Synergistic infection of two viruses MCMV and SCMV increases the accumulations of both MCMV and MCMV-derived siRNAs in maize

The co-infection of Maize chlorotic mottle virus (MCMV) and Sugarcane mosaic virus (SCMV) can cause maize lethal necrosis. However, the mechanism underlying the synergistic interaction between these two viruses remains elusive. In this study, we found that the co-infection of MCMV and SCMV increased the accumulation of MCMV. Moreover, the profiles of virus-derived siRNAs (vsiRNAs) from MCMV and SCMV in single- and co-infected maize plants were obtained by high-throughput sequencing. Our data showed that synergistic infection of MCMV and SCMV increased remarkably the accumulation of vsiRNAs from MCMV, which were mainly 22 and 21 nucleotides in length. The single-nucleotide resolution maps of vsiRNAs revealed that vsiRNAs were almost continuously but heterogeneously distributed throughout MCMV and SCMV genomic RNAs, respectively. Moreover, we predicted and annotated dozens of host transcript genes targeted by vsiRNAs. Our results also showed that maize DCLs and several AGOs RNAs were differentially accumulated in maize plants with different treatments (mock, single or double inoculations), which were associated with the accumulation of vsiRNAs. Our findings suggested possible roles of vsiRNAs in the synergistic interaction of MCMV and SCMV in maize plants.

against various viruses while an ago1 mutant is less susceptible to Tobacco rattle virus (TRV), to which ago4 mutant is more susceptible 9,21,[23][24][25] . Recent studies have also revealed that AGO5 plays a role in antiviral RNA silencing 26 . One model states that vsiRNAs are recruited into specific AGO complexes to function in antiviral silencing, which is preferentially, but not exclusively, dictated by their 5ʹ -terminal nucleotides 2,12,22,27,28 . AGO1, 2, 3, 5, 7 and 10 can bind to vsiRNAs and exhibit in vitro slicer activity 29 . Moreover, AGO1, 2, 3, 4, 5 and 9 can all bind to siRNAs derived from viruses or viroids in vivo, and AGO10 has been demonstrated to associate with siR-NAs derived from VSR-deficient Turnip mosaic virus (TuMV) 18,28,30,31 . Interestingly, it has recently been reported that AGO18, a member of a monocot-specific AGO protein clade, confers broad-spectrum virus resistance in rice by sequestering a host microRNA and is induced in virus-infected tissues 32 . One or more of RDR1, RDR2, and RDR6 have been shown to be involved in antiviral silencing by amplification of secondary vsiRNAs and exhibit specificity in targeting viral genome sequences 10,11,18,20,21,33 .
In addition to targeting viral RNAs, vsiRNAs have been predicted to target host mRNAs at post-transcriptional level using bioinformatics and a few studies have provided the experimental evidence [5][6][7][8][33][34][35] . It has been reported that the Arabidopsis At1g76950 mRNA can be down-regulated by vsiRNA derived from the Cauliflower mosaic virus 35S leader sequence and At1g30460 and At2g16595 mRNAs were specifically cleaved by vsiRNAs from Tobacco mosaic virus (Cg) 5,33 . In addition, two research groups simultaneously demonstrated that the Y-satellite of CMV produced a vsiRNA that could specifically and directly silence the ChlI gene in Nicotiana benthamiana and induce yellow symptoms 7,8 . Moreover, the chloroplastic heat-shock protein 90 mRNAs were targeted by the siRNA containing the pathogenic determinant of a chloroplast-replicating 34 . By degradome analysis and 5ʹ RACE, several host mRNAs were proved to be silenced by vsiRNAs from Grapevine fleck virus and Grapevine rupestris stem pitting-associated virus in a sequence-specific manner 35 . Recently, the tomato callose synthase genes were reported to be silenced by a small RNA derived from the virulence-modulating region of the Potato spindle tuber viroid 36 .
Maize chlorotic mottle virus (MCMV) in the genus Machlomovirus of the family Tombusviridae can infect various crops and lead to typical symptoms, such as mild mosaic, severe stunting, and leaf necrosis [37][38][39] . Maize lethal necrosis (MLN) is caused by the synergistic infection between MCMV and Maize dwarf mosaic virus (MDMV), Wheat streak mosaic virus (WSMV) or Sugarcane mosaic virus (SCMV), leading to serious yield losses in maize (Zea mays L.) [40][41][42] . The effect reported for these synergisms is a dramatic increase in MCMV concentrations in mix-infected plants compared with single-infected plants 40,41 . However, the mechanism underlying the synergistic interaction between MCMV and SCMV remains elusive. In this study, we found that the synergistic infection of MCMV and SCMV increased the accumulation of MCMV. Moreover, we obtained the profiles of vsiRNAs from SCMV and MCMV in singly and doubly infected maize plants by high-throughput sequencing, respectively. The characters of vsiRNAs were analysed and the target genes of some vsiRNAs were predicted and annotated. In addition, we investigated the gene expression of maize DCLs and several AGOs through the characterization of the mRNA accumulation in phosphate buffer (mock), SCMV, MCMV or co-inoculated (S + M) maize plants.

Synergistic infection of MCMV and SCMV increases the accumulation of MCMV.
To understand the synergistic interaction of MCMV and SCMV, maize seedlings at the third leaf stage were inoculated with phosphate buffer (Mock), SCMV, MCMV and both viruses (S + M), respectively. The first systemically infected leaves of co-infection became significantly chlorotic at 9 days post inoculation (dpi) and developed necrotic areas at 10 dpi while the leaves of SCMV or MCMV single infection showed mosaic or chlorotic symptom at 9 or 10 dpi, respectively (Fig. 1A,B). Total RNA was isolated from the systemically infected leaves for Northern blotting and quantitative reverse transcription (RT)-PCR (qRT-PCR) analyses at 9 dpi. The accumulation level of MCMV genomic RNAs was higher in mix-infected leaves than that in single-infected leaves (Fig. 1C), while the SCMV RNA was slightly decreased (Fig. 1D). The expression levels of SCMV and MCMV CP were quantified by Western blotting. The results showed that MCMV CP accumulated to higher levels in mix-infected leaves as did the genomic RNAs (Fig. 1E). However, there was no obvious difference in the expression level of SCMV CP between mix-infected and single-infected leaves (Fig. 1F). Taken together, synergistic infection increased the expression levels of both the genomic RNAs and coat protein of MCMV, but decreased the accumulation of SCMV RNA.

Profiles of vsiRNAs in SCMV-, MCMV-or co-inoculated maize plants.
To obtain the profiles of vsiRNAs produced during viral infections, total RNA was extracted from the systemically infected leaves of maize plants inoculated with buffer (Mock), SCMV, MCMV and S + M at 9 dpi, respectively. The cDNA libraries of small RNAs generated from the total RNA were analysed by high-throughput sequencing on the Illumina Solexa platform yielding around 10,000,000 sequences for each library (Table 1). Reads ranging from 18-to 36-nt were mapped to the viral genomes within 2 mismatches in both sense and antisense orientations. A total of 1,255,641 and 6,740,592 vsiRNA reads were obtained in MCMV-and SCMV-inoculated maize plants, respectively, and 2,044,540 MCMV-derived siRNA (M-vsiRNA) and 6,520,905 SCMV-derived siRNA (S-vsiRNA) reads were obtained in S + M inoculated maize plants (Table 1). Further analysis showed that 22-and 21-nt M-vsiRNAs accumulated to higher levels in S + M inoculated maize plants than that in MCMV inoculated maize plants, which accounted for 87.06% and 46.51% of total M-vsiRNAs, respectively ( Fig. 2A). Moreover, the percentage of 22-nt M-vsiRNAs was higher than that of 21-nt M-vsiRNAs in both MCMV and S + M inoculated maize plants (23.73% compared with 22.79%, and 49.07% compared with 37.98%, respectively) ( Fig. 2A). The majority of S-vsiRNAs were 21 nt and 22 nt in length in both SCMV and S + M inoculated maize plants, representing 94.19% and 93.54% of total S-vsiRNAs, respectively, and the 21-nt S-vsiRNAs accumulated more than 22-nt S-vsiRNAs (  Previous studies have demonstrated that the 5ʹ -terminal nucleotides of small RNAs mainly determined the recruitment by specific AGOs 27,28 . To explore the potential interactions between vsiRNAs and distinct AGO complexes, the nucleotide at the 5ʹ -terminal position in vsiRNA sequences was analysed. There was an A preference existing in the 5ʹ -terminal nucleotides of M-vsiRNAs in MCMV and S + M inoculated maize plants, which were mainly recruited by AGO2 and/or AGO4, although the percentage was slightly decreased in the co-infected maize plants (32.61% compared with 35.28%) (Fig. 4A). Like in M-vsiRNAs, A was the most abundant at the 5ʹ -terminal nucleotide of S-vsiRNAs and there was no obvious difference in SCMV and S + M inoculated maize plants (Fig. 4B). Interestingly, vsiRNAs with 5ʹ -terminal U, which were mainly loaded into AGO1, accounted for a low percentage (17-19%) of total vsiRNAs (Fig. 4).   and CP coding regions and the 3ʹ -UTR region (Fig. 5E). There were no obvious differences in the distribution  Prediction and annotation of the putative target genes of vsiRNAs. The putative target genes of vsiRNAs were predicted using the MiRnada, which was an algorithm for finding the targets of miRNAs 43 . In our study, only these vsiRNAs with high abundant reads (top 50 derived from either genomic strand or complementary strand) were selected to predict target genes in co-infected maize plants (Supplementary Table S1). Thousands of target genes were predicted and only those whose scores were not less than 180 were presented (Supplementary Table S2). Moreover, we selected the putative target genes whose scores were not less than 170 for further analyses. To understand the putative roles of the predicted target genes of vsiRNAs in maize, we conducted Gene Ontology (GO) analysis using Blast2GO in terms of biological processes, cellular components, and molecular functions 44 . A total of 1969 and 1560 predicted target genes of M-vsiRNAs and S-vsiRNAs were annotated by GO analysis, respectively. The percentage of the predicted target genes with a particular category in the total GO-annotated target genes was indicated in Fig. 6. Interestingly, the major categorized groups of the predicted target genes of vsiRNAs were similar for both MCMV and SCMV. 'Metabolic process' was the most highly represented group under the biological process category, followed by 'regulation of transcription, DNA-templated' , 'protein phosphorylation' and 'oxidation-reduction process' . For the cellular component category, 'membrane' , 'nucleus' and 'integral component of membrane' were the most highly represented groups. With regard to the molecular function category, 'binding' groups were the most highly represented, mainly including ' ATP binding' , 'nucleotide binding' , 'metal ion binding' and 'DNA binding' . To demonstrate whether or not the predicted target genes with a particular category were enriched, we used the percentage of the genes with specific category in all GO-annotated maize genes as the control.
To get a better understanding of the special biochemical pathways for the predicted target genes of vsiRNAs, we assigned them based on the KEGG database using BLASTx 45 . A total of 816 and 674 predicted target genes of M-vsiRNAs and S-vsiRNAs were aligned in the KEGG database, respectively, and the percentage of the predicted target genes with a specific category were shown (Fig. 7). The metabolism pathway contained most of the predicted target genes of vsiRNAs for MCMV and SCMV, in which the most frequently represented pathway was 'Metabolic pathways' , followed by 'Biosynthesis of secondary metabolites' . To understand the enrichment of the predicted target genes in a particular category, the percentage of the genes with specific category in all KEGG-annotated maize genes was used as the control. Northern blotting confirmed that the co-infection of SCMV and MCMV increased the accumulation of M-vsiRNAs. To confirm the impact of synergistic infection on the production of vsiRNAs, total RNA was extracted to analyse the accumulation of vsiRNAs from different genome positions of both (+ )-and (− )-sense strands of SCMV and MCMV by Northern blotting, respectively. The results showed that the vsiRNAs could be detected, although the signal of M831 (− ) was weak, which demonstrated the existence of vsiRNAs (Fig. 8). The M-vsiRNAs from S + M co-infected maize plants accumulated more than those from MCMV infected maize plants for M-vsiRNAs derived from both MCMV genomic and complementary strands (Fig. 8A,B), although the results of high-throughput sequencing revealed that the level of M4330 (+ ) was decreased (Fig. 5C,E). Moreover, for all selected M-vsiRNAs from S + M co-infected maize plants, there was a preference for 22-nt while almost equivalent 21-and 22-nt M-vsiRNAs accumulated in MCMV infected maize plants except for M4375 (+ ), which had a 22-nt preference (Fig. 8A,B). However, there were no obvious differences in the accumulation of S-vsiRNAs selected between SCMV and S + M infected maize plants (Fig. 8C). For S693 (+ ) and S4541 (+ ) from the genomic strand of SCMV, 21-and 22-nt vsiRNAs accumulated almost equally as well as S5017 (− ) while S109 (− ) had a preference for 22 nt, of which both derived from the complementary strand (Fig. 8C). The results obtained by Northern blotting revealed that synergistic infection increased  showed that the expression of DCL2 mRNA was significantly increased in the singly and doubly infected maize plants (Fig. 9A). However, other DCLs mRNAs were down-regulated in SCMV, MCMV and S + M infected maize plants except for DCL3b in SCMV infection (Fig. 9A). Interestingly, S + M infection had the significant influence on the expression of maize DCLs mRNAs: the highest expression level of DCL2 was detected and other DCLs accumulated to the lowest levels (Fig. 9A). In maize AGO1 homologs, the expression of AGO1a and AGO1b were up-regulated while that of AGO1c was unchanged in SCMV infected maize plants (Fig. 9B). However, the MCMV and S + M infection decreased the expression of AGO1c and had no obvious effect on AGO1a and AGO1b (Fig. 9B). Moreover, AGO2a and AGO18a were significantly up-regulated by viral infections, especially in S + M infected maize plants, which accumulated to the highest levels (Fig. 9C,D). These data suggested that viral infections differentially modified the expression of components involved in antiviral RNA silencing pathway.

Discussion
In plants, synergistic interactions between independent viruses in mixed infections have been well documented 46 , but the mechanism underlying these interactions remains elusive. In our study, the expression levels of MCMV genomic RNAs and CP were increased in S + M co-infected maize plants compared with that in MCMV infected maize plants (Fig. 1C,E), in agreement with the results of previous reports 40, 41 . It has been demonstrated that HC-Pro, the silencing suppressor encoded by potyviruses, could enhance the pathogenicity and accumulation of other heterologous viruses [46][47][48] . Moreover, the synergistic infection of WSMV and MCMV was independent of WSMV HC-Pro, which was not a silencing suppressor 48,49 . However, the effects of SCMV HC-Pro as well as WSMV P1 on the synergistic infections remain to be investigated, which have been proved to function as suppressors of RNA silencing 50,51 .
RNA silencing is a conserved surveillance mechanism in the defence against viruses in plants, which can trigger the production of vsiRNAs in virus-infected plant cells. In this study, the profiles of vsiRNAs from SCMV and MCMV in singly and doubly infected maize plants were obtained to understand the role of RNA silencing in the The expression levels were determined by qRT-PCR at 9 dpi. Three independent experiments were conducted with at least 3 biological replicates each and the data were analysed using a two-sample t-test. Bars represented the grand means ± SD. Lowercase letters indicate significant difference (P-value < 0.05).
Scientific RepoRts | 6:20520 | DOI: 10.1038/srep20520 synergistic interaction between SCMV and MCMV in maize plants. In the SCMV singly or doubly (with MCMV) infected maize plants, S-vsiRNAs accounted for more than half of total small RNAs (Table 1), similar to the results of our previous report 52 . However, the accumulation level of M-vsiRNAs was lower compared with endogenous small RNAs within a library, accounting for 14.75-19.49% of total small RNAs (Table 1). Further analysis of S + M library suggested that there was a preference to SCMV RNAs for RNA silencing, which accumulated more S-vsiRNAs than M-vsiRNAs ( Table 1). The nonsense-mediated decay (NMD) was reported to recognize and eliminate viral RNAs with internal termination codons and long 3′ -UTRs, but it had no effect on the potyvirus 53 . NMD, as a general virus restriction mechanism in plants, might compete for MCMV RNA substrates with RDR and decrease the accumulation of M-vsiRNAs in MCMV singly and doubly (with SCMV) infected maize plants.
In effect, saturation of NMD by increasing amounts of viral RNAs may constitute a switch for RDR action and secondary RNA silencing during viral infection 53 . In co-infected maize plants, the increased accumulation levels of M-vsiRNAs might be the results of processing the increased accumulation of MCMV RNAs by RNA silencing.
In positive-strand RNA virus-infected plants, DCL4-dependent 21-nt vsiRNAs are predominant than DCL2-dependent 22-nt vsiRNAs, which accumulated to higher levels in the absence of DCL4 11,[15][16][17] . In the MCMV singly and doubly infected maize plants, 22-nt M-vsiRNAs accumulated to higher levels than 21-nt M-vsiRNAs ( Fig. 2A), which might be the results of increased accumulation levels of maize DCL2 mRNAs and decreased levels of DCL4 (Fig. 9A). However, the majority of S-vsiRNAs were 21 nt in length in the SCMV singly and doubly infected maize plants (Fig. 2B), indicating that DCL4 played a major role in the biogenesis of S-vsiRNAs and had a preference for processing SCMV RNA, although the accumulation levels of maize DCL4 mRNAs were down-regulated (Fig. 9A). In the S + M infected maize plants, 22-nt S-vsiRNAs accumulated to higher levels while 21-nt S-vsiRNAs were decreased compared with that in SCMV infected maize plants (Fig. 2B), which associated with the changed accumulation level of maize DCL4 and DCL2 mRNAs. The accumulation levels of both 22-nt vsiRNAs and maize DCL2 mRNAs were increased in the singly and doubly infected maize plants, indicating that DCL2 played an important role in the production of vsiRNAs, which supported the model that cooperative interaction between DCL4 and DCL2 was necessary during systemic antiviral silencing 11,19,26 . As a result of viral infections, the decreased expression levels of maize DCL1, DCL3a and DCL3b mRNAs might affect the accumulations of miRNAs and the methylation of DNA and/or histone of maize, however, their roles in the defence against RNA viruses remain to be investigated.
For a long time, the dsRNA replication intermediates were thought to be the major origin of vsiRNAs from positive-strand RNA viruses. However, it has been reported that the vsiRNAs had a positive sense strand bias by high-throughput sequencing, suggesting that vsiRNAs originated predominantly from highly structured single-stranded viral RNAs 4,54 . Our results, in agreement with our previous report 52 , demonstrated that almost equal amount of (+ )-and (− )-sense S-vsiRNAs existed in SCMV-infected maize plants (Fig. 3B), indicating that most of the S-vsiRNAs were likely generated from dsRNA precursors. In MCMV infected maize plants, the (+ )-sense M-vsiRNAs accumulated more than those from the (− )-sense strand (Fig. 3A), suggesting that the majority of M-vsiRNAs were derived from MCMV genomic RNAs. Nevertheless, recent study has revealed that genomic viral RNAs might sequester complementary vsiRNAs during gel electrophoresis 55 , and the sequestration could be decreased by the fully-denaturing formaldehyde polyacrylamide gel electrophoresis (FDF-PAGE) 56 . By applying FDF-PAGE, the predominant precursor of vsiRNAs was demonstrated to be a long dsRNA, however, whether this conclusion is relevant to the origin of M-vsiRNAs remains to be studied.
The small RNAs are loaded into distinct AGO-containing RISCs to function, which are mainly directed by the 5′ -terminal nucleotide 27,28 . For example, AGO1 preferentially associates with small RNAs that have a 5′ U, AGO2 and AGO4 have a 5′ A preference, and AGO5 mainly associates with small RNAs that begin with a 5′ C. For M-vsiRNAs and S-vsiRNAs, A was the most abundant nucleotide at the 5ʹ -end (Fig. 4), suggesting that these vsiRNAs might be mainly recruited by AGO2 and/or AGO4. Interestingly, the accumulation levels of maize AGO2 mRNAs were induced in singly and doubly infected maize plants (Fig. 9C), which further increased the possibility that maize AGO2 participated in antiviral defence. Previous reports have shown that AGO1 played a dominant role in the defence against RNA viruses 21,23 . However, the accumulation levels of maize AGO1a and AGO1b mRNAs remained unchanged even AGO1c was decreased in MCMV or S + M infected maize plants, although SCMV infection slightly induced the expression of AGO1a and AGO1b mRNAs (Fig. 9B). Moreover, the vsiRNAs from SCMV and MCMV with a 5′ -terminal U accounted for a small proportion (Fig. 4). These data suggested that the AGO1 might play a less important role than AGO2 in the defence against SCMV and MCMV as the results obtained by recent studies 25,26,31 . In addition, the presence of large amounts of vsiRNAs with 5′ -terminal G or C revealed that other AGOs might also be recruited to form specific RISCs and involved in antiviral defence, which were reported to bind siRNAs from viruses or viroids 18,[28][29][30][31] .
Recent research demonstrated that AGO18, a member of a monocot-specific AGO protein clade, played a role in antiviral defence by sequestering miR168 and was induced in virus-infected tissues 32 . In our study, we detected the expression level of maize AGO18a gene, a homolog of rice AGO18, by qRT-PCR in buffer (Mock), SCMV, MCMV and S + M inoculated maize plants (the expression level of maize AGO18b was almost undetectable in maize leaves 57 ). The results showed that the accumulation of AGO18a mRNA was significantly induced after viral infections, especially in MCMV and S + M co-infected maize plants (Fig. 9D). We also found that miR168 level was up-regulated in SCMV infected maize plants ( Supplementary Fig. S1), in addition to the results that AGO1a and AGO1b mRNAs were up-regulated (Fig. 9A), suggesting that miR168 could be sequestered by AGO18a as reported previously 32 . Interestingly, the accumulation of miR168 had no obvious change in MCMV or S + M infected maize plants in which the AGO1a, b, c mRNAs were not induced (Fig. 9B), suggesting that the significantly induced AGO18a might be involved in antiviral defence by other modes of action, such as influencing the function of other miRNAs associated by AGO18 32 . However, the antiviral roles of AGO18 remain to be elucidated in maize plants, especially in MCMV and S + M co-infected maize plants.  58 . MCMV was prepared from the full-length cDNA clone (pMCM41) provided by Dr Kay Scheets. Crude extracts from SCMV or MCMV-infected maize leaf tissues were prepared as described previously 59 , which were then equally mixed as the source of co-infection while equal volume of phosphate buffer was added for single infection, respectively.
Small RNA sequencing and Bioinformatics analyses. At approximately 9 dpi, before the leaves showed necrosis symptoms as shown in Fig. 1A, the systemically infected leaves were harvested and maintained at − 80 °C. With each treatment, the systemically infected leaves of at least 15 maize seedlings were pooled for small RNA sequencing. Total RNA was treated as described 52 and subjected to Solexa/Illumina sequencing by SBC (Shanghai Biotechnology Corporation, Shanghai, China).
After excluding low quantity reads and 5ʹ -and 3ʹ -adaptor contaminants, the raw reads were obtained. Small RNAs of 18-36 nt in length were extracted and only the sequences identical or complementary to viral genomic sequences within 2 mismatches were recognized as vsiRNAs for further analysis. Small RNA sequences were analysed as described 52 .
Target Gene Prediction and Analysis. The MiRnada program was used to predict maize mRNAs targeted by vsiRNAs from SCMV and MCMV in co-infected maize plants 43 . The criteria used were as described previously 52 . The predicted target genes were assigned to various GO and KEGG classifications as reported previously 44,45 .

RNA blot analysis.
Approximately 40 μ g of total RNA was prepared for small RNA blot analysis, and the blots were probed and washed as previously reported 52 . Probe sequences used for small RNA blot analysis were shown in Supplementary Table S3.
About 2 μ g of total RNA was used to detect MCMV by Northern blot analyses. Northern blots were performed with [α -32 P] dCTP randomly-labelled cDNA probes, which were from 3ʹ -terminal 712 nucleotides (3721-4432) of MCMV genome. Random Primer DNA Labelling Kit Ver.2 was used for labelling the probes as instructed by the manufacturer (Takara Bio Inc., Dalian, China). Blots were hybridized at 65 o C overnight using hybridisation buffer (Sigma, USA) and washed as instructed by the manufacturer.
Western blot assay. The protein extraction and Western blot assay were performed as described previously 60 . SCMV CP polyclonal antibody was used at dilution of 1:5000. MCMV CP antibody was kindly provided by Prof. Xueping Zhou (Zhejiang University) and used at dilution of 1:8000.
Quantitative Real-time RT-PCR. Total RNA was extracted using TRIzol reagent as instructed by the manufacturer (Invitrogen, USA) and treated with RNase-free DNase I (Takara Bio Inc., Dalian, China). About 2 μ g of total RNA was used to synthesize the first-strand cDNA with an oligo (dT) primer and the qRT-PCR was performed as previously reported 60 . Maize UBI (ubiquitin) was used as an internal standard. The sequence information of maize DCLs and AGOs was reported in two references 52,57 . The information of the primers used in the qRT-PCR experiments were listed in Supplementary Table S4. Three independent experiments were performed with biological and technical replicates.