MANF Promotes Diabetic Corneal Epithelial Wound Healing and Nerve Regeneration by Attenuating Hyperglycemia-Induced Endoplasmic Reticulum Stress

Patients with diabetes often present with decreased corneal wound healing, reduced corneal innervation, and a loss of corneal sensitivity (1–9). These complications of diabetes in the cornea are known as diabetic keratopathy (DK), which can lead to severe visual impairment (10). Increasing evidence suggests that endoplasmic reticulum (ER) stress is one of the pathogenic mechanisms leading to type 1 diabetes, type 2 diabetes (11–13), and diabetic peripheral neuropathy (14). ER stress impairs insulin synthesis and causes β-cell dysfunction and apoptosis in patients with diabetes (13,15). In addition, ER stress is involved in the pathogenesis of nervous system complications in diabetes (16). Thus, ER stress can potentially serve as a target for treating diabetes. For diabetic ocular diseases, multiple studies have shown that ER stress influences diabetic retinopathy by mediating retinal cell death and inflammation (17,18). However, to date, no reports have described the relationship between ER stress and DK.

Mesencephalic astrocyte-derived neurotrophic factor (MANF), a member of the novel neurotrophic factor family, is widely distributed in mammalian neural and nonneural tissues. In cells, MANF is mostly localized in the ER and plays a crucial role in maintaining ER homeostasis (19). MANF has a noncanonical ER retention sequence and is structurally and functionally distinct from any other canonical neurotrophic factor; thus, MANF is classified as a noncanonical neurotrophic factor (20). Emerging evidence suggests that MANF plays beneficial roles in a broad range of diseases. The protective effect of MANF has been suggested to depend on its ability to lessen ER stress (21–23). MANF could exert neuroprotective effects in various models, including Parkinson disease, myocardial infarction, and cortical neurons from patients after an ischemic stroke (24–27). Recently, Xu et al. (22) demonstrated that MANF protects neuronal cells from amyloid β-peptide–induced neurotoxicity by attenuating ER stress. MANF also plays a key role in diabetes (28). MANF was essential for the proliferation and survival of pancreatic β-cells, which is associated with the amelioration of ER stress (29,30). MANF deficiency leads to progressive postnatal loss of β-cells and insulin-dependent diabetes in mice (29,31). Hence, MANF is a potential therapeutic target for treating diabetes. Because MANF can rescue neuronal loss in several diseases of the nervous system, we are interested in the MANF induction profile and its potential significance in DK. In this study, we found that both mouse and human corneas express MANF. Our data reveal that recombinant human MANF (rhMANF) plays a beneficial role in improving delayed wound healing and nerve regeneration in diabetic corneas by reducing apoptosis induced by ER stress.

Male C57BL/6 mice (6–8 weeks old) were purchased from the Vital River Laboratory Animal Technology Co., Ltd. (Beijing, China), and maintained in the animal center of Shandong Eye Institute. All animal experiments were approved by the ethics committee of Shandong Eye Institute and conducted in accordance with the Association for Research in Vision and Ophthalmology Statement for the Use of Animals in Ophthalmic and Vision Research. Type 1 diabetes was induced by intraperitoneal injection of a low dose of streptozotocin (STZ) 50 mg/kg (Sigma-Aldrich, St. Louis, MO) for 5 consecutive days. Blood glucose levels were monitored using a OneTouch Basic glucometer (LifeScan, Johnson & Johnson, Milpitas, CA). Diabetic mice were used after 16 weeks of STZ injection, at which point the blood glucose values were 26.41 ± 0.87 mmol/L in diabetic mice and 6.27 ± 0.10 mmol/L in normal mice.

Mouse central corneal epithelium (2.5-mm diameter) was removed with an Algerbrush II corneal rust ring remover (Alger Co., Lago Vista, TX) after anesthesia under a stereoscopic microscope. Wound healing progress was monitored by fluorescence staining for epithelial defects, and photographs were taken with a slit lamp microscope. The corneas were either snap frozen in Tissue-Tek optimum cutting temperature (OCT) compound for immunofluorescence staining, or corneal epithelia cells were collected for real-time quantitative PCR (qRT-PCR), Western blot, and ELISA analyses.

Subconjunctival injection is routinely used in ophthalmology clinics to treat ocular diseases because it enables injected materials to slowly diffuse into the cornea with minimal systemic effects. Mice were injected subconjunctivally with 5 μL of solution per injection. Anesthetized mice were injected subconjunctivally with rhMANF (10 μg/mL) (R&D Systems, Minneapolis, MN) or 4-phenylbutyrate (4-PBA) (20 mmol/L), a classic ER stress inhibitor (Sigma-Aldrich) 24 h before, 0 h, and 24 h after wounding. For Akt inhibition, normal and diabetic mice were injected subconjunctivally with triciribine (an Akt inhibitor of cellular phosphorylation/activation of Akt1/2/3 by targeting an Akt effector molecule, 0.65 μg/eye) (Sigma-Aldrich) (32) and mouse Akt-specific small interfering RNA (siRNA) or nonspecific control siRNAs (Horizon Discovery Company, London, U.K.). The Akt inhibitor was injected 48 h and 24 h before, 0 h, and 24 h after wounding. Mouse Akt-specific siRNA or nonspecific control siRNAs were injected 24 h and 4 h before and 24 h postwounding. Mouse MANF-specific siRNA or nonspecific control siRNAs were injected 24 h and 4 h before and 24 h after wounding. The siRNAs were injected at concentrations of 20 μmol/L.

Corneal sensitivity was measured according to our previous descriptions (1) with a Cochet-Bonnet esthesiometer (Luneau Ophtalmologie, Chartres Cedex, France) in unanesthetized normal and diabetic mice before and every other day after scratching the epithelium until 7 days. The longest filament length with a positive response was considered as the corneal sensitivity threshold, which was verified at least four times.

The mRNA expression of MANF in normal and diabetic corneal epithelium at 0, 6, 24, and 48 h postwounding were detected. Total RNA was extracted from collected mouse corneal epithelium using NucleoSpin RNA kits (Takara, Dalian, China). cDNAs were synthesized using PrimeScript RT Master Mix (Takara). qRT-PCR was carried out using ChamQ Universal SYBR qPCR Master Mix (Vazyme Biotech Co., Ltd., Nanjing, China) and the 7500 Real-Time PCR System (Applied Biosystems, Foster City, CA). The thermocycling conditions used were 30 s at 95°C followed by 40 two-step cycles (10 s at 95°C and 30 s at 60°C). The sequences of the primers used for amplifying mouse β-actin were 5′-ACT GCC GCA TCC TCT TCC T-3′ (forward) and 5′-TCA ACG TCA CAC TTC ATG GA-3′ (reverse). The sequences of the primers used for amplifying mouse MANF were 5′-CTG CGG CCA GGA GAC TG T-3′ (forward) and 5′-CAA CCG ATT CTC TTT GCC TCT T-3′ (reverse). The quantification data were analyzed with Sequence Detection System software (Applied Biosystems), using β-actin expression as an internal control.

Western blot analysis was performed using standard protocols, as described (1). Total proteins were extracted from mouse corneal epithelium lysates in radioimmunoprecipitation assay buffer. Samples (20 μg) were run on 12.5% SDS-PAGE gels and transferred to polyvinylidene fluoride membranes (Millipore, Billerica, MA). The membranes were probed with antibodies against MANF (Sigma-Aldrich), nerve growth factor (NGF), brain-derived neurotrophic factor (BDNF), Akt, phospho-Akt (p-Akt) (Affinity Biosciences), CHOP, X-box binding protein 1 spliced (XBP-1s), caspase 12 (Cell Signaling Technology), Bcl-2 (Proteintech), and Bax (Abcam) followed by incubation with horseradish peroxidase–conjugated secondary antibodies (1:3,000; Absin Bioscience, Inc., Shanghai, China). Nonmuscle β-actin (Cell Signaling Technology) was detected as a loading control. The bands were visualized with the Enhanced SuperSignal Chemiluminescent Substrate (Thermo Fisher Scientific), and the images were acquired using a Kodak Image Station 4000R Pro.

Mouse eyeballs were snap frozen in Tissue-Tek OCT compound (Sakura Finetek, Tokyo, Japan). Seven-micrometer-thick sections were cut and mounted to poly-l-lysine–coated glass slides, fixed in 4% paraformaldehyde for 15 min, permeabilized with 0.1% Triton X-100 for 10 min, and blocked with 5% BSA for 1 h at room temperature. The samples were stained with antibodies against MANF (Sigma-Aldrich) and followed by fluorescein-conjugated secondary antibodies (1:100; Beijing Zhongshan Jinqiao Biotechnology Co., Ltd., Beijing, China). All staining was observed under an Eclipse TE2000-U microscope (Nikon, Tokyo, Japan) after counterstaining with DAPI. The corneoscleral rings of healthy donor human corneal grafts, of which the central parts were used for penetrating keratoplasty, and whole donor human corneas from age-matched patients with diabetes were used for the staining of MANF. The method of human corneal immunofluorescence staining was the same as that of mouse. This study was approved by the ethics committee of Shandong Eye Institute and strictly adhered to the guidelines of the Declaration of Helsinki.

Mouse eyeballs were collected and fixed in Zamboni fixative for 2 h, with the cornea dissected around the scleral-limbal region. The cornea was blocked in PBS containing 0.1% Triton X-100, 2% goat serum, and 2% BSA for 2 h and subsequently incubated in the same buffer with an Alexa Fluor 488–conjugated neuronal class III β-tubulin antibody (Merck Millipore, Darmstadt, Germany) overnight at 4°C. After washing each cornea six times, the flat mounts were examined under a confocal microscope (Zeiss, Rossdorf, Germany), with the cornea cut into six petals.

Diabetic mice and age-matched normal mice were subjected to 2.5-mm central corneal epithelial debridement. Mouse corneal epithelial cells were collected at 48 h after wounding. Corneal epithelial cells from six corneas of each group were homogenized in 250 μL of cold PBS and analyzed using a mouse MANF ELISA kit (Aviva Systems Biology, San Diego, CA) according to the manufacturer’s procedures. Total protein concentrations were calculated using the Bicinchoninic Acid Kit (Beyotime, Shanghai, China).

Statistical analyses were performed using SPSS version 19.0 software (IBM Corporation, Chicago, IL). Data are presented as the mean ± SD. Experiments with two treatments and/or conditions were analyzed using a two-tailed Student t test. Experiments with more than two groups were analyzed using one-way ANOVA, and if the experiments involved two independent variables (e.g., normal and diabetic, unwounded and wounded), a two-way ANOVA was used (8). Differences were considered statistically significant at P < 0.05. Experiments were repeated at least three times to ensure reproducibility.

The data sets and resource generated and/or analyzed during the current study are available from the corresponding author upon reasonable request.

MANF is widely expressed in mouse tissues throughout embryonic development and in adulthood (33). To determine the MANF expression profile in normal and diabetic corneal epithelium during corneal wound healing, MANF expression in epithelial cells from unwounded and wounded corneas of both normal and diabetic mice was detected by qRT-PCR, ELISA, Western blotting, and immunofluorescence staining. As shown in Fig. 1A, qRT-PCR assays revealed that the level of MANF mRNA was significantly upregulated at 6 h postwounding and then downregulated and decreased to a lower level than that in normal corneal epithelium at 48 h postwounding. Of note, the expression of MANF in diabetic corneal epithelium was significantly lower than that in normal corneal epithelium at any time point. Furthermore, we found that MANF protein level was increased significantly at 6 h postwounding in both normal and diabetic corneal epithelium compared with unwounded corneal epithelium; however, lower protein expression of MANF was detected in diabetic corneal epithelium than in normal corneal epithelium (Fig. 1B and C). The ELISA results showed that MANF protein concentrations in normal and diabetic corneal epithelium decreased significantly from 432.6 ± 23.8 pg/mg and 308.0 ± 16.4 pg/mg, respectively, in unwounded corneal epithelium to 345.3 ± 21.2 pg/mg and 147.2 ± 19.8 pg/mg, respectively, in regenerated corneal epithelium after 48 h postwounding (Fig. 1D). Moreover, Western blot analysis confirmed the expression patterns determined by qRT-PCR and ELISA, with lower MANF protein expression in diabetic mice than in normal mice, both in intact and in regenerated (48 h postwounding) corneal epithelium (Fig. 1E and F). Immunofluorescence staining of mice showed that MANF was broadly localized in the corneal epithelium, stroma, and endothelium layers of normal corneas but was lower in diabetic corneas. However, corneal epithelial injury (48 h postwounding) caused an apparent decrease of corneal MANF expression, especially in diabetic corneas (Fig. 1G). In addition, we observed the MANF expression in both normal and diabetic human corneal epithelium by immunofluorescence staining. We determined a higher MANF expression in normal human corneal epithelium than in diabetic human corneal epithelium (Fig. 1H).

Given its low expression in wounded corneal epithelium at 48 h, the potential role of MANF in corneal epithelial wound healing under normal and diabetic conditions were detected. The central corneal epithelium was scraped from the eyes of age-matched normal and diabetic mice, with or without exogenous administration of MANF. In preliminary experiments, we confirmed that the optimum concentration of MANF was 10 μg/mL (Supplementary Fig. 1). MANF supplementation significantly accelerated epithelial wound closure both in normal and in diabetic corneas (Fig. 2).

We next assessed the effect of rhMANF on corneal nerve regeneration, finding that MANF also dramatically promoted nerve regeneration in both normal and diabetic corneas (Fig. 3A and B). Consistent with its effects on nerve regeneration, MANF significantly improved corneal sensation recovery in normal and diabetic mice (Fig. 3C). We further found that MANF upregulated the canonical neurotrophic factors NGF and BDNF, which played important roles in corneal nerve regeneration (2,34) in regenerated diabetic corneal epithelium (Fig. 3D–F).

MANF is mainly localized in the ER, and the protective effect of MANF has been proposed to depend on its ability to dampen ER stress (21–23). To explore the underlying mechanisms whereby MANF promotes corneal epithelial repair and corneal nerve regeneration, we investigated whether MANF regulated ER stress in the corneal epithelium during corneal epithelial wound healing. The ER stress markers transcription factor XBP-1s and CHOP were elevated in regenerated diabetic corneal epithelium compared with normal corneal epithelium, suggesting that excessive ER stress was induced by hyperglycemia in diabetic mice (Fig. 4A–C). As expected, the levels of XBP-1s and CHOP were alleviated by MANF administration both in normal and in diabetic corneal epithelium, and the effect in diabetic cornea was more pronounced (Fig. 4A–C). CHOP is known to act as a main mediator of ER stress–induced apoptosis (35); therefore, we further detected the protective effect of MANF on ER stress–mediated apoptosis in regenerated corneal epithelial cells. The results showed that cleaved caspase 12, a caspase that activates ER stress–mediated apoptosis (36), was heightened in diabetic mice but was significantly inhibited by MANF (Fig. 4D and E). We also found that expression of the apoptosis regulator Bax was decreased by MANF in both normal and diabetic corneal epithelium (Fig. 4D and F). In contrast, the expression of Bcl-2, a cellular protein that inhibits apoptosis, decreased significantly in the diabetic corneal epithelium but was increased by MANF both in normal and in diabetic corneal epithelium (Fig. 4D and G). These results suggest that MANF suppresses hyperglycemia-induced ER stress and ER stress–mediated apoptosis during corneal epithelial wound healing.

To clarify the role of ER stress during corneal wound healing, 4-PBA was injected subconjunctivally, and its effects on corneal epithelial closure and nerve regeneration were assessed. In preliminary experiments, we confirmed that the optimum concentration of 4-PBA was 20 mmol/L (Supplementary Fig. 2). It was observed that 4-PBA accelerated reepithelialization after corneal epithelial injury in both normal and diabetic corneas when observed at 24, 36, and 48 h postscraping (Fig. 5A and B). Similar to MANF, 4-PBA also ameliorated normal and diabetic corneal nerve regeneration (Fig. 5C and D). Moreover, 4-PBA improved corneal sensitivity, especially in diabetic corneas (Fig. 5E), and decreased the levels of cleaved caspase 12 and Bax while increasing Bcl-2 expression in normal and diabetic corneas (Supplementary Fig. 3). These findings indicate that ER stress exerts a negative effect during corneal epithelial wound healing and hyperglycemia-induced ER stress, which may contribute to delayed corneal epithelial wound healing and impaired nerve regeneration in diabetic corneas.

To elucidate the signaling mechanism underlying the MANF-dependent promotion of corneal epithelial wound healing and nerve regeneration, we investigated the effect of MANF on the activation of Akt, which is a well-known signaling factor that plays important roles in tissue repair. Western blot analysis showed that p-Akt levels decreased in diabetic corneal epithelial cells and significantly increased in both normal and diabetic regenerated corneal epithelium following MANF exposure (Fig. 6A and B). Inhibiting Akt by subconjunctival injection of an Akt inhibitor, triciribine, reversed the effects of MANF in promoting wound healing in normal and diabetic corneal epithelial cells (Fig. 6C and D) and nerve regeneration (Fig. 6E and F). The data were confirmed by an Akt knockdown experiment using Akt-specific siRNA (Supplementary Fig. 4). We also found that 4-PBA markedly upregulated p-Akt levels in regenerated normal and diabetic corneal epithelium (Fig. 7A and B). Akt suppression or knockdown also inhibited the stimulatory effect of 4-PBA on diabetic corneal epithelial wound healing (Fig. 7C and D and Supplementary Fig. 5A and B) and nerve regeneration (Fig. 7E and F and Supplementary Fig. 5C and D). Therefore, MANF and 4-PBA may accelerate healing of the corneal epithelium and nerve repair by activating Akt signaling.

To further confirm the critical role of MANF in corneal wound healing and nerve regeneration, siRNA was used to knock down MANF expression. Western blot analysis showed significant MANF protein downregulation in normal and diabetic corneas after subconjunctival injection of MANF siRNA (Fig. 8A and B). Corneal epithelium scratching experiments indicated that MANF knockdown inhibited epithelial wound closure and extended the healing time in both normal and diabetic corneas (Fig. 8C and D). Whole-mount staining of β-tubulin revealed that MANF siRNA impaired nerve regeneration during wound healing in both normal and diabetic corneas (Fig. 8E and F).

Diabetes is an important chronic disease, affecting multiple people and many different organs. The main clinical manifestations of patients with DK are the delayed corneal epithelial wound healing and decreased corneal sensitivity (37). It is well documented that diabetes causes damage to corneal epithelial wound healing and nerve regeneration (1–9). However, the related mechanisms are not completely understood and lack of effective clinical treatment methods at present. In this study, we investigated the roles of MANF and its associated mechanisms in corneal epithelial wound healing and nerve regeneration. We found that MANF was expressed in mouse and human corneas. We also detected the expression changes of MANF during corneal epithelial wound healing and found that under the stress of corneal epithelial injury, the mRNA level of MANF was upregulated in an earlier stage and then decreased, eventually to below normal levels. Interestingly, the expression of MANF in diabetic corneal epithelium was markedly lower than that in normal corneal epithelium. Subsequently, by Western blot detection of MANF expression in corneal epithelium at 6 h and 48 h postwounding, we also found similar changes in pattern at the protein level. It has been reported that MANF plays an important role in neurodegenerative diseases, such as Alzheimer disease and Parkinson disease (19,22,24). According to these results, we speculate that MANF is upregulated as a stress response factor in the early stage of corneal epithelial wound healing in the emergency of injury; in the subsequent process of corneal epithelial wound repair, MANF acts more as a neurotrophic factor to promote corneal epithelial and nerve repair. However, diabetic corneal epithelium showed an insufficient supply of MANF. Research has found that MANF is indispensable for the proliferation and survival of pancreatic β-cells and that MANF deficiency in mice leads to diabetes (28,29,31). In the retina, MANF was shown to induce the repair of damaged retina in flies and mice (38). In the cornea, we found that supplementation with rhMANF ameliorated epithelial wound healing, restored corneal sensitivity, and increased nerve regeneration in both diabetic and nondiabetic corneas. We found that MANF promoted classic neurotrophic factors BDNF and NGF expression, which may represent one of the mechanisms whereby MANF plays a protective role. We also tested whether MANF played a role in corneal wound healing using a complementary approach: siRNA-mediated downregulation. MANF-specific siRNA was injected subconjunctivally. MANF downregulation resulted in larger wound sizes and decreased nerve regeneration.

MANF expression is mainly localized to the ER (39) and can help to maintain the function of the ER (23,40). ER stress is reported to orchestrate major neurological syndromes, such as Alzheimer disease and Parkinson disease (41,42). Pathogenic ER stress also leads to neurodegenerative diseases, diabetes, and diabetic peripheral neuropathy (14,43). Our results indicated that diabetic corneas undergo more severe ER stress than normal corneas after wounding. This strong ER stress may be caused by long-term hyperglycemic stimulation. Supplementation with rhMANF effectively relieved ER stress in diabetic corneas. We also used ER inhibitors to further verify the effects of ER stress on corneal wound healing. The results revealed that the inhibition of ER stress ameliorated corneal epithelial wound healing and promoted nerve regeneration in diabetic mice.

Our findings revealed that MANF plays a protective role in corneal wound healing and nerve regeneration by inhibiting ER stress. We further explored the mechanism of tissue damage induced by ER stress. ER stress plays a role in inducing proteotoxic apoptosis (44). In diabetic retinopathy, ER stress induces various inflammatory and apoptotic pathways, leading to retinal cell death during prolonged diabetic conditions that exacerbate retinopathy (18,45). In diabetic corneas, cells undergo more severe cellular apoptosis after wounding. ER stress induces caspase 12–mediated apoptotic cell death (46). Our Western blot results showed high expression of cleaved caspase 12 in diabetic corneas at 48 h after removing the epithelium. However, low expression of the cleaved caspase 12 protein was observed in normal corneas. The application of rhMANF and 4-PBA downregulated cleaved caspase 12 and Bax expression while upregulating Bcl-2 expression in both diabetic and normal corneas. The results indicated that diabetic mouse corneas undergo more serious apoptosis after wounding, while MANF reduced apoptosis in both diabetic and normal corneas. In normal mice, corneal epithelial debridement wounds may act as a kind of stimulator of ER stress. At 48 h postwounding, a low degree of apoptosis existed in normal corneas, which explains why MANF was effective in normal corneas. Hence, we suggest that MANF is an effective factor in suppressing corneal apoptosis induced by ER stress in diabetes.

It is well known that intracellular signal-transduction pathways play important roles in maintaining cellular functions. Previously, it was reported that ER damage is responsible for insulin resistance in HepG2 cells, while enhancing p-Akt restored insulin activity (47). Our previous findings confirmed that p-Akt promotes the migration of corneal epithelial stem/progenitor cells and repair of the corneal epithelium (32). Here, we show that low levels of p-Akt were present in diabetic corneas. Adding rhMANF and 4-PBA upregulated p-Akt production after wounding in both diabetic and normal corneas. In addition, inhibition or knockdown of Akt delayed epithelial wound healing and nerve regeneration rescued by rhMANF and 4-PBA. These results suggest that the Akt signaling pathway plays an important role in rhMANF-induced protection.

Taken together, our study demonstrates, for the first time, that mouse and human corneas express MANF. Exogenous supplementation of MANF promotes diabetic corneal epithelial wound healing and nerve regeneration by attenuating hyperglycemia-induced ER stress through the Akt signaling pathway. In light of the projected incidence of diabetes, which is expected to increase tremendously by the next decade, it will be important to explore strategies that may prevent diseases associated with diabetes (48). Corneal wound healing and nerve regeneration in patients with diabetes is a challenge that needs to be faced. NGF has been proven effective in clinical treatment for neurotrophic keratitis (49,50), suggesting that neurotrophic factors may have clinical application prospects. Here, our findings suggest that exogenous application of MANF may provide a potential strategy for ameliorating delayed corneal epithelial wound healing and impaired nerve regeneration in patients with DK. Of course, whether MANF can really play a role in the clinic requires much more work in the future.

Acknowledgments. The authors thank Ya Li, Jie Li, Yangyang Zhang, Xiaomin Liu, and Lianghui Zhao (Shandong Eye Institute) for help with the animal tissue collection, statistical analysis, and instrument use.

Funding. We acknowledge support from the National Natural Science Foundation of China (81670828, 81570820, 81800805, 81530027), the Innovation Project of Shandong Academy of Medical Sciences, and the Academic promotion programme of Shandong First Medical University (2019ZL001). L.Y. is partially supported by the Taishan Scholar Project of Shandong Province (tsqn20161059).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. Xiaoc.W. designed and performed the experiments and wrote the manuscript. W.L. contributed to the sample collection and data analysis. Q.Z. was responsible for study design and reviewed the manuscript. J.L., X.Q., T.L., X.Z., and S.L. contributed to the sample collection. Xiaol.W. and J.Z. were responsible for antibody selection. D.L. contributed to the data analysis. L.Y. and L.X. designed the experiments and revised the manuscript. L.Y. and L.X. are the guarantors of this work and, as such, had full access to all data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.