Coordinate regulation of the expression of SdsR toxin and its downstream pphA gene by RyeA antitoxin in Escherichia coli

In Escherichia coli, SdsR and RyeA, a unique pair of mutually cis-encoded small RNAs (sRNAs), act as toxin and antitoxin, respectively. SdsR and RyeA expression are reciprocally regulated; however, how each regulates the synthesis of the other remains unclear. Here, we characterized the biosynthesis of the two sRNAs during growth and investigated their coordinate regulation using sdsR and ryeA promoter mutant strains. We found that RyeA transcription occurred even upon entry of cells into the stationary phase, but its apparent expression was restricted to exponentially growing cells because of its degradation by SdsR. Likewise, the appearance of SdsR was delayed owing to its RyeA-mediated degradation. We also found that the sdsR promoter was primarily responsible for transcription of the downstream pphA gene encoding a phosphatase and that pphA mRNA was synthesized by transcriptional read-through over the sdsR terminator. Transcription from the σ70-dependent ryeA promoter inhibited transcription from the σS-dependent sdsR promoter through transcriptional interference. This transcriptional inhibition also downregulated pphA expression, but RyeA itself did not downregulate pphA expression.


Regulation of sdsR and RyeA biosynthesis.
To determine the effects of SdsR on the biosynthesis of RyeA, or vice versa, we constructed sdsR and ryeA promoter mutant strains in which their −10 elements were inactivated by changing each of them to CTCGAG. We then examined levels of SdsR and RyeA in the promoter mutant cells during growth (Figs 3 and S1). As expected, SdsR and RyeA were not detected in the sdsR and ryeA promoter mutant strains, respectively (Fig. 3A). In wild-type cells, RyeA was expressed during the early exponential phase, and its expression sharply decreased in the mid-exponential phase. The sdsR promoter mutation increased RyeA synthesis during the exponential phase and allowed its continued synthesis in the stationary phase. On the other hand, SdsR expression started in the late-exponential phase in wild-type cells, but appeared at an earlier stage in ryeA promoter mutant cells (Fig. 3B). Since the larger ryeA covers the entire sdsR, the whole 104-nt SdsR sequence can base-pair with RyeA. We performed analysis of mutual degradation to determine whether interaction between two sRNAs leads to degradation of each other (Fig. S2). We ectopically co-expressed both SdsR and RyeA, but expression of SdsR was increased at a fixed expression level of RyeA, or vice versa. We found that when one is overexpressed, degradation of the other is facilitated, suggesting that this extensive base-pairing leads to degradation of both sRNAs. Because SdsR and RyeA are degraded each other, it is likely that the early appearance of SdsR in ryeA promoter mutant cells and the prolongation of RyeA expression to the stationary phase in sdsR promoter mutant cells are attributable to the absence of mutual degradation of the two sRNAs. The lack of RyeA degradation products of about 60 nt in sdsR promoter mutant cells further supports reciprocal regulation of the two sRNAs through degradation of each other. This mutual degradation could serve to further restrict RyeA expression to the exponential phase and SdsR expression to the stationary phase. The regions +341 to +132 (containing the ryeA promoter) and −79 to +11 (containing the sdsR promoter) were cloned into pKK232-8 to generate ryeA-CAT and sdsR-CAT plasmids, respectively. Total cellular RNA extracts were prepared from either MG1655 cells containing the ryeA-CAT or sdsR-CAT plasmid grown for 2 h, 6 h, or 10 h at 37 °C. The 32 P-labeled primer CAT_R was used to analyze ryeA-CAT and sdsR-CAT fusion transcripts. Primer extension products were analyzed on a 5% polyacrylamide sequencing gel containing 8 M urea. The DNA ladders (G, A, T and C) were prepared by dideoxy sequencing using the template plasmid DNA and the same primer. Loading amounts are indicated below the lanes. The transcription start nucleotides are indicated by arrows. (C) RACE analysis. The 5′ or 3′ RACE products (primary PCR products or nested PCR products) were analyzed on 2% agarose gels. Predicted RACE products were indicated by a, b, c, and d. For 5′ RACE, E. coli RNA pyrophosphatase (RppH)-treated RNA and untreated RNA were used. M, 100 bp size markers; R, RyeA; S, SdsR.
www.nature.com/scientificreports www.nature.com/scientificreports/ The tight growth phase-specific control of SdsR and RyeA may be needed for SdsR-or RyeA-mediated regulation of certain genes in specific growth phases. To identify such genes, we performed an RNA-seq analysis of ryeA and sdsR promoter mutant cells as well as wild-type cells. For this RNA-seq analysis, cultures were sampled in both the exponential phase (3 h) and stationary phase (8 h), and genes in sdsR or ryeA promoter mutant cells that showed a greater than 2-fold change and a p-value < 0.05 were selected for further characterization. We identified 17 genes (7 upregulated and 10 downregulated) in the exponential phase and 5 genes (2 upregulated and 3 downregulated) in the stationary phase for ryeA promoter mutant cells (Tables S1 and S2). In contrast, sdsR promoter mutant cells showed many changed genes: 171 genes (110 upregulated and 51 downregulated) in the exponential phase and 55 genes (33 upregulated and 22 downregulated) in the stationary phase (Tables S3-5). Selected genes were confirmed by qRT-PCR analysis (Table S6), which showed that RNA-seq data were reliable. A Gene Ontology (GO) analysis showed that many of the genes that were altered by the sdsR promoter mutation, either downregulated or upregulated, were related to transport function and encoded membrane proteins (Fig. S3), an observation consistent with previous RNA-seq analysis of SdsR-overexpressing cells 18 .
As expected, the RNA-seq data revealed upregulation of RyeA and SdsR in sdsR and ryeA promoter mutant cells, respectively (Fig. 4A,B). We also examined the RNA-seq data to see how expression of known SdsR target genes is affected by the sdsR or ryeA promoter mutation (Table S7). The sdsR promoter mutation increased tolC and yhcB expression by about 2-fold in the stationary phase and exponential phase, respectively. In contrast, the ryeA promoter mutation led to a slight decrease in tolC and yhcB expression, suggesting that RyeA can modulate SdsR-mediated gene regulation. However, mutS, another target for SdsR 18,22,23 , did not show gene expression profiles that would be expected from sdsR or ryeA promoter mutant cells, implying that it may not be a primary target of SdsR.
Interestingly, the RNA-seq data showed that the pphA gene, downstream of sdsR, was downregulated in sdsR promoter mutant cells and upregulated in ryeA promoter mutant cells (Fig. 4C). An analysis of pphA expression by quantitative reverse transcription-polymerase chain reaction (qRT-PCR) and semi-quantitative RT-PCR confirmed the same downregulation in sdsR promoter mutant cells and upregulation in ryeA promoter mutant cells (Fig. 4D,E). These data indicate that pphA expression may be regulated by the sdsR/ryeA regulatory circuit.
pphA is co-transcribed with sdsR from the sdsR promoter. The previously reported pphA promoter, located about 180 bp upstream of the sdsR promotor, is known as a heat-shock promoter 30 . Therefore, transcription from this promoter should pass through the sdsR/reyA locus to transcribe the pphA gene; furthermore, the sdsR promoter can generate pphA transcripts. To determine which promoter makes the greater contribution to pphA expression, we constructed various transcriptional sdsR-lacZ fusions and analyzed their transcriptional activities using LacZ assays (Fig. 5). The promoter region containing a DNA fragment 200 bp upstream and 100 bp downstream (−379 to −170 relative to the 5′ end of SdsR) from the previously known pphA transcription site showed little LacZ activity at 37 °C and even at 42 °C. However, a promoter region extended to include the sdsR/ryeA locus (+222) exhibited high LacZ activity at 37 °C that was not significantly further increased at 42 °C. Therefore, it is unlikely that the previously reported heat inducible pphA promoter is responsible for the observed LacZ activity of the −379/+222 construct. Furthermore, the sdsR promoter mutation sharply decreased the LacZ activity of sdsR(−379+222)-lacZ, indicating that the sdsR promoter is a major contributor to pphA transcription. The −329 to +222 promoter region was fused to the CAT gene in plasmid pKK232-8 and the 5′ ends of the fused mRNA, transcribed in vivo, were analyzed by primer extension analysis (Fig. S4). Most extension products were from sdsR transcripts, and no extension products from the reported pphA promoter were detected. These data, taken together with in vitro data showing read-through transcripts traversing the sdsR terminator (Fig. 2B), suggest that the sdsR promoter rather than the previously reported pphA promoter is responsible for pphA transcription.
Effects of ryeA transcription on expression of SdsR and sdsR-pphA dicistronic mRNA. Next, we examined whether ryeA transcription affected pphA expression using the ryeA promoter mutant. To avoid possible effects of chromosomally expressed SdsR and RyeA, we constructed the sdsR-lacZ fusions in an sdsR/ ryeA-knockout background. The LacZ activity of the sdsR(−379/+222)-lacZ fusion was increased by about 4-fold by the ryeA promoter mutation (Fig. 5). This increase could be attributable to the absence of RyeA (acting in trans) or ryeA transcription itself (acting in cis). To discriminate between these two possibilities, we analyzed LacZ activity following ectopic expression of RyeA. To determine conditions for ectopically expressing RyeA at levels comparable to those generated by the endogenous ryeA promoter, we varied the concentration of isopropyl β-D-1-thiogalactopyranoside (IPTG) used to induce expression (Fig. 6A). We found that RyeA levels induced by  www.nature.com/scientificreports www.nature.com/scientificreports/ 0.005 mM IPTG were comparable to those produced by the ryeA promoter. Induction of RyeA caused no decrease in LacZ activity (Fig. 6B), suggesting that the absence of RyeA is not responsible for the increase in LacZ activity. Therefore, it is likely that ryeA transcription, not RyeA, inhibits pphA expression by reducing read-through transcription from the sdsR promoter. Then we examined how ryeA transcription affects the read-through transcription during the growth using the sdsR(−379/+222)-lacZ fusion. We found that the ryeA transcription represses LacZ expression at all growth phases and delays its expression 3 h to the stationary phase (Fig. S5). To further confirm that read-through transcripts are increased by the ryeA mutation, we inserted a Brevibacterium albidum tRNA Arg sequence between the sdsR sequences and the CAT coding sequence in the sdsR-CAT fusion constructs and examined exogenous tRNA expression as well as SdsR in an sdsR/ryeA-knockout background (Fig. 7). The reason that we used heterologously expressed B. albidum tRNA Arg , was because the tRNA was previously shown to be metabolically stable in E. coli and detectable by Northern blot analysis without cross-hybridization with E. coli tRNAs 36 . The ryeA promoter mutation caused an increase in tRNA Arg expression, confirming that the ryeA mutation increases read-through transcription. Taken together, these data show that pphA expression is inhibited by ryeA transcription itself, and not by RyeA. www.nature.com/scientificreports www.nature.com/scientificreports/ To examine whether the increase in read-through transcripts was caused by reduced termination efficiency at the sdsR terminator, we introduced a terminator mutation into the sdsR(−79/+222)-lacZ fusion (Fig. 8). This terminator mutation did not affect the ryeA promoter mutation-induced increase in LacZ activity, suggesting that the increase in read-through transcripts caused by the ryeA promoter mutation does not result from reduced termination efficiency. Therefore, it seems likely that transcription from the ryeA promoter interferes with transcription of sdsR.
Because it is known that transcriptional interference can occur in DNA constructs with convergent promoters 37 , we examined the existence of transcriptional interference with respect to ryeA transcription using in vitro transcription assays using pSdsR(−379/+222) carrying the ryeA promoter and its ryeA promoter mutant derivative, pSdsR(−379/+222ryeAP m ) (Fig. 2). For these assays, we used different ratios of Eσ 70 and Eσ S . As expected, the increased ratio of Eσ S /Eσ 70 increased SdsR and its read-through transcripts (SdsR-rnpB) with the decrease of RyeA (Fig. 2B,C), which may be reminiscent of their reciprocal synthesis during growth. The ryeA promoter mutation increased both SdsR and read-through transcripts, producing the highest increase in levels at an Eσ 70 to Eσ S ratio of 1:9 (Fig. 2B), which may be similar to that in the stationary phase in vivo.
Transcriptional interference can be due to promoter occlusion, colliding RNA polymerases, or transcription-induced changes in DNA supercoiling that affect initiation of transcription. Promoter occlusion is not likely because the two promoters are far apart. We tested whether positive supercoiling induced by transcription from the ryeA promoter could affect transcription from the sdsR promoter. For this purpose, we treated cells with novobiocin, a gyrase subunit B inhibitor. Growth of E. coli cells was inhibited by novobiocin with a 50% inhibitory concentration (IC 50 ) of 100 μg/ml (Fig. S6A). We examined effects of ryeA transcription on sdsR  www.nature.com/scientificreports www.nature.com/scientificreports/ transcription in the presence of novobiocin at 100 μg/ml (Fig. S6B). Since SdsR levels could be affected by RyeA in trans, we analyzed the novobiocin effects in LacZ activity from sdsR(−379/+222)-lacZ fusion rather than those in SdsR levels. Novobiocin had little effect on the increase of LacZ activity by the ryeA mutation, suggesting that transcription-induced changes in DNA supercoiling contribute little to the observed transcriptional interference. On the other hand, when we examined the ryeA and sdsR promoter activities during growth using lacZ transcriptional fusions, the ryeA and sdsR promoters showed transcription activities in all the growth phases, although ryeA promoter is more active in the exponential phase than the sdsR promoter (Fig. S7). These results altogether suggest the possibility that two oppositely transcribing RNA polymerases collide in the cell. Therefore, it seems likely that ryeA transcription inhibits transcription from the sdsR promoter through transcriptional interference, which may occur by RNA polymerase collision sometime between RNA polymerase binding to the

Discussion
In this study, we first defined the transcription units and biogenesis of sdsR and ryeA in more detail, showing that 3′ ends of both SdsR and RyeA correspond to their own transcription termini. We identified three transcription initiation start sites from the ryeA promoter. The primary ryeA transcripts are trimmed at the 5′ end to form processed RNAs with 5′ monophosphate whose 5′ end corresponds to the most downstream transcription start, which can generate RyeA of 272 nt with the longest 3′ end. On the other hand, the 5′ end of SdsR, which is the same as the sdsR transcription initiation site, retains triphosphate. This may explain that RyeA has a much shorter half-life than SdsR in the cell 18 because RNA with 5′ monophosphate is more vulnerable to degradation than RNA with 5′ triphosphate 38 . The transcription of sdsR terminates at base positions ranging from +100 to +102 relative to its own transcription start site, generating the longest SdsR (102 nt) in the cell.
RyeA is expressed in the exponential phase, whereas SdsR expression starts in late-stationary phase. Since we showed here that RyeA and SdsR are almost exclusively transcribed in vitro by Eσ 70 and Eσ S , respectively, acting at their respective promoters, the growth-dependent regulation of RyeA and SdsR should be mediated by the sigma factor selectivity of the two promoters. Promoter mutation analyses revealed that mutual degradation of SdsR and RyeA also contributes to discrete growth phase-dependent regulation of the expression of www.nature.com/scientificreports www.nature.com/scientificreports/ each sRNA. Furthermore, RyeA transcription can interfere with SdsR transcription, generating less SdsR when RyeA transcription occurs. As a result of this biosynthetic pathway, more restricted stationary phase expression of SdsR can be achieved through the coupled action of degradation of SdsR by RyeA expressed in the exponential phase and inhibition of SdsR transcription through transcriptional interference by RyeA transcription from the σ 70 -dependent ryeA promoter.
RNA-seq analysis showed that expression of pphA, encoding a phosphatase, downstream of sdsR was decreased by the sdsR promoter mutation and increased by the ryeA promoter mutation, suggesting that pphA expression is controlled by the sdsR/ryeA regulatory circuit. We showed that pphA mRNA is transcribed as an sdsR-pphA dicistronic RNA from the σ S -dependent sdsR promoter rather than from the previously reported heat-inducible pphA promoter. Thus, pphA mRNA may not be transcribed from the heat-inducible promoter, as suggested previously 30 . However, pphA expression differs from SdsR expression in that it is not affected in trans by RyeA although the ryeA transcription represses pphA expression. Therefore, pphA could be expressed at an earlier growth phase than SdsR. Trans-acting sRNAs usually downregulate expression of multiple target genes by inhibiting translation initiation of mRNA or by inducing degradation 5 . Our RNA-seq analysis showed that the sdsR promoter mutation upregulated more than 100 genes, but the ryeA promoter mutation led to only a few upregulated genes, suggesting that a major function of RyeA is to downregulate SdsR as a cis-encoded sRNA. The finding that the well-known SdsR target gene tolC 22 was repressed by the ryeA promoter mutation and activated by the sdsR mutation supports this possibility.
Our study showed that pphA expression begins with transcription from the sdsR promoter, and requires read-through transcription over the sdsR terminator and continued transcription into the pphA coding region. RNA-seq data showed that there was a considerable amount of read-through RNA over the sdsR terminator, but the amount of RNA continued to decrease with progression into the pphA open reading frame. Therefore, it is likely that other regulatory systems, such as premature termination or translational control, in addition to the sdsR/ryeA regulatory circuit are involved in pphA regulation. This remains to be determined in the future.
Antisense RNAs are expressed through convergent transcription, and their expression promotes transcriptional interference 37 . Transcriptional interference can come into play through various mechanisms, depending on the promoter strength and transcription velocity 39 . In this context, the sdsR/ryeA transcription unit is interesting because transcription levels from the ryeA and sdsR promoter vary during growth, leading to different transcriptional interference levels depending on growth conditions. It is likely that the sdsR/ryeA region at which transcriptional interference between two RNA polymerases occurs depends on the growth phase. SdsR transcription can be affected by convergent ryeA transcription at both initiation and elongation steps because the entire www.nature.com/scientificreports www.nature.com/scientificreports/ sdsR transcription unit is encompassed by the ryeA gene. Like SdsR expression, pphA expression is also affected by transcriptional interference by ryeA transcription.
To summarize, we herein characterized regulation of the expression of the toxin SdsR, including its downstream pphA gene, and the antitoxin RyeA (Fig. 9). SdsR expression is tightly regulated during growth through transcription from the σ S -dependent sdsR promoter, RyeA-mediated degradation of SdsR, and transcriptional interference from the σ 70 -dependent ryeA promoter. On the other hand, downstream pphA expression, like that of SdsR, is also under control of both sdsR and ryeA promoters, but is unaffected by RyeA. Therefore, the sdsR/ ryeA regulatory circuit plays a critical role in tightly controlling growth-dependent expression of SdsR toxin and pphA to ensure that they are not expressed during the exponential phase. experimental procedures strains, plasmids, and oligonucleotides. All strains and plasmids used in this study are listed in Table 2.
All primers and oligonucleotides used in this study are shown in Table 3. To generate sdsR and ryeB promoter mutant strains, each −10 element was changed to 'CTCGAG' using scarless mutagenesis, as described previously 40 . A series of lysogen-containing lacZ transcriptional fusion constructs was prepared. Briefly, various promoter regions around ryeA/sdsR were amplified, and the resulting fragment was cloned between the EcoRI and BamHI sites of the pRS1553 vector to generate lacZ transcriptional fusion plasmids. Lysogens were constructed by transforming E. coli strain DJ480 with the various fusion plasmids and transfecting with λRS468 to construct the corresponding lacZ fusion lysogens. Single-copy integration was confirmed by PCR 41 . For point mutations in sdsR-lacZ fusions, site-directed mutagenesis was performed, as described previously 18 . DJ480ΔsdsR/ryeA strain was generated from MG1655ΔsdsR/ryeA by P1-mediated transduction 42 and confirmed by sequence analysis of the amplified, knocked-out region. pRyeA and pSdsR carrying a pBR322 origin ectopically express IPTG-inducible RyeA and SdsR, respectively 43 . Plasmid pAKA-ara containing a pACYC184 origin was constructed by cloning the pBAD-AraC DNA sequence into the AvaI/EcoRI sites of plasmid pAKA, and used as a cloning vector to generate pRyeA-ara and pSdsR-ara, as described previously 44 , which express arabinose-inducible RyeA and SdsR, respectively. To generate sdsR-or ryeA-CAT fusion plasmids, sdsR or ryeA promoter-containing DNA fragments were obtained via PCR amplification of genomic DNA. The resulting PCR products were digested with BamHI/HindIII and ligated into pKK232-8. To generate sdsR-tRNA Arg -CAT fusion plasmids, a B. albidum tRNA Arg sequence 36,45 was amplified by PCR and inserted immediately upstream of the CAT gene in the sdsR-CAT fusion plasmids. Template DNA plasmids for in vitro transcription were prepared by cloning sdsR or ryeA promoter-containing DNA fragments into the HindIII/EcoRI site of plasmid pLS16 35 . The oligonucleotides employed are listed in Table 3. primer extension. An SdsR +59R primer (5′-GCT CTT GGG AGA GAG CCG-3′) was 5′ end-labeled with [γ-32 P]ATP using T4 polynucleotide kinase. The primer was then used to analyze sdsR(−379/+131)-CAT fusion transcripts. Total cellular RNA was isolated from cells carrying the sdsR(−379/+131-CAT fusion plasmid. The labeled primer was used for primer extension analysis, as previously described 46 .  In vitro transcription by E. coli RNA polymerase. In vitro transcription using E coli RNA polymerase was performed as described previously 46 . E. coli RNA polymerase holoenzyme Eσ 70 was purchased from New England Biolabs. σ S (RpoS) was purified using the plasmid described previously 46 . Eσ S was reconstructed by combining the core enzyme (New England Biolabs) and σ s in a molar ratio of 1:2. After 1 h reaction at 37 °C, the reaction was terminated by adding phenol/chloroform. Northern blotting. E. coli cells were grown overnight in LB broth in the presence of appropriate antibiotics. Overnight cultures were diluted 1:100 in fresh LB medium and further cultured at 37 °C. Total cellular RNA was extracted at the desired time points using the acidic hot-phenol method, as described previously 35 . RNA was generated in vitro using the T7 RiboMAX Express Large Scale RNA Production System (Promega). Northern blot analysis was carried out as described previously 35 . Briefly, 5-10 μg of total RNA was fractionated on a 5% polyacrylamide gel containing 7 M urea and electrotransferred to a Hybond-XL membrane (Amersham Biosciences). Membranes were hybridized with 32 P-labeled DNA probes in PerfectHyb Plus Hybridization Buffer (Sigma Aldrich) and analyzed using an FLA 7000 Image Analyzer (Fuji).
Mutual degradation assays of sdsR and RyeA. MG1655ΔsdsR/ryeA cells were co-transformed with a plasmid pair of pSdsR-ara/pRyeA or pRyeA-ara/pSdsR. Overnight cultures of the transformed cells were diluted (1:100) into LB medium containing ampicillin (100 μg/ml) and tetracycline (10 μg/ml), and grown at 37 °C for 2 h. Arabinose of 0.2% was added into the culture and cells were grown for 5 min or 25 min. Then 0.1 mM IPTG was added into each culture and cells were further grown further. Aliquots of the cell culture were taken in intervals and their RyeA and SdsR contents were analyzed by Northern blotting RNA-seq sample preparation and analysis. E. coli MG1655 wild-type, ryeA promoter mutant, and sdsR promoter mutant cells were grown at 37 °C overnight. After diluting 1:100, overnight cultures were grown for 3 h (exponential phase) and 8 h (stationary phase); total RNA was extracted at each point. All preparation procedures were as previously described 18 . All RNA sequencing and alignment procedures were conducted by ChunLab. Ribosomal RNA was depleted using a Ribo-Zero rRNA removal kit (Epicentre) according to the manufacturer's instructions. Libraries for Illumina sequencing were generated using a TruSeq Stranded mRNA Sample Prep kit (Illumina) according to the manufacturer's protocol. RNA sequencing was performed on the Illumina HiSeq 2500 platform using single-end 50-bp sequencing. Sequence data for the reference genome were retrieved from the NCBI database. Quality-filtered reads were aligned to the reference genome sequence using Bowtie2.
Transcript abundance was measured as Relative Log Expression (RLE). To screen for mRNAs whose levels differed by more than 2-fold versus wild-type cells, we filtered for mRNAs with EdgeR p-values < 0.05. For GO analyses, the 171 genes and 55 genes found to be up-or downregulated in the exponential phase and stationary phase, respectively, by the sdsR promoter mutation were sorted according to their GO categories (http://amigo. geneontology.org) 48 .
LacZ assay. Three colonies for each strain were cultured overnight in LB medium with/without ampicillin (100 μg/ml), after which overnight cultures were diluted 1:100 and cultured in fresh medium containing 0.02% arabinose and 1 mM IPTG. Cultures were incubated for 2 h for exponential phase and 10 h for stationary phase. LacZ activity was assayed as described previously 49 .

Name
Sequence 5′-3′ Use  www.nature.com/scientificreports www.nature.com/scientificreports/ semi-quantitative Rt-pCR and qRt-pCR. RNA extraction and DNase treatment were performed as described above for RNA-seq. cDNA was synthesized using M-MLV reverse transcriptase (Enzynomics) with specific primers for semi-quantitative RT-PCR. qRT-PCR was performed on an Exicycler 96 system (Bioneer) using Prime Q-master mix (Genet Bio). Expression levels of each mRNA of interest were normalized to that of the gapA gene 50 . All experiments were performed according to the manufacturer's instructions.

Data Availability
RNA-seq raw data for this study have been deposited in the National Center for Biotechnology Information Gene Expression Omnibus (GEO) and are accessible through the GEO Series accession number GSE122921.