Yun Xu, Department of Neurology, Affiliated Drum Tower Hospital of Nanjing University Medical School, 321 ZhongShan Road, Nanjing City, Jiangsu Province, 210008, China. Tel./fax: +86 25 8310 5208; e-mail:firstname.lastname@example.org or
Renxiang Tan, Institute of Functional Biomolecules, State Key Laboratory of Pharmaceutical Biotechnology, Nanjing University, Nanjing City, Jiangsu Province, 210093, China. Tel./fax: +86 25 8368 6559; e-mail: email@example.com
Increasing evidence demonstrates that amyloid beta (Aβ) elicits mitochondrial dysfunction and oxidative stress, which contributes to the pathogenesis of Alzheimer's disease (AD). Identification of the molecules targeting Aβ is thus of particular significance in the treatment of AD. Hopeahainol A (HopA), a polyphenol with a novel skeleton obtained from Hopea hainanensis, is potentially acetylcholinesterase-inhibitory and anti-oxidative in H2O2-treated PC12 cells. In this study, we reported that HopA might bind to Aβ1–42 directly and inhibit the Aβ1–42 aggregation using a combination of molecular dynamics simulation, binding assay, transmission electron microscopic analysis and staining technique. We also demonstrated that HopA decreased the interaction between Aβ1–42 and Aβ-binding alcohol dehydrogenase, which in turn reduced mitochondrial dysfunction and oxidative stress in vivo and in vitro. In addition, HopA was able to rescue the long-term potentiation induction by protecting synaptic function and attenuate memory deficits in APP/PS1 mice. Our data suggest that HopA might be a promising drug for therapeutic intervention in AD.
Neurotoxicity resulting from increased levels of intracellular amyloid beta (Aβ), predominantly Aβ1–42 and its aggregates, plays an important role in cognitive impairment in Alzheimer's disease (AD; Cleary et al., 2005; Muirhead et al., 2010). Recent evidence indicates that mitochondrial dysfunction is a critical feature of Aβ-induced neurotoxicity, that is, reduced brain metabolism, oxidative stress and calcium dysregulation, collaborate with each other to form a deadly mitochondrial spiral (Mutisya et al., 1994; Smith et al., 2000). Aβ-binding alcohol dehydrogenase (ABAD) is a direct molecular link between Aβ and mitochondrial dysfunction. Specifically, the interaction between Aβ and ABAD decreases the activity of ABAD and causes mitochondrial dysfunction in patients with AD and transgenic mouse models of AD (Takuma et al., 2005). An ABAD decoy peptide with the capability of specifically disrupting ABAD–Aβ interaction suppresses Aβ-mediated mitochondrial toxicity (Yao et al., 2011), suggesting that this interaction may be a promising target for therapeutic intervention in AD.
Certain molecules, such as curcumin, are able to bind to Aβ directly and inhibit its aggregation due to three key structural features – two phenolic end groups, the proper substitution pattern of these aromatic rings and the appropriate length and flexibility of the linker region (Reinke & Gestwicki, 2007; Parachikova et al., 2010). Hopeahainol A (HopA), with a molecular formula of C28H16O8, is a potentially acetylcholinesterase (AChE)-inhibitory and anti-oxidative polyphenol in H2O2-treated PC12 cells with a novel carbon skeleton isolated from Hopea hainanensis in our laboratory (Ge et al., 2008; Nicolaou et al., 2010). Its total synthesis and bioactivity confirmation are accomplished recently (da Shi et al., 2009). Motivated by the fact that HopA shared the polyphenolic feature and motifs with curcumin, we examined whether HopA could bind to Aβ1–42 directly using molecular dynamic simulation and binding assays. In this study, we demonstrated for the first time that HopA was able to decrease amyloid burden and inhibit the detrimental ABAD–Aβ1–42 interaction. In addition, HopA-mediated neuroprotective effects in AD models were determined both in vivo and in vitro.
HopA might interact with Aβ1–42 directly and protects neurons from Aβ1–42 toxicity in vitro
To examine whether HopA was able to inhibit Aβ1–42 assembly, several experiments were conducted. First, to determine whether HopA interacted with Aβ1–42, we performed an in vitro binding assay. The signal from HopA at different concentrations (10, 20, 30, 40, 50, and 60 μm) was monitored in the absence or presence of Aβ1–42 (0.1 mm) by high performance liquid chromatography (HPLC), while other conditions were kept constant. As shown in Fig. 1A, the concentration of HopA recovered in the supernatant increased in proportion to the amount added, whether or not Aβ1–42 was present. However, Aβ1–42 markedly reduced HopA recovery, which indicated that HopA might bind to Aβ1–42 directly. Next, transmission electron microscopy (TEM) analysis was performed to determine whether HopA could inhibit the formation of Aβ1–42 fibrils in vitro. When incubated for 6 days at 37 °C, Aβ1–42 (50 μm) formed extensive mature fibrils compared with coculture of Aβ1–42 and HopA (Fig. 1B2–B4). To determine whether HopA could also be inhibitory when added after initial aggregation events, 50 μm of Aβ1–42 was incubated for 3 days at 37 °C and then incubated for another 3 days at 37 °C with 0.1 or 1 μm HopA (Fig. 1B5–B6). Similar results were obtained, indicating that HopA decreased fibril formation even after initial Aβ1–42 aggregation had already occurred.
HopA was able to cross the blood–brain barrier quickly in wild-type B6 mice (Fig. S1), and the ability of HopA inhibiting amyloid plaque formation was examined using Congo red staining in APP/PS1 mice. The intensity of Congo red staining in cortex and hippocampus was significantly reduced after treatment with 4 mg kg−1 day−1 HopA for 30 days (cortex: 0.68 ± 0.06% vs. 0.35 ± 0.03%, P <0.01; hippocampus: 0.51 ± 0.04% vs. 0.30 ± 0.03%, P <0.01, Fig. 1C1). In addition, the immunohistochemistry data indicated that Aβ1–42 deposition in the cortex of APP/PS1 mice was also suppressed by HopA (0.93 ± 0.10% vs. 0.57 ± 0.03%, P <0.05, Fig. 1C2).
Finally, the molecular dynamics (MD) simulation, a widely used computational tool to investigate protein–ligand interaction, was performed to investigate the HopA–Aβ1–42 interaction. Four HopA binding sites including Glu3, Asp7, Glu22, and Ala42 were found in Aβ1–42 (Fig. 1D1–D2). Correspondingly, four different HopA–Aβ1–42 complexes were formed (Fig. 1D3–D6). In addition, Phe4 was another potential binding site for HopA.
To understand the dominant binding mode in the complexes, the contributions from electrostatic and van der Waals interactions were also displayed. As shown in Table S1, Ala42 at the C-terminal of Aβ1–42 was the most favorable binding site for HopA with interaction energy of −175.19 kcal mol−1. Meanwhile, the electrostatic attraction (−164.56 kcal mol−1) was much stronger than the van der Waals interaction (−3.96 kcal mol−1), which demonstrated that the binding of HopA to Aβ1–42 at Ala42 might be predominately governed by electrostatic attraction. For a more comprehensive understanding of the interaction, the geometry of the complex of HopA with Aβ1–42 was shown in Fig. 1D. As expected, there were hydrogen bonding interactions between the hydroxyl oxygen atom of HopA and the carboxyl of Ala42 in Aβ1–42. The hydrogen bonding length (d(O−)H···O) and angle (∠O–H···O) were 1.72 and 172.4°, respectively.
To investigate whether HopA could protect against Aβ1–42-induced neurotoxicity, primary cortical neurons were incubated with Aβ1–42 for 24 h in the presence or absence of HopA. As shown in Fig. S2A, the viability of neurons was decreased in Aβ1–42-treated neurons and HopA (2 and 4 μm) significantly reduced Aβ1–42-induced cell death (P <0.05). The neuroprotective effects were also confirmed by PI staining and LDH assay (Fig. S2B,C).
HopA reduces the interaction between Aβ1–42 and ABAD
As shown in Fig. 2A, we confirmed that Aβ1–42 colocalized with ABAD in Aβ1–42-treated neurons by immunostaining, which was also demonstrated by immunoprecipitation in APP/PS1 mice and Aβ1–42-treated neurons (Fig. 2C,D). Treatment with Aβ1–42 reduced the activity of ABAD in primary neurons, while this effect could be partially reversed by HopA (Fig. 2B). The intensity of the ABAD–Aβ complex was significantly decreased after treating with HopA for 30 days in APP/PS1 mice, indicating that HopA decreased the ABAD–Aβ1–42 interaction. In addition, the expression of ABAD was increased in APP/PS1 mice compared with B6 mice, while it was not altered in APP/PS1 mice treated with or without HopA (Fig. 2C). Similar results were observed in Aβ1–42-treated neurons with HopA treatment (Fig. 2D).
HopA protects against mitochondrial dysfunction
Aβ1–42 was demonstrated to interact with ABAD in the mitochondria and to induce mitochondrial dysfunction and oxidative stress (Lustbader et al., 2004). To examine whether HopA could preserve mitochondrial function, mitochondrial membrane potential (Δψ) was measured using JC-1, a cationic lipophilic fluorescent probe. Mitochondrial complex IV activity, release of cytochrome c and caspase-3 activity were also assessed. As shown in Fig. 3A, a significant loss of Δψ was found in neurons treated with Aβ1–42, which could be alleviated after HopA treatment (P <0.05). In addition, mitochondrial complex IV activity was significantly increased in Aβ1–42-treated neurons following HopA treatment (Fig. 3B, P < 0.05). No changes appeared in the activities of complex I, complex II, and complex III in Aβ1–42-treated neurons with or without HopA (data not shown). HopA treatment could also decrease the release of cytochrome c and activity of caspase-3 (Fig. 3C,D, P < 0.05).
HopA suppresses oxidative stress in vitro and in vivo
Mitochondria were a primary site of reactive oxygen species (ROS) generation, and Aβ1–42-induced mitochondrial dysfunction could result in the accumulation of ROS. As shown in Fig. 4A, Aβ1–42 induced the accumulation of ROS in primary neurons and HopA could significantly suppress the ROS production (P <0.05).
To further characterize HopA's ability to inhibit oxidative stress in the brain of APP/PS1 mice, markers of lipid peroxidation (4-hydroxynonenal, 4-HNE), protein oxidation (3-nitrotyrosine, 3-NT), and DNA oxidation (8-hydroxy-2′-deoxyguanosine, 8-OHdG) were examined. As shown in Fig. 4B, 4-HNE levels in the cortex and hippocampus of APP/PS1 mice were significantly increased by 122.3% (P <0.01) and 29.2% (P <0.01), respectively, relative to age-matched wild-type B6 mice. 4-HNE levels in APP/PS1 mice treated with HopA were reduced 19.6% in the cortex (P <0.05) and 11.4% in the hippocampus (P =0.18) of those in mice treated with vehicle. Similarly, the levels of 3-NT were reduced 44.7% (P <0.01, hippocampus) and 56.8% (P <0.01, cortex), and 8-OHdG were reduced by 30.1% (P <0.01, hippocampus) and 22.3% (P <0.01, cortex) after treatment of HopA (Fig. 4C,D).
HopA rescues synaptic dysfunction in APP/PS1 mice
To examine the basal properties of the Schaffer collaterale-CA1 synapse, input/output (I/O) function was evaluated by plotting the slope of excitatory postsynaptic potential (EPSP) against the stimulus intensity at 0.1–0.9 mA. As shown in Fig. 5A, the I/O curve in slices obtained from APP/PS1 mice (B = 0.68; n = 12 slices/6 mice) was significantly decreased compared to control mice (B = 1.12; P <0.01, n = 12 slices/6 mice). The treatment with HopA (4 mg kg−1, i.p.) for 30 days could markedly rescue the synaptic dysfunction in APP/PS1 mice (B = 0.93; P <0.05 vs. vehicle-treated APP/PS1 mice, n = 12 slices/6 mice). We further examined the induction of long-term potentiation (LTP), a cellular model of learning and memory, using a high-frequency stimulus (HFS) protocol (100 pulses at 100 Hz). The results showed that HFS evoked a stable potentiation of EPSP slopes in control mice (151.47 ± 7.15% at 55–60 min post-HFS; n = 14 slices/8 mice; Fig. 5B-i). However, the same protocol failed to induce the increase in EPSP slopes in APP/PS1 mice (100.43 ± 6.18%; n = 14 slices/8 mice; Fig. 5B-ii). Importantly, the treatment with HopA could partially recover the LTP induction in APP/PS1 mice (126.35 ± 3.95%; n = 14 slices/8 mice; Fig. 5B-iii), although it did not affect the amplitude of LTP in control mice (148.72 ± 4.17%; n = 14 slices/8 mice; Fig. 5B-iv). The results indicated that HopA treatment could rescue the LTP induction by protecting synaptic function from Aβ1–42-induced damage.
HopA attenuates the memory impairment in APP/PS1 mice
The effect of HopA (4 mg/kg, i.p. for 30 days) on learning and memory was investigated using the Morris water maze test. As shown in Fig. 6A, the mean escape latency of APP/PS mice was significantly increased compared with wild-type B6 mice (P <0.01), while HopA-treated APP/PS1 mice showed significant improvements compared with APP/PS1 mice after the training periods (P <0.05). In addition, the searching distance of APP/PS1 mice was significantly increased compared with B6 mice (P <0.01), while HopA-treated APP/PS1 mice showed significant improvements compared with APP/PS1 mice after the training periods (P <0.05, Fig. 6B). No significant differences of swimming speed were observed among all the groups (P >0.05, Fig. 6C). On the fifth day, the platform was removed and the probe trail was conducted. The number of platform crossings by the HopA-treated APP/PS1 mice was significantly higher than that of APP/PS1 mice (P <0.05, Fig. 6D). Moreover, HopA-treated APP/PS1 mice spent more time in the target quadrant than APP/PS1 mice (Fig. 4E), and the percent distance in the target quadrant of HopA-treated APP/PS1 mice was also significantly increased compared with APP/PS1 mice (P <0.05, Fig. 6F). Thus, these results demonstrated that HopA treatment significantly improved the memory deficits in APP/PS1 mice.
This study demonstrated for the first time that (i) HopA might be able to bind to Aβ1–42 directly through four potential amino acid residues and decrease the formation of amyloid plaque; (ii) HopA decreased the interaction between Aβ1–42 and ABAD, which was demonstrated to induce mitochondrial dysfunction as well as oxidative stress both in vivo and in vitro; and (iii) HopA protected synaptic function and attenuated memory deficits in APP/PS1 mice.
The accumulation of Aβ-forming senile plaques is one of the main pathological hallmarks of AD. Aβ aggregation is neurotoxic and may be the primary toxic species in AD (Muirhead et al., 2010). Therefore, prevention of Aβ aggregation and attenuation of its neurotoxicity have been the focus of AD therapies. This study suggested that HopA might bind to Aβ1–42 directly through four negatively charged residues, Glu3, Asp7, Glu22, and Ala42, which was possibly due to its functional groups, such as the hydroxyl group, and aromatic rings (Reinke & Gestwicki, 2007). Clinical trials showed that immunization with Aβ42 could clear the amyloid plaques in patients with AD, but it did not prevent progressive neurodegeneration (Holmes et al., 2008). Actually, intracellular Aβ accumulation seems to be more closely related to AD symptoms onset and progression than extracellular Aβ deposition, which results in amyloid plaque formation (Cavallucci et al., 2012). Our results have demonstrated that HopA not only decreased amyloid load, but also suppressed the interaction of Aβ1–42 and ABAD, which led to partial recovery of the mitochondrial function and decrease in oxidative damage in vitro and in vivo. Aβ–ABAD interaction induces mitochondrial dysfunction and oxidative stress (Lustbader et al., 2004; Seo et al., 2011), which triggers a series of events including increased generation of ROS, DNA fragmentation, lactate dehydrogenase and cytochrome c release, diminished COX IV activity, and decreased ATP levels. Finally, it results in high caspase-3 activity, apoptosis, and severe impairment in spatial learning and memory (Marques et al., 2009). The expression of ABAD was increased in transgenic APP/PS1 mice, which was consistent with previous reports (Yan et al., 1997; He et al., 2002; Lustbader et al., 2004). We speculated that abundant Aβ led to the dysfunction of ABAD, which was induced to maintain the homeostasis in AD brains, and eventually became a cellular cofactor for Aβ-induced cell stress. In addition, our study also demonstrated that HopA could not only bind to the C-terminal of Aβ1–42 directly and decrease the ABAD–Aβ1–42 interaction, but also restored ABAD activity without changing its expression. Interestingly, an ABAD decoy peptide (ABAD-DP) alleviates cognitive dysfunction in Tg mAPP/ABAD mice by restraining the ABAD–Aβ interaction (Yao et al., 2011), which indicates that ABAD–Aβ interaction might be an effective target for the treatment of AD. Furthermore, HopA protected the synaptic function and LTP induction in APP/PS1 mice, which was accompanied by the improvement of spatial memory performance.
In addition, HopA is a revisable inhibitor of AChE with an IC50 value of 4.33 ± 0.17 μm in vitro (Ge et al., 2008) and also exerts an AChE inhibitory effect in APP/PS mice (data not shown). Thus, HopA inhibited both Aβ1–42 aggregation and AChE, which contributed to the therapeutic effects on AD. Besides, AChE is a potent amyloid-promoting factor and promotes the aggregation of Aβ through its peripheral anionic site (PAS; Alvarez et al., 1997). The binding assay and TEM analysis indicated that HopA might bind to Aβ1–42 directly and inhibit the self-induced Aβ1–42 aggregation, but whether HopA could also inhibit the AChE-induced Aβ1–42 aggregation remains to be investigated.
In conclusion, HopA is a novel polyphenol extracted from the bark of the stems of Hopea hainanensis. The compound might target Aβ1–42 directly and decrease the ABAD–Aβ1–42 interaction. Furthermore, HopA alleviates mitochondrial dysfunction induced by Aβ1–42 and its downstream cascades, suggesting that it is a promising candidate for the treatment of AD.
Preparation of HopA
HopA was extracted and isolated from Hopea hainanensis as previously described (Ge et al., 2008). HopA was dissolved in DMSO for in vitro studies and dosed at 10% (v/v) DMSO and 90% (v/v) saline for in vivo studies.
APP/PS1 mice identification and treatment
The experimental protocol was approved by the Institutional Animal Care and Use Committee of Nanjing University. We made every effort to minimize the number of mice used and their suffering. APP/PS1 mice expressing chimeric mouse/human amyloid precursor protein (APP) with a double mutation (K670N and M671L) and human presenilin 1 (PS1) with a deletion of exon 9 were obtained from Model Animal Research Center of Nanjing University. Ten-month-old APP/PS1 mice and age-matched wild-type C57BL/6 mice were used. Both male and female mice (1:1) were used to determine the HopA-mediated neuroprotective effects and were randomly divided into four groups including normal (wild-type C57BL/6 (B6) mice), AD (transgenic APP/PS1 mice), B6 mice treated with HopA (4 mg kg−1 day−1, peritoneal injection, 30 days), and APP/PS1 mice treated with HopA (4 mg kg−1 day−1, peritoneal injection, 30 days) groups. Investigators were blind to mouse genotypes and treatments until completion of the whole study.
Primary cortical neuron culture and treatment
Primary cortical neuron culture was carried out as described (Xu et al., 2006). Briefly, neurons were isolated from each embryonic cortex (E15–E17 wild-type B6 mouse) and maintained in Neurobasal media supplemented with B27 (Invitrogen, Carlsbad, CA, USA) and 25 nm glutamine at 37 °C in a humidified 5% CO2 incubator. Over 95% purity of neurons was confirmed. The cells at day 8 were treated with 2 μm Aβ1–42 and HopA at different concentrations (0.1, 0.5, 1, 2, and 4 μm) for 24 h or with vehicle (control).
The reaction mixture contained 50 μL HopA (10–60 μm) and 50 μL Aβ1–42 (0.1 mm, diluted in PBS) in a final volume of 100 μL. The control group contained 50 μL HopA (10–60 μm) and 50 μL PBS. The mixture was incubated at 37 °C for 1 h. The free HopA was separated by centrifugation for 15 min at 16 000 g and detected by HPLC. Similar experiment was repeated by adding Aβ42–1 without activity to exclude HopA that was reduced rather than binding with Aβ1–42.
Transmission electron microscopy analysis
Transmission electron microscopy analysis was used to determine whether HopA could inhibit Aβ1–42 formation and disaggregate preformed fibrils as described previously (Yang et al., 2005). More information can be found in Methods in Data S1.
Congo red staining
Brain sections (20 μm) were incubated in 0.5% Congo red (Sigma-Aldrich, St. Louis, MO, USA) in an alkaline alcoholic saturated sodium chloride for 30 min at room temperature and then incubated in a freshly prepared alkaline alcoholic saturated sodium chloride (2.5 mm NaOH in 80% alcohol) for 20 s at room temperature. And the sections were stained with hematoxylin for 2 min and dehydrated through 95% alcohol and 100% alcohol, cleared in xylene and cover slipped with Permount (Fisher Scientific, Lafayette, CO, USA). The total area of Congo red staining in the cortex and hippocampus was measured using NIH imagej software (http://rsbweb.nih.gov/ij/) and expressed as% of total area for each slice.
Brain section (20 μm) incubated with anti-Aβ1–42 monoclonal antibody (1:100; Abcam, Cambridge, MA, USA) overnight and then avidin–biotin–HRP complex was applied for 1 h at room temperature. Samples were stained with DAB (Vector Laboratories, Orton Southgate, Peterborough, UK), then washed and mounted. Total area of Aβ1–42 staining in the cortex was measured using NIH imagej software (http://rsbweb.nih.gov/ij/) and expressed as% of total area for each slice.
Primary cortical neurons were washed with PBS and fixed with 4% paraformaldehyde for 15 min. The cells were incubated with anti-Aβ1–42 monoclonal antibody (1:100; Abcam) and anti-ABAD polyclonal antibody (1:200; Abcam) in blocking buffer overnight at 4 °C and then incubated with the secondary goat anti-mouse and goat anti-rabbit antibody (1:200; Invitrogen) for 45 min. Fluorescent images were taken using a fluorescent microscope and analyzed with the slide book software (Intelligent Imaging Innovations, Washington St Denver, CO, USA).
Simulation of HopA–Aβ1–42 interactions
To investigate the interaction between HopA and Aβ1–42, the MD simulations was used in the FF03 force field within amber9 program (Hongjiankexin Communication, Haidian District, Beijing, China). More information can be found in Methods in Data S1.
Primary cortical neurons or mouse cortex were lysed and harvested. The lysate was pre-incubated with protein G–Sepharose beads (Sigma-Aldrich) and then centrifuged. Anti-Aβ antibody and protein G–Sepharose beads were incubated with the supernatant and then centrifuged, washed and boiled for analysis of proteins by Western blot. More information can be found in Methods in Data S1.
Analysis of the mitochondrial function
Mitochondrial membrane potential was measured using JC-1 (Sigma-Aldrich). The activities of mitochondrial complex IV and ABAD were determined by the commercial kit (Genmed Scientifics Inc, Pudong New District, Shanghai, China). The activity of caspase-3 was measured by the colorimetric assay kits (Keygen BioTech, Nanjing, China) according to the manufacturer's instruction. More information can be found in Methods in Data S1.
Oxidative stress measurement
Intracellular ROS and 4-HNE, 3-NT and 8-OHdG in hippocampus and cortex were measured using kits according to the manufacturer's instruction (Genmed Scientifics Inc). More information can be found in Methods in Data S1.
Electrophysiological studies were performed as described previously (Chen et al., 2010). The detail information can be found in Methods in Data S1.
Morris water maze test
The Morris water maze test was performed as described previously (Vorhees & Williams, 2006). The detail information can be found in Methods in Data S1.
Data were presented as mean ± standard error of the mean (SEM). All statistical calculations were performed by spss 17.0 (SPSS Inc., Chicago, IL, USA). Group differences in the escape latency, searching distance and swimming speed during the MWM test were analyzed using two-way analysis of variance (anova) with repeated measures followed by Bonferroni multiple comparison test with day and treatment as the sources of variation. The data in 55–60 min post-HFS were analyzed using two-way anova followed by Bonferroni's post hoc. All other data were analyzed with a one-way anova followed by Bonferroni's post hoc. A statistically significant difference was set at P <0.05.
This work was supported by the National Nature Science Foundation of China (81200839, 30971010, 30670739, 81171085 & 30821006), the Natural Science Foundation of Jiangsu Province of China (BL2012013), the State Key Laboratory of Pharmaceutical Biotechnology of Nanjing University (KF-GN-200901), Key Project of Nanjing Municipal Bureau of Health (ZKX08030), the Medical Leading Talent and Innovation Team Project of Jiangsu Province (LJ201101), and Science Development Foundation of Nanjing Medical Foundation (2011NJMU272). We thank Brad Peterson and Kunlin Jin for writing modification.
Conflict of interest
The authors declare no conflict of interest.
Y.X. designed the experiments, wrote and edited the manuscript. R.X.T. provided the HopA sample and edited the manuscript. X.L.Z, L.Y., R.L., L.L.L., L.Q. C.C., and D.N.G. performed experiments. L.C. performed electrophysiology. H.M.G. reisolated the sample. S.J and S.Z. performed LC-MS measurement. N.J. provided the computational assessment. All authors contributed to the data analysis and the manuscript preparation.