Amyloid β‐induced elevation of O‐GlcNAcylated c‐Fos promotes neuronal cell death

Abstract Alzheimer's disease (AD) is an age‐related neurodegenerative disease characterized by progressive memory loss resulting from cumulative neuronal cell death. O‐linked β‐N‐acetyl glucosamine (O‐GlcNAc) modification of the proteins reflecting glucose metabolism is altered in the brains of patients with AD. However, the link between altered O‐GlcNAc modification and neuronal cell death in AD is poorly understood. Here, we examined the regulation of O‐GlcNAcylation of c‐Fos and the effects of O‐GlcNAcylated c‐Fos on neuronal cell death during AD pathogenesis. We found that amyloid beta (Aβ)‐induced O‐GlcNAcylation on serine‐56 and 57 of c‐Fos was resulted from decreased interaction between c‐Fos and O‐GlcNAcase and promoted neuronal cell death. O‐GlcNAcylated c‐Fos increased its stability and potentiated the transcriptional activity through higher interaction with c‐Jun, resulting in induction of Bim expression leading to neuronal cell death. Taken together, Aβ‐induced O‐GlcNAcylation of c‐Fos plays an important role in neuronal cell death during the pathogenesis of AD.

modification that reflects glucose metabolism is disrupted in the brains of patients with AD (Alfaro et al., 2012;Forster et al., 2014;Wang et al., 2017;Zhu, Shan, Yuzwa, & Vocadlo, 2014). The O-GlcNAc modification, regulated by O-GlcNAc transferase (OGT, an enzyme that adds O-GlcNAc to proteins from Uridine diphosphate N-acetylglucosamine (UDP-GlcNAc)) and O-GlcNAcase (OGA, and enzyme that removes O-GlcNAc from proteins), is a dynamic process, affected in various human diseases as well as many biological processes (Bond & Hanover, 2015;. A recent studies on proteomic analysis using the brains of AD patients and 3xTg-AD mice showed altered abundance of O-GlcNAc peptides in AD (Alfaro et al., 2012;Forster et al., 2014;Wang et al., 2017).
Therefore, disrupted O-GlcNAc modification is considered to be important for understanding the pathophysiology of AD. However, whether dysregulated O-GlcNAc cycling might be resulted from altered activity or substrate specificity of OGT and OGA in AD remains unclear. In addition, the role of O-GlcNAc modification in cell death associated with AD has not been studied so far. Although c-Fos is known to be an O-GlcNAcylated protein (Tai, Khidekel, Ficarro, Peters, & Hsieh-Wilson, 2004)

| Aβ increases c-Fos O-GlcNAcylation as well as the protein of c-Fos
Several reports show that the expression c-Fos is increased in the brains of patients with AD, and that c-Fos can be O-GlcNAcylated (Marcus et al., 1998;Tai et al., 2004). Since O-GlcNAc cycling is altered in patients with AD (Forster et al., 2014;Liu et al., 2009;Zhu et al., 2014), we hypothesized that c-Fos O-GlcNAcylation might be dysregulated in AD and play a role in its pathogenesis. To determine whether c-Fos O-GlcNAcylation is altered under conditions simulating AD, we pulled down O-GlcNAcylated proteins using wheat-germ-agglutinin (WGA)-conjugated agarose beads (that wellknown to bind to O-GlcNAc) in the brains of 5xFAD mice, a model of AD, and Aβ-treated primary cortical neurons and SH-SY5Y cells.
Although the levels of c-Fos were increased in 5xFAD as reported previously (Marcus et al., 1998), we observed that c-Fos O-GlcNAcylation was also increased in 5xFAD compared to wild-type littermates, which was confirmed when the levels of pulled-down c-Fos (O-GlcNAcylated c-Fos) were normalized to the input level of c-Fos (Figure 1a-c). We also found that both c-Fos and c-Fos O-GlcNAcylation levels were increased by Aβ in primary neurons and SH-SY5Y cells (Figure 1d-i). In addition, when removing O-GlcNAc using β-hexosaminidase (β-HEX), an O-GlcNAc-specific exoglycosidase that catalyzes the hydrolysis of β-D-N-acetylglucosamine residues, WGA-pulled-down c-Fos were remarkably decreased in the brains of 5xFAD mice and in the presence of Aβ, although the total levels of c-Fos levels were not changed (Supporting Information Figure S1). Especially, increased WGA-pulled-down c-Fos by Aβ were reduced even lower than vehicle group by β-HEX reaction, though, the input level of c-Fos was not changed (Supporting Information Figure S1b). These data support that c-Fos O-GlcNAcylation was increased by Aβ.
Furthermore, to exclude the possibility of c-Fos induction by Aβ and confirm an increase in c-Fos O-GlcNAcylation, we performed immunoprecipitation of overexpressed exogenous c-Fos using an anti-GFP antibody (Figure 1j,k). The increase in signals using CTD110.6 antibody, that specifically detects O-GlcNAc, clearly showed that c-Fos O-GlcNAcylation was increased by Aβ.
In addition, we proved that there were no nonspecific bindings during pull down assay and immunoprecipitation assay. To confirm nonspecific binding to beads, cell lysates were incubated with agarose beads, because we use WGA-coated-agarose beads when performing WGA-pull down assay, and there were no signals (2nd lane in Figure 3a and 1st lane in Supporting Information Figure S1b).
Competitive assay using exogenously added GlcNAc to cell lysates, to confirm nonspecific binding to WGA, showed that O-GlcNAcylated c-Fos binds specifically to WGA (3rd and 4th lane in Figure 3a).
In addition, no signal of WGA-pulled-down EGFP and no signal of

GlcNAcase and c-Fos
Next, we wondered how Aβ dysregulates O-GlcNAc cycling . Forster et al. (2014) showed that the proteins of approximately 50 -60 kDa and 25 kDa in size show increased O-GlcNAcylation, while the proteins of over 70 kDa in size show reduced O-GlcNAcylation in the brains of patients with AD (Forster et al., 2014). This suggested that O-GlcNAc cycling in AD was altered. Thus, increased O-GlcNAcylation on c-Fos, a 55 kDa protein, may be a result of altered O-GlcNAc cycling in AD. It has been reported that the interaction between OGT and its substrate (ATP synthase 5A) is altered by Aβ . Therefore, we wondered whether O-GlcNAc modifying enzymes (OGT or OGA) are involved in increased c-Fos O-GlcNAcylation by Aβ. Immunoprecipitation with GFP antibody, followed by Western blotting with c-Fos and OGT showed that the interaction between OGT and c-Fos was not changed by Aβ (Figure 2a,b). However, the interaction between OGA and c-Fos was decreased by Aβ using immunoprecipitation with GFP and Western blotting with c-Fos and OGA (Figure 2c,d). To confirm the less interaction between OGA and c-Fos by Aβ, imaging with structured illumination microscopy (SIM), which is a super-resolution microscopy, was performed.
It showed that less merged signals between c-Fos and OGA in the presence of Aβ (Figure 2e,f). In addition, we also observed that the interaction between c-Fos and OGA was decreased in the brains of 5xFAD mice compared to wild-type littermate (Figure 2g,h). This suggests that the ability of OGA to interact with c-Fos is altered by Aβ, resulting in increased O-GlcNAcylated c-Fos in the cells. Tai et al. (2004)   Probably, because of sequence characteristics of c-Fos protein, there might be not enough ionization on the peptide around S56 and S57 residues. From these data, we carefully suggested that both S56 and S57 of c-Fos are the first identified O-GlcNAc sites on c-Fos.

| O-GlcNAcylation of c-Fos promotes cell death in the presence of Aβ
Several studies show that c-Fos is involved in cell death under cytotoxic conditions because it regulates the transcription of apoptotic genes (Chen et al., 2015;Fernandez et al., 2005;Gillardon et al., 1996;Whitfield et al., 2001). Aβ is a well-known cytotoxic protein that induces neurite atrophy, synaptic dysfunction, and neuronal cell death (Kumar et al., 2015;Querfurth & LaFerla, 2010). Thus, we hypothesized that c-Fos may be involved in cell death in response to

| Aβ-induced c-Fos O-GlcNAcylation regulates transcriptional activity of c-Fos and the expression of Bim
In the experiments illustrated in Figure 5 Recently, OGA inhibition has been evoked as one of the therapeutic strategies for AD Zhu et al., 2014). This was supported by several studies showing that AD-like pathologies, such as memory impairment, hyperphosphorylated tau, neurofibrillary tangles, and Aβ plaques, were improved using OGA inhibitor for modulation of autophagy and the O-GlcNAcylation of tau, and Nicastrin, one of γ-secretase components (Kim et al., 2013;Graham et al., 2014;Yuzwa, Cheung, Okon, McIntosh, & Vocadlo, 2014;Yuzwa, Shan, et al., 2014;Zhu et al., 2018). However, several researches suggested that down regulation of OGA caused proteotoxicity resulted from accumulation of aggregate-prone proteins such as huntingtin, Aβ and α-synuclein Wani et al., 2017). Moreover, elevating O-GlcNAc by OGA inhibition or OGT overexpression induced mitochondrial dysfunction (Gawlowski et al., 2012;Tan, Villar, & E L, Lu J, Selfridge JE, Artigues A, Swerdlow RH, Slawson C, 2014). In addition, we demonstrated that increased O-GlcNAcylation on c-Fos elevated apoptosis in AD condition. These contradictory reports suggest that modulating O-GlcNAc cycling rather than inhibiting OGA might be important therapeutic strategy for AD. Liu et al. (2009) showed that the overall O-GlcNAcylation was reduced in the brains of patients with AD (Liu et al., 2009). However, another report showed that the proteins of approximately 50 -60 kDa and 25 kDa in size show increased O-GlcNAcylation, while the proteins of over 70 kDa in size show reduced O-GlcNAcylation in the brains of patients with AD (Forster et al., 2014). In addition, a recent study on proteomic analysis using the brains of AD patients showed increased abundance of O-GlcNAc peptides . Another proteomic study using 3xTg-AD mice reported that some of the O-GlcNAcylated proteins present only in 3xTg-AD mice (Alfaro et al., 2012). These data suggested that O-GlcNAc cycling in AD was altered rather than simply reduced. Thus,

increased O-GlcNAcylation on c-Fos may be a result of altered O-
GlcNAc cycling in AD.
The regulation of O-GlcNAc cycling seems to be complicated and the levels and activities of O-GlcNAc modifying enzymes seemed not to be changed in AD . Nevertheless, there are several possible regulatory mechanisms. First, the interactions between O-GlcNAc modifying enzymes (OGT or OGA) and their substrates might be altered. It has been reported that the interaction between OGT and its substrate (ATP synthase 5A) is altered by hindering inhibition of Aβ . We also observed that the interaction between OGA and c-Fos was reduced. Second, there F I G U R E 6 O-GlcNAcylation of c-Fos at S56 and S57 regulates the transcriptional activity of c-Fos and the expression of Bim in the presence of Aβ. SH-SY5Y (a, b, d-g) and HEK293T (c) cell lines were treated with Aβ (at doses indicated) for 24 hr. (a, b) The interaction between c-Fos and c-Jun was analyzed by performing immunoprecipitation. SH-SY5Y cells were transfected with EGFP-c-Fos-WT or EGFP-c-Fos-S56A-S57A. EGFP-c-Fos was immunoprecipitated using an anti-GFP antibody and probed with c-Jun. Representative immunoblot images are shown in (a) and quantitative graph in (b) (n = 5). (c) AP-1 luciferase assay was performed in tag free-c-Fos-WT or tag free-c-Fos-S56A-S57A in AP-1 luciferase vector co-transfected HEK293T cells (n = 4). (d) Relative Bim mRNA expression levels were measured in tag free-c-Fos-WT or tag free-c-Fos-S56A-S57A transfected SH-SY5Y cells (n = 5). (e-g) Relative expression levels of Bim and cleaved caspase-3 were measured using western blotting in EGFP-c-Fos-WT or EGFP-c-Fos-S56A-S57A transfected SH-SY5Y cells (n = 5). Representative immunoblot images are shown in (e). Quantitative graphs showing the relative expression of Bim and cleaved caspase-3 ((f) and (g), respectively). Data are shown as mean ± SEM. # p < 0.05, ## p < 0.01, ### p < 0.001 among c-Fos-WT or c-Fos-S56A-S57A transfected groups (one-way ANOVA, Bonferroni posthoc test), *p < 0.05, **p < 0.01, ***p < 0.001 between c-Fos-WT and c-Fos-S56A-S57A groups (two-way ANOVA, Bonferroni posthoc test). (h) Schematic diagram of the sites and the role of c-Fos O-GlcNAcylation in the presence of Aβ. O-GlcNAc sites of c-Fos are S56 and S57. c-Fos O-GlcNAcylation is increased by Aβ, which is mediated by a decreased interaction between c-Fos and OGA. O-GlcNAcylation of c-Fos increases its stability and the interaction with c-Jun, consequently elevating its transcriptional activity to induce the expression of Bim, an apoptotic protein, in the presence of Aβ. Therefore, O-GlcNAcylation of c-Fos promotes neuronal cell death in the presence of Aβ. α-tub: αtubulin; IP: Immunoprecipitation; S56A-S57A: c-Fos-S56A-S57A; Veh: Vehicle; WT: c-Fos-WT were reports that several kinases regulate substrate specificities of OGT and OGA (Nagel & Ball, 2014). For example, p38 MAPK affects O-GlcNAcylation of neurofilament H in neuroblastoma cells by interacting with OGT and not affecting phosphorylation (Cheung & Hart, 2008). Although these phenomena have not been tested in AD condition, it is known that kinases are dysregulated in AD (Dolan & Johnson, 2010;Perluigi, Barone, Domenico, & Butterfield, 2016).
Thus, kinases might be involved in altered O-GlcNAc cycling such as disrupted OGA and c-Fos interaction in AD. Third, phosphorylation might affect O-GlcNAcylation by competitive or synergistic manner (Bond & Hanover, 2015). A recent study revealed that the phosphorylation/O-GlcNAcylation interplay motif, (pS/pT)P(V/A/T)(gS/gT), and around O-GlcNAcylation sites of tau resemble to this motif (Leney, Atmioui, Wu, Ovaa, & Heck, 2017). However, S56 and S57 containing peptide of c-Fos is different from that motif, and there has been no report about phosphorylation around or on S56 and S57 of c-Fos so far. Thus, c-Fos O-GlcNAcylation may be regulated differently from tau. Although we described possible O-GlcNAc regulatory mechanisms, the precise regulatory mechanism of decreased OGA binding to c-Fos in response to Aβ and altered O-GlcNAc cycling in AD needs to be further studied. In addition, it requires further investigation whether O-GlcNAcylation on substrates affect the ability of interaction between O-GlcNAc modifying enzymes and substrates or not.
We revealed that c-Fos O-GlcNAcylation at S56 and S57 can improve its stability. It has been reported that N-terminal region of c-Fos is a destabilizer (Ferrara et al., 2003;Gomard et al., 2008), although the specific sequence of this region is not clear. It is also known that phosphorylation at S32 of c-Fos by ERK5 inhibits its degradation by the proteasome (Sasaki et al., 2006). Thus, a potential reason that c-Fos O-GlcNAcylation can regulate its stability may be that S56 and S57 sites of c-Fos might be present within N-terminal destabilizer. In contrast, O-GlcNAcylation at S56 and S57 might affect the phosphorylation on S32 of c-Fos. We also demonstrated that the interaction between c-Fos and c-Jun was increased by c-Fos O-GlcNAcylation in the presence of Aβ. Although S56 and S57 are located far from DNA binding-(amino acids 139-160) or leucine zipper-domains (amino acids 165-193) (Eferl & Wagner, 2003), O-GlcNAcylation at S56 and S57 might affect the cross-talk with transcriptional co-factors, such as CBP, CRTC1, BAF complex, by altering the three-dimensional configuration of these components (Bannister & Kouzarides, 1995;Canettieri et al., 2009;Vierbuchen et al., 2017). However, the precise underlying mechanisms need further investigation. Several AP-1 target genes under cytotoxic stimuli have been described, such as Fas ligand (FasL), Fas, and Bim (Chen et al., 2015;Shaulian & Karin, 2002;Whitfield et al., 2001). We tested the levels of these proteins in our experiments, and observed that those of Dysregulated glucose metabolism is involved in several human diseases such as diabetes mellitus and AD (Bond & Hanover, 2015).
Here, we demonstrated altered c-Fos O-GlcNAcylation in AD and revealed that the function of c-Fos O-GlcNAcylation is to increase the stability of c-Fos, stimulating its interaction with c-Jun and the transcriptional activity followed by an induction of Bim, an apoptotic gene, expression therefore directly promoting cell death in the presence of Aβ. These results provide an insight into the effects of dysregulated O-GlcNAc cycling in AD pathophysiology. Therefore, modulating O-GlcNAc cycling rather than inhibiting OGA might be a potential therapeutic strategy for AD.

| Animals
Eight-month-old 5xFAD mice (Tg6799; B6SJL-Tg (APPSwFlLon, PSEN* M146L*L286V) 6799Vas/J, stock number 006554, Jackson Labs, Bar Harbor, ME, USA) overexpressing human amyloid precursor protein 695 with three mutations (Swedish, Florida, and London) and human presenilin 1 with two mutations (M146L and L286V) under transcriptional control of the murine Thy-1 promoter and wild-type littermate (B6/SJL) were used for brain tissue analysis. Animal experiments were performed in accordance with the Principle of Laboratory Animal Care (NIH publication No. 85-23, revised 1985) and the Animal Care and Use Guidelines of Seoul National University, Seoul, Korea. All experimental protocols were approved by Institutional Animal Care and Use Committee (IACUC) at Seoul National University.

| Primary neuronal culture
Primary cortical neuronal cultures were prepared as previously described (Jung, An, Hong, Kim, & Mook-Jung, 2012). In brief, brain tissue of Sprague-Dawley rat embryos (E18) was dissected (KOA-TECH, Korea), the brains were trypsinized in Hank's Balanced Salt Solution (HBSS; WelGENE, Korea), and dissociation was performed in NeuroBasal medium (Gibco, USA) supplemented with B27 (Gibco, USA) and penicillin/streptomycin (Sigma, USA). Dissociated neurons were plated on poly-D-lysine (Sigma, USA) coated dishes. Half of the culture medium was replaced with fresh medium every three days.

| Fluorescence imaging
For analyzing OGA and c-Fos interaction, we performed immunocytochemistry as previously described . Briefly, EGFPc-Fos-WT transfected cells were plated on coverslips and treated by Aβ. After fixed by 4% paraformaldehyde (BIOSESANG, Inc., Korea), cells were incubated with anti-OGA (Sigma, USA) and GFP (OSE00001G, Osenses, Australia) antibodies, which were diluted in 0.5% Tx-100, 1% goat serum in PBS, overnight at 4°C. After labeling with fluorescent-labeled secondary antibodies (Invitrogen, USA) for 1 hr at room temperature, cells were imaged by SIM (Nikon N-SIM, Nikon Instruments, Inc., Japan). Images were processed and analyzed by NIS-E software (Nikon Instruments, Inc., Japan). For imaging c-Fos level, native EGFP signals were imaged by confocal microscopy (Olympus FV10i;Olympus,Japan and LSM710;Carl Zeiss,Germany) in living cells which were transfected by EGFP-c-Fos-WT and treated by CHX or MG132.

| TUNEL assay
For measuring cell death, EGFP-c-Fos transfected cells were fixed by 4% paraformaldehyde after drug treatment, and permeabilized by 0.1% Tx-100 in PBS. After incubated with TUNEL reaction mixture (TMR Red; Roche, Germany) for 1 hr at 37°C in dark, cells were imaged by fluorescent microscopy (EVOS FL Auto2; Thermo Fisher Scientific, USA). Images were processed and analyzed by Celleste software (Thermo Fisher Scientific, USA).

| Calcein-AM assay
For measuring cell viability, cells were incubated with 1 μM of Calcein-AM (Invitrogen, USA) in DMEM for 1 hr at 37°C. After changing the medium to PBS, fluorescent signals were measured using luminometer (excitation at 490 nm and emission at 520 nm).

| MTS assay
For measuring cell viability, cells were treated with MTS solution (Promega, USA) in cell culture medium. After 1 hr incubation at 37°C, signals were measured using a spectrophotometer at 490 nm.

| Luciferase assay
The cells transfected with AP-1-luc plasmid were lysed in passive lysis buffer (Promega, USA). Equal amounts of cell lysates were mixed with a luciferase assay reagent (Promega, USA), and then the measurement was performed using luminometer. These processes were performed according to manufacturer's instructions (Promega, USA).

| Statistical analysis
Data were analyzed using unpaired t-tests or one-way analysis of variance (ANOVA) or two-way ANOVA with Bonferroni posthoc tests for multiple comparisons. p values <0.05 were considered as statistically significant. All data are shown as mean ± SEM.