, SERP1 functions to promote the formation of γ-secretase complex in lipid raft through protein stabilization of the subcomplex. Especially, the last four amino acid residues (Arg63-Met-Gly-Met66) in the SERP1 C terminus were critical for its function, but this sequence was not observed in other γ-secretase activators, including GPR3 (9) and γ-secretase activating protein (GSAP) (30). At this moment, the mechanism by which SERP1 stabilizes APH1A/NCT proteins is not clear. There are conserved residues in the N terminus of SERP1 among different species (27), and the Lys16 in the conserved region is predicted as an ubiquitination site. Thus, ubiquitination of the lysine residue may play a role in the stabilization of the γ-secretase complex. However, further studies are needed to show how these separated regions of SERP1, the N terminus and C terminus, function coordinately to regulate the γ-secretase complex and how SERP1 promotes γ-secretase complex formation in lipid raft. Moreover, there are single-nucleotide polymorphisms that map to an adjacent region on chromosome 3, the location of SERP1 gene, but there is no report showing a genome-wide association study implicating SERP1 until now.

Second, SERP1 shows distinct effects on the cleavage of APP and Notch by regulating the interaction between γ-secretase complex and substrates. Whereas it is needed to clarify how SERP1 confers the ability to distinguish APP and Notch on γ-secretase, this selectivity of γ-secretase on the substrates may be attributable to regulating docking of γ-secretase with the substrates. It has been revealed that Aβ generation by γ-secretase occurs in lipid rafts and Notch cleavage occurs in nonlipid rafts (47, 48). Increase of γ-secretase complex in lipid raft by SERP1 might thus promote APP-favorable processing (Fig. 7). In addition, because recent studies revealed that the extracellular domain (ECD) of NCT is thought to play a critical role in blocking γ-secretase substrate (6, 49), we predict that the SERP1 C terminus facing the luminal side of ER interacts with the ECD of NCT and this interaction may regulate substrate blocking by ECD of NCT. Moreover, protein modification of NCT or APH1A by SERP1, such as glycosylation as seen in major histocompatibility complex class II–associated invariant chains and glucagon-like peptide 1 receptor (28), might also play a role in the substrate selection.

Fig. 7 A proposed model on APP cleavage regulation by SERP1.

Increased SERP1 under ER stress interacts with the APH1A/NCT subcomplex and stabilizes it, leading to increased abundance of γ-secretase complex and its abundance incorporation into lipid rafts. We propose that increased γ-secretase complex in lipid rafts enables more interaction with APP than with Notch, thereby promoting the generation of Aβ.

Epidemiological studies show that diabetes is a critical risk factor for late-onset AD (20, 21, 24). Because insulin plays an important role in maintaining normal brain function (50) and insulin-degrading enzyme (IDE) can degrade Aβ (51), most of in vitro and in vivo studies have focused on showing correlations among insulin, IDE, Aβ degradation, and AD. Several studies demonstrated that hyperglycemia is associated with an increased risk of dementia (52) and increases Aβ levels by promoting Aβ generation in vitro and in vivo (23, 53). Despite increasing evidence, the molecular mechanism underlying the relationship between hyperglycemia and AD pathogenesis remains unexplored. Here, we propose that hyperglycemia- or ER stress–induced SERP1 may explain the high risk of AD in people with diabetes. Because hyperglycemia provokes ER stress (40, 54), the notions that SERP1 is up-regulated under ER stress (35), high-glucose culture condition (55), and by X-box binding protein 1 (XBP-1) (26) support that SERP1 could be a mediator connecting ER stress to γ-secretase activation. Accordingly, treatment with phenylbutyric acid, a molecular chaperone, reduces amyloid plaques and rescues cognitive behavior in AD model mice (56). SERP1 KO is not lethal in mice but shows decreased viability because of glucose tolerance deficiency (55). In the future, investigations using SERP1 tissue-specific KO mice would give more clues to the role of SERP1 in AD progression as it relates to diabetes.

In conclusion, our findings indicate that SERP1 is a previously unknown γ-secretase activator that regulates γ-secretase assembly and substrate selectivity. Given that SERP1 is up-regulated to stimulate γ-secretase under ER stress and in a diabetic mouse model, these findings suggest that SERP1 is a missing link that connects ER stress/diabetes to AD pathology, providing an avenue for new preventive and therapeutic strategies against AD progression coincident with diabetes.


Genome-wide functional screening using cDNA

Genome-wide functional screen was performed as described previously (25). HEK293T cells were cotransfected with pC99-TetOn, pTRE-GFP, monomeric RFP vector, and either control vector (pCtrl) or each cDNA for 18 hours and then treated with doxycycline (100 ng/ml; Sigma-Aldrich) for another 12 hours. The AICD–reverse Tet-controlled transactivator (rtTA) is generated by endogenous γ-secretase; then, it is transported into the nucleus for inducing GFP expression. The putative positive cDNA clones that markedly increased green fluorescence under fluorescence microscope (Olympus) were isolated. The cDNAs encoding membrane proteins were generated by RT-PCR analysis and subcloned into mammalian expression vectors or purchased from OriGene Inc. (25). After the primary screening, the secondary screening was performed using luciferase reporter assay. HEK293T cells were cotransfected with pC99-GVP, pUAS-luciferase, pβ-galactosidase, and experimental control genes with or without 1 μM Compound E. After 24 hours, cell extracts were analyzed by luciferase assay following the manufacturer’s instructions (Promega). The cDNA library was prepared as previously described (37, 57), and the data are provided in data file S1.

Antibodies and Western blot analysis

The following antibodies were used: anti-tubulin (TUBA) (Sigma-Aldrich, T6074), anti-actin (ACTB) (Sigma-Aldrich, A1978), anti-HA (HA hybridoma), anti-FLAG (Sigma-Aldrich, S7425), anti-GFP (Santa Cruz Biotechnology, sc-8334), anti-V5 (Sigma-Aldrich, S2540), anti-Aβ (4G8) (GeneTex, GTX109283), anti-BIP (Santa Cruz Biotechnology, sc-33757), anti–Caveolin-1 (CAV1) (Abcam, ab2910), anti–PS1-NTF (Santa Cruz Biotechnology, sc-7860), anti–APP-CTF (Sigma-Aldrich, A8717), anti-NCT (Sigma-Aldrich, N1660), anti-NICD (Novus Biologicals, NB200-251), anti-Tim23 (BD Biosciences, 611222), anti-Calnexin (CANX) (Santa Cruz Biotechnology, sc-11397), anti–glial fibrillary acidic protein (GFAP; Cell Signaling Technology, 80788S), anti-synaptophysin (SYP) (Santa Cruz Biotechnology, sc-17750), and anti-APH1A and anti-PEN2 (gift from A. Takashima, RIKEN Brain Science Institute, Japan) antibodies. Anti-SERP1 antibody (homemade) was generated by following standard immunization procedure using purified SERP1 protein from Escherichia coli. Western blot analysis was performed using standard techniques. Cells were lysed with sample buffer [10% glycerol, 2% SDS, 62.5 mM tris-HCl, and 2% β-mercaptoethanol (pH 6.8)]. For the preparation of membrane protein, cells were solubilized in 1% CHAPS buffer containing protease inhibitor cocktail and centrifuged at 10,000g for 10 min. The soluble supernatants were subjected to SDS-PAGE, and the separated proteins were transferred to polyvinylidene fluoride membrane (ATTO, AE-6667-P) using a Bio-Rad semidry transfer unit (Bio-Rad). Immunoblot analysis was then performed and visualized by the enhanced chemiluminescence method.

Plasmid constructions

Human SERP1 was subcloned into pcDNA3-HA, pEGFP, monomeric RFP (Invitrogen), and pET-28a vector. Human APH1A and SC100 were subcloned into pEGFP and monomeric RFP vector (Invitrogen). pNCT-V5, pPS1-Myc, and pAPH1A-FLAG were previously described (58). SERP1 deletion mutants were generated by subcloning the PCR products [ΔN (SEC61G N1-32 + SERP1 TMC35-66)/ΔNTM (SEC61G N1-61 + SERP1 C62-66)/ΔC (SERP1 NTM1-61 + SEC61G C62-68)]. For SERP1 shRNA, target sequence was cloned into pSuper-neo vector (pshRNA, Oligoengine). The sequences for the construction of pSuper-neo-shSERP1 are as follows: forward, 5′-GAT CCC CGC ACT TAG CTT AAA TTA CAT TCA AGA GAT GTA ATT TAA GCT AAG TGC TTT TTA-3′ and reverse, 5′-AGC TTA AAA AGC ACT TAG CTT AAA TTA CAT CTC TTG AAT GTA ATT TAA GCT AAG TGC GGG-3′. For SERP1 single-guide RNA, target sequence was cloned into lentiCRISPR v2 vector (Addgene, no. 52961). The target sequences for SERP1 are as follows: forward, 5′-CAC CGC TTG GCG ACG TTG CCG CGC T-3′ and reverse, 5′-AAA CAG CGC GGC AAC GTC GCC AAG C-3′.

Cell culture and DNA transfection

SY5Y-APPswe cells and HEK-APP695 cells are described previously (59, 60). HEK293T, HeLa, PS 1/2 DKO MEF, CHO-7PA2, and stable cells were cultured in Dulbecco’s modified Eagle medium (HyClone, SH30243.01) supplemented with 10% fetal bovine serum (HyClone, SH30919.03) and gentamicin (50 μg/ml; Gibco, 15750-060) at 37°C under 5% CO2 (v/v) condition. HeLa/shCtrl and HeLa/shSERP1 stable cells were selected with G418 at 2 mg/ml (GoldBio, G418-5) and maintained with G418 at 1 mg/ml for experiments. SERP1 KO MEF cells were selected with puromycin (1 μg/ml). According to the manufacturer’s instructions, transfection was performed using PolyFect (Qiagen, 1015586) or polyethylenimine (PEI) (Sigma-Aldrich, 764647).

Luciferase reporter γ-secretase activity assay

Luciferase reporter γ-secretase activity assay was as described (25). HEK293T cells were cotransfected with pC99-GVP, pUAS-luciferase, pβ-galactosidase, and experimental control genes with or without 1 μM Compound E. After 24 hours, cell extracts were analyzed by luciferase assay following the manufacturer’s instructions (Promega). The luciferase activity was normalized by β-galactosidase activity for transfection efficiency.

Enzyme-linked immunosorbent assay

According to the manufacturer’s instructions, Aβ in the conditioned media of HEK/APPswe cells, SH-SY5Y/APPswe cells, or brain extracts were measured using sandwich ELISA kits with anti-Aβ antibodies (Invitrogen and IBL).

In vitro AICD generation assay

In vitro AICD generation assay was performed as described previously with minor modification (61). The harvested cells were lysed by sonication in buffer A [50 mM Hepes (pH 7.4), 150 mM NaCl, 5 mM 1,10-phenanthroline monohydrate, 2 mM EDTA, and protease inhibitor cocktail]. The homogenate was centrifuged at 1000g for 10 min, and the supernatant was further centrifuged at 10,000g for 30 min. The membrane fraction in pellets was washed once with buffer A. The final membrane pellet was resuspended with buffer A, and the protein amounts in samples were measured using Bradford assay (Bio-Rad). The same amount of protein was incubated at 37°C for 12 hours with fluorogenic C99 substrate (Invitrogen) or purified C99-FLAG protein.

Blue native–polyacrylamide gel electrophoresis

BN-PAGE was performed as described previously (9, 62). The membrane protein was solubilized in BN-PAGE buffer [1% digitonin, 20% glycerol, and 25 mM bis-tris (pH 7.0)] for 60 min on ice and centrifuged at 100,000g for 30 min. The remaining supernatant was used for BN-PAGE. The same volume of soluble protein was separated by BN-PAGE at 4°C and transferred into polyvinylidene difluoride membrane. The transferred blot was washed in destaining solution (distilled water:methanol:acetic acid, 6:3:1) for 15 min and analyzed with Western blotting.

Immunoprecipitation assay

General immunoprecipitation was performed using 1% Triton X-100 buffer (50 mM tris, 150 mM NaCl, 2 mM EDTA, 0.1% SDS, 1% Triton X-100, 0.5% sodium deoxycholate, and protease inhibitor cocktail). Immunoprecipitation of γ-secretase protein complex was performed using 1% CHAPS buffer [25 mM Hepes (pH 7.4), 150 mM NaCl, 2 mM EDTA, 1% CHAPS, and protease inhibitor cocktail] or 1% digitonin buffer [25 mM Hepes (pH 7.4), 150 mM NaCl, 2 mM EDTA, 1% digitonin, and protease inhibitor cocktail]. Cell lysates were prepared by sonication in Triton X-100, digitonin, or CHAPS buffer. After brief centrifugation, the supernatants were incubated with anti-HA, anti-GFP, anti-SERP1, anti-PS1, or anti-APH1A antibody at 4°C overnight and then pulled down using Protein G Sepharose beads (GE Healthcare).

Sucrose gradient fractionation

Sucrose gradient fractionation was performed as described previously (63). Cells were suspended in buffer S [25 mM tris (pH 7.4), 150 mM NaCl, 2 mM EDTA, 1% CHAPS, and protease inhibitor cocktail] and homogenized by 10 passages through a 25-gauge needle. After centrifugation at 10,000g for 30 min, soluble lysates were adjusted to a final concentration of sucrose (45%) and transferred to a 10-ml ultracentrifuge tube. Then, a discontinuous sucrose gradient was formed by sequentially layering 35% sucrose (3.2 ml) and 5% sucrose (3.2 ml), and the tubes were subjected to ultracentrifugation at 100,000g for 16 hours in Beckman SW 32.1 Ti rotor at 4°C with no brakes. Twelve 0.8-ml fractions were collected from the top to the bottom of the gradient, and the same volume of each fraction was analyzed by Western blotting.

Glycerol velocity gradient fractionation

HEK293T cells were homogenized with homogenization buffer [5 mM Hepes (pH 7.4), 1 mM EDTA, 250 mM sucrose, and protease inhibitor cocktail], and supernatant was prepared by centrifugation at 800g. After centrifugation at 100,000g for 1 hour at 4°C, the pellet was solubilized with CHAPS buffer [2% CHAPS, 50 mM tris (pH 7.5), 2 mM EDTA, and 150 mM NaCl]. The soluble lysates were centrifuged again at 100,000g for 30 min, and the supernatants were subjected to glycerol gradient centrifugation. The same amount of protein extracts was applied to the top of 10 to 40% (w/v) linear glycerol gradient and centrifuged for 15 hours at 100,000g and 4°C using a Beckman SW 32.1 Ti rotor. Each fraction was collected from the top to the bottom of the gradient, and the same volume of each fraction was analyzed by Western blotting.

Overlay assay

Overlay assay was performed as described previously (63). SERP1 protein (10 μg), which was expressed in E. coli and purified, was resolved by SDS-PAGE and transferred to nitrocellulose membranes. After blocking, the membranes were incubated at 4°C for 12 hours with whole-brain lysates of 3-month-old C57BL/6 mice. Membranes were then processed to Western blotting as mentioned above.

Mouse brain extraction

Cortices and hippocampi from C57BL/6J mice and 3x Tg-AD mice were sonicated in ice-cold tris-buffered saline containing 20 mM tris-Cl (pH 7.4), 150 mM NaCl, 1% NP-40, and protease inhibitors. The homogenates were clarified by centrifugation at 4°C, aliquoted, and stored at −70°C. For the assays, 20 μg of supernatants was used for quantitation of Aβ using an ELISA kit (IBL) and analyzed by SDS-PAGE and Western blotting. For BN-PAGE analysis, mouse brains were sonicated in ice-cold phosphate-buffered saline, and then, the crude membrane fractions were obtained by centrifugation at 4°C. The pellets were resolved by BN-PAGE.

Preparation of human brain samples

Hippocampal tissues of patients with AD (Braak V and VI; aged 71 to 93 years; postmortem intervals, 2 to 16 hours) and age-matched control were supplied by the Harvard Brain Tissue Resource Center (McLean Hospital, Boston). Hippocampal tissues of patients with AD were homogenized in ice-cold tris-buffered saline containing of 20 mM tris-Cl (pH 7.4), 150 mM NaCl, and protease inhibitors. The homogenates were clarified by centrifugation at 4°C, aliquoted, and stored at −70°C. The supernatants were resolved by SDS-PAGE.


Immunohistochemical studies of brain sections from patients with AD were described previously (64). The hippocampi of patients with AD were fixed in paraformaldehyde (PFA) for 48 hours before serial coronal sections and cut on a freezing microtome. The sections were placed on the slides and refixed in 4% PFA. Samples were incubated in 0.1% Triton X-100 for 20 min and then in 1% bovine serum albumin for 1 hour before the application of a primary antibody. Blocked sections were incubated with anti-SERP1 (1:200), anti-APH1A (1:300), anti-GFAP (1:200), or anti-SYP (1:200) antibodies; washed; and incubated with fluorescein isothiocyanate–conjugated and tetramethyl rhodamine isothiocyanate–conjugated secondary antibodies (the Jackson laboratory). Samples were washed, mounted, and observed under a Zeiss LSM700 microscope.

STZ induction of diabetes

Male C57BL/6J mice and 3x Tg-AD mice (8 to 12 weeks old) were given daily by intraperitoneal injections with STZ (60 mg/kg body weight in 0.1 M citrate buffer) (Sigma-Aldrich) for 5 days to induce diabetes. Control mice were injected with buffer alone. Blood glucose levels were examined 7 days after the final STZ injection by measuring glucose concentration with a glucometer (Accu-Chek instant, Boehringer Mannheim Corporation). Mice with blood glucose levels greater than 300 mg/dl were considered diabetic. At the time of harvest, 19 days after the first STZ injection, blood glucose measurements were again obtained, mice were humanely euthanized, and brains were harvested for analyses. All experiments involving animals were performed according to the protocols approved by the Seoul National University Institutional Animal Care and Use Committee guidelines.

Statistical analysis

All experiments were performed in triplicate parallel instances and repeated at least three times. Statistical analyses were carried out using the Microsoft Office 2016 Excel software package (Microsoft Corporation). Mean values were compared using unpaired t tests.


Fig. S1. SERP1 expression influences AICD generation.

Fig. S2. SERP1 decreases the interaction of PS1 and NΔE.

Fig. S3. SERP1 regulates the amounts of γ-secretase subunits.

Fig. S4. SERP1 interacts with γ-secretase complex.

Fig. S5. The subcomplex of γ-secretase is dissociated by 1% Triton X-100.

Fig. S6. SERP1 colocalizes with APH1A.

Fig. S7. The C terminus of SERP1 is critical for γ-secretase activity.

Fig. S8. ER stress regulates the amount of SERP1 via IRE1 pathway.

Fig. S9. SERP1 and γ-secretase complex are up-regulated in diabetic mouse brains.

Fig. S10. Hyperglycemia increases the abundance of SERP1 and γ-secretase complex in SH-SY5Y cells.

Fig. S11. SERP1 abundance is increased in the parietal lobes of patients with Braak stage VI AD.

Fig. S12. Immunohistochemical detection of SERP1, SYP, or GFAP in the hippocampi of patients with AD.

Table S1. Top hits from the genome-wide functional screen.

Data file S1. Genome-wide functional screen data.


Acknowledgments: We thank A. Takashima (RIKEN Brain Science Institute, Japan) for the anti-PEN2 and anti-APH1A antibodies, D. J. Selkoe (Harvard Medical School, MA) for CHO and 7PA2 cells, and W. Araki (National Institute of Neuroscience, Japan) for SH-SY5Y-APPswe cells. AD tissues were provided from the Harvard Brain Tissue Resource Center of McLean Hospital, MA and the tissue bank of Boston University School of Medicine. Funding: This work was supported by the CRI grant (NRF-2019R1A2B5B03070352) and a Bio & Medical Technology Development Program of the National Research Foundation (NRF-2017M3A9G7073521) funded by the Ministry of Education, Science, and Technology and by a grant (to Y.-K.J.) funded by the Alzheimer’s Association (USA). Author contributions: S.J. and Y.-K.J. conceived and planned the experiments. S.J., J. Hyun, J.N., and J. Han carried out the experiments. S.-H.K., J.P., Y.O., Y.G., and S.M. contributed to sample preparation. D.-G.J. supported materials for in vivo studies. S.J. and J. Hyun contributed to the interpretation of the results. S.J. and Y.-K.J. took the lead in writing the manuscript. All authors provided critical feedback and helped shape the research, analysis, and manuscript. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper or the Supplementary Materials.

Stay Connected to Science Signaling

Navigate This Article