Disrupting the Interaction of nNOS with CAPON Prevents the Reinstatement of Morphine Conditioned Place Preference

Abstract

Drug abuse is a dramatic challenge for the whole society because of high relapse rate. Environmental cues are crucial for the preference memory of drug abuse. Extinction therapy has been developed to inhibit the motivational effect of drug cues to prevent the reinstatement of morphine abuse. However, extinction therapy alone only forms a new kind of unstable inhibitory memory. We found that morphine conditioned place preference (CPP) extinction training increased the association of nitric oxide synthase (nNOS) with its carboxy-terminal PDZ ligand (CAPON) in the dorsal hippocampus (dHPC) significantly and blocking the morphine-induced nNOS–CAPON association using Tat-CAPON-12C during and after extinction training reversed morphine-induced hippocampal neuroplasticity defect and prevented the reinstatement and spontaneous recovery of morphine CPP. Moreover, in the hippocampal selective ERK2 knock-out or nNOS knockout mice, the effect of Tat-CAPON-12C on the reinstatement of morphine CPP and hippocampal neuroplasticity disappeared, suggesting ERK2 is necessary for the effects of Tat-CAPON-12C. Together, our findings suggest that nNOS–CAPON interaction in the dHPC may affect the consolidation of morphine CPP extinction and dissociating nNOS–CAPON prevents the reinstatement and spontaneous recovery of morphine CPP, possibly through ERK2-mediated neuroplasticity and extinction memory consolidation, offering a new target to prevent the reinstatement of drug abuse.

ERKs, hippocampus, morphine CPP, nNOS–CAPON, synaptic plasticity

Introduction

Opioids are widely used in clinic as an analgesic with strong effect, but their application is restricted by addiction. Drug abuse has become a major social issue worldwide, as the number of drug abusers has increased steadily in the last decade (S.A.M.H.S.A. 2018). Individuals with opioid use disorders often reinstate preference memory into drug-seeking behavior after recalling memories linked to the drug use experience, as drug-related environmental cues form an associative memory (Thompson and Ostlund 1965). Extinction therapy, a means of suppressing the expression of original drug-associated memories, has been developed to inhibit the motivational effect of drug cues to prevent reinstatement of drug preference memory, in which addicts are repeatedly exposed to drug-associated cues without given drugs of abuse (Xue et al. 2014). However, extinction does not erase the already formed preference memory, but forms a new kind of unstable inhibitory memory (Bouton 1993). Improving extinction efficacy has been used as a strategy to treat substance use disorders and suppress the reinstatement of drug memory.

When morphine is given to mice in the drug-related environment, mice can form context-associated conditioned place preference (CPP) (Kim et al. 1996; Solecki et al. 2009). Long-term potentiation (LTP) is impaired and expression of the GluN1 and GluN2B N-methyl-D-aspartic acid receptor (NMDAR) subunits are increased at the postsynaptic density (PSD) in the hippocampus after the extinction of morphine CPP (Portugal et al. 2014), implicating alterations in synaptic plasticity and glutamatergic transmission in the reinstatement of morphine CPP. Extracellular signal-regulated kinase (ERK) is a downstream component of the NMDAR signal transduction pathway (Karpova et al. 2013). Although ERK has stage and brain specificity in the process of regulating preference behavior, extinction memory depends on ERK signaling in the hippocampus (Selcher et al. 2001; Haghparast et al. 2014; Pahlevani et al. 2014). NMDAR activation also induces association of neuronal nitric oxide synthase (nNOS) with its carboxy-terminal PDZ ligand (CAPON), an event causing ERK dysfunction (Zhu et al. 2014). Disrupting nNOS–CAPON interaction inhibits the NMDAR-mediated excitotoxicity, promotes the synaptic plasticity and facilitates fear extinction consolidation by increasing the phosphorylation of ERK (Zhu et al. 2014; Ni et al. 2019; Qin et al. 2020). We thus wonder whether nNOS–CAPON interaction regulates the morphine preference memory. Here, we show that chronic treatment with morphine up-regulates nNOS–CAPON association and causes ERK dysfunction in the hippocampus. Dissociating nNOS–CAPON during and after extinction training prevents the reinstatement and spontaneous recovery of morphine preference memory through up-regulating ERK2 phosphorylation and promoting structural and functional neuroplasticity in the hippocampus, suggesting a target to prevent the preference memory of opioid use disorders to reinstate into drug-seeking behavior.

Materials and Methods

Animals

Young adult (6–8 week-old) C57BL/6 mice were purchased from Mode Animal Research Center of Nanjing University. Young adult (6–8 week-old) male homozygous nNOS-deficient mice (B6; 129S4-Nos1^tm1Plh/J, knockout (KO), 002633) and their wild-type (WT) controls of similar genetic background (B6129SF2/J, 101045); ERK2^flox/flox mice (B6; 129-Mapk1^tm1Gela/J, 019112), their WT controls of similar genetic background (C57BL/6 J, 000664); homozygous nNOS-Cre mice (B6; 129-Nos1^tm1(cre)Mgmj/J) and homozygous Rosa-EYFP mice (B6; Cg-Gt(ROSA)26Sor^{tm3(CAG-EYFP)Hze}/J) were purchased from the Jackson Laboratory and maintained at Model Animal Research Center of Nanjing University at a controlled temperature (20 ± 2 °C) on a 12 h light/dark cycle with water and food ad libitum. All animal experiments were conducted in accordance with the Institutional Animal Care and Use Committee of Nanjing Medical University.

Drugs and Recombinant Viruses

Morphine hydrochloride from Northeast Pharm was dissolved in saline and used by subcutaneous injection (s.c.) at dose of 5–20 mg/kg body weight. Tat-CAPON-12C and Tat-CAPON-12C/A22D were prepared in our laboratory (Zhu et al. 2014) and were microinjected into the hippocampus at the concentration of 50 nM (Zhu et al. 2014). The dose for intraperitoneal injection (i.p.) of Tat-CAPON-12C was 3 mmol/kg body weight (Zhang et al. 2018). Dil, a lipophilic membrane stain, was microinjected into that hippocampus (C1036, Beyotime) at the concentration of 5 μM. Recombinant viruses AAV-EF1α-tTA-WPRE-pA; AAV-TRE3g-CRE-2α-EYFP-WPRE-pA and its control AAV-EF1α-tTA-WPRE-pA; AAV-TRE3g-EYFP-WPRE-pA were purchased from Obio Technology (Shanghai, China). Viral titers were ~3 × 10¹² particles per ml for all viruses.

Cannulae Implantation

Mice were anesthetized with 4% isoflurane and placed on a stereotaxic apparatus. During the surgery of craniotomy, the mice were maintained with 1–2% isoflurane. We implanted 1 millimeter stainless-steel guide cannulae (RWD Life Science) above the hippocampus (AP –2.0 mm, ML ±1.3 mm, DV –1.6 mm). The cannulae were secured and fixed to skull using dental cement. The mice recovered at least 1 week before behavior experiments.

Conditioned Place Preference

Training and testing of morphine conditioning CPP were performed in a 3-chamber apparatus as previous study (Portugal et al. 2014). The time mice spent and the total distance mice walked in the chambers were recorded by a video camera and analyzed using the Topscan software (Clever Sys, Inc.). On the pretest day, the mice were allowed to prefer for three chambers during 15 min, and preferred chamber and non-preferred chamber of each mouse were recorded. In the formation stage, the mice were randomly divided into morphine and saline group and then received 4 consecutive days conditioning sessions with saline and daily escalating dose of morphine (5, 10, 15, 20 mg/kg) treatment, respectively. For morphine group, mice were injected with morphine and placed in their nonpreferred chamber for 1 h in the morning, and 5 h later, mice were injected with saline and placed in their preference chamber for 1 h in the afternoon. For saline group, mice were injected with saline only and placed in their non-preferred chamber for 1 h at am and in their preference chamber for 1 h at pm, with a 5 h interval. Extinction of morphine CPP was scheduled 24 h after formation testing. Mice in all groups received 7 consecutive days extinction training with saline treatment (s.c.). For morphine group, the mice received saline injections and were placed in morphine-associated chamber for 1 h at am, and 5 h later, the mice received saline injections and were placed in the control chamber for 1 h at pm. For saline group, mice were placed as that in the formation stage of morphine CPP. We treated mice at 5 min after daily training with i.p. or 1 h after daily training with intra-dHPC microinjection of Tat-CAPON-12C or its control (Tat-CAPON-12C/A22D). For ERK2^flox/flox mice, we treated them with i.p. Tat-CAPON-12C for 4 consecutive days after the retrieval of extinction. In the reinstatement test, mice were placed in CPP apparatus with a priming dose of morphine (5 mg/kg, s.c.), and we recorded the time the mice spent in each chamber and the total distance the mice walked in 15 min. After each CPP test and training, the compartments were cleaned with 75% alcohol to eliminate the residual smell of mice. Preference scores were calculated by the time mice spent in the morphine-paired chamber during formation, extinction or reinstatement test minus the time mice spent in the pretest.

Western Blot

One hour after 7 days of extinction training, hippocampus tissue was extracted and homogenized in the lysis buffer (50 mM Tris, 150 mM NaCl, 1 mM EDTA, 0.1% SDS, 1% NP-40, 0.5% Sodium deoxycholate, and protease inhibitor). Samples for western blot were treated as previously described (Zhou et al. 2010). Primary antibodies used were: rabbit anti-nNOS (Thermo Fisher Scientific, 1:1000), rabbit anti-CAPON (Abcam, 1:2000), rabbit anti-ERK (Cell Signaling Technology, 1:3000), rabbit anti-p-ERK (Cell Signaling Technology, 1:1000), and mouse anti-GAPDH (KangChen Biotech, 1:8000). Secondary antibody was HRP-linked anti-rabbit/mouse IgG antibody (MULTI SCIENCE, 1:8000). The signal of protein was scanned by the western lightning gel imaging system (BIO RAD) and the intensity of the blots was quantified using the Image J software.

Coimmunoprecipitation

One hour after 7 days of extinction training, hippocampus tissue was extracted and homogenized in the lysis buffer (50 mM Tris, 150 mM NaCl, 1 mM EDTA, 0.1% SDS, 1% NP-40, 0.5% Sodium deoxycholate, and protease inhibitor). Samples for Coimmunoprecipitation were treated as previously described (Zhou et al. 2010). The protein samples were mixed with the primary antibody of rabbit anti-nNOS and incubated at 4 °C for 8 h. Protein G-magnetic beads (Millipore) were pretreated with supernatant and PBS at 4 °C, and then the pretreated magnetic beads and the target mixed sample were added for incubation at 4 °C overnight. Proteins were analyzed by western blot. Primary antibodies used were: rabbit anti-nNOS, rabbit anti-CAPON. Secondary antibody was HRP-linked anti-rabbit IgG antibody.

Immunofluorescence

Mice were anesthetized with isoflurane and perfused with PBS containing 4% paraformaldehyde. The brain tissues were fixed in 4% paraformaldehyde overnight and slices were prepared at 40 μm thickness in PBS. Slices were rinsed 3 times in PBS containing 0.1% Triton X-100 (PBST) for 10 min and blocked in PBST with 10% fetal bovine serum at room temperature for 2 h. The sections were incubated with rabbit anti-c-Fos primary antibody (1:500; Synaptic Systems) or anti-ERK2 primary antibody (1:400; abcam) in 0.1% PBST at 4 °C overnight. Next day, sections were rinsed 3 times in PBST and incubated with goat anti-rabbit Cy3 (1:200; Jackson ImmunoResearch Labs) at room temperature for 2 h. After 3 times of PBST rinses, sections were incubated with Hoechst 33258 at room temperature for 10 min to label the nuclei. Images were scanned using a confocal microscope (LSM 510, Zeiss).

Golgi-Cox Staining

Golgi staining was performed according to the instructions of the FD Golgi stain™ kit (FD NeuroTechnologies). Mouse brain was removed at 1 h after 7 days of extinction training. According to the user manual, the brains were first placed in mixed solutions for 2 weeks. Then the brains were transferred to solution C and left in the dark for 1 week at room temperature. The sections were cut in PBS with a thickness of 200 μm using a vibratome (Leica) and stained. To calculate spine density of Golgi-stained neurons in the hippocampal CA1, the exact length of the dendritic segment was calculated and the number of spines along that length was counted using Image J software. The density of dendritic spines is calculated by dividing the number of spines by the length of dendritic spines. Six neurons randomly for each sample were measured, and the average was regarded as the final value of one sample.

Electrophysiology

One hour after 7 days of extinction training, mouse brain was removed. Field electrophysiology LTP, pair-pules ratio (PPR) and input/output (I/O) curve in the schaffer to CA1 circuit from acute 350-mm transverse hippocampal slices were recorded as previously described (Gao et al. 2010). The hippocampal slices were cut using a vibratome at 4 °C in oxygenated (95% O₂/5% CO₂) artificial cerebrospinal fluid (ACSF) that contained 75 mM sucrose, 87 mM NaCl, 2.5 mM KCl, 1.25 mM NaH₂PO₄, 21.4 mM NaHCO₃, 0.5 mM CaCl2, 7 mM MgCl₂, 20 mM glucose and were maintained at 32 °C for 1 h in a holding chamber with oxygenated ACSF that contained 119 mM NaCl, 2.5 mM KCl, 1 mM NaH₂PO₄, 26.2 mM NaHCO₃, 2.5 mM CaCl2, 1.3 mM MgSO₄, 11 mM glucose. Next, the slices were incubated at room temperature for 1 h before recording. Data acquisition and analysis were performed using Clampex and Clampfit 8.2 (Molecular Devices).

Statistics Analysis

Data were analyzed using SPSS software, and illustrations were created using GraphPad Prism software. Comparisons between two groups were made by a two-tailed Student’s t-test. The difference among multiple groups were compared with one-way ANOVA, two-way ANOVA and two-way repeated measures ANOVA followed by Bonferroni post hoc test. All data were expressed as the mean ± SEM. The significance level for all tests was set at P < 0.05.

Results

Association of nNOS with CAPON in the Dorsal Hippocampus Contributes to the Reinstatement of Morphine CPP

To test whether nNOS–CAPON association is implicated in the reinstatement of morphine CPP, we used a 3-chamber apparatus (Portugal et al. 2014) to measure the preference scores in formation, extinction and priming tests of mice CPP model (Fig. 1A). In the formation test, morphine-treated mice in the CPP procedure displayed a significantly increased time spent in the morphine conditioning chamber, compared with vehicle-treated mice, suggesting a morphine preference memory (Fig. 1B). After extinction training, preference scores were very low in both morphine- and vehicle-treated mice, suggesting a successful extinction (Fig. 1B). In priming test, the mice with chronic morphine exposure displayed significantly increased preference score and locomotor activity, compared with vehicle, suggesting a reinstatement of morphine CPP and locomotor sensitization (Fig. 1B and C).

Figure 1

Establishment of morphine CPP model and the activation of dCA1 in the extinction phase of morphine CPP. (A) Design of the experiments for (B) and (C). (B), Reinstatement of morphine CPP by a priming dose of morphine (5 mg/kg, s.c.) at 24 h after extinction training (two-way repeated measures ANOVA, n = 11, time: F_2,22 = 13.938, P = 0.000, group: F_(1,11) = 34.583, P = 0.000, interaction: F_(2,22) = 13.649, P = 0.000, ^***P < 0.001). (C) Morphine exposure increases the locomotor activity at 24 h after extinction training (two-way repeated measures ANOVA, n = 11, time: F_(2,22) = 28.239, P = 0.000, group: F_(1,11) = 27.612, P = 0.000, interaction: F_(2,22) = 25.537, P = 0.000,^***P < 0.001). (D) Design of the experiments for (E) and (F). (E) Images showing expression of c-Fos in the dCA1 of mice with (right) or without extinction training (left). Scale bar, 100 μm. (F) Extinction training of morphine CPP increases the c-Fos expression in dCA1 (two-tailed t test, n = 5, t₍₈₎ = 10.977, P = 0.000, ^***P < 0.001). (G) The time course of morphine preference score during extinction training (two-way repeated measures ANOVA, n = 12, time: F_(7,77) = 4.832, P = 0.000, group: F_(1,11) = 11.736, P = 0.006, interaction: F_(7,77) = 4.635, P = 0.000, ^***P < 0.001, ^**P < 0.01). (H) The time course of locomotor activity during extinction training (two-way repeated measures ANOVA, n = 12, time: F_(7,77) = 0.994, P = 0.442, group: F_(1,11) = 1.312, P = 0.276, interaction: F_(7,77) = 0.390, P = 0.905). NS: saline, Mor: morphine, For: formation, N.S.: no statistics significance.

Synaptic plasticity of the hippocampus plays a critical role in the extinction of morphine CPP (Nicoll and Malenka 1991; Portugal et al. 2014; Fakira et al. 2016). We thus test whether the dorsal hippocampus (dHPC) is activated by extinction learning in our morphine CPP model. We measured c-Fos expression in the dHPC 1 h after the first day of extinction training (Fig. 1D). As expected, extinction training increased the number of c-Fos⁺ cells in the dCA1 of mice significantly, compared with that of mice without extinction training (Fig. 1E and F), suggesting an activation of dHPC. Moreover, to test whether the extinction training procedure affects the rate of extinction learning, we detected preference score and locomotor activity after each extinction session. As shown in Figure 1G, the mice with morphine exposure acquired a significant drug-related preference memory, and the morphine CPP gradually decreased over time and approached baseline level after 7 days of extinction training, suggesting a successful extinction of morphine CPP, while control mice showed no significant preference during extinction learning, indicating that the manipulation of extinction procedure itself does not affect the temporal dynamics of extinction learning. Moreover, no significantly different locomotor activity was observed between groups during extinction learning (Fig. 1H).

Our previous studies showed that nNOS–CAPON association affects synaptic plasticity in the hippocampus and cortex (Zhu et al. 2014; Ni et al. 2019). We thus test whether extinction training in morphine CPP model affects nNOS–CAPON association in the dHPC. We detected nNOS–CAPON complex in the dHPC by coimmunoprecipitation. As shown in Figure 2A and B, extinction training significantly increased nNOS–CAPON association in the dHPC in morphine-treated mice, compared with that in vehicle-treated mice, suggesting an extinction-induced nNOS–CAPON association in the dHPC in morphine CPP model. Next, we examined whether the extinction-induced nNOS–CAPON association in the dHPC affects the retention of preference extinction memory. We treated mice at 5 min after daily extinction training with i.p. of Tat-CAPON-12C, a peptide selectively dissociating CAPON from nNOS (Zhu et al. 2014), or its control Tat-CAPON-12C/A22D, and detected nNOS and CAPON protein levels and nNOS–CAPON complex in the dHPC at 1 h after the last drug injection and measured the preference scores of mice in formation, extinction and priming sessions (Fig. 2C). Tat-CAPON-12C reversed extinction-induced increase in nNOS–CAPON complex (Fig. 2D), but did not affect nNOS and CAPON levels (Fig. 2E) in the dHPC in morphine CPP model, compared with Tat-control (Tat-CAPON-12C/A22D). Although Tat-CAPON-12C did not affect morphine preference in formation and extinction tests, it significantly reduced preference score in priming test, compared with morphine- or morphine + Tat-control group (Fig. 2F), suggesting that dissociating nNOS–CAPON in the dHPC inhibits the reinstatement of morphine CPP. Moreover, Tat-CAPON-12C did not affect morphine-induced increase in locomotor activity in priming test (Fig. 2G).

Figure 2

Association of nNOS–CAPON in the dHPC facilitates the reinstatement of morphine CPP. (A) Design of the experiments for (B). (B) Immunoblots (left) and bar graph (right) showing nNOS–CAPON complex levels in the dHPC at 1 h after extinction training (one-way ANOVA, n = 5, F_(2,12) = 8.356, P = 0.007, ^**P < 0.01). (C) Design of the experiments for (D–G). (D) Immunoblots (left) and bar graph (right) showing nNOS–CAPON complex levels in the dHPC at 1 h after extinction training (one-way ANOVA, n = 9, F_(3,32) = 6.541, P = 0.001, ^**P < 0.01, ^##P < 0.01). (E) Immunoblots showing expression of nNOS and CAPON in the dHPC at 1 h after extinction training. (F) Tat-CAPON-12C prevents the reinstatement of morphine CPP but has no effect on formation of morphine CPP (two-way repeated measures ANOVA, n = 12, time: F_(2,22) = 19.449, P = 0.000, group: F_(3,33) = 8.224, P = 0.000, interaction: F_(6,66) = 4.132, P = 0.001, ^***P < 0.001, ^##P < 0.01). (G) Tat-CAPON-12C does not affect morphine-induced increase in locomotor activity (two-way repeated measures ANOVA, n = 12, time: F_(2,22) = 107.285, P = 0.000, group: F_(3,33) = 11.216, P = 0.000, interaction: F_(6,66) = 10.363, P = 0.000, ^***P < 0.001). (H) Design of the experiments for (I–K). (I) Image showing dil dye in the dCA1 to confirm microinjection site of drugs. (J) Intra-dCA1 microinjection of Tat-CAPON-12C prevents the reinstatement of morphine CPP but has no effect on formation of morphine CPP (two-way repeated measures ANOVA, n = 10, time: F_(2,18) = 3.636, P = 0.047, group: F_(2,18) = 4.747, P = 0.022, interaction: F_(4,36) = 2.706, P = 0.045, ^*P < 0.05, ^#P < 0.05). (K) Intra-dCA1 microinjection of Tat-CAPON-12C does not affect morphine-induced increase in locomotor activity (two-way repeated measures ANOVA, n = 10, time: F_(2,18) = 27.677, P = 0.000, group: F_(2,18) = 9.022, P = 0.002, interaction: F_(4,36) = 13.681, P = 0.000, ^***P < 0.001). NS: saline, Mor: morphine, WB: western blot, Co-IP: coimmunoprecipitation.

To further confirm this finding, we performed the morphine CPP paradigm as above and microinjected Tat-CAPON-12C or its control though implanted cannulae into dCA1 at 1 h after daily extinction training to dissociate local nNOS–CAPON (Fig. 2H). Before the microinjection of drugs, we infused a dil dye into the dHPC to confirm the microinjection site of drugs (Fig. 2I). Similar with systemic administration of drugs, intra-dHPC microinjection of Tat-CAPON-12C significantly decreased preference score in priming test, compared with Tat-CAPON-12C/A22D, although two groups had similar preference scores in the formation and extinction tests (Fig. 2J), and Tat-CAPON-12C did not affect morphine-induced increase in locomotor activity (Fig. 2K).

Many researchers report that nNOS is primarily expressed in inhibitory interneuron (Itzhak and Ali 2006; Christenson Wick et al. 2019). It has been known that hippocampal CA1 pyramidal neurons are crucial for spatial and contextual memory (Bloss et al. 2018). To know why the nNOS–CAPON association in interneurons produces so strong effect on the reinstatement of morphine CPP, we generated nNOS-Cre:ROSA-EYFP mice by crossing nNOS-Cre transgenic mice to ROSA26-flox-stop-flox-EYFP reporter (ROSA-EYFP) mice. In ROSA-EYFP mice, YFP expression is blocked by a loxP flanked STOP cassette. In nNOS-Cre:ROSA-EYFP mice, Cre-mediated excision of the STOP cassette limits YF expression only to nNOS-expressing cells. Surprisingly, almost all dCA1 neurons are nNOS-expressing neurons (Fig. S1A–C).

To test whether the inhibitory effect of Tat-CAPON-12C on the reinstatement of morphine CPP depends on dHPC nNOS, nNOS KO and WT mice (Fig. S1D) were conditioned in a morphine CPP paradigm (Fig. S1E). We microinjected Tat-CAPON-12C or its control through implanted cannulae into dCA1 at 1 h after daily extinction training (Fig. S1E). After morphine conditioning, both nNOS KO and WT mice showed similar preference scores in formation and extinction tests, and Tat-CAPON-12C did not affect the preference scores, compared with control (Fig. S1F), suggesting that deletion of nNOS and dissociating nNOS–CAPON do not affect the formation and extinction learning of morphine CPP. In priming test, however, Tat-CAPON-12C significantly decreased the time spent in the morphine-paired chamber in WT mice but not in nNOS KO mice, compared with its control Tat-CAPON-12C/A22D (Fig. S1F), suggesting that the drug inhibits the reinstatement of morphine CPP through targeting nNOS. Moreover, nNOS deletion or treatment with Tat-CAPON-12C did not affect morphine-induced increase in locomotor activity (Fig. S1G). Thus, nNOS–CAPON association is necessary for the effect of Tat-CAPON-12C on the retention of morphine extinction memory. Collectively, morphine-induced nNOS–CAPON association in the dHPC during extinction learning facilitates the reinstatement of morphine CPP and dissociating nNOS–CAPON prevents the reinstatement of morphine CPP.

Association of CAPON with nNOS in the dHPC Negatively Regulates Neuroplasticity

Synaptic plasticity, including structural and functional plasticity, is generally accepted as the principal implementation of information storage in neural systems (Kandel et al. 2014), especially LTP-mediated synaptic strength in the hippocampus being the basis of learning and memory (Herron et al. 1986; Lisman et al. 2018). To investigate the effect of morphine and nNOS–CAPON blocker on the structural plasticity of hippocampus in morphine CPP model during extinction training, we performed the morphine CPP paradigm and treated mice with Tat-CAPON-12C as Figure 2C and measured dendritic spine density of dCA1 pyramidal neurons by Golgi-Cox staining at 1 h after the last Tat-CAPON-12C injection (Fig. 3A). We found that chronic morphine exposure caused significant decrease in the spine density of dCA1 neurons, compared with its vehicle saline, and treatment with Tat-CAPON-12C reversed the morphine-induced decrease in the spine density of dCA1 neurons (Fig. 3B and C). Thus, chronic morphine exposure reduces synaptic structural plasticity of dHPC during morphine CPP extinction and this pathological variation is reversed by dissociating CAPON from nNOS.

Figure 3

Tat-CAPON-12C facilitates hippocampal neuroplasticity of mice with extinction training after chronic morphine abuse. (A) Design of the experiments for (B–G). (B) Representative images showing the dendritic spine in CA1 of each group mice (top) and high-magnification images (middle and bottom) from selected areas in the top and middle images. Scale bar, 50 μm. (C) Bar graph showing the density of dendritic spines at 1 h after extinction training (one-way ANOVA, n = 5, F_(3.16) = 38.078, P = 0.000, ^***P < 0.001, ^###P < 0.001). (D–G) Tat-CAPON-12C increases the hippocampal synaptic functional plasticity. Morphine CPP extinction treated with or without Tat-CAPON-12C has no effect on basal synaptic transmission (D) and presynaptic transmitter release E in hippocampus. (F) Left, electrophysiological response obtained from glass electrode placed in the hippocampus in vitro. Right, plot of baseline normalized fEPSP and in vitro electrically evoked responses following 2 electrical theta burst stimulation (TBS). (G) Bar graph showing the change in last 15 min from baseline of hippocampal LTP (one-way ANOVA, n = 8–9, F_(3,30) = 64.945, P = 0.000, ^***P < 0.001, ^###P < 0.001). (H) Bar graph showing percentage decline in the last 15 min from the initial hippocampal LTP (one-way ANOVA, n = 8–9, F_(3,30) = 9.140, P = 0.000, ^**P < 0.01, ^###P < 0.001). NS: saline, Mor: morphine, N.S.: no statistics significance.

Morphology and density of spine reflect the maturation and reorganization of excitatory circuits (Moyer and Zuo 2018). Accordingly, we tested whether nNOS–CAPON association affects the dHPC functional plasticity in morphine CPP model. We recorded the electrophysiological changes in presynaptic transmitter release, basal synaptic transmission and LTP in the dHPC at 1 h after the last Tat-CAPON-12C injection (Fig. 3A), and found that all groups had similar PPF ratios and I/O curves, indicating that neither morphine exposure nor Tat-CAPON-12C treatment changes presynaptic transmitter release and basal synaptic transmission (Fig. 3D and E). However, the mice with chronic morphine exposure had significantly impaired hippocampal LTP, compared with saline-treated mice, and treatment with Tat-CAPON-12C reversed the effect of morphine on hippocampal LTP (Fig. 3F and G). Moreover, morphine exposure caused an increase in % of LTP decay and Tat-CAPON-12C reversed the morphine-induced LTP decay (Fig. 3H). Thus, dissociating nNOS form CAPON improves the morphine-induced impairment of hippocampal neuroplasticity during extinction.

Association of nNOS with CAPON Negatively Regulates Morphine Extinction Memory and Neuroplasticity through Inhibiting ERK2

NMDAR is an important regulator of synaptic plasticity (Kukushkin and Carew 2017). ERK is a downstream component of the NMDAR signal transduction pathway (Karpova et al. 2013) and the phosphorylation of ERK2 is involved in the regulation of LTP and dendritic spine plasticity (Futter et al. 2005; Bosch et al. 2014), contributing to the preservation of morphine extinction memory. However, NMDARs activation also induces nNOS–CAPON association, an event causing ERK dysfunction (Zhu et al. 2014) To test whether the regulation of morphine extinction memory by nNOS–CAPON depends on ERK2 function in the dHPC, we measured the phosphorylation of ERKs at 1 h after 7 days of extinction training (Fig. 4A). We found that phosphorylated ERK (pERK) in the dHPC of mice subjected to chronic morphine exposure was significantly downregulated after morphine CPP extinction training, compared with saline-treated mice (Fig. 4B–D). Moreover, the morphine exposure caused a decrease in pERK2 but not pERK1 in the dHPC (Fig. S2). Dissociating CAPON from nNOS with Tat-CAPON-12C after extinction training reversed morphine-induced ERK dysfunction in the dHPC, compared with its control, and neither morphine nor Tat-CAPON-12C affected ERK expression in the dHPC (Fig. 4B–D).

Figure 4

Uncoupling nNOS–CAPON rescues the decrease of ERK phosphorylation induced by chronic treatment of morphine and conditional deletion of ERK2 in the dCA1. (A) Design of the experiments for (B–D). (B) Representative immunoblots and (C, D) bar graph showing ERK (one-way ANOVA, n = 9, F_(3,32) = 0.072, P = 0.975, P > 0.05 between groups) and p-ERK levels (one-way ANOVA, n = 9, F_(3,32) = 6.457, P = 0.006, ^**P < 0.01, ^##P < 0.01) in the dHPC at 1 h after CPP extinction session. (E) Design of the experiments for (F) and (G). (F) Top, representative immunoblots showing ERK2 knockdown in the dHPC of ERK2^flox/flox mice. Bottom, bar graph showing ERK2 expression in the dHPC of ERK2^flox/flox mice (two-tailed t test, n = 6, t₍₁₀₎ = 7.658, P = 0.000, ^***P < 0.001). (G) Images showing expressions of ERK2 and EYFP in the dCA1 of ERK2^flox/flox mouse infected by AAV-Cre or control AAV (left) and high-magnification images (right) from selected areas in the left images. NS: saline, Mor: morphine, Dox: doxycycline.

Next, we used a cocktail of tTA and TRE-CRE viruses to conditionally delete ERK2 in the dCA1. A mixed solution of an AAV expressing tetracycline (Tet) transcriptional activator (AAV-EF1α-tTA-WPRE-pA) and an AAV encoding Tet-responsive element and Cre recombinase (AAV-TRE3g-Cre-2α-EYFP-WPRE-pA) was infused into the dCA1 of ERK2^flox/flox mice 28 d before morphine CPP paradigm (Fig. 4E). Doxycycline diet (On) causes tTA to fall off TRE, thereby the expression of downstream Cre recombinase is switched off. When doxycycline is removed from diet (Off), tTA and TRE bind to initiate expression of Cre recombinase. To examine whether the cocktail viruses conditionally delete ERK2 in the dCA1, we performed western blot analysis and immunofluorescence staining at 24 h after doxycycline off. The time of doxycycline off we chose is based on that the half-life of doxycycline in the brain of C57Bl/6 mice is 3.9 h (Lucchetti et al. 2019), and usually, ~ 99% of drug is eliminated from body after 6 half-lives (23.4 h). In the hippocampal selective ERK2 knock-out mice (AAV-tTA:AAV-TRE3g-Cre), the level of ERK2 in the dorsal hippocampus was significantly reduced (Fig. 4F). Immunofluorescence staining showed that ERK2 was almost completely deleted in the infected neurons in the dCA1 of ERK2^flox/flox mice injected by the cocktail viruses (AAV-tTA:AAV-TRE3g-Cre), while ERK2 was normally expressed in the infected neurons in the dCA1 of ERK2^flox/flox mice receiving the control cocktail viruses (AAV-tTA:AAV-TRE3g) (Fig. 4G).

In the experiments above, Tat-CAPON-12C was given shortly after daily extinction training, its inhibitory effect on the restatement of morphine CPP can be interpreted as enhanced extinction consolidation/retention or extinction learning. Because our previous study showed that ERK2 does not affect within-session fear extinction learning (Qin et al. 2020), Tat-CAPON-12C may inhibit the restatement of morphine CPP through enhancing extinction consolidation and retention. To test this notion, we detected extinction memory at 1 h after the last day extinction training, and 5 min later, we treated mice with i.p. Tat-CAPON-12C (3 mg/kg/d) or its control for 4 consecutive days (Fig. 5A). To ensure that expression of Cre recombinase can be initiated immediately before drug treatment, we removed doxycycline from diet at 24 h before using Tat-CAPON-12C and tested the reinstatement of morphine CPP at 24 h after the last drug injection (Fig. 5A). Behavioral tests showed that all groups of mice displayed high preference score after morphine CPP conditioning and negligible preference score after morphine CPP extinction training (Fig. 5B). For the morphine-treated mice not receiving the cocktail viruses, Tat-CAPON-12C decreased preference score significantly compared with Tat-CAPON-12C/A22D; for the morphine-treated mice receiving the cocktail viruses, Tat-CAPON-12C decreased preference score significantly in WT (AAV-tTA:AAV-TRE3g) but not in the hippocampal selective ERK2 knock-out mice (AAV-tTA:AAV-TRE3g-Cre) (Fig. 5B), suggesting that ERK2 in the dCA1 is necessary for the inhibitory effect of nNOS–CAPON blocker used during extinction consolidation and retention on the reinstatement of morphine CPP. Moreover, the cocktail viruses, doxycycline intake and selective ERK2 knock-out in the hippocampus had no effect on locomotor activity, because all groups of mice had similar total distance in each preference test (Fig. 5C). Thus, dissociating nNOS–CAPON may facilitate the consolidation and retention of morphine extinction memory through ERK2 signaling, thereby inhibiting the reinstatement of morphine CPP.

Figure 5

Association of nNOS with CAPON negatively regulates morphine extinction memory and neuroplasticity through inhibiting ERK2. (A) Design of the experiments for (B–G). (B) The inhibitory effect of dissociating CAPON from nNOS on morphine CPP reinstatement depends on ERK2 two-way repeated measures ANOVA, n = 12, time: F_(2,22) = 27.193, P = 0.000, group: F_(3,33) = 4.498, P = 0.009, interaction: F_(6,66) = 5.774, P = 0.000, ^**P < 0.01, ^###P < 0.001. (C) The knockout of ERK2 has no effect on the locomotor activity of mice (two-way repeated measures ANOVA, n = 12, time: F_(2,22) = 108.188, P = 0.000, group: F_(3,33) = 1.195, P = 0.327, interaction: F_(6,66) = 0.33, P = 0.919). (D) Left, electrophysiological response obtained from glass electrode placed in the hippocampus in vitro. Right, plot of baseline normalized fEPSP and in vitro electrically evoked responses following 2 electrical stimuli (TBS). (E) Bar graph showing the change in last 15 min from baseline of hippocampal LTP (two-way ANOVA, n = 6–7, ERK2: F_(2,22) = 109.817, P = 0.000, tat: F_(1,22) = 161.347, P = 0.000, ^***P < 0.001, ^###P < 0.001). (F, G) Tat-CAPON-12C and knockout of ERK2 have no effects on basal synaptic transmission shown by I/O curve and presynaptic transmitter release shown by PPF recording in the dHPC. (H) Bar graph showing percentage decline in the last 15 min from the initial hippocampal LTP (two-way ANOVA, n = 6–7, ERK2: F_(2,22) = 8.741, P = 0.002, tat: F_(1,22) = 11.819, P = 0.002, interaction: F_(3,22) = 11.059, P = 0.000, ^**P < 0.01, ^##P < 0.01). (I) Top, electrophysiological response obtained from glass electrode placed in the hippocampus in vitro. Bottom, plot of baseline normalized fEPSP and in vitro electrically evoked responses following 2 electrical TBS. (J) Bar graph showing the change in last 15 min from baseline of hippocampal LTP (n = 3–4, t₍₅₎ = 18.11, P = 0.000). NS: saline, Mor: morphine, Dox: doxycycline, N.S.: no statistics significance.

To investigate whether the upregulation of neuroplasticity by nNOS–CAPON blocker depends on ERK2 function in the dHPC, we performed the same morphine CPP paradigm as above and recorded the electrophysiology at 24 h after the last drug injection (Fig. 5A). For the morphine-treated mice not receiving the cocktail viruses, Tat-CAPON-12C significantly increased the slope of dHPC LTP, compared with Tat-CAPON-12C/A22D, suggesting an improvement of neuroplasticity, for the morphine-treated mice receiving the cocktail viruses, Tat-CAPON-12C significantly enhanced dHPC LTP in WT (AAV-tTA:AAV-TRE3g) but not in the hippocampal selective ERK2 knock-out mice (AAV-tTA:AAV-TRE3g-Cre) (Fig. 5D and E). Thus, dissociating nNOS–CAPON enhances neuroplasticity of dHPC through ERK2 signaling in dCA1. Moreover, selective ERK2 knock-out in the hippocampus and drug administration had no effect on presynaptic transmitter release and basal synaptic transmission as indicated by PPF and I/O curve records (Fig. 5F and G). Moreover, Tat-CAPON-12C reduced morphine-induced LTP decay significantly compared with control, in hippocampal selective ERK2 knock-out mice, however, the effect of Tat-CAPON-12C disappeared (Fig. 5H). Thus, ERK2 is necessary for the enhancement of hippocampal LTP by dissociating nNOS–CAPON. To test the stability of late-LTP, we performed a long-term recording and found that late-LTP could last up to 120 min in saline group at least (Fig. 5I–J), suggesting stability of recording. Similar to 60 min electrophysiology recording (Fig. 3F), in the mice with chronic morphine exposure, LTP decayed to and remained at baseline level after 15 min (Fig. 5I–J).

Dissociating CAPON from nNOS Inhibits the Return of Morphine Preference Memory

As dissociating CAPON from nNOS inhibited the reinstatement of morphine CPP by enhancing the consolidation/retention of morphine extinction memory, we detected whether nNOS–CAPON blocker prevents the return of remote morphine preference memory. We trained mice in a spontaneous recovery of morphine paradigm (Fig. S3A). Morphine-treated mice with Tat-CAPON-12C or Tat-CAPON-12C/A22D injection spent significantly more time staying in drug-paired chamber than saline-treated mice after morphine conditioning, suggesting that nNOS–CAPON association has no effect on morphine CPP formation (Fig. S3B). All group of mice had negligible preference score (Fig. S3B), showing similar extinction learning. After CPP extinction session, mice were kept in their home cages for 30 d to allow the remote preference memory spontaneously recover (Fig. S3A). For the mice subjected to chronic morphine exposure and treated with Tat-CAPON-12C/A22D during extinction learning, the preference score in the spontaneous recovery test was similar to that in the formation test, indicating a substantial spontaneous recovery of morphine CPP. Interestingly, the mice subjected to chronic morphine exposure and treated with Tat-CAPON-12C during extinction learning displayed significantly lower preference score than that control mice did (Fig. S3B), suggesting an inhibition of spontaneous recovery of morphine CPP. For locomotor activity, there was no significant difference in total distance between groups during all tests (Fig. S3C). Thus, dissociating CAPON from nNOS prevents the spontaneous recovery of morphine CPP at remote time point.

Discussion

Understanding the relationship between hippocampal dependence of opioids and environmental stimuli is an important and unclear part of addiction. Drug-related environmental stimuli can cause craving, which in turn contributes sustaining drug use or cause relapse (Daglish et al. 2001; Rukstalis et al. 2005). How to interrupt the connection between environment stimuli and drugs is urgent for addiction treatment. Morphine CPP is a behavioral model to investigate context-dependent preference memory of drugs, and CPP can show extinction when the animals are exposed to the environment stimuli repeatedly in the absence of the effect of drugs (Tzschentke 1998). Unfortunately, with the successful extinction training, a spontaneous recovery of extinguished contextual preference can occur after the experimental animals are placed in their home cages for a period of time. Moreover, after morphine CPP conditioning, mice are trained to form the extinction memory successfully, but the preference score of mice will return to a higher level before extinction retrieval under the action of a priming dose of morphine. These phenomena indicate that extinction learning does not damage and erase the already formed drug preference memory, but forms a new and unstable extinction memory to inhibit preference memory (Morrow and Flagel 2016). Interestingly, we here find a new way to consolidate the extinction memory to prevent the reinstatement and spontaneous recovery of morphine preference memory.

Previous studies showed that morphine increases hippocampal GluN2B expression after morphine CPP extinction session and ifenprodil, a selective GluN2B antagonist, inhibits the reinstatement of morphine CPP administrated before morphine priming test (Portugal et al. 2014). Similarly, Girardi and colleagues find that administration of ifenprodil during extinction training facilitates the extinction of morphine CPP (Girardi et al. 2020). Furthermore, despite successful extinction training, the excessive GluN2B induced by long-term morphine abuse leads to decreases in hippocampal plasticity, including hippocampal LTP and dendritic spine density (Portugal et al. 2014; Fakira et al. 2016). Our previous work showed that uncoupling nNOS–CAPON, a downstream signal of GluN2B, enhances hippocampal plasticity and a consequent anxiolytic-like phenotype (Zhu et al. 2014). Interestingly, we find here that uncoupling nNOS–CAPON in the hippocampus also consolidates extinction memory of morphine preference to prevent the reinstatement of CPP. In addition, recent studies suggested that morphine-induced apoptosis and adult neurogenesis in the hippocampus positively regulates contextual memory retention, which is implicated in the eventual context-related relapse (Razavi et al. 2014; Fan et al. 2018). And nNOS, a downstream molecular of GluN2B, mediates neurotoxicity and alterations in synaptic plasticity (Zhang et al. 2018; Wang et al. 2019; Aarts et al. 2020). In our study, there is a possibility that dissociating CAPON from nNOS may improve hippocampal plasticity by inhibiting GluN2B-mediated neurotoxicity, and thereby consolidates morphine extinction memory.

ERK is a downstream component of the NMDAR signal transduction pathway activated during memory formation and is involved in the consolidation of memories (Schafe et al. 2000; Karpova et al. 2013; Gupta-Agarwal et al. 2014; Jahrling et al. 2014; Karunakaran et al. 2016). In morphine CPP extinction, hippocampal ERKs activation is seldom studied. Our previous study showed that the uncoupling nNOS–CAPON up-regulates ERK–CREB–BDNF signaling in the hippocampus and prefrontal cortex (mPFC) (Zhang et al. 2018; Qin et al. 2021), improves stroke recovery and facilitates the neuroplasticity (Ni et al. 2019). In this study, we found that, in the hippocampal selective ERK2 knock-out mice, dissociating CAPON from nNOS did not inhibit the reinstatement of morphine CPP and had no effect on the recovery of hippocampal LTP after morphine abuse, suggesting that hippocampal ERK2 is necessary for the effects of nNOS–CAPON blocker on the reinstatement of morphine CPP and neuroplasticity. Several studies have indicated that ERK2 has a crucial role for long-term memory (Mazzucchelli et al. 2002; Satoh et al. 2007, 2011). ERK2 but not ERK1 is activated following direct synaptic activation (English and Sweatt 1996), and the selective activation of ERK2 is necessary for learning (Thomas et al. 1994; Selcher et al. 2001). After the induction of hippocampal LTP, ERK2 expression is significantly increased while ERK1 is not influenced (Thomas et al. 1994). Moreover, the loss of ERK1 has no effect on the context-associated learning while the ERK2-deficient mice show an impairment in long-term memory (Selcher et al. 2001; Satoh et al. 2007). ERK2 mRNA expression is apparently decreased in the hippocampus after morphine CPP extinction phase (Ma et al. 2014), and chronic morphine treatment causes apparent decreases in pERK2 in mPFC (Pang et al. 2016). We found that morphine CPP extinction caused ERK2 dysfunction in the dHPC, and by selectively deleting hippocampal ERK2, we demonstrated that the dHPC ERK2 was crucial for the role of nNOS–CAPON in regulating the consolidation of associative extinction memory, suggesting the importance of dHPC ERK2 in preventing addiction relapse.

Extinction does not erase the already formed preference memory, but forms a new kind of unstable inhibitory memory (Bouton 1993). Thus, the consolidation of extinction memory is crucial for preventing the return of preference memory. Enhanced LTP facilitates the consolidation and retention of extinction memory (Marek et al. 2011; Xiong et al. 2014; Bentefour et al. 2018). The LTP induction promotes spine head enlargement and the formation and stabilization of new spines (Bosch et al. 2014; Chidambaram et al. 2019). New spines can persist for days or months (Holtmaat et al. 2005) and are critical for the preservation of extinction memory (Lai et al. 2012). It is well known that hippocampus is indispensable in preference extinction (Fan et al. 2018; Farahimanesh et al. 2018; Nazari-Serenjeh et al. 2020). Hippocampal LTP has been considered an important mechanism of learning and memory reflecting synaptic plasticity (Neves et al. 2008). We found that LTP was impaired in the hippocampus after the extinction of morphine CPP and nNOS–CAPON contributed to the morphine-induced hippocampal LTP impairment and spines reduction, and ERK2 is necessary for the enhancement of hippocampal LTP by dissociating nNOS–CAPON. Recently, we found that blocking nNOS–CAPON after extinction facilitates the retention of extinction memory through ERK–CREB–BDNF pathway and ERK-mediated acetylation of histone H3 (Qin et al. 2021). Thus, the negative regulation of neuroplasticity by nNOS–CAPON may lead to the reinstatement of morphine CPP through weakening the consolidation of extinction memory, in which, underlying mechanisms may involve ERK–CREB–BDNF pathway. Additionally, LTP contains a stimulation-labile phase of short-term potentiation (STP, or transient LTP, t-LTP) that decays into stable LTP. STP and LTP have dramatically different consequences during high-frequency synaptic transmission, such that STP modulates the frequency response of synaptic transmission whereas LTP maintains the synaptic fidelity (Volianskis et al. 2013). Interestingly, we found that chronic morphine exposure displayed a normal STP but impaired LTP after 7 days of extinction training. Consistent with our findings, it has been reported that STP is normal during morphine extinction stage (Portugal et al. 2014). As indicated by paired-pulse facilitation (PPF) ratio in the hippocampus, chronic morphine exposure did not affect presynaptic transmitter release after 7 days of extinction training, which may explain why morphine exposure displays normal STP. Moreover, in the hippocampal selective ERK2 knock-out mice subjected to chronic morphine exposure, treatment with nNOS–CAPON blocker only affected LTP but not STP, suggesting that ERK2 is not implicated in the regulation of STP. Indeed, it is reported that ERKs inhibitor U0126 did not interfere with basal neurotransmission and a form of presynaptic short term plasticity (Vara et al. 2009).

Behavioral sensitization is defined by the enhanced locomotor response following repeated, intermittent exposure to some specific drugs (Paulson et al. 1991). The animals show more robust locomotor sensitization when re-exposed to drug-paired rather than un-paired environment (Kalivas and Stewart 1991), implicating drug cues in locomotor sensitization. The neural circuits underlying behavioral sensitization and reinstatement are similar, in which, dopamine projections from the ventral tegmental area (VTA) to the nucleus accumbens (NAc) and glutamate projections from the mPFC to the NAc are essential. However, in the sensitization, hippocampus impacts the mesolimbic dopamine system via inputs to the VTA, while in the reinstatement, hippocampus affects the mesolimbic dopamine system via inputs to the NAc rather than the VTA (Steketee and Kalivas 2011). Our findings that uncoupling nNOS–CAPON in the dHPC can prevent reinstatement of morphine CPP but not morphine-induced locomotor sensitization may be due to that nNOS–CAPON association in the dHPC affects hippocampus-NAc but not hippocampus-VTA circuit, which needs to be confirmed by further study. Additionally, it is reported that behavioral sensitization is closely related with opiate receptors (Spanagel 1995; Shippenberg et al. 2001). The nNOS–CAPON blocker has no effect on locomotor sensitization, possibly because it does not affect opiate receptors. The selectivity of nNOS–CAPON for the reinstatement of morphine CPP suggests a possibility that psychostimulant effect brought by morphine may be retained while relapse into drug seeking behavior is inhibited, revealing a very prospective treatment way for addiction relapse.

In sum, excessive hippocampal GluN2B induced by chronic morphine exposure facilitates nNOS–CAPON interaction, which in turn, prevents the consolidation of extinction memory via reducing ERK2-mediated neuroplasticity, thereby leading to the reinstatement of morphine preference. Dissociating CAPON from nNOS upregulates ERK2 phosphorylation, enhances hippocampal neuroplasticity and the consolidation of associative extinction memory, thereby preventing the reinstatement of morphine preference. CREB–BDNF signaling may be the downstream of ERK2-mediated hippocampal neuroplasticity (Fig. S4). Thus, nNOS–CAPON can be a potential target for addiction relapse treatment in clinical practice.

Funding

National Natural Science Foundation of China (31530091, 82090042, 81870912); National Key Research and Development Program of China (2016YFC1306703); Science and Technology Program of Guangdong (2018B030334001).

Notes

Conflict of Interest: The authors declare no competing financial interests.

Author notes

Authors Xiaolin Kou and Jiayun Xian contributed equally to this work