Dye-decolorizing peroxidases in Irpex lacteus combining the catalytic properties of heme peroxidases and laccase play important roles in ligninolytic system

Background The white rot fungus Irpex lacteus exhibits a great potential in biopretreatment of lignocellulose as well as in biodegradation of xenobiotic compounds by extracellular ligninolytic enzymes. Among these enzymes, the possible involvement of dye-decolorizing peroxidase (DyP) in lignin degradation is not clear yet. Results Based on the extracellular enzyme activities and secretome analysis, I. lacteus CD2 produced DyPs as the main ligninolytic enzymes when grown in Kirk’s medium supplemented with lignin. Further transcriptome analysis revealed that induced transcription of genes encoding DyPs was accompanied by the increased expression of transcripts for H2O2-generating enzymes such as alcohol oxidase, pyranose 2-oxidase, and glyoxal oxidases. Meanwhile, accumulation of transcripts for glycoside hydrolase and protease was observed, in agreement with abundant proteins. Moreover, the biochemical analysis of IlDyP2 and IlDyP1 confirmed that DyPs were able to catalyze the oxidation of typical peroxidases substrates ABTS, phenolic lignin compounds DMP, and guaiacol as well as non-phenolic lignin compound, veratryl alcohol. More importantly, IlDyP1 enhanced catalytic activity for veratryl alcohol oxidation in the presence of mediator 1-hydroxybenzotriazole, which was similar to the laccase/1-hydroxybenzotriazole system. Conclusions The results proved for the first time that DyPs depolymerized lignin individually, combining catalytic features of different peroxidases on the functional level. Therefore, DyPs may be considered an important part of ligninolytic system in wood-decaying fungi. Electronic supplementary material The online version of this article (10.1186/s13068-018-1303-9) contains supplementary material, which is available to authorized users.


Background
Lignin is the second most abundant constituent of lignocellulosic biomass, amounting to 15-30% by weight or up to 40% by energy [1]. The degradation of lignin represents a key step for carbon recycling in the land ecosystems, as well as a critical issue for cost-effective lignocellulosic biofuels and bio-based chemicals [2]. However, due to the complex and random phenylpropanoic polymeric structure, lignin is highly recalcitrant toward chemical and biological degradations [3], resulting in lignocellulosic waste and environment pollution. White rot fungi, a large group of wood-decaying basidiomycetes, are able to completely decompose lignin into carbon dioxide and water by extracellular ligninolytic enzymes, which include an array of heme peroxidases and oxidases [4]. Among them, the heme peroxidases, such as manganese peroxidase (MnP), versatile peroxidase (VP), lignin peroxidase (LiP), and laccase (Lac) have been considered to play important roles in lignin degradation [5][6][7].
Dye-decolorizing peroxidase (DyP) is a member of the novel heme peroxidase family (DyP-type peroxidase superfamily), showing no homology to classic fungal heme peroxidases including MnP, VP, and LiP [8,9]. So far only eleven fungal DyPs have been purified and characterized [10]. Compared with classic fungal heme peroxidases, the specific feature of all DyP is the ability to oxidize synthetic high redox potential dyes of the anthraquinone type [8]. DyP can oxidize phenolic compounds, such as 2,6-dimethoxyphenol and guaiacol [7]. Recently, there are a few reports about its catalytic ability to non-phenolic lignin model compound veratryl alcohol (VA) and Mn 2+ , which is attributed to high redox potential peroxidase LiP/VP and MnP, respectively [7,10,11]. These findings indicate that DyP might be an important part of ligninolytic system in white rot fungi, although biological roles of DyP are ambiguous in terms of different substrate specificities.
Irpex lacteus is a white rot fungus with a significant potential for various biotechnological applications such as bioremediation of organopollutants in water and soil environments and biopretreatment of lignocellulose [12,13]. Its biotechnological applications were attributed to the extracellular ligninolytic enzymes, including MnP, LiP, laccase-like, and DyP [14][15][16][17][18]. Our preliminary work demonstrated I. lacteus CD2 could degrade all kinds of lignocellulose and dyes [13][14][15]. Genome analysis reveals that I. lacteus CD2 has seven mnp genes, two lip genes, and four dyp genes, without lac gene [14]. Compared with MnP, the main ligninolytic enzyme of I. lacteus, DyP is scarcely known for catalytic properties and substrate specificities, especially in lignin degradation. Herein, the main ligninolytic enzymes of I. lacteus CD2 grown in lignin medium were determined, combining extracellular enzyme activities and secretome analysis. Furthermore, the mechanisms of lignin degradation by the main ligninolytic enzymes DyPs were elucidated using transcriptomics and biochemical analysis.

Major extracellular proteins and ligninolytic enzymes of I. lacteus in lignin medium
The SDS-PAGE analysis showed there were seven main bands within 30-75 kDa in extracellular proteins of I. lacteus in lignin medium (Fig. 1a). At the same time, extracellular ligninolytic enzyme activities were detected (Fig. 1b). There was no significant difference between total ligninolytic enzymes activity and manganese-independent peroxidases (MiPs) activity, which suggested MiPs were the main extracellular ligninolytic enzymes of I. lacteus CD2 in lignin medium, and MnP activity might be weak or negligible. DyPs are one of MiPs and showed the similar trend to that of MiPs in I. lacteus CD2 grown in lignin medium (Fig. 1a), suggesting that DyPs might be the main extracellular ligninolytic enzymes of I. lacteus CD2 in lignin medium. MiPs including DyP were rapidly induced since lignin and lignin-derived aromatic compounds were the most efficient inducers of ligninolytic enzymes [19]. In this study, MiPs activities increased rapidly on day 3 and obtained peaks on day 5, with maximal activities of 144.9 U/L and 41.1 U/L against ABTS and RB19, respectively (Fig. 1a). The change of protein contents was consistent with extracellular ligninolytic enzymes activities over time. In order to determine the corresponding MiPs produced in lignin degradation, the seven gel bands on day 3 in SDS-PAGE were recovered and subjected to nano LC/ MS-MS analysis. The results confirmed that DyPs were the main extracellular ligninolytic enzymes of I. lacteus in lignin medium (Table 1, Additional file 1). In parallel, I. lacteus secreted different oxidoreductases including aryl alcohol oxidase (AAO), copper radical oxidase, and cellobiose dehydrogenase (CDH) ( Table 1), which can generate H 2 O 2 required for the action of extracellular peroxidases [20].
In addition to ligninolytic enzymes, some proteins involving in fungal growth, such as glycosidase hydrolases (including chitinases, glucanases, mannosidases and so on) and proteases were identified in the extracellular proteins. Chitinases, glucanases, and mannosidases are involved in hyphal cell wall biosynthesis [21]. Proteases such as polyporopepsin, aspartic protease, and glutaminase are implicated in protein degradation and supplying nitrogen for fungal growth [22,23]. Glycoside hydrolases are essential for cell wall synthesis and cell wall integrity, involving in protein maturation and transport, such as N-linked glycoproteins processing or carbohydrate structural degradation [24].

Comparative transcriptome analysis of I. lacteus grown in lignin and glucose medium
A total of 10,167 genes were determined from the transcriptomes of four samples: LIG3d (3 days in lignin), GLU3d (3 days in glucose), LIG6d (6 days in lignin) and GLU6d (6 days in glucose). To identify key genes and pathways associated with lignin degradation, four pairwise comparisons were performed, including LIG3d versus GLU3d, LIG6d versus GLU6d, LIG6d versus LIG3d , GLU6d versus GLU3d. The results indicated that 4603 and 3816 genes showed at least twofold differences in comparisons of groups LIG3d versus GLU3d, and LIG6d versus GLU6d, respectively, whereas only 300 genes exhibited not less than twofold differences in the comparison of group LIG6d versus LIG3d. Moreover, GO-enrichment analysis between LIG6d versus GLU6d and LIG6d versus LIG3d revealed that these differential expression genes were significantly enriched in lignocellulose-degrading process, including lignin, carbohydrate, polysaccharide, hemicellulose, hydrogen peroxide, and phenylpropanoid metabolic/catabolic process (Fig. 2a). Besides, genes encoding oxidoreductase, heme binding, peroxidase, monooxygenase, and glycoside hydrolase (GH) were also significantly enriched (Fig. 2b). These results were in agreement with that of protein analysis. Since white rot fungi were able to produce different heme peroxidases in synergy with oxidases, and each one might contribute in different ways to the final degradation of lignin [25]. The synergetic effects of different enzymes play vital roles in depolymerizing lignin by I. lacteus CD2.

Ligninolytic enzyme system
Various lignin-degrading peroxidases were significantly differentially expressed along with H 2 O 2 -generation enzymes in I. lacteus in lignin medium (Fig. 3). The transcripts encoding DyPs were highly expressed in LIG3d and LIG6d, which were in agreement with extracellular peroxidases activities and protein profiles. Compared with in GLU3d, IlDyP1 and IlDyP2 in LIG3d were induced by 22.7-and 77.3-fold, then decreased by 1.5-and 3.6-fold in LIG6d, respectively. IlDyP4 gene was upregulated by 9.5-fold and further increased by 1.7fold in LIG6d. Also, one LiP and four MnP transcripts were abundant in LIG3d, especially for IlMnP2. Notably, although the respective activities were not detected in the extracellular culture filtrates, MnP and LiP isoenzymes were still expressed to different extents in different stages. The similar cases were also reported in Pleurotus ostreatus that MnP and Lac with high transcription levels were not found in extracellular proteins, which might be inefficient secretion and the action of specific proteases [26,27]. In addition, transcripts for one alcohol oxidases (AOX) gene, one pyranose-2-oxidase (POX) gene, and two glyoxal oxidase (GLOX) genes were significantly accumulated in LIG3d. AOX gene was upregulated by 2405.1-fold in LIG3d and decreased by 1.4-fold in LIG6d. The membrane-anchored AOX, proposed to be an extracellular source of H 2 O 2 during lignin degradation [20,28], was the most abundant transcript in H 2 O 2 -generation enzymes. POX gene transcript also accumulated in LIG3d and LIG6d. POX was not found in extracellular proteins in some studies, which was attributed to its location in the periplasmic space and associated with membranous materials [29,30]. Besides, IlGLOX1 and IlGLOX2 were also induced by 135.9-and 14.8-fold in LIG3d, respectively. These suggested that AOX, POX and GLOX were implicated in providing H 2 O 2 required for DyP activity.

Carbohydrate metabolism, nitrogen metabolism and related enzymes
In addition to oxidoreductases, genes encoding carbohydrate active enzymes (CAZYs) as well as other proteins were upregulated in lignin medium. 47 CAZYs genes of glycoside hydrolase, carbohydrate esterase (CE) or polysaccharide lyase (PL) families had significantly different transcript abundance in LIG3d relative to GLU3d (Additional file 2). The GHs mainly consisted of cellulose-degrading enzymes (GH6), hemicellulose-degrading enzymes (GH10/GH43/GH51), and pectin-degrading enzymes (GH28/GH78/GH88). These enzymes might be to hydrolyze a small amount of carbohydrate remained in lignin, and required for fungal growth. In particular, genes related to CAZYs including trehalase, chitinase, and mannosidase were upregulated. These enzymes might liberate carbon from major storage compound trehalose and be involved in fungal hyphal cell wall biosynthesis for growth [21,31]. Numerous genes involved in mobilizing and recycling nitrogen were also expressed, including oligopeptide transporter, nucleoside transporter, acetamidase, amino acid permease, amine oxidase, arginase, aspartokinase, methionine synthase, nitrilase, and proteases (Additional file 3). In accordance with the abundance in the extracellular culture filtrates, polyporopepsins, and aspartic protease genes were early induced by lignin (LIG3d). Notably, polyporopepsin gene 0925.239 had the most expression level among extracellular proteases. In addition, acetamidases and amino acid permease genes were The above results suggested that DyPs were the main extracellular ligninolytic enzymes of I. lacteus in lignin medium, while in lignocellulose medium MnPs played key roles [14]. Therefore, DyPs were purified from lignin medium for studying their characterization and catalytic properties.

Purification and characterization of DyPs from I. lacteus in lignin medium
Two DyP isoenzymes were purified to homogeneity from the liquid cultures of I. lacteus grown in alkali lignin medium for 3 days (Fig. 4). The molecular mass of IlDyP2 and IlDyP1 were about 74 kDa and 72 kDa, respectively. Like other classic heme peroxidases (e.g., MnP, VP, and LiP), DyPs also contain heme group judging from the absorbance at 410 nm [5,16,32]. The Rz (A 410 /A 280 ) ratios of purified DyPs were 2.5 and 1.0, respectively. Meanwhile, in order to identify corresponding DyP isoenzymes expressed in I. lacteus CD2, purified DyPs were separated from SDS-PAGE to digested for peptide mass fingerprinting analysis. Based on the number of unique peptides matched in DyPs from I. lacteus, IlDyP2 and IlDyP1 corresponded to genes 0808.621 and 0821.105, respectively ( Table 2).
The optimal pH for IlDyP2 and IlDyP1 was determined in pHs ranging from 2 to 7 among six different substrates, including high redox potential dyes (RB19 and RB5), lignin model compounds (DMP, guaiacol, and VA) and ABTS (Fig. 5). IlDyP2 and IlDyP1 showed the same optimal pH 4.0 for oxidation of RB19 and guaiacol, in agreement with DyP for RB19 from I. lacteus CCBAS 238 [16]. IlDyP2 showed acidic optima for guaiacol (67%, 95%, and 57% of residual activities at pH 2, pH3, and pH 5, respectively), while IlDyP1 was found to show less acidic optima (9%, 52%, and 90% of residual activities at pH 2, pH3, and pH 5, respectively). However, only IlDyP2 could oxidize RB5 in the narrow pH range from 3.0 to 4.0, which was slightly different from DyP in I. lacteus CCBAS 238 with optimal pH 3 and retaining 60% of residual activity at pH 2 for RB5. The optimal pH of 3.0 and 4.0 for IlDyP2 and IlDyP1 against VA and ABTS, respectively, and pH 4 for IlDyP1 could retain 80% of residual activity against VA and ABTS, while DyP in I. lacteus CCBAS 238 had optimal pH 2 for VA and pH 3 for ABTS, and retained 10% and 60% of residual activities at pH 4 for VA and ABTS, respectively. IlDyP2 and IlDyP1 showed the same optimal pH 3.0 for DMP, in accordance with DyP in I. lacteus CCBAS 238 [16]. IlDyP2 showed acidic optima of DMP (74%, 81%, and 55% of residual activity at pH 2, pH4, and pH 5, respectively), while IlDyP1 was found to show less acidic optima (14%, 98% and 76% of residual activity at pH 2, pH 4, and pH 5, respectively). The optimal values of DyPs in this study were very near to those of DyP from Auricularia auricula-judae (pH 3.5 for RB19, pH 2.5 for VA, and pH 3.0 for DMP, ABTS, and RB5) [33]. Liers et al. [7] observed that all DyPs tested showed rather acidic pH optima (pH 1.4-2.5) for oxidizing nonphenolic aromatics.
These differences in biochemical properties of DyP isoenzymes might be attributed to their divergent evolutionary origin. Furthermore, DyP isoenzymes had complementary effects on different pH values, the combination of IlDyP2 and IlDyP1 isoenzymes might result in wider pH range for efficient lignin degradation.
The catalytic properties of IlDyP2 and IlDyP1 for classic fungal heme peroxidases substrates ABTS, DMP, and guaiacol were presented in Table 3. IlDyP2 and IlDyP1   11:302 exhibited significant differences in catalytic efficiencies for different substrates. K m of IlDyP1 toward three substrates was higher than that of IlDyP2, suggesting that IlDyP2 had higher affinity for three substrates. IlDyP2 and IlDyP1 had similar activity for ABTS and guaiacol, while IlDyP2 had higher activity for DMP than for IlDyP1. IlDyP2 had sixfold and threefold higher catalytic efficiencies than IlDyP1 for ABTS and guaiacol. This suggested that IlDyP2 had better substrate fitting at the oxidation site, which was in accordance with the lack of activity of IlDyP1 on the recalcitrant dye RB5 [10]. In comparison with other typical fungal heme peroxidases, IlDyP2 and IlDyP1 displayed higher catalytic efficiencies for the oxidation of ABTS and DMP [34,35].
In addition, the data shown in Fig. 5 revealed that the substrate specificities of IlDyP2 and IlDyP1 were similar to that of classic fungal heme peroxidases. IlDyP2 and IlDyP1 were both able to decolorize the high redox potential anthraquinone dye RB19, which is the common feature of all DyP-type peroxidases [8]. Notably, IlDyP2 had the ability to decolorize the recalcitrant dye RB5, while IlDyP1 could not decolorize RB5. Although IlDyP2 and IlDyP1 could not oxidize Mn 2+ to Mn 3+ , they could catalyze the oxidation of typical peroxidases substrates ABTS as well as phenolic lignin compounds DMP and guaiacol. Moreover, they exhibited low activity to oxidize non-phenolic lignin compound VA through peroxidase activity, belonging to the classic high redox potential peroxidases LiP and VP [2,36]. It was also reported that other fungal DyPs can oxidize VA with very low activities [7,35]. DyP in I. lacteus CCBAS 238 showed low activities for VA and RB5 [16]. In this study, we observed that mediator 1-HBT could promote VA degradation by IlDyP1. This phenomenon is similar to the oxidation of a non-phenolic lignin model compound by the laccase/1-HBT redox system [6,37]. With 1-HBT as mediator, the oxidation rate of VA to veratraldehyde increased by 29.4% (Fig. 6). Considering their catalytic versatility, IlDyP2 and IlDyP1 might be a part of the alternative biocatalytic system for lignin degradation by I. lacteus, combining the catalytic properties of heme peroxidases and laccase.  Some proteins in DyP family have low amino acid sequence identity (lower than 15%), and significant differences in catalytic efficiency (k cat /K m ) with a few orders of magnitude [38]. Bacterial DyPs possess a lower oxidizing ability than fungal DyPs, oxidizing only less recalcitrant phenolic lignin model compounds and monophenolic substrates [39]. However, Chen et al. found DyP from Thermomonospora curvata showed high catalytic efficiency with ABTS, close to that of fungal DyPs [40]. Except for dye decolorization, DyP from Raoultella ornithinolytica OKOH-1 can directly decolorize melanin, and immobilization can improve its activity and stability [41]. Whereas only a limited number of DyPs were purified and characterized. Further studies are still needed to assess the precise physiological roles and catalytic properties of DyPs, including fungal DyPs.

Conclusions
Irpex lacteus CD2 grown in lignin medium secreted DyPs as the main extracellular ligninolytic enzymes. Transcriptomics analysis revealed that DyPs-and H 2 O 2 -generating enzymes including AOX, POX, and GLOX were coordinately expressed for efficient lignin degradation. Moreover, IlDyP2 and IlDyP1 could catalyze the oxidation of typical peroxidases substrates ABTS, phenolic lignin compounds DMP, and guaiacol as well as non-phenolic lignin compound VA. IlDyP1 could enhance the oxidation of non-phenolic lignin compound VA in the presence of mediator 1-HBT, the same as with Lac. These results proved for the first time that DyPs might depolymerize lignin when lacking classic heme peroxidases such as MnP, LiP, and Lac. DyPs could display different catalytic features of different peroxidases to different substrates, combining the catalytic properties of classic heme peroxidases and Lac. Therefore, DyPs may form an important constituent of the ligninolytic system in wooddecaying fungi.

Strain and culture conditions
Irpex lacteus CD2 [14] was maintained at 4 °C on potato dextrose agar plate. The inoculum was precultured in potato dextrose broth for 7 days at 28 °C, then 10% (v/v) inoculum was transferred into the modified Kirk's medium, and shaken at 150 rpm. The Kirk's medium contained: alkali lignin (or glucose) as the sole carbon source, 10 g/L; ammonium tartrate, 0.

Enzymatic assays
Total ligninolytic enzyme activities were measured by monitoring the oxidation of ABTS (ε 420 = 36,000 M −1 cm −1 ) at 420 nm, in 50 mM sodium tartrate buffer (pH 4.0) containing 1 mM ABTS, 1 mM Mn 2+ , and 0.1 mM H 2 O 2 . Manganese-independent peroxidase activity was also determined by ABTS oxidation in the absence of Mn 2+ . DyP activity was assayed by the decolorization of an anthraquinone dye RB19 (ε 595 = 10,000 M −1 cm −1 ) at 595 nm. The reaction was performed in the same buffer containing 125 μM RB19 and 0.1 mM H 2 O 2 . One unit of enzyme activity was defined as the amount of enzyme that oxidized 1 μmol of ABTS or RB19 per minute at 25 °C.

Secretome analysis
The extracellular enzymes of I. lacteus CD2 grown in alkali lignin at different periods of time were collected and concentrated by 80% ammonium sulfate [14]. The concentrated proteins were separated by one-dimensional SDS-PAGE, and the main bands on the third day were excised from the gel, digested with trypsin, and identified by nano LC-MS/MS. The peptides were separated in a reverse-phase C18 column, 0.18 mm × 100 mm, 5 μm particle size (Thermo). The mobile phases were A Fig. 6 Oxidation of the non-phenolic lignin model compound, veratryl alcohol, in the absence or the presence of the mediator 1-HBT by IlDyP2 and IlDyP1 at 30 °C for 24 h (water) and B (acetonitrile) containing 0.1% formic acid [22]. The flow rate was maintained at 300 nL/min. The phase B gradient was started at 3%, followed by a linear gradient to 8% in 1 min, 8-40% in 5 min, 40-85% in 1 min, and held there for 1 min. All MS/MS spectra were searched using PEAKS software against I. lacteus CD2 protein database using the following criteria [42]: enzyme trypsin; fixed modification of cysteine (+ 57.02 Da); and variable modification of methionine (+ 15.99 Da).

Transcriptome analysis
Irpex lacteus CD2 was grown in the modified Kirk's medium containing lignin or glucose as carbon source. The total RNA was extracted from mycelia on days 3 and 6 using the TRIZOL reagent (Invitrogen, Waltham, MA) according to the manufacturer's instructions. The total RNA was sent to Annoroad Genomics (Beijing, China) for sample preparation and sequencing. All samples were in duplicate. The cDNA was synthesized and prepared for sequencing using the Illumina mRNA-Seq Sample Prep Kit (San Diego, CA). The samples were run on independent lanes, and paired-end sequences of 150 bp were obtained at 4 Gb clean data for each sample using the Illumina Hiseq 2500. The raw reads were trimmed and filtered using Trimmomatic software to remove adapters and low-quality bases [43]. Then clean reads were assembled into transcripts using TopHat and Cufflinks with the I. lacteus CD2 genome as a Ref. [14,44]. All sequences of transcripts were extracted from reference sequence using gffread from cufflinks pipeline. The gene expression levels were conducted using the fragments per kilobase of exon per million fragments (FPKM) mapped method [45], and read counts were analyzed for differential expression using DESeq with a q value < 0.05 [46].

Purification and characterization of IlDyPs
The liquid cultures of I. lacteus CD2 grown in alkali lignin for 3 days were collected and concentrated by 80% ammonium sulfate at 4 °C. 20 mM sodium acetate buffer (pH 5.0) was used to dissolve the pellets and dialyzed using 30-kDa cutoff membrane. Then IlDyPs were purified using a HiTrap Q HP anion exchange column (GE Health, Fairfield, CT) pre-equilibrated with the same acetate buffer. The IlDyPs were eluted with a linear gradient of 0-1.0 M NaCl, and fractions containing active enzymes were pooled after SDS-PAGE. Meanwhile, the bands were excised and identified by peptide mass fingerprinting.
To determine the optimal pH, 50 mM sodium tartrate buffers with pH ranging from 2.0 to 7.0 were used for all substrates including ABTS, DMP, guaiacol, VA, RB19, and RB5 at 25 °C. The maximum activities of IlDyP2 and IlDyP1 were considered to be 100%. For catalysis properties, the reactions were performed in optimal pH at 25 °C using 50-4000 μM substrates by monitoring corresponding oxidation products. The nonlinear least square fitting method was used to calculate the K m , k cat , and k cat /K m parameters of IlDyP2 and IlDyP1 using the GraphPad Prism 5 software.
1-HBT was used as the mediator in evaluating the abilities of IlDyP1 and IlDyP2 to oxidize the non-phenolic lignin compound VA. The oxidation of VA was performed in 50 mM sodium tartrate buffer (pH 4.0 or pH 3.0) containing 1 mM VA, 1 mM 1-HBT, 0.1 mM H 2 O 2 , and 0.5 U/mL IlDyP1 or IlDyP2, without 1-HBT or DyP as corresponding control. The reaction proceeded at 30 °C. After 24 h, the reaction products were analyzed by HPLC using a reverse-phase C18 column, 4.6 mm × 250 mm, 5-m particle size (Waters XTerra). The isocratic elution condition was performed with 55% methanol containing 0.1% formic acid at a flow rate of 1 mL/min. The elution peaks were monitored at 310 nm.
Authors' contributions XZ, XS, and BY conceived and designed the experiments. XQ performed the experiments. XQ, HL, and FM analyzed the data. XQ, FM, and XS wrote the manuscript. All authors read and approved the final manuscript.