Bidirectional Regulation of AdpAch in Controlling the Expression of scnRI and scnRII in the Natamycin Biosynthesis of Streptomyces chattanoogensis L10

AdpA, an AraC/XylS family protein, had been proved as a key regulator for secondary metabolism and morphological differentiation in Streptomyces griseus. Here, we identify AdpAch, an ortholog of AdpA, as a “higher level” pleiotropic regulator of natamycin biosynthesis with bidirectional regulatory ability in Streptomyces chattanoogensis L10. DNase I footprinting revealed six AdpAch-binding sites in the scnRI–scnRII intergenic region. Further analysis using the xylE reporter gene fused to the scnRI–scnRII intergenic region of mutated binding sites demonstrated that the expression of scnRI and scnRII was under the control of AdpAch. AdpAch showed a bi-stable regulatory ability where it firstly binds to the Site C and Site D to activate the transcription of the two pathway-specific genes, scnRI and scnRII, and then binds to other sites where it acts as an inhibitor. When Site A and Site F were mutated in vivo, the production of natamycin was increased by 21% and 25%, respectively. These findings indicated an autoregulatory mechanism where AdpAch serves as a master switch with bidirectional regulation for natamycin biosynthesis.


INTRODUCTION
The secondary metabolic process in Streptomyces is regulated by a complex regulatory network involving pathway-specific, pleiotropic, and global regulators which respond to a variety of physiological and environmental condition alterations (van Wezel and McDowall, 2011;Liu et al., 2013). The best characterized is the A-factor regulatory cascade in which AdpA is the most important transcriptional factor for the secondary metabolism (Horinouchi, 2002;Ohnishi et al., 2005). In early culture stages, the transcription of adpA in Streptomyces griseus is repressed by ArpA, the receptor protein for A-factor (Onaka and Horinouchi, 1997). When A-factor reaches a critical concentration, it binds to ArpA and confers the conformational change of ArpA (Ohnishi et al., 1999). This results in dissociation of ArpA from the adpA promoter, in turn switching on the expression of adpA (Ohnishi et al., 1999). The induced AdpA then activates the transcription of various genes related to secondary metabolism such as strR, the pathway-specific regulatory genes for streptomycin in S. griseus (Retzlaff and Distler, 1995;Tomono et al., 2005).
Typically, AdpA is regarded as an activator for downstream regulated genes, except itself which is proved to be negatively auto-regulated by binding to its own promoter region (Kato et al., 2005b;Hara et al., 2009). The molecular mechanism of transcriptional activation begins as a dimer of AdpA binds to the target sites with consensus sequences which then recruit RNA polymerase to the promoter for transcriptional initiation (Yamazaki et al., 2004;Kato et al., 2005a). For different target genes, AdpA showed a different number of binding sites in the promoter regions. For example, there are two AdpA-binding sites in the promoter of strR , whereas there are three AdpA-binding sites for regulation of ssgA (Yamazaki et al., 2003a). However, the precise regulation mechanism how the AdpA binds to multiple sites to activate transcription has not been experimentally determined. Based on the importance of AdpA in the biosynthesis of the secondary metabolism, it is necessary to elucidate details of its regulatory mechanisms.
Natamycin, an antifungal polyene macrolide antibiotic, is synthesized by a type I polyketide synthase gene cluster. Previous analysis of the gene cluster of natamycin in Streptomyces chattanoogensis L10 revealed the existence of 17 open-reading frames, including two pathway-specific genes, scnRI and scnRII (Du et al., 2011a). These two genes showed high sequence identity to pimR and pimM of Streptomyces natalensis, respectively (Antón et al., 2007;Santos-Aberturas et al., 2012). Gene disruption of scnRI resulted in a large decrease in the expression of biosynthetic genes, indicating its role as a pivotal activator for the biosynthesis of natamycin (Du et al., 2011a). scnRII, adjacent but divergently transcribed transcriptional regulatory genes, was shown to act as a second positive regulator for natamycin production (Du et al., 2009). We also had proved that AdpA ch controls the production of natamycin, but the detailed relationship among AdpA ch , ScnRI, and ScnRII had not been well characterized (Du et al., 2011a).
Here, we reveal the sophisticated regulatory characteristics of AdpA ch in the natamycin biosynthesis of S. chattanoogensis L10. AdpA ch acts as a "higher level" pleiotropic regulator for transcription of the two divergently transcribed pathway-specific genes, scnRI and scnRII. In this regulatory process, AdpA ch shows a bi-stable regulatory ability, where it firstly acts as an activator, then a repressor. Moreover, natamycin production was enhanced by mutating the AdpA ch -binding sites which had an inhibitory effect. This work not only advances the understanding of detailed regulatory mechanism of AdpA, but also provides a potential target for the enhancement of other antibiotic production levels by manipulating the regulatory network.

RESULTS
AdpA ch Identified as a "Higher Level" Pleiotropic Regulator for Natamycin Biosynthesis In our previous study, the biosynthetic gene cluster of natamycin has been cloned and characterized in S. chattanoogensis L10. Within this there are two divergently transcribed genes, scnRI and scnRII, encoding proteins that resemble pathway-specific regulators (Du et al., 2009(Du et al., , 2011a. Although the functions of these two regulators have been well characterized, an important question remains as to whether there are multiple levels of control in the biosynthesis of natamycin. Based on our previous study that AdpA ch affected the transcription of these two pathway-specific genes (Du et al., 2011a), we speculated that AdpA ch may act as a "higher level" pleiotropic regulator for regulating the natamycin biosynthesis.
To test this hypothesis, electrophoretic mobility shift assays (EMSAs) were applied. As shown in Figure 1, retardation was readily detected upon the addition of 50 pM AdpA ch with the probe RI-RII, while the addition of 50-to 100-fold excess of unlabeled specific PCR product reduced the proportion of the labeled promoter-containing fragment (Figure 1). These data clearly demonstrate that AdpA ch could specifically bind to the scnRI-scnRII intergenic region and could control the expression of these two pathway-specific genes.
FIGURE 1 | AdpA ch binds to the DNA sequence of the intergenic promoter region between scnRI and scnRII. Lanes 1-3, DNA probe with AdpA ch protein 0, 50, and 100 pM, respectively. Lanes 4 and 5, 50-and 100-fold excess of unlabeled specific PCR product was added into binding reactions.
Frontiers in Microbiology | www.frontiersin.org DNase I Footprinting Assay Reveals Six AdpA ch -Binding Sites in the scnRI-scnRII Intergenic Region To identify the exact DNA sequences that AdpA ch protected in the scnRI-scnRII intergenic region, DNase I footprinting assays, in absence or presence of purified recombinant AdpA ch , were performed. In our previous studies, we had determined the transcription start site (TSS) of the two pathway-specific genes, scnRI and scnRII (Du et al., 2011a). As seen in Figure 2A, at a lower AdpA ch protein concentration of 100 pM, the DNA strands of the scnRI-scnRII intergenic region showed two protected regions, Site C and Site D, extending from positions −69 to −44 and −106 to −74 relative to the TSS of scnRI. When increasing the protein concentration to 500 pM, another four protected regions (Sites A, B, E, and F) were observed. With respect to the scnRI TSS, the AdpA ch -binding Site A locates at positions +8 to +54, Site B at positions −20 to +2, Site E at positions −161 to −114, and Site F at positions −283 to −259 ( Figure 2B). The six AdpA ch -binding sites were spread over the scnRI-scnRII intergenic region. Notably, Site A was located downstream of the scnRI TSS, while Site B overlapped the −10 region of the scnRI promoter. Site F was located downstream of the scnRII TSS, and Site E overlapped the −35 region of the scnRII promoter. This data suggest that AdpA ch might have a negative regulatory ability for the expression of these two pathway-specific genes. Additionally, the results from the DNase I footprinting assay also reveal that AdpA ch may have higher affinity to Site C and Site D than to the others.
The Consensus AdpA ch -Binding Sequence in the AdpA ch -Binding Sites The orthologs of AdpA ch identified in S. griseus and S. coelicolor have been reported to have the consensus binding sequence, 5-TGGCSNGWWY-3 (S: G or C; W: A or T; Y: T or C; N: any nucleotide) (Yamazaki et al., 2004). After alignment of these six protected regions, we also found that there were highly conserved AdpA ch -binding sequences in each binding site ( Figure 3A). To further study the roles of these consensus sequences in the AdpA ch -binding ability, EMSAs were carried out using the probes containing either the sequences of wild-type (wt) binding sites or the mutated sites ( Figure 3A). As shown in Figure 3B, no binding shift was detected for the mutated sites A-F when compared with their corresponding wt targets. Taken together, these data demonstrated that AdpA ch indeed FIGURE 2 | DNase I footprinting assay for determination of the AdpA ch -binding sites. (A) A 5 -FAM-labeled probe pRI-RII was used in the DNase I footprinting assay with 0, 100, and 500 pM purified AdpA ch , respectively. The protected regions are underlined. (B) Nucleotide sequences of the scnRI-scnRII intergenic region showing the predicted AdpA ch -binding sites. The TSS is marked by a bent arrow, the AdpA ch -binding sites are underlined, and the −10 and −35 regions are overlined. has six binding sites in the scnRI-scnRII intergenic region and the consensus sequence is essential for the binding activity of AdpA ch .

AdpA ch Has Differing Affinities for Different Binding Sites
In the DNase I footprinting analysis, Site C and Site D were occupied with a lower concentration of AdpA ch than the other sites. This suggests that there may be affinity differences for AdpA ch between the six binding sites. To test this possibility, competitive EMSAs with 50-to 100-fold excess of unlabeled fragments of six AdpA ch -binding sites were used to compete with each labeled fragment. As shown in Figure 4A, 100-fold excess of unlabeled S B (Site B) and S F (Site F) could not completely abolish AdpA ch complex formation with the labeled probe S A (Site A). However, the same amount of unlabeled S C (Site C), S D (Site D), and S E (Site E) outcompeted the labeled probe S A . This result indicated that AdpA ch binds to Site A more tightly than Site B and Site F, but less tightly than Site C, Site D, and Site E. Following this way, we could conclude that Site B has less affinity for AdpA ch than others, except for Site F (Figure 4B), which was the weakest affinity among the six binding sites (Figure 4F), and Site D was the strongest affinity of these six sites ( Figure 4D). The affinity of Site E for AdpA ch was between that of Site C and Site A (Figures 4A,C,E). Therefore, we determined the affinity of AdpA ch to different binding sites in the following order: Site Promoter-Probe Assays of the AdpA ch -Binding Sites in the scnRI-scnRII Intergenic Region The binding sites of AdpA ch in the scnRI-scnRII intergenic region were adjacent to either the scnRI or the scnRII start codon. This raised the possibility that this intergenic region might harbor a bidirectional promoter allowing AdpA ch to regulate transcriptions of the divergently transcribed flanking genes, scnRI and scnRII (Figure 2B). To investigate the promoter activities of the two pathway-specific genes with each of the AdpA ch -binding sites, we used the promoter-probe plasmid pIJ8601 carrying the xylE gene, encoding catechol 2,3-dioxygenase, as the reporter. As shown in Figure 5A, the transcriptional profiles of scnRI were severely decreased when the AdpA ch -binding Site C and Site D were mutated. Conversely, its transcriptional activity was increased when Site A and Site B were mutated and remained almost unchanged when Site E and Site F were mutated. For the promoter activity of scnRII, we did not detect any consistent differences when Sites A, B, and C were mutated, but mutation in the Sites D and E resulted in a large decreases of up to 70 and 40%, respectively, compared to those of the wt. The mutation in Site F resulted in a statistically significant increase FIGURE 4 | Comparison of the relative affinity of AdpA ch with different binding sites. Labeled probes S A , S B , S C , S D , S E , and S F contained the fragment of Sites A-F as shown in Figure 3A, respectively. Probes S A , S B , S C , S D , S E , and S F also contained the fragment of Sites A-F as shown in Figure 3A, respectively, but they are unlabeled. The amount of AdpA ch protein used was 100 pM.
( Figure 5B). These findings indicated that expressions of scnRI and scnRII are both under the control of AdpA ch , which has a completely different regulatory ability (activation or inhibition) when binding to different binding sites.

Effect of Mutated AdpA ch -Binding Sites on Natamycin Production in Vivo
There have been some reports where effects upon DNA-binding sites were found in vitro that failed to be exhibited in vivo. In order to test this possibility and reveal the function of the six AdpA ch -binding sites in natamycin biosynthesis in vivo, a series of mutants were constructed as described in Experimental procedures. As shown in Figure 6A, compared to the WT strain, the level of natamycin production in the R-mA (mutation in Site A) and R-mF (mutation in Site F) had increased by 21 and 25%, respectively. However, the constructed strains of R-mC (mutation in Site C), R-mD (mutation in Site D), and R-mE (mutation in Site E) showed up to 31, 42 and 15% reductions, respectively. The natamycin production of R-mB (mutation in Site B) mutant exhibited almost no change. This finding indicated that the AdpA ch -binding Sites A and F play negative roles for natamycin biosynthesis, while the functions of the Sites C, D, and E were positive. Quantitative real-time PCR (qRT-PCR) analysis showed that the promoting effect of site mutation on natamycin production was due to alteration of the pathway-specific genes at the transcriptional level ( Figure 6B).

DISCUSSION
Streptomyces spp. have developed complicated mechanisms to adapt to altered circumstances (Santos-Beneit et al., 2009;Yu et al., 2012). Among these mechanisms, the multiple levels of regulation in controlling the expression of the genes responsible for the formation of the secondary metabolism are drawing increased attention. In this study, we focused on the regulatory network of natamycin biosynthesis in S. chattanoogensis L10, FIGURE 5 | Promoter activities of scnRI (A) and scnRII (B) with the effect of mutations in the AdpA ch -binding sites. The strains were grown in YEME medium for 24 h, and catechol dioxygenase activity was calculated as the change of catechol quantity (mmol) per minute. Error bars correspond to the standard error of the mean of four culture replicates. * Indicates significant differences between promoter mutants and promoter wt (P < 0.05).
an industrial strain for natamycin production. In our previous study, we determined that gamma-butyrolactones (GBLs) serve as quorum-sensing signaling molecules for activating natamycin production in S. chattanoogensis L10 (Du et al., 2011b), and ScnRII acts as a positive regulator by directly binding to the promoters of natamycin biosynthetic genes (Du et al., 2009) where ScnRI acts as a positive regulator for the transcription of scnRII (Du et al., 2011a). However, the deletion of scnRI did not result in a complete halt of the transcription of scnRII (our unpublished data). This is quite different from the function of PimR in S. natalensis where the deletion of pimR almost completely destroys the transcription of pimM (Antón et al., 2004;Santos-Aberturas et al., 2012). As the regulation of antibiotic biosynthesis involves numerous transcription factors (McKenzie and Nodwell, 2007;van Wezel and McDowall, 2011), participation of other regulator(s) is possible, in the regulation of scnRII.
With AdpA ch being able to regulate the expression of both of the pathway-specific genes, scnRI and scnRII, it provides a possible explanation that there is a coordinate regulation in FIGURE 6 | (A) The effect of mutated AdpA ch -binding sites on the natamycin production in vivo. The strains were grown in YEME medium for 96 h. Vertical error bars correspond to the standard error of the mean of four replicated cultures. (B) Real-time RT-PCR analysis of the scnRI and scnRII transcript in the wt strain and mutated AdpA ch -binding sites strain. The expression level of scnRI and scnRII is presented relative to the wt sample from 24 h, which was arbitrarily assigned a value of 1. The transcription of hrdB was assayed as an internal control. Error bars were calculated by measuring the standard deviation among three replicates of each sample. * Indicates significant differences between AdpA ch -binding site mutants and wt (P < 0.05).
controlling expression of scnRII by AdpA ch and ScnRI. This regulatory pattern may occur in following steps. Firstly, AdpA ch binds to the scnRI-scnRII intergenic region and activates both transcription of scnRI and scnRII. Then ScnRI also binds to the scnRI-scnRII intergenic region which, in turn, promotes the transcriptional level of scnRII. However, these two genes were not completely controlled by AdpA ch . Trace expression of scnRI was observed in the adpA ch mutant, and then ScnRI would promote the transcription of scnRII (Du et al., 2011a). Notably, a certain amount of AdpA ch is required for binding to the scnRI-scnRII intergenic region (∼50 pM). This is why we did not detect the shifted band with low concentration AdpA ch (∼1 pM) in the binding reaction of our previous study (Du et al., 2011a).
In most cases, AdpA acts as an activator for the target genes, except for itself where it exhibits an autorepression (Kato et al., 2005b). In this study, we concluded from promoterprobe assays in vivo that AdpA ch could not only regulate both pathway-specific genes, but also displayed completely opposite regulatory abilities in control of them. The AdpA ch -binding Site C and Site D were involved in activating the transcription of scnRI, while AdpA ch binding to Sites A and B resulted in repression. For the promoter activity of scnRII, mutation in the Site C and Site D resulted in a decrease of transcriptional profiles, while a mutation in the Site F led to a statistically significant increase. A similar phenotype was observed in S. ansochromogenes where transcription of sanG decreased when Site I and Site V were mutated but increased when other three AdpA-L-binding sites were mutated (Pan et al., 2009). However, when combinations of binding site mutations were carried out, the promoter activities were not in accordance with our predictions. For example, mutations in both Sites E and F reduced the transcriptional level of scnRII (data not shown). Based on the short distances between the AdpA ch -binding sites which are spread over the scnRI-scnRII intergenic region, there may be complicated interactions between different AdpA ch dimmers to explain this.
With further analysis using competitive gel shift assays, we could conclude that AdpA ch binds to Sites A-F with the following affinities: Site D > Site C > Site E > Site A > Site B > Site F (Figure 4). These data are consistent with the footprinting assay where the regions of Site C and Site D were previously protected at a lower AdpA ch protein concentration (Figure 2A). This gives a hint that the regulatory ability of AdpA ch may occur in a growth phase-dependent manner. In the early stage, AdpA ch firstly binds to the Site C and D to recruit RNA polymerase to the promoter and initiates the transcription of scnRI and scnRII. This in turn triggers natamycin production (Figure 7). When AdpA ch is accumulated to a certain critical level, it will bind to other binding sites located near the TSS. A DNA loop may be formed via the interaction between different AdpA ch dimers, thus preventing RNA polymerase from access to the promoter of the pathway-specific genes (Figure 7). Reduced transcription of the pathway-specific genes will result in a low rate of natamycin production.
The discovery of this bidirectional regulation of AdpA ch in the control of natamycin biosynthesis reveals an artful adaptive mechanism in microbial cells. Microorganisms produce molecules with antibiotic activity and expel them into the environment, presumably enhancing their ability to compete with their neighbors (Berdy, 2005;Hopwood, 2007). However, most of these molecules are toxic to the producer (Mak et al., 2014;Moody, 2014). Mechanisms must exist to ensure that antibiotic production reaches a reasonable level. The proposed model of AdpA ch in Figure 7 may provide a fresh mechanistic insight into how S. chattanoogensis controls the production level of natamycin via AdpA ch . However, further work will be needed to prove the proposed model and the detailed mechanism of how AdpA ch responds to the signal of natamycin. In all, the complicated regulatory network involving AdpA ch , ScnRI, and ScnRII helps advance our understanding of the molecular regulation mechanisms of antibiotic biosynthesis and provides an effective strategy to help improve yields in industrial strains.

Media, Plasmids, Strains, and Growth Conditions
All plasmids and bacterial strains used in this study are listed in Table 1. General techniques for the manipulation of nucleic acids and bacterial growth were carried out according to the standard protocols as previously described (Kieser et al., 2000). Escherichia coli DH5α was the general cloning host. Vectors used were pSET152, pIJ8660, pTA2. S. chattanoogensis L10 strains were grown at 28 • C on YMG agar for sporulation and at 30 • C FIGURE 7 | Proposed model of AdpA ch regulation for scnRI and scnRII transcription. The AdpA ch -binding sites are spread over the scnRI-scnRII intergenic region, and the presumptive manner of AdpA ch -binding forms a dimmer. Stage 1: AdpA ch binds to the Site C and Site D when the concentration of AdpA ch is low in the early growth stage. Stage 2: When the concentration AdpA ch reaches a high level, AdpA ch binds to Site A, Site B, Site E, and Site F, most of which located downstream of the −35 sequences of the corresponding genes. A DNA loop may be formed via interaction between different AdpA ch dimmers, thus preventing RNA polymerase from access to the promoter. amplification. The probes A (295 bp), B (281 bp), C (294 bp), D (282 bp), E (288 bp), F (284 bp), mA (295 bp), mB (281 bp), mC (294 bp), mD (282 bp), mE (288 bp), and mF (284 bp) were prepared following the above-mentioned method. In the EMSAs assay, 1 ng of the probe was incubated with varying quantities of AdpA ch , at 25 • C for 30 min in the buffer (20 mM Tris, pH 7.5, 5% glycerol, 0.01% BSA, 50 µg ml −1 sheared sperm DNA). For the competition assay, 100 times of excessive un-labeled probes and non-specific DNA were added to the reaction buffer, respectively. Reactions were displayed on 5% acrylamide gels for separation in 0.5× TBE buffer. EMSA gels were then electro-blotted onto the nylon membrane and UV-fixed by UV crosslinker. Labeled DNA was detected with streptavidin-HRP and BeyoECL plus (Beyotime, China) as described by the manufacturer.

DNase I Footprinting Assay
DNase I footprinting assay was performed as previously described (Mao et al., 2009). Firstly, AdpA ch protein was ultra-filtered with YM-10 (Millipore) for 10 kD cut-off and eluted in 20 mM Tris buffer, pH 7.5. Then, FAM-labeled probe was amplified using 5 -(6-FAM)-labeled M13 universal primers from plasmid pT-RI-RII, followed by gel recovery. About 50 ng of fluorescently labeled probe was added to the reaction mixture to a final volume of 50 µl. After binding of the AdpA ch protein to 5 -(6-FAM)-labeled probe (30 • c, 30 min), 0.01 U of DNase I (Promega) was added for 1 min at 30 • C, followed with equal volume of 100 mM EDTA to stop the reactions and extracted by phenol/chloroform. After precipitation with 40 µg of glycogen, 0.75 M ammonium acetate (NH 4 Ac), and ethanol, the digested DNA mixture was loaded into ABI 3130 DNA sequencer with Liz-500 DNA marker (MCLAB). DNA sequencing ladder was prepared according to Thermo Sequenase Dye Primer Manual Cycle Sequencing Kit (USB).

Alterations of the Consensus Sequence for AdpA ch -Binding Sites
The consensus sequence of AdpA ch -binding sites A-F was replaced by the sequence of EcoRV restriction sequence sites using overlapping primers (Supplementary Table S1). The PCR product was then cloned into a pTA2-vector (TOYOBO). The resulted plasmids were used as template for PCR to amplify mutated probes using 5 -biotin-labeled M13 universal primers, and the binding ability was measured by EMSAs.

Construction and Analysis of Transcriptional Fusions to the xylE Reporter Gene
For xylE fusions, the xylE gene was PCR amplified with the primers xylE-F and xylE-R. This fragment was digested with NdeI and NotI, and introduced into the likewise-digested pIJ8660 (Sun et al., 1999) to construct pIJ8601. To probe scnRIp and scnRIIp activities with the mutation of AdpA ch -binding sites, the wt and mutated promoter regions were amplified by PCR using upstream primers carrying a BamHI site listed in Supplementary  Table S1. These promoter fragments were cloned into BamHI-cut pIJ8601 and transferred by conjugation into S. chattanoogensis L10. Plasmid-containing strains were grown on YEME medium for 24 h. Cell pellets from 1 ml culture samples were kept on ice and measured immediately. Assays of catechol 2,3-dioxygenase were performed as previously described (Kieser et al., 2000).
Mutational Analysis of the AdpA ch -Binding Sites on Natamycin Biosynthesis The 1.8 kb DNA fragment containing the sequence of scnRI-scnRII intergenic region was amplified by PCR using primers scnRI-F and scnRII-R. The resulted 1.8 kb sequence was used as template to amplify the DNA fragment for construction of mutated AdpA ch -binding sites in vivo using overlapping primers (Supplementary Table S1), then PCR product was purified and ligated into pKC1139. The resulting plasmids containing DNA fragment of mutated sites was conjugated by E. coli ET12567/pUZ8002 into S. chattanoogensis L10. The mutants were selected by replica plating for apramycin-sensitive colonies and they were used as template for PCR with primer pairs RI-RII-F and RI-RII-R. The amplified sequences were digested with EcoRV to confirm the mutants.

Determination of Natamycin Production by HPLC Analysis
Natamycin production was confirmed by HPLC analysis with the Agilent 1100 HPLC system. HC-C 18 column (5 µm, 4.6 by 250 mm) was used with UV detector set at 303 nm. Mobile phase and gradient elution process were as described previously (Du et al., 2009).

AUTHOR CONTRIBUTIONS
PY, Q-TB, and Y-LT performed the experiments. X-MM assisted with the primary data analysis. Y-QL supervised the project and revised the manuscript. All authors reviewed the manuscript.