Novel Long Noncoding RNA lnc-URIDS Delays Diabetic Wound Healing by Targeting Plod1

Diabetic foot is one of the major complications of diabetes, leading to ulcerations and amputations of the lower extremity. Based on epidemiologic studies, as many as one in four patients with diabetes will develop a diabetic foot ulcer (DFU) in their lifetime (1). The most serious, feared, and costly consequence of DFU, but also the most common, is limb amputation, which is 10–20 times more frequent in patients with type 2 diabetes mellitus (T2DM) than those without (2–4). Impaired wound healing is a common cause of DFU and limb amputations. Despite the high incidence and severity of these clinical consequences, the mechanism by which T2DM impairs wound healing remains unclear.

The healing of a wound is a complex process involving well-orchestrated biological and molecular events, such as cell migration and proliferation and extracellular matrix (ECM) deposition and remodeling (5). Dermal fibroblasts have been shown to play a critical role in this biological process. During wound repair, fibroblasts proliferate and migrate to the wound site and reform the ECM, and then the collagen in the dermis is remodeled (6). In diabetic skin, however, the proliferation and migration of fibroblasts was decreased, the collagen fibers were denatured and fractured, and the ratio of collagen I/III was decreased (7,8), which may be the significant pathophysiological changes of diabetic skin during wound healing.

Increasing evidence has shown that epigenetic changes contribute to the etiology of diabetes and its associated complications (9,10). Environmental factors such as hyperglycemia can use epigenetic mechanisms and set a “metabolic memory” phenomenon, resulting in alterations of gene expression, which may lead to the development of diabetic complications (10,11). Noncoding RNAs play an important role in epigenetic regulation (12,13). Long noncoding RNAs (lncRNAs) are a cluster of noncoding RNAs with a length >200 nucleotides that are involved in a variety of biological processes and diseases (14,15). Several studies have demonstrated the roles of lncRNAs in diabetic complications, including diabetic nephropathy, diabetic retinopathy, and diabetic cardiomyopathy (16–18). However, the expression profile and functions of lncRNA in the diabetic foot remain largely unknown.

In this study, we used lncRNA microarray to identify lncRNAs with differential expression between diabetic skin tissue and nondiabetic skin tissue. We found a novel lncRNA, MRAK052872 (named lnc UpRegulated in Diabetic Skin [lnc-URIDS]), that was upregulated in diabetic skin. Through construction of an lncRNA-mRNA coexpression network and enrichment analyses of its coexpressed mRNAs, we predicted that lnc-URIDS might participate in the process of wound healing. Therefore, we studied the functional roles of lnc-URIDS in wound healing under diabetic conditions both in vitro and in vivo and explored the mechanism through which lnc-URIDS regulates diabetic wound healing.

Sprague-Dawley rats were intraperitoneally injected with 60 mg/kg streptozotocin (Sigma-Aldrich, St. Louis, MO) to induce diabetes. Rats were housed for 4 weeks before being anesthetized, and skin tissue was harvested for microarray analysis. Microarray expression profiling was performed using Arraystar 4×44K Rat LncRNA Array v2.0 (Agilent Technologies, Santa Clara, CA) by KangChen Bio-tech (Shanghai, China). Genes with fold change ≥2.0 and a P value <0.05 were considered differentially expressed. The lncRNA-mRNA coexpression analysis was based on calculation of the Pearson correlation coefficient (PCC) between the expression levels of lncRNA and mRNA. For each lncRNA, coexpressed mRNAs were identified based on a PCC ≥0.99. Gene Ontology (GO) biological processes enrichment analysis was performed using DAVID Bioinformatics Resources.

The 5′-rapid amplification of cDNA ends (RACE) and 3′-RACE analyses were performed to determine the transcriptional initiation and termination sites of lnc-URIDS and the full length of the transcript using a SMARTer RACE cDNA Amplification Kit (Clontech Laboratories, Inc., Palo Alto, CA) according to the manufacturer’s instructions. The PCR primers used in RACE are shown in Supplementary Table 1.

Total RNA was extracted using TRIzol reagent (Invitrogen, Carlsbad, CA) according to the manufacturer’s instructions. Relative mRNA expression levels were normalized to the expression of the housekeeping gene β-actin (ACTB). The primers used in real-time PCR are shown in Supplementary Table 1.

Western blot analysis was performed as previously described (19). The antibodies used in Western blot analysis are listed in Supplementary Table 2.

Rat dermal fibroblast cell lines (CRL-1213) were purchased from ATCC (Manassas, VA). Rat primary dermal fibroblasts were isolated by Dispase II (Roche) from skin tissue of newborn Sprague-Dawley rats. All cells were cultured in DMEM containing 1,000 mg/L glucose (Gibco, Gaithersburg, MD) and supplemented with 10% FBS (Gibco) and 1% penicillin-streptomycin (Gibco) at 37°C in a humidified atmosphere containing 5% CO₂. Cells were treated with 200 μg/mL advanced glycation end products (AGEs)–BSA or BSA (Merck Millipore, Temecula, CA) for 12 h.

siRNAs were designed and synthesized by Synbio Technologies (Suzhou, China). Cells were transfected with siRNA or nontargeting siRNA using Lipofectamine RNAiMAX (Invitrogen) according to the manufacturer’s instructions. Rat dermal fibroblasts were infected with lnc-URIDS shRNA lentivirus and control lentivirus (GenePharma, Shanghai, China) at a multiplicity of infection of 10 and screened with a medium supplemented with 2 µg/mL puromycin (Thermo Fisher Scientific). Microarray analysis for the mRNA expression profiles in lentivirus-mediated lnc-URIDS knockdown cells was performed using Rat 4×44K Gene Expression Microarrays (Agilent Technologies). Rat primary dermal fibroblasts were infected with adenovirus-expressing lnc-URIDS and control adenovirus (Hanbio, Shanghai, China) at a multiplicity of infection of 100.

Skin tissue sections were deparaffinized, rehydrated, and deproteinated using proteinase K (20 μg/mL) at 37°C for 25 min. After being prehybridized at 37°C for 1 h, sections were hybridized with digoxigenin-labeled probes for lnc-URIDS (8 ng/μL) at 37°C overnight. Slides were washed and blocked in BSA for 30 min. Then, the sections were incubated with alkaline phosphatase–conjugated anti-digoxigenin antibody (Jackson ImmunoResearch Laboratories, West Grove, PA) at 37°C for 40 min and detected by BCIP/NBT reagent (Boster Bio, Pleasanton, CA). Sections were counterstained with Nuclear Fast Red (Servicebio Technology Co., Ltd., Wuhan, China). Images were observed with an inverted light microscope (Nikon, Tokyo, Japan).

RNA fluorescence in situ hybridization (FISH) assay was performed with the Fluorescence In Situ Hybridization Kit (RiboBio Co., Guangzhou, China) according to the manufacturer’s instructions. Images were observed with a confocal laser-scanning microscope (Carl Zeiss, Oberkochen, Germany).

Rat collagen I and III in the supernatant of cultured fibroblasts were measured with ELISA kit (Cusabio Technology LLC, Wuhan, China). Wound-healing assay was measured by using a wound-healing culture insert (ibidi GmbH, Martinsried, Germany) according to the manufacturer’s instructions. For the transwell assay, cells were resuspended in a serum-free medium and added to the upper face of a transwell chamber (24 wells, 8 μmol/L pore size). Images of five random fields were captured by a light microscope (Nikon) from each membrane, and the number of migrated cells was counted.

Eight-week-old male Sprague-Dawley rats (weighing 250–300 g) were purchased from the Laboratory Animal Center of the Sun Yat-sen University. A skin wound was made as previously described (19). To explore the effect of lnc-URIDS on wound healing in normal rats, full-thickness excisional wounds were created on the dorsum of the nondiabetic Sprague-Dawley rats. Rats were randomly divided into three groups (n = 6/group) receiving 1) 100 μL PBS (blank), 2) 100 μL of 10¹¹ plaque-forming units (PFU)/mL control adenovirus (Ad-NC), and 3) 100 μL of 10¹¹ PFU/mL adenovirus-expressing lnc-URIDS (Ad-lnc-URIDS). To study the effect of lnc-URIDS on diabetic wounds, rats were intraperitoneally injected with 60 mg/kg streptozotocin and allowed to manifest hyperglycemia for 4 weeks before making a cutaneous wound. Diabetic rats were randomly divided into three groups (n = 4–6/group) receiving 1) 100 μL PBS (blank), 2) 100 μL of 10¹¹ PFU/mL control adenovirus (Ctrl RNA interference [RNAi]), and 3) 100 μL of 10¹¹ PFU/mL adenovirus-expressing lnc-URIDS shRNA (lnc-URIDS RNAi). Respective treatments were injected once around the wounds (four injection sites 5 mm from the edge of wound, 25 μL/site) immediately after wounding. Images of the wounds were digitally photographed at days 0, 4, 7, and 10 postwounding. The skin specimens excised 10 days postwounding were stained with hematoxylin and eosin (HE), Masson trichrome, and Sirius Red. Wound length and wound areas were measured using ImageJ software, and wound-closure rate was calculated. Immunohistochemistry was performed as previously described (20). All animal studies were approved by the Animal Research Committee of Sun Yat-sen University (Guangzhou, China).

Biotinylated lnc-URIDS were transcribed in vitro with the TranscriptAid T7 High Yield Transcription Kit (Thermo Fisher Scientific) followed by RNase-free DNase I treatment at 37°C for 15 min. Biotinylated RNAs were added to washed streptavidin-coupled Dynabeads (Invitrogen) and further incubated at 4°C for 2 h. Whole-cell lysates from fibroblasts were incubated with RNAbeads for 2 h at 4°C. Beads were washed three times and then boiled in loading buffer for 10 min. The samples were separated by 12% SDS-PAGE gel, and protein bands were analyzed by mass spectrometry (MS). ProteinSimple capillary electrophoresis immunoassay was performed to confirm the interaction between target protein and lnc-URIDS in fibroblasts.

Demographic information and laboratory parameters were recorded for patients with T2DM with or without foot ulcer (Supplementary Table 3). The grade of DFU was assessed using the Wagner classification. The study conformed to the ethical principles outlined in the Declaration of Helsinki and was approved by Ethics Committee of Sun Yat-sen Memorial Hospital, Sun Yat-sen University (Guangzhou, China).

Quantitative data were presented as the mean ± SD from at least three independent experiments. Comparisons between the two groups were performed with the Student t test. For data that are not normally distributed or data that do not have the same variance, variables were log transformed before statistical analysis. Multiple-group comparisons were analyzed with one-way ANOVA followed by a least significant difference post hoc test. A P value <0.05 was considered statistically significant. All statistical analyses were performed with SPSS 24.0 (IBM, Armonk, NY).

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

To identify the functional lncRNAs involved in diabetic wound healing, the differential expression profiles of lncRNAs and mRNAs in diabetic skin and nondiabetic skin were analyzed through microarray and shown in the heat maps (Fig. 1A and B). Based on the preliminary screening of gene expression abundance, lncRNA length, coding capability, relationship of lncRNA and its nearby coding gene, and conservation (details of the criteria are shown in Supplementary Table 4), we identified seven upregulated lncRNAs: MRAK032934, MRAK051933, MRAK052872, MRAK080592, uc.309-, uc.334-, and uc.436- (Fig. 1A). These were selected for further study.

In order to predict the function of these lncRNAs, an lncRNA-mRNA coexpression network among these seven upregulated lncRNAs and dysregulated mRNAs was performed, and the biological functions of these lncRNAs were predicted through GO biological processes enrichment analyses of their coexpressed mRNAs (Fig. 1C and Supplementary Fig. 1). We found only lncRNA MRAK052872-coexpressed protein-coding genes showed significant enrichment in the process of wound healing (GO:0042060) (Supplementary Table 5). Interestingly, MRAK052872 was predicted to be involved in the pathophysiological changes during wound healing, such as inflammatory response, immune response, cell proliferation, and migration (Fig. 1D). Therefore, we identified MRAK052872 as a candidate lncRNA associated with wound healing and named it lnc-URIDS. Real-time PCR analysis validated that the expression levels of lnc-URIDS were significantly upregulated in diabetic skin (Fig. 1E). In patients with T2DM, lnc-URIDS levels in serum were dramatically increased in patients with Wagner IV foot ulcer (Supplementary Fig. 2).

lnc-URIDS is located on chromosome Xq31 in rats and was found to be a 1497-nucleotide transcript with a poly(A) tail comprised of only one exon (Fig. 2A and B). Sequence analysis using ORF Finder failed to predict an open reading frame >200 bp (Supplementary Fig. 3A). In addition, the coding potential calculator score of lnc-URIDS was −1.04, supporting the fact that lnc-URIDS had no protein-coding potential (Supplementary Fig. 3B). Although in general, lncRNAs are poorly conserved across species, genome sequences alignment revealed that rat lnc-URIDS had homologs in humans (Supplementary Fig. 4A and B). The UCSC Genome Browser showed that lnc-URIDS was conserved among placental mammals (Fig. 2B). We also performed RACE to identify the full length of human homologous sequences of lnc-URIDS in human fibroblasts (Supplementary Fig. 4C). Multiple sequence alignment of lnc-URIDS and the preservation of sequence are shown in Supplementary Fig. 4D.

To validate whether lnc-URIDS was expressed in multiple tissues, we examined the expression of lnc-URIDS in the heart, liver, spleen, lung, kidney, and pancreas in nondiabetic and diabetic rats. As shown in Fig. 3A, lnc-URIDS was found to be widely expressed in multiple organs and tissues. However, the expression of lnc-URIDS was only significantly increased in the skin tissue of diabetic rats. ISH assay revealed that lnc-URIDS was localized in the dermal layer of skin and mainly expressed in the dermal fibroblasts (Fig. 3B). The expression and subcellular location of lnc-URIDS in fibroblasts treated with AGEs were further assayed. As shown in Fig. 3C, the expression levels of lnc-URIDS were significantly increased in rat dermal fibroblasts treated with AGEs-BSA. FISH assay demonstrated that lnc-URIDS was predominantly distributed in the cytoplasm of fibroblasts, and its expression was markedly upregulated when cells were exposed to AGEs-BSA (Fig. 3D).

To further explore the functional roles of lnc-URIDS in dermal fibroblasts, we used lentivirus-mediated RNAi to establish stable lnc-URIDS knockdown dermal fibroblasts (lnc-URIDS RNAi) and control cells (Ctrl RNAi) and performed mRNA microarray to measure gene expression profiling after lnc-URIDS knockdown (Fig. 4A and B). Results from microarray analysis showed that a total of 151 mRNAs were identified to be differentially expressed (with fold change ≥2.0 and P < 0.05) in lnc-URIDS knockdown fibroblasts, of which 75 were upregulated and 76 were downregulated. The GO biological processes enrichment analysis revealed that the dysregulated genes in lnc-URIDS knockdown cells showed a notable enrichment in the process of “wound healing” (GO:0042060) and “response to wounding” (GO:0009611) (Fig. 4C), suggesting that lnc-URIDS regulated the expression of genes associated with wound healing. The mRNA levels of differentially expressed genes that were enriched in wound-healing biological process in the microarray assay were further verified by real-time PCR (Supplementary Fig. 5).

Impaired migration of fibroblasts is one of the most notable pathological changes during diabetic wound healing. To examine the effect of lnc-URIDS on cell migration, we performed wound healing and transwell assay. The results showed that the migration of AGEs-treated fibroblasts was significantly increased after lnc-URIDS knockdown (Fig. 4D and E). These data indicated that lnc-URIDS knockdown ameliorated phenotypes associated with diabetic dermal fibroblasts in vitro.

We further investigated the effect of lnc-URIDS on wound healing in vivo. As shown in Fig. 5A and B, wound healing was significantly delayed in the Ad-lnc-URIDS group compared with Ad-NC group (day 10: 85.93 ± 3.39% in Ad-lnc-URIDS group vs. 94.95 ± 1.51% in Ad-NC group; P < 0.01). Rats treated with Ad-lnc-URIDS exhibited a wider wound compared with rats treated with Ad-NC at day 10 postwounding (Fig. 5A and B). Moreover, Masson staining and Sirius Red staining demonstrated that the deposition of collagen was sparse and the ratio of collagen I/III was decreased in the Ad-lnc-URIDS treatment group in comparison with Ad-NC groups (Fig. 5C and D).

To explore whether lnc-URIDS regulated diabetic wound healing, a well-established animal model was used to measure cutaneous wound healing in diabetes. The results showed that inhibition of lnc-URIDS significantly accelerated diabetic wound healing (day 10: 90.81 ± 4.04% in lnc-URIDS RNAi group vs. 84.60 ± 2.09% in Ctrl RNAi group; P < 0.01) (Fig. 6A and B). Suppressed lnc-URIDS contributed to a small-length wound (Fig. 6A and B) and increase in deposition of collagen and the ratio of collagen I/III in diabetic skin (Fig. 6C and D). In addition, the expression of lnc-URIDS in skin decreased during the wound healing period and was significantly upregulated in diabetic rats compared with nondiabetic rats (Fig. 6E).

To further investigate the mechanism by which lnc-URIDS regulates wound healing and collagen deposition, we performed an RNA pulldown assay to identify the lnc-URIDS binding proteins in rat primary dermal fibroblasts. RNA-associated proteins were resolved on SDS-PAGE gel (Fig. 7A), followed by MS to search for lnc-URIDS binding proteins. Procollagen-lysine, 2-oxoglutarate 5-dioxygenase 1 (Plod1), a critical enzyme that catalyzes the hydroxylation of certain lysine residues in collagen α chains, was identified as an lnc-URIDS–interacting protein. The interaction was also verified by Western blot analysis (Fig. 7B). Knockdown of Plod1 by siRNA significantly suppressed the protein levels and secretion of collagen I in primary dermal fibroblasts (Fig. 7C–E). However, no significant difference was found in the levels of collagen III after Plod1 knockdown. Therefore, the ratio of collagen I/III was decreased. Overexpression of lnc-URIDS showed no effect on the mRNA levels of Plod1 (Supplementary Fig. 6), but resulted in a decrease in Plod1 protein levels (P = 0.057) (Fig. 7F and G). The expression and secretion of collagen I was also reduced after overexpression of lnc-URIDS (Fig. 7F–H). Similarly, collagen III levels were not changed, and the ratio of collagen I/III was decreased. The expressions of Plod1 in the skin tissue of rats were also analyzed (Supplementary Fig. 7). Overexpression of lnc-URIDS in nondiabetic rats inhibited the expression of Plod1 protein. In diabetic skin, the protein levels of Plod1 were increased after lnc-URIDS knockdown (Supplementary Fig. 7A and B). However, lnc-URIDS overexpression or knockdown did not alter the mRNA levels of Plod1 in skin (Supplementary Fig. 7C). These results indicated that lnc-URIDS negatively regulated the protein expression of Plod1 in skin wounds.

To test whether lnc-URIDS affected Plod1 protein stability, rat primary dermal fibroblasts were treated with proteasome inhibitor (MG132) or protein synthase inhibitor (cycloheximide [CHX]) with different incubation times. As shown in Fig. 7I, CHX treatment significantly decreased Plod1 protein expression (Fig. 7I, left panel). The addition of MG132 in the vector group restored Plod1 protein expression to a level that was comparable to that in untreated cells, but only partially rescued Plod1 protein expression after lnc-URIDS overexpression (Fig. 7I, right panel). These data suggested that the interaction between lnc-URIDS and Plod1 reduced Plod1 protein stability, which might be a key factor in how lnc-URIDS affects wound healing.

In the current study, we identified a novel lncRNA, lnc-URIDS, that regulated cutaneous wound healing in diabetes. We found that lnc-URIDS was upregulated in diabetic skin and dermal fibroblasts treated with AGEs. Inhibition of lnc-URIDS promoted migration of AGEs-treated dermal fibroblasts in vitro and accelerated diabetic wound healing in vivo. Moreover, lnc-URIDS regulated collagen production and deposition by targeting Plod1 and contributed to the delay of wound healing. Our findings indicated that lnc-URIDS played an important role in the pathogenesis of diabetic wound healing.

Recently, lncRNAs are emerging as critical regulators of physiological as well as pathological processes. Several studies have demonstrated the roles of lncRNA in diabetes and its complications. lncRNAs functioned in concert with transcription factors and regulated β-cell–specific transcriptional networks (21,22). The regulatory roles of lncRNAs were also observed in diabetic nephropathy (16), diabetic cardiomyopathy (18), diabetic neuropathic pain (23), and diabetes-induced microvascular dysfunction (17). In the current study, we investigated the functional roles of lncRNAs in diabetic wound healing. Our study identified that lnc-URIDS was associated with diabetic wound healing by performing lncRNA microarray and constructing of the lncRNA-mRNA coexpression network. Recently, several studies also demonstrated that lncRNA was involved in diabetic wound healing. A study found that extracellular vesicle–mimetic nanovesicles transporting lncRNA-H19 would be a potential treatment for diabetic wounds (24). The authors observed that extracellular vesicle–mimetic nanovesicles with a high content of lncRNA-H19 neutralized the regeneration-inhibiting effect of hyperglycemia and significantly accelerated the healing processes of chronic wounds. Another study demonstrated the specific role of lncRNA MALAT1 in wound healing (25). MALAT1 promoted fibroblast activation and wound healing in diabetic mice via activation of the hypoxia-inducible factor-1α signaling pathway. In the current study, we identified a novel lncRNA (lnc-URIDS) and investigated its roles in diabetic wound healing in vitro and in vivo. lnc-URIDS knockdown enhanced migration of dermal fibroblasts under diabetic conditions (AGEs treatment). In an animal model, subcutaneous injection of adenovirus vector expressing lnc-URIDS shRNA accelerated diabetic wound healing and increased collagen content in diabetic skin tissues.

Production and disposition of ECM is essential for the repair of a wound (5,26). During wound healing, fibroblasts produce ECM mainly composed of collagen, which provides a supporting structure for cell proliferation and migration, restore the function and integrity of the skin tissue, and sustain tissue resilience and strength (5,26–28). Thus, the increase and stabilization of collagen fibers are essential during wound healing. Collagen type I and type III are the most abundant collagens in the dermis of skin (29). During the formation and remodeling phases of wound healing, the amount and proper ratio of synthesized and deposited collagen type I and type III are of great impact on connective tissue quality (5,30,31). A reduced collagen I/III ratio was associated with change in fibril diameter and reduction in stability of connective tissue, which might contribute to the weakness of skin (8,32). In our previous study, we found that the collagen fibers were denatured and fractured, and the ratio of collagen I/III was decreased in diabetic skin tissue (7,8). In the current study, we observed that inhibition of lnc-URIDS by subcutaneous injection of adenovirus increased type I collagen content and the ratio of collagen I/III, accompanied by increased wound closure in diabetic skin tissues. However, the mechanism remains unclear. Our further results showed that lnc-URIDS interacted with Plod1, which encodes lysyl hydroxylase 1, a critical enzyme that catalyzes the formation of hydroxylysine in the process of collagen synthesis, assembly, and cross-linking (33,34). It is known that hydroxylysine is one of the basic components of collagen fiber structure. Moreover, it has been reported that the mutations in lysyl hydroxylase prevent the formation of stable hydroxylysine-derived cross-links (34). Several studies have shown that PLOD1 was expressed in multiple tissues, and its expression levels among various tissues did not vary significantly, which was consistent with the steady levels of lysine hydroxylation in the triple-helical domain of type I collagen molecules, indicating that PLOD1 might be responsible for the hydroxylation of lysine residue in collagen I (35). In the current study, we observed that the knockdown of Plod1 in fibroblasts resulted in a significant decrease in protein levels and secretion of collagen I. Therefore, the ratio of synthesized and secreted collagen I/III was decreased after Plod1 knockdown. Overexpression of lnc-URIDS reduced the protein level but not the mRNA level of Plod1, suggesting that lnc-URIDS regulated the expression levels of Plod1 at the posttranscriptional level. It has been demonstrated that lncRNAs could interact with protein and regulate its stability (36,37). Therefore, we further investigated whether Plod1 protein stability was regulated by lnc-URIDS. Our results demonstrated that lnc-URIDS could bind and decrease the stability of Plod1 protein, resulting in a dysregulation of collagen production and deposition.

In summary, the findings in our study have uncovered a novel lncRNA, lncRNA-URIDS, that was highly expressed in diabetic skin and was involved in the process of wound healing in diabetes. lnc-URIDS targets Plod1 and results in a decreased protein stability of Plod1, which leads to the dysregulation of collagen production and deposition and delay in wound healing. The findings of the current study might offer some insight into the potential mechanism for the delay of wound healing in diabetes. lnc-URIDS may be a promising therapeutic target for diabetic foot.

Funding. This study was funded by the National Natural Science Foundation of China (81870571, 81670764, and 81900752) and Science and Technology Planning Project of Guangdong Province, China (2014A020212161 and 2014A020212069).

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

Author Contributions. M.H. and Y.W. designed and performed the experiments, analyzed the data, and wrote the manuscript. C.Y. designed the methods, interpreted the results, and provided intellectual insights. X.W., W.W., L.Z., and T.Z. performed the experiments and collected the data. J.Z., C.W., and G.L. contributed to the discussion and reviewed and edited the manuscript. L.Y. and M.R. design the experiments, discussed analyses and their interpretation, revised the manuscript, and approved the final version of the manuscript. L.Y. and M.R. are the guarantors of this work and, as such, had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.