nNOS–CAPON interaction mediates amyloid‐β‐induced neurotoxicity, especially in the early stages

Summary In neurons, increased protein–protein interactions between neuronal nitric oxide synthase (nNOS) and its carboxy‐terminal PDZ ligand (CAPON) contribute to excitotoxicity and abnormal dendritic spine development, both of which are involved in the development of Alzheimer's disease. In models of Alzheimer's disease, increased nNOS–CAPON interaction was detected after treatment with amyloid‐β in vitro, and a similar change was found in the hippocampus of APP/PS1 mice (a transgenic mouse model of Alzheimer's disease), compared with age‐matched background mice in vivo. After blocking the nNOS–CAPON interaction, memory was rescued in 4‐month‐old APP/PS1 mice, and dendritic impairments were ameliorated both in vivo and in vitro. Furthermore, we demonstrated that S‐nitrosylation of Dexras1 and inhibition of the ERK–CREB–BDNF pathway might be downstream of the nNOS–CAPON interaction.

The carboxy-terminal PDZ ligand of nNOS (NOS1AP, also named CAPON), a scaffolding protein of nNOS that indirectly binds to NMDARs, is a risk factor in many neurological diseases, such as schizophrenia, autism, bipolar disorder, post-traumatic stress disorder, and depression (Candemir et al., 2016;Courtney, Li & Lai, 2014). CAPON forms a complex with nNOS and transmits NO signals to other proteins related to excitotoxicity when NMDARs are activated Zhu et al., 2015). CAPON also regulates dendritic morphology, dendrite patterning, and dendritic spine development (Candemir et al., 2016;Carrel et al., 2009;Richier et al., 2010). Both excitotoxicity and synaptic dysfunction are major causes of Alzheimer's disease. Thus, we speculated that nNOS-CAPON interaction might be an important downstream signaling pathway of abnormal NMDAR activation in Alzheimer's disease.
We previously reported that increased nNOS-CAPON interaction induced anxiety-related behaviors via decreasing dendritic spine density and weakening prosurvival signals . In this study, we found that nNOS-CAPON interaction was increased in both amyloid-b 1-42 -treated primary cultured neurons in vitro and in the hippocampus of APP/PS1 mice. Blocking nNOS-CAPON interaction rescued neuron damage, the decrease in dendritic spines, memory loss, and prosurvival signals impaired by amyloid-b. Therefore, our work presents a potential downstream signal of amyloid b-mediated dysfunction of NMDARs.
2 | RESULTS 2.1 | nNOS-CAPON interaction was increased in primary cultured hippocampal neurons treated with Ab and in the hippocampus of APP/PS1 mice As a potential ligand for NMDARs, Ab may activate NMDARs and, consequently, increase nNOS-CAPON interaction, causing pathological changes in neurons (Candemir et al., 2016;Courtney et al., 2014). Here, we hypothesized that Ab-induced nNOS-CPAON interaction was associated with Ab neurotoxicity. Therefore, we APP/PS1 mice represent one of the most important animal models of Alzheimer's disease, and APP/PS1 mice show memory loss with age. nNOS-CAPON interaction was detected in the hippocampus of 4-month-old and 9-month-old APP/PS1 mice. Compared with aged-matched background mice, APP/PS1 mice exhibited increased nNOS-CAPON interaction in the hippocampus (Figure 1c, p < .05).
Therefore, nNOS-CAPON interaction was increased in the Alzheimer's disease model in vitro and in vivo. Subsequently, we determined whether increased nNOS-CAPON interaction contributed to Ab-mediated neurotoxicity. We constructed a peptide, named TAT-CAPONi, to interfere with the nNOS-CAPON interaction. TAT-CAPONi comprises a cell-penetrating peptide (TAT) and the 12 amino acids of the C terminal of CAPON. TAT-CAPONi selectively blocks nNOS-CAPON binding . The Ala 22 Asp mutation of TAT-CAPONi renders it incapable of binding to nNOS (Jaffrey, Snowman, Eliasson, Cohen & Snyder, 1998). Thus, we Fibril-Ab 1-42 , another toxicity form of Ab 1-42 , could also induce increased nNOS-CAPON interaction in vitro (Fig. S1c), and TAT-CAPONi showed protective effects on neurotoxicity induced by fibril-Ab 1-42 (10 lmo/L) in vitro (Fig. S1d). Together, these findings suggest that Ab 1-42 causes an increase in nNOS-CAPON interaction, and nNOS-CAPON interaction contributes to neuronal damage induced by Ab 1-42 .
2.2 | Blocking nNOS-CAPON interaction rescues memory deficits in vivo, especially in the early stage Injection of fibril-Ab 1-42 into the lateral ventricles of mice can induce Alzheimer's-like symptoms (Shen, Yan & He, 2016). We pretreated C57BL/6J mice with TAT-CAPONi or TAT-CAPONi/A22D (3 nmol/g weight, i.v.) and injected fibril-Ab 1-42 into the lateral ventricles of mice after 45 min. Mice were treated with peptides at a dose of 3 nmol/g weight/day for 3 days, and their memory abilities were tested using the Morris water maze. TAT-CAPONi, TAT-CAPONi/ A22D, or vehicle was administered to mice daily during the period of the Morris water maze test (Figure 2a). Injection of fibril-Ab 1-42 into the lateral ventricles of mice induced memory deficit (Figure 2b). There was no difference in the swimming speed of mice between groups (Fig. S2b). Administration of TAT-CAPONi but not TAT-CAPONi/A22D prevented the memory deficit ( Figure 2b). To confirm the protective effects of uncoupling the nNOS-CAPON interaction, we repeated the above experiment with a small-molecule blocker of nNOS-CAPON binding, Zlc002 (30 mg kg À1 day À1 , i.v.) . The results showed that Zlc002 rescued the memory loss of mice injected with Ab ( Figure 2c).
To exclude other potential pathways influenced by TAT-CAPONi, we administered TAT-CAPONi to nNOSKO mice and then injected fibril-Ab 1-42 into the lateral ventricles of the nNOSKO mice using the same protocol as above. nNOSKO mice did not benefit from the administration of TAT-CAPONi in the Morris water maze test ( S2c). Moreover, TAT-CAPONi was able to across the blood brain barrier (Fig. S3).
In summary, blocking nNOS-CAPON interaction rescued the memory loss in the in vivo Alzheimer's disease model.

| Blocking nNOS-CAPON interaction rescues dendritic impairments induced by Ab
Synaptic dysfunction and spine loss, but not neuron loss, are related to memory deficits in patients in the early stage of Alzheimer's disease and in young APP/PS1 mice (Herms & Dorostkar, 2016). Therefore, the beneficial effects observed upon blocking nNOS-CAPON interaction in the 4-month-old APP/PS1 mice might be due to rescue of synaptic function. To confirm this hypothesis, we evaluated morphological changes associated with dendritic and synaptic functions. First, we evaluated the PSD95 and Synapsin I densities in the hippocampus of APP/PS1 mice by immunofluorescence. PSD95 is a specific postsynaptic membrane protein, and Synapsin 1 is a specific presynaptic membrane protein. Therefore, the densities of PSD95 and Synapsin I partly reflect the potency of synaptic connections (Goetzl et al., 2016). The results showed that the densities of PSD95 and Synapsin I in the hippocampus of 4-month-old APP/PS1 mice were significantly lower than those of the background group, while F I G U R E 1 nNOS-CAPON interaction increased after exposure to oligo-Ab both in vitro and in vivo. (a) Summary of the role of nNOS-CAPON interaction in Abinduced neurotoxicity. (b) Administration of oligo-Ab 1-42 influenced the nNOS-CAPON-Dexras1 interaction in vitro (n = 6, * p < .05 compared to the control group, # p < .05 compared to the oligo-Ab 1-42 group). (c) Interaction of nNOS-CAPON-Dexras1 increased in both 4month-old and 9-month-old APP/PS1 mice compared to age-matched wild-type mice (n = 5, * p < .05 compared to the control group, # p < .05 compared to the APP/PS1 group) the administration of TAT-CAPONi increased the densities of both PSD95 and Synapsin I in the hippocampus of APP/PS1 mice (Figure 3a). Second, we evaluated the density of dendritic spines in the hippocampus by Golgi staining. The density of spines in the hippocampus was reduced in 4-month-old APP/PS1 mice compared with that in background mice, and TAT-CAPONi increased the spine density ( Figure 3b). Finally, we used a primary neuron model to confirm whether the change in dendrites was directly related to Ab.
Twenty-four hours after neurons were treated with oligo-Ab 1-42 , the densities of PSD95 and Synapsin 1 were significantly decreased, and the lengths and number of branch points of dendrites were also decreased ( Figure 3c,d). Therefore, blocking nNOS-CAPON interaction with TAT-CAPONi rescues dendritic impairments induced by Ab in vitro.

| ERK-CREB-BDNF pathway is involved in the effects of blocking nNOS-CAPON interaction
To explore the molecular mechanisms underlying the dendritic morphology changes after nNOS-CAPON blocking, we examined ERK phosphorylation, a kinase that influences synaptic maturation and stability (Collingridge, Peineau, Howland & Wang, 2010). TAT-F I G U R E 2 Blocking nNOS-CAPON interaction rescued memory loss in vivo. (a) Schedule implemented for the intracerebroventricular injection model. (b) After the injection of fibril-Ab 1-42 , mice showed memory loss in the Ab group, while administration with TAT-CAPONi rescued memory loss. However, mice did not benefit from the administration of TAT-CAPONi/A22D. (n = 14 per group in Morris water maze, * p < .05 compared to the control group, # p < .05 compared to the fibril-Ab 1-42 group). (c) Administration of Zlc-002, a small molecule that blocks nNOS-CAPON interaction, rescued memory loss induced by Ab 1-42 (n = 13 per group, * p < .05 compared to the control group, # p < .05 compared to the fibril-Ab 1-42 group) . (d) Blocking nNOS-CAPON interaction with TAT-CAPONi could not rescue memory loss induced by Ab 1-42 in nNOSKO mice (n = 14 per group). (e) Blocking nNOS-CAPON interaction rescued memory loss in 4-month-old APP/PS1 mice (n = 12 per group, * p < .05 compared to background mice, # p < .05 compared to the APP /PS1 group) F I G U R E 3 Blocking nNOS-CAPON interaction rescued dendritic changes both in vivo and in vitro. (a) Administration of TAT-CAPONi increased the density of Synapsin I and PSD95 in the CA1, the CA3 and the hilus (5 mice per group, * p < .05 compared to the age-matched C57BL/6J group, # p < .05 compared to the APP/PS1 group). (b) Administration of TAT-CAPONi increased the density of spines in different regions of the hippocampus in APP/PS1 mice (n = 6, * p < .05 compared to the age-matched C57BL/6J group, # p < .05 compared to the APP/PS1 group). (c,d) Administration of oligo-Ab 1-42 decreased the density of Synapsin I and PSD-95 double-labeling puncta in neurons, which indicated a loss of connections between neurons, but increased density of Synapsin I and PSD-95 double-labeling puncta was found after the administration of TAT-CAPONi (n = 24 per group, * p < .05 compared to the control group, # p < .05 compared to the oligo-Ab 1-42 group). (e, f) Administration of TAT-CAPONi rescued the length of dendrites in vitro (n = 30, * p < .05 compared to the control group, # p < .05 compared to the oligo-Ab 1-42 group) CAPONi significantly increased pERK abundance in the hippocampus

| DISCUSSION
The amyloid cascade hypothesis is one of the most prominent theories in Alzheimer's disease research, while Ab 1-42 and/or its oligomers are thought as the major toxic components. Clinical evidence suggests that increases in Ab 1-42 and its oligomers are associated with disease, and laboratory evidence suggests that Ab 1-42 oligomers induce abnormal synaptic function and neuronal death (Wilcox, Lacor, Pitt & Klein, 2011). Based on the amyloid cascade hypothesis, studies of Alzheimer' disease have generally focused on two common treatment strategies: 1. decreasing the level of Ab, especially toxic forms; and 2. protecting neurons or maintaining neuronal function.
Inhibition/activation of secretases of Ab generation in the amyloidogenic pathway was shown to significantly decrease the level of Ab (Parsons & Rammes, 2017), but some forms of Ab were reported to protect the brain from infection and inflammation (Kumar et al., 2016). In physiological conditions, some transmembrane receptors and signaling proteins are also substrates of alpha-, beta-, and gamma-secretase, key enzymes in the amyloidogenic pathway, and thus interference with the activation of secretases causes serious side effects (Parsons & Rammes, 2017). Therefore, strategies aimed to decrease the total level of Ab may be debatable. Neutralizing toxic forms of Ab with antibodies seems to be a more direct approach than interfering with secretases, but the outcomes of clinical trials using antibody drugs showed less beneficial effects than expected (Freskgard & Urich, 2017). Drugs are also developed to modulate Ab aggregation, but a more detailed understanding of the molecules that retard aggregate formation and of the structure-activity relationships are needed (Alam, Siddiqi, Chturvedi & Khan, 2017).
Therefore, many problems must be solved before drugs can be used to treat Alzheimer's disease by decreasing Ab or its toxic forms.
Another strategy is to reduce the toxicity of Ab based on its tar-  ). Nonetheless, no nNOS inhibitor has yet to be approved for entry into clinical trials.
In the present research, we focused on nNOS-CAPON interaction, which is a downstream signal of nNOS and NMDARs. According to previous studies, increased nNOS-CAPON interaction mediates excitotoxicity, overload of iron, and stress-related depressive behaviors in the brain (Courtney et al., 2014). Under physiological conditions, CAPON regulates dendrite patterning and dendritic spine development of neurons, and Dexras1 plays important roles in the circadian clock and the responses to fluid deprivation and salt loading (Carrel et al., 2009;Cheng et al., 2004;Greenwood et al., 2016;Richier et al., 2010;Van Gelder, 2004). These functions are associated with neither nNOS-CAPON interaction nor nNOS. Moreover, nNOS-CAPON interaction blockers used in our experiments did not affect the resting membrane potential of neurons, the F I G U R E 4 The beneficial effect of blocking the nNOS-CAPON interaction might be due to rescue of the ERK-CREB-BDNF pathway. (a) pression of spinophilin and BDNF and decreased phosphorylation of ERK and CREB in the hippocampus, and these changes were recovered by TAT-CAPONi (n = 6, * p < .05 compared to the 4-month-old C57BL/6J mouse group, # p < .05 compared to the 4-month-old APP/PS1 mice group). (b) Beneficial effects could be observed in vitro and were abolished with an ERK1/2 inhibitor but not with a PKA inhibitor or a TrkB inhibitor (n = 6, * p < .05 compared to the control group, # p < .05 compared to the oligo-Ab group, & p < .05 compared to the oligo-Ab+TAT-CAPONi group) ZHANG ET AL. | 7 of 12 interaction of PSD95 with NMDARs, or the interaction of CAPON with synapsins . Therefore, blocking nNOS-CAPON interaction might induce fewer side effects than the direct intervention of NMDARs or nNOS. The ERK-CREB-BDNF pathway is important for maintaining the survival and function of neurons in Alzheimer's disease, and increased activation of this pathway might slow the development of Alzheimer's disease (Kamat et al., 2016).
The beneficial effects of blocking nNOS-CAPON interaction come from the recovery of dendrites and the ERK-CREB-BDNF pathway, and thus, this approach is different from blocking NMDARs, such as with memantine, which decreases the influx of Ca 2+ and excitotoxicity (Ferreira-Vieira, Guimaraes, Silva & Ribeiro, 2016). The difference of plaque deposition between 4-month-old APP/PS1 mice than in 9-month-old APP/PS1 mice (Fig. S4a) might be the reason that blocking nNOS-CAPON interaction showed beneficial effects in 4-month-old APP/PS1 mice but not in 9-month-old APP/PS1 mice.

| Ethical statements of animal research
Homozygous male nNOS-deficient mice (B6; 129S4-Nos1 tm1Plh , nNOS À/À ; The Jackson Laboratory) and male wild-type (WT) mice with a similar genetic background (B6129SF1, nNOS +/+ ) were F I G U R E 5 The nNOS-CAPON interaction mediated S-nitrosylation of Dexras1, which plays important roles in the neurotoxicity induced by Ab. (a) Increased S-nitrosylation of Dexras1 was detected in vitro and in vivo, but the increase could be rescued by administration of TAT-CAPONi (n = 6 in vitro, n = 4 in the model in vivo, * p < .05 compared to the control group, # p < .05 compared to the oligo-Ab 1-42 group or the APP/PS1 group). (b) Overexpression of Dexras1-C11S, in which an S-nitrosylation site is lost, decreased the neuronal injury induced by Ab 1-42 (n = 6, * p < .05 compared to the LV-GFP group, # p < .05 compared to the LV-GFP+ oligo-Ab 1-42 group). (c) Overexpression of Dexras1-C11S also rescued memory loss in the in vivo intracerebroventricular injection model (n = 12 per group, * p < .05 compared to the control group, # p < .05 compared to the fibril-Ab 1-42 group) maintained in the Model Animal Research Center of Nanjing University. The nNOS À/À and nNOS +/+ embryos were used for cell cultures. Male C57BL/6J-TgN (APP/PS1) and male WT mice (C57BL/ 6J) were purchased from Beijing HFK Bio-technology Co. Ltd. APP/ PS1 mice express chimeric amyloid precursor protein (APPswe) encoding the Swedish mutations K595N/M596L, and human presenilin 1 PS1-dE9 (deletion of exon 9) controlled by independent mouse prion protein promoter elements. All experimental protocols using animals were approved by the Institutional Animal Care and Use Committee of Nanjing Medical University.

| Cell cultures
Primary neurons were isolated from the hippocampus of mouse embryos at embryonic day 15 (E15), as previously described (Zhang et al., 2010). Cultures were maintained in neurobasal media supplemented with B27, penicillin, streptomycin, and L-glutamine. Cells were cultured on polyornithine-coated tissue culture dishes (3.5 cm in diameter) at a density of 1 9 10 4 cells/cm 2 (morphological analysis) or 1 9 10 5 cells/cm 2 (morphological analysis). Cells were grown for 7-14 d in vitro (DIV) and used for specific experiments, as indicated below. The proportion of b-III-tubulin + cells at 10 DIV was $ 92%.
All cultures were maintained in an incubator (HERAcell 150, Thermo Fisher Scientific) with a humidified atmosphere of 95% air and 5% CO 2 at 37°C.

| Preparation of Ab 1-42 oligomers and fibrils
Oligomerized Ab 1-42 was prepared as previously reported (Brkic et al., 2015). Briefly, Ab 1-42 was dissolved at 1 mg/ml in hexafluoroisopropanol (HFIP; Sigma-Aldrich), and the HFIP was removed after 1 hr. The peptide film was resolved at 1 mg/ml in DMSO (Sigma-Aldrich). The solution was diluted to 100 lmol/L with Ham's F-12 medium without glutamine. Then, the solution was allowed to stand for 1 hr at 25°C followed by centrifuging at 14,000 9 g for 10 min at 4°C to remove any insoluble aggregates.
For preparation of fibrils, the dried hexafluoro-2-propanol film of Ab 1-42 (see above) was dissolved in sterile PBS to yield a 3 lg/ll solution and incubated for 7 days at 37°C (De Felice et al., 2007).

| Recombinant virus production and infection
The recombinant lentivirus LV-GFP-Dexras1-C11S or its control LV-GFP was generated as we previously described .

| Coimmunoprecipitation
Lysis and coimmunoprecipitation of cultures and tissues were performed as we previously described (Zhou et al., 2010). Cultured neurons or hippocampal tissues were lysed in 50 mmol/L Tris-HCl (pH 7.4) in buffer containing 150 mmol/L NaCl, 1 mmol/L EDTA-Na, 1% NP-40, 0.02% sodium azide, 0.1% SDS, 0.5% sodium deoxycholate, 1% PMSF, 1& aprotinin, 1& leupeptin, and 0.5& pepstatin A. The lysates were centrifuged at 12,000 9 g for 15 min at 4°C. The supernatant (200 ll) was preincubated for 1 h at 4°C with 25 ll of protein G-Sepharose beads (Sigma-Aldrich) and then centrifuged to remove proteins that adhered nonspecifically to the beads and to obtain the target supernatant for the following IP experiment. Protein G-Sepharose beads were incubated with rabbit anti-nNOS (1:200) for 3-4 h. The antibody-conjugated protein G-Sepharose beads and the target supernatant were added for incubation overnight at 4°C. Immune complexes were isolated by centrifuging and washing five times with 0.05 mol/L HEPES buffer (pH 7.1) containing 0.15% Triton X-100, 0.15 mol/L NaCl, and 0.1 9 10 À3 mol/L sodium orthovanadate. Bound proteins were eluted by heating at 100°C in loading buffer. Proteins were analyzed by immunoblotting using rabbit anti-CAPON (1:500) or rabbit anti-nNOS (1:1000).

| Morris water maze
The spatial cognitive performance of mice was evaluated by the Morris water maze. The Morris water maze protocol has been described in detail in our previous report (Li et al., 2009). In brief, the mouse was placed in opaque water of a circular swimming pool and trained to locate the hidden platform 0.5 cm under the surface of the water. During the training to find the hidden platform, mice were allowed to swim for a maximum of 60 s in the pool for each trial. One block of four trials per day was performed for 5 consecutive days. On the sixth day, mice performed one 60 s retention probe test during which the platform was removed from the pool.
During retention, the number of crossings of the platform location and the time spent in the target quadrant were measured.

| Biotin-switch assay
This assay was performed in the dark, as previously described (Fang et al., 2000). Briefly, cells were lysed in HEN buffer (250 mmol/L HEPES, 1 mmol/L EDTA, and 100 mmol/L neocuproine) adjusted to contain 0.4% CHAPS. Samples were homogenized, and free cys- propionamide; 1 mmol/L; Sigma-Aldrich], proteins were incubated at room temperature for 1 hr. After separation using an SDS-PAGE gel in nonreducing loading buffer, biotinylated proteins were detected by immunoblotting using rabbit antibiotin. Alternatively, biotinylated proteins were resuspended in 250 ll of HENS buffer plus 500 ll of neutralization buffer (20 mmol/L HEPES, 100 mmol/L NaCl, 1 mmol/L EDTA, 0.5% Triton X-100) and precipitated with 50 ll of prewashed avidin-affinity resin beads (Sigma-Aldrich) at room temperature for 1 h. The beads were washed five times at 4°C using neutralization buffer containing 600 nmol/L NaCl. Biotinylated proteins were eluted using 30 ll of elution buffer (20 mmol/L HEPES, 100 mmol/L NaCl, 1 mmol/L EDTA, 100 mmol/L b-mercaptoethanol) and heated at 100°C for 5 min in reducing SDS-PAGE loading buffer.

| Golgi-Cox staining
Fresh brains that had not undergone perfusion or fixation were used for Golgi-Cox staining to show subtle morphological alterations in neuronal dendrites and dendritic spines. Golgi-Cox staining was performed with an FD Rapid GolgiStain Kit (FD NeuroTechnologies) according to the user manual. Briefly, the brains were first placed in impregnation solution for 2 weeks followed by 2 days in a 30% sucrose solution. Then, they were cut into 100 lm coronal sections using a vibratome (World Precision Instruments) and stained. For morphological analysis, 10 random neurons from each sample were measured, and the average was regarded as the final value of one sample.

| Immunofluorescence
The details of immunofluorescence for brain sections and cultured cells have been described . Briefly, brain slices were fixed in 4% paraformaldehyde (30 lm) and cultured cells were blocked with blocking solution (10% serum of donkey, 0.2% Triton X-100). After washing with PBS, samples were incubated with primary antibodies overnight at 4°C. Then, samples were washed with PBS and incubated with secondary antibodies (2 hr at room temperature). After washing with PBS, samples were counterstained with Hoechst 33258 (Sigma-Aldrich) to label the nuclei and mounted. Fluorescence was visualized by confocal microscopy (LSM 700, Zeiss).
For the immunofluorescence analysis of brain sections from TAT-CAPONi-treated mice and control mice, brains were obtained 45 min after the injection (TAT-CAPONi or vehicle) and fixed with 4% paraformaldehyde. After sectioning, brain slices (30 lm) were treated with goat anti-mouse IgG (H+L) (1:200) at 30°C for 1 hr after blocking with blocking solution. After washing with PBS, samples were incubated with anti-HIV1 tat antibody (Abcam, ab63957, 1:100, overnight at 4°C). Then, samples were washed with PBS and incubated with secondary antibodies (2 hr at room temperature).
After washing with PBS, samples were counterstained with DAPI (Sigma-Aldrich) to label the nuclei and mounted. Fluorescence was visualized by confocal microscopy.

| Congo red staining
Brain slices (30 lm) fixed in 4% paraformaldehyde were used for Congo red staining. For Congo red staining, the sections were stained in Congo red (0.5% in methanol with 20% glycerol) for 20 min, differentiated in alkaline alcohol solution, counterstained with Gill's hematoxylin for 30 s, and finally mounted with neutral gum. Congo red-stained plaques were visualized by microscopy.

| Statistical analysis
Comparisons among multiple groups were made with one-way ANOVA (one factor) or two-way ANOVA (two factors) followed by Scheff e's post hoc test. Comparisons between two groups were made with two-tailed Student's t test. Data were presented as the mean AE SEM, and p < .05 was considered statistically significant.
Investigators were blinded to the group allocation when assessing the outcomes.

AUTHOR CONTRIBU TI ONS
Yu Zhang participated in designing the study, supervising the analysis, and writing the manuscript. Zhu Zhu, Hai-Ying Liang, and Lei Zhang carried out the analysis and participated in designing the study. Huan-Yu Ni participated in the analysis. Qi-Gang Zhou participated the progress of the revision and helped to write the manuscript. Chun-Xia Luo and Dong-Ya Zhu participated in the study design, coordinating the study, and drafting and finalizing the manuscript.