The Vagus Nerve Attenuates Fulminant Hepatitis by Activating the Src Kinase in Kuppfer Cells

Authors


  • Y. L., Z. X. and Y. Y. contributed equally to this work.

Abstract

The parasympathetic nervous system has been known to modify innate immune responses. In animal models, acetylcholine (Ach) released from the distal ends of nerves has been shown to inhibit inflammatory responses such as endotoxic shock, pancreatitis, intestinal inflammation, etc. However, its role in LPS-induced fulminant hepatitis remains to be elucidated. Here, we demonstrate that the vagus nerve acts as a suppressor in the liver after challenge with LPS plus D-gal. The vagus nerve acts through the α7 AchR expressed on the surface of Kupffer cells, inhibiting the production of the proinflammatory cytokines TNF and IL-6. A mechanism study revealed that the suppressive effect of Ach may occur through the activation of Src kinase and subsequent inhibition of the Myd88 signal pathway. Our study has suggested a suppressive role of vagus nerve in the modulation of liver inflammatory responses, which should be noticed during clinical massive hepatectomy and liver transplantation. The nicotinic anti-inflammatory pathway may also be a potential target for sepsis after liver transplantation.

Introduction

The innate immune system is pivotal to host health as the first line of defence against invading pathogens and tissue damage [1]. However, its responses must be tightly controlled because excessive inflammation can lead to tissue damage. Many negative regulators and pathways that restrict excessive inflammatory responses have been identified [2, 3]. Recent studies strongly suggest that innate immune responses are under substantial neuronal control [4, 5]. In addition to the sympathetic nervous system, which is already known to regulate local immune responses [6], the parasympathetic nervous system, especially the vagus nerve, is increasingly recognized as a regulator of immune inflammation [4, 7]. In general, the afferent vagus system acts through the hypothalamic–pituitary–adrenal axis to regulate the inflammatory responses, but the efferent vagus nerve also exerts immunomodulatory functions [4, 5]. Stimulation of the efferent vagus nerve decreases the serum production of proinflammatory cytokines, such as tumour necrosis factor (TNF), interleukin-6 (IL-6), IL-1β and high-mobility group box 1 (HMGB1) etc. and thus alleviates endotoxemia and endotoxin shock in rodents [7, 8]. Subsequent studies suggest that the efferent vagus nerve also exerts an anti-inflammatory effect in animal models such as the pancreatitis, arthritis, peritonitis and colitis [9-12]. The efferent vagus nerve also attenuates Fas-induced liver apoptosis through alpha7 nicotinic acetylcholine receptor (nAChR) [13]. However, the effect of the vagus nerve on acute hepatitis and its molecular mechanisms has not been fully examined till now.

The anti-inflammatory effect of the vagus nerve is mediated by the interaction of Ach, the principal neurotransmitter of the vagus nerve, with cholinergic nicotinic receptors. Two nAchR are involved: the α7 homopentamer expressed by human monocyte-derived macrophages and mouse peritoneal macrophages [14], and the α4β2 heteropentamer expressed by alveolar macrophages [15]. The liver has been suggested as an innate immune organ continuously exposed to lipopolysaccharide (LPS) and other pathogenic components from the gastrointestinal tract via the hepatic portal vein [16-18]. Infection, injury, malnutrition, autoimmune responses and other factors can lead to liver inflammation or immunopathological liver injury [19]. Liver non-parenchymal cells such as sinusoidal endothelial cells, dendritic cells and Kupffer cells have been shown to respond to LPS challenge, especially the Kupffer cells. Whether the efferent vagus nerve suppresses the inflammatory response by Kupffer cells in the liver and the associated receptors has not been verified.

Furthermore, the liver receives parasympathetic innervation from the hepatic branch of the left vagus nerve. After radical clinical procedures such as massive hepatectomy or liver transplantation, the liver loses parasympathetic signals from the hepatic branch of the vagus nerve. In these patients, how this ‘vagotomy’ will influence the liver microenvironment and the local inflammatory response remains largely unexplored, which may be much important for the susceptibility to pathogen infections of these patients.

In this study, we demonstrate that the vagus nerve suppresses inflammatory responses in the liver during LPS-induced hepatitis. The Ach released from the efferent vagus nerve may act through the α7 AchR on Kupffer cells and inhibit the production of proinflammatory cytokines such as TNF, IL-6 etc. Furthermore, Ach may exert an inhibitory effect on Kupffer cells by activating Src kinase phosphorylation and subsequently suppressing the activation of TLR4 downstream signal pathways. Our study suggests a suppressive role of the vagus nerve on the modulation of liver inflammatory responses, which may have clinical implications for patients undergoing massive hepatectomy or liver transplantation.

Materials and methods

Mice and cell culture

Balb/c mice were from Joint Ventures Sipper BK Experimental Animals Co. All mice were bred in specific pathogen-free conditions. All animal experiments were performed in accordance with the National Institute of Health Guide for the Care and Use of Laboratory Animals, with the approval of the Scientific Investigation Board of Second Military Medical University, Shanghai.

Reagents

LPS (0111:B4), nicotine and α-BGT were from Sigma. Antibodies against Erk phosphorylated at Thr202-Tyr204 (E10), to Jnk phosphorylated at Thr183-Tyr185 (G9), to p38 phosphorylated at Thr180-Tyr182 (9211), to IKKα/β phosphorylated at Ser176-Ser180 (16A6), to IκBα (4812), src (2108), p-src phosphorylated at Tyr416 (100F9) were from Cell Signaling Technology. Antibodies specific to β-actin (sc-130656) were from Santa Cruz.

Endotoxin-induced fulminant hepatitis model and drug administration in vivo

Fulminant hepatitis in mice was established by i.p. injection of LPS (5 μg/kg body weight) plus D-gal (400 mg/kg body weight) as previously described [20]. Six hours post-injection, the serum and liver tissue were collected. Serum TNF and ALT were detected. Liver histology was performed with H&E staining. Some liver sections were further homogenized for TNF concentration in the liver. In some experiments, the mice were given 10 mg/kg of body weight Gdcl3 30 h and 6 h prior to LPS challenge. Nicotine (1 mg/kg body weight) or α-BGT (1 nmol) was administrated i.v. after LPS was injected. Control animals were injected with 200 μl of saline.

Left cervical vagotomy

Mice were anesthetized and subjected to left cervical vagotomy or sham surgery 3 days before induction of hepatitis as described previously [11]. A ventral cervical midline incision was used to expose the left cervical vagus trunk, which was ligated and divided. Subsequently, the skin was closed. In sham-operated animals, the nerve was exposed and isolated from surrounding tissue but not transected.

Isolation and culture of Kupffer cells from the liver

The liver was digested followed by differential centrifugation using Percoll to isolate Kupffer cells as previously described [21]. Briefly, the liver was excised and minced after collagenase perfusion of the portal vein. The liver slurry was filtered through gauze mesh, washed with culture media and centrifuged two times at 600 g for 5 min to remove hepatocyte fraction. The non-parenchymal cells were further purified using a discontinuous Percoll gradient of 25 and 50% Percoll. Purified non-parenchymal cells were washed and cultured in media at 37 °C for one-half hour before washing and incubating in tissue culture media overnight. These cells were ∼80% pure for Kupffer cells as estimated by their ability to ingest latex beads. Cell viability was always >90% as assessed by Trypan Blue.

RNA interference

Small interfering RNA targeting Src was from Genepharma (Shanghai, China). SiRNA duplexes were transfected into mouse Kupffer cells using INTERFERin Reagent from Polyplus according to the standard protocol. Cells were cultured for an additional 36 h before LPS stimulation.

RNA quantification & quantitative real-time PCR

Total RNA was extracted with TRIzol reagent (Invitrogen, Carlsbad, CA) following the manufacturer's instructions. RNA concentrations were determined with a NanoDrop instrument (NanoDrop Technologies, Wilmington, DE). Quantitative real-time PCR (Q-PCR) analysis was performed by Light-Cycler (Roche, Indianapolis, IN) and SYBR RT-PCR kit (Takara) as described previously [22]. The primers used for TNF were Tnf forward 5′-AAAAGCAAGCAGCCAACCAG-3′, reverse 5′-TGCC ACAAGCAGGAATGAGAA-3′; the relative expression level of mRNAs was normalized by the level of β-actin expression in each sample.

Immunoblot

Cells were lysed with cell lysis buffer (CST) supplemented with protease inhibitor cocktail (Calbiochem). Protein concentrations of the extracts were measured with BCA assay (Pierce). Samples were separated by SDS-PAGE and were blotted onto polyvinyldifluoride membranes (Millipore). Membranes were blocked and were incubated overnight with the appropriate antibodies in 1% BSA and 0.1% Tween-20 in Tris-buffered saline. Secondary antibodies were incubated for 1 h after washing the membrane for 3 times.

Cytokine detection

TNF, IL-6 and MIP2 in the supernatants, serum or liver homogenate were measured as described [7] with ELISA Kits (R&D Systems).

Statistical analysis

The statistical significance of comparisons between two groups was determined with Student's t-test. P values of less than 0.05 were considered statistically significant.

Results

Previous vagotomy exaggerates hepatitis

Intraperitoneal administration of LPS plus D-galactosamine (D-gal) elicited fulminant hepatitis in Balb/c mice, as indicated by the profound increase in the serum concentration of alanine aminotransferase (ALT) and histopathologic changes in the liver showing marked infiltration of neutrophils and focal area of necrosis (Fig. 1A, B). To investigate the relative contribution of the vagus nerve on fulminant hepatitis, we cut the left cervical vagus nerve 3 days prior to the administration of LPS plus D-gal, as the hepatic branch of the vagus nerve originates from the left cervical vagus nerve. A comparable sham surgical procedure was also performed on the control mice in which the vagus nerve was isolated but not transected. Left cervical vagotomy significantly increased the serum ALT level and the histopathologic liver damage (Fig. 1A, B), suggesting for a protective role of vagus nerve. Considering that TNF levels play an important role in LPS-induced hepatitis, we further detected TNF concentration by ELISA. LPS-induced hepatitis was associated with strongly elevated levels of serum TNF and liver homogenate TNF, which were even higher in mice that underwent left cervical vagotomy (Fig. 1C, D). Furthermore, after induction of fulminant hepatitis, mice that underwent vagotomy surgery had lower survival rates than control group mice (Fig. 1E). These data indicated that previous vagotomy exaggerated the hepatitis and that the vagus nerve may function as an anti-inflammatory factor.

Figure 1.

Left cervical vagotomy exaggerates fulminant hepatitis in mice. (A–D) Mice with left cervical vagotomy (VGX) or sham operated (sham) or without treatment (NT) were challenged with LPS/D-gal, and the serum and liver were collected after 6–8 h. Serum ALT level (A) and TNF concentration (C) were determined. The liver tissues were used for pathologic analysis (B) and further homogenized for liver TNF concentration (D). (E) The survival rate of vagotomy mice and sham control after LPS/D-gal challenge (n = 10). Data shown are mean ± SD (A, C, D) or representative of at least 3 independent experiments with same results (B, E). *P < 0.05. Arrows indicated the liver necrosis and infiltration of lymphocytes.

Blockade of nicotinic receptors augments whereas stimulation of nicotinic receptors attenuates hepatitis

Taking into account the vigorous effect of vagotomy on LPS-induced hepatitis, we next set to determine whether pharmacological manipulation of nicotinic receptors influences hepatitis severity. Hepatitis was induced after intraperitoneal injection of PBS, nicotine or unlabelled α-bungarotoxin (α-BGT), a selective antagonist for α7 nicotinic acetylcholine receptor [23]. In accordance with the vagotomy, α-BGT augmented LPS-induced hepatitis, displaying elevated levels of serum and liver homogenate TNF compared with that in the PBS-injected group (Fig. 2A, B). Accordingly, the nicotine treatment attenuated the severity of hepatitis, with reduced serum ALT and liver homogenate TNF levels (Fig. 2A, B).

Figure 2.

Pharmacological manipulation of nicotinic receptors affects fulminant hepatitis. (A) Serum ALT level from mice pretreated with nicotine or α-BGT and then challenged with LPS/D-gal. (B) Liver tissues in A were collected, and TNF concentrations in the homogenate were determined by ELISA assay. Data shown are means ± SD from 3 independent experiments. *P < 0.05.

Kupffer cells are needed for the anti-inflammatory function of the vagus nerve

To determine whether Kuppfer cells are involved in the anti-inflammatory function of the vagus nerve, we inactivated Kupffer cells by injection of gadolinium chloride (Gdcl3) before LPS plus D-gal challenge [24, 25]. Gdcl3 treatment significantly reduced the liver damage, displaying suppressed serum ALT and liver TNF levels compared with that in control mice (Fig. 3A, B). Moreover, the promotion of the inflammatory response in the liver by left cervical vagotomy was also abrogated by Gdcl3 pretreatment (Fig. 3A, B). The inhibitory effect of nicotine was also attenuated with Gdcl3 pretreatment (Fig. 3C), suggesting for a non-redundant role of Kuppfer cells in the effect of the vagus nerve and nicotine.

Figure 3.

Involvement of Kupffer cells in the function of vagus nerve. (A) Serum ALT levels in vagotomy treated or sham control-operated mice, with or without Gdcl3 pretreatment before LPS/D-gal challenge. (B) Liver TNF concentration from mice in A. (C) Serum ALT levels in nicotine treated mice, with or without Gdcl3 pretreatment before LPS/D-gal challenge. (D) Western blot analysis of α7-AchR in Kupffer cells and RAW264.7. Data shown are means ± SD or representative from 3 independent experiments. *P < 0.05.

Previous studies suggest that the anti-inflammatory effect is mediated by an interaction of acetylcholine, the principal neurotransmitter of the vagus nerve, with macrophage cholinergic nicotinic receptors expressing the α7 subunit [14]. We first determined whether Kupffer cells express the α7-AChR. Western blot assay revealed that Kupffer cells expressed comparable α7-AChR as RAW264.7 cells (Fig. 3D). These data suggested that Kupffer cells were not only responsible for the LPS-induced hepatitis, but may also attribute to the anti-inflammatory function of the vagus nerve and nicotine.

Nicotine attenuates Kupffer cells activation through α7-AchR

We next investigated whether the vagus nerve could attenuate LPS-induced Kupffer cell activation in vitro. Kupffer cells were isolated from mouse liver and stimulated with LPS with or without different concentrations of nicotine. As shown in Fig. 4A, activation of the nAchR with nicotine indeed suppressed the production of proinflammatory cytokines (TNF, IL-6, etc.) from Kupffer cells in a dose-dependent manner. The suppressive effect was also seen at the transcription level, as real-time quantitative PCR analysis revealed reduced mRNA level of these cytokines (Fig. 4B). Moreover, in the competition study, the addition of an α7 nicotinic receptor antagonist, α-BGT, abrogated the suppressive effect of nicotine (Fig. 4C), confirming that nicotine inhibited Kupffer cells activation through the α7-AchR. Other study also suggested that in alveolar macrophages, the α4β2 AchR instead of the α7 homopentamer AchR was responsible for the suppressive effect of the vagus nerve. However, disruption of α4β2 AchR by specific antagonist dihydro-β-erythroidine (DHβE) [26] did not influence the inhibitory effect of nicotine in Kupffer cells (data not shown).

Figure 4.

Nicotine suppresses Kupffer cell activation in vitro. (A) Cytokines in the supernatant from Kupffer cells after LPS stimulation. Nicotine was added with LPS at gradient concentrations. (B) Real-time quantitative PCR of Tnf mRNA in Kupffer cells treated as in A. (C) TNF concentration in the supernatant from Kupffer cells stimulated with LPS plus 1 μm nicotine, and in some groups α-BGT was added at gradient concentrations (0.01–10 μm). (D) Western blot assay of TLR4 downstream signal molecules in Kupffer cells after LPS stimulation, with or without nicotine. Data shown are means ± SD (A–C) or representative (D) from 3 independent experiments. *P < 0.05.

Because LPS mainly induces the production of proinflammatory cytokines through TLR4 and the downstream Myd88 pathway, we further examined the activation of the Myd88 signal pathway by Western blot assay. Consistent with the reduced cytokine production, nicotine treatment also inhibited the activation of IKKα/β, I-κBα, p65, p38, ERK and JNK (Fig. 4D). These data indicated that the vagus nerve/nicotine may suppress the activation of Kupffer cells by inhibiting the activation of Myd88 signal pathways, thus resulting in reduced cytotoxic cytokine production.

Src kinase is responsible for reduced activation of Myd88 pathway in Kupffer cells

Previous studies identified various negative regulators of the Myd88 pathway, including SHP-2, Src, A20 and IRAK-M [22, 27-30]. To determine which of these regulators are involved in the suppressive function of nicotine, we screened these molecules by siRNA technology. Silencing of Src kinase strongly abrogated the effect of nicotine on Kupffer cell TNF production, while other molecules had no or minor effect (Fig. 5A and data not shown). PP1, a Src kinase inhibitor, was then used to further identify the role of Src in nicotine-suppressed LPS activity. As shown in Fig. 5B, blocking Src kinase with PP1 abrogated the suppression of Myd88 pathway activation by nicotine treatment,as evidenced by PP1 pretreatment increased nuclear translocation of p65 and AP-1. TNF production also showed no differences with or without nicotine treatment (Fig. 5C). Ex vivo analysis of Kupffer cells from LPS plus D-gal challenged liver also revealed reduced activation of Src kinase in the vagotomy group (Fig. 5D). These data suggested that Src kinase plays a non-redundant role in the suppressive effect of the vagus nerve/nicotine on LPS-induced Kupffer cell activation.

Figure 5.

Src kinase contributes to nicotine-mediated suppression of Myd88 pathway. (A) TNF concentration in the supernatant of LPS-stimulated Kupffer cells in the presence of nicotine after silencing of Src kinase with specific small interfering RNA. (B) Western blot analysis of Myd88 pathway activation in Kupffer cells after PP1 addition. (C) Supernatant TNF concentration of LPS-stimulated Kupffer cells in the presence of nicotine after PP1 blockade. (D) Phosphorylation of src kinase in Kupffer cells ex vivo from hepatitis mice with vagotomy or not. Data shown are means ± SD (C) or representative (A, B, D) from 3 independent experiments. *P < 0.05

Discussion

The parasympathetic nervous system has recently been recognized as an important anti-inflammatory factor via the interaction of neurotransmitter Ach with its nicotinic receptors on macrophages [4, 5, 7]. Here, we show for the first time that the blockade of the nicotinic anti-inflammatory effect through left cervical vagotomy or administration of its inhibitor leads to exaggerated inflammatory cytokines production, and worsens the LPS-induced hepatitis. Our data suggest that the efferent vagus nerve plays an important role in the regulation of the inflammatory responses during fulminant hepatitis.

The receptor subtypes of Ach are also expressed on immune cells in addition to the brain and motor end-plate [4, 8, 14]. The inhibitory effect of efferent the vagus nerve has been implicated in various disease models more than one decade ago. The first report is in experimental endotoxemia [7]. Direct electrical or pharmacological vagus nerve stimulation reduced serum TNF levels and prevented shock, and bilateral cervical vagotomy augmented serum TNF levels and sensitized animals to the lethal effects of endotoxin [5, 7, 31]. Subsequently, the anti-inflammatory effect of vagus nerve has been demonstrated in sterile hemorrhagic shock [32, 33], ischaemia/reperfusion injury [34, 35], pancreatitis [11], peritonitis [12] and inflammatory bowel disease [9]. Our report here is in line with previous studies and extended the suppressive effect of the vagal reflex to a well-established fulminant hepatitis. These results also indicated that the efferent vagus nerve can inhibit various sterile and infective inflammation, which suggested another potential target for the control of inflammatory diseases.

Pervious and current studies consistently demonstrated the macrophages as the responsible cell type targeted by acetylcholine released from the efferent vagus nerve and other nicotinic receptor agonists such as nicotine. Macrophages are derived from bone marrow precursors and blood monocytes, and they are localized in peripheral tissues constituting the mononuclear phagocyte system after maturation [36]. Kupffer cells are the resident macrophages in the liver and have both similarities and heterogeneities with macrophages in other tissues (lung, spleen, skin, etc.) [36, 37]. In this study, we have also determined whether Kupffer cells are responsible for the inhibitory effect of vagus nerve or nicotine. Depletion of Kupffer cells not only attenuated the inflammatory responses in the liver but also abrogated the increased TNF production via left cervical vagotomy, suggesting a non-redundant role of Kupffer cells. In addition, analogous to murine peritoneal macrophages and human monocyte-generated macrophages, nicotine can suppress the proinflammatory cytokines production from Kupffer cells ex vivo. However, other cells such as dendritic cells (DCs) and neutrophils also express nicotinic AchR [38, 39]. The involvement of these cells in the suppression of fulminant hepatitis by vagus nerve cannot be excluded. Given the scarcity of hepatic DCs [40, 41] and recruitment of neutrophils into the liver after inflammation [42, 43], we can at least conclude that Kupffer cells mainly contribute to the inhibitory effect of vagus nerve and nAchR agonists.

Our experiments have not provided conclusive evidence that Ach released from the efferent vagus nerve acts directly on AchR on Kupffer cells in the liver and attenuates hepatitis, as left cervical vagotomy not only damages the liver function but also other organs such as lung, heart and etc. A selective hepatic branch vagotomy would be more helpful to elucidate the exact role of vagus nerve on liver inflammation and damage. However, we should also take into account that even selective hepatic branch vagotomy also influence the physiologic function of organs other than the liver, such as the pancreas.

The best way to stimulate the vagus efferent transmission is electric stimulation. However, this is really a technically challenge for us to repeat it stably in each mice. More importantly, there is no data showing the duration of the electric stimulation effect. Thus, we choose an alternative approach to manipulate the Ach receptors. As previous studies suggested the anti-inflammatory effect of vagus nerve was mainly through Ach release, we chose nicotine, an AchR agonist, to mimic the effect of vagus nerve. Our data suggested an anti-inflammatory effect of nicotine cholinergic pathway. However, this alternative approach could not exclude indirect effects of systemic usage of nicotine, for example the side effect on circulation and metabolism. But our study have suggested a potential therapeutic effect of nicotine or other cholinergic drugs in liver inflammation.

There are two nicotinic acetylcholine receptors (nAchR) demonstrated to be involved in the anti-inflammatory effect of the vagus nerve: the α7 homopentamer expressed by mouse peritoneal macrophages [14] and the α4β2 heteropentamer expressed by alveolar macrophages [15]. In Kupffer cells, blockade of the α7-AchR abrogated the suppressive effect of nicotine, while disruption of α4β2 AchR by specific antagonist DHβE had no effect. As found in this study, previous studies also suggested that the activation of α7 homopentamer nAChRs inhibits activation of the transcription factor NF-κB. Furthermore, we have identified Src kinase as being responsible for the reduced activity of NF-κB. Stimulation of α7-AchR promoted the phosphorylation of Src kinase, and inhibition of Src kinase activity impeded the suppressive effect of nicotine on Kupffer cell cytokine production after LPS stimulation. The detailed mechanisms by which Src kinase restricts NF-κB activation are not examined in our study. However, recent reports suggest that Src kinase may inhibit TLR downstream signalling via phosphorylation of Myd88, leading to its subsequent ubiquitination and degradation [28].

Liver transplantation has enjoyed dramatic success as a treatment option for patients suffering from chronic end-stage liver diseases [44]. It also serves as a definitive treatment for certain genetic conditions such as familial amyloidosis and primary oxalosis, and as a potential curative therapy in selected cases of primary liver cancer [44-46]. While the outcome of liver transplantation improves along with our understanding of the transplant immunity and along with the development of new immune response modulators, sepsis or systemic inflammatory response syndrome (SIRS) remains the leading cause of early post-operative mortality. After liver transplantation surgery, excessive inflammatory response can lead to graft failure or multi-organ failure. Whether the vagotomy that occurs as part of the liver surgery participates in the immune system overstimulation requires further elucidation.

In the study presented here, we demonstrate that the vagus nerve or nicotinic anti-inflammatory pathway is an essential regulator of inflammation during experimental hepatitis. Ach released from the efferent vagus nerve may act through Kupffer cell α7 AchRs and inhibit production of proinflammatory cytokines in the liver. Furthermore, Ach may exert its inhibitory effect on Kupffer cells by activating Src kinase phosphorylation and subsequently suppressing the activation of TLR4 downstream signal pathways. Our study increases understanding of the suppressive effect of the vagus nerve on fulminant hepatitis. The nicotinic anti-inflammatory pathway may be a future target for the treatment of hepatitis, especially for massive hepatectomy and liver transplantation.

Acknowledgment

This work was supported by Grants from Sponsored by Shanghai Rising-Star Program (13QA1405000), the Youth Research Foundation of Shanghai Public Health Bureau (2009Y063) and the Key Project of the ‘Twelfth Five-year Plan’ for Medical Science Development of PLA (BWS12J027). We thank Dr. Chaofeng Han for technical assistance and Dr. Sheng Xu for helpful discussion.

Disclosures

The authors have no financial conflicts of interests.