A double-edged sword in antiviral defence: ATG7 binding dicer to promote virus replication

Abstract

RNA interference (RNAi) and autophagy are two pivotal biological processes that regulate virus replication. This study explored the complex relationship between autophagy and RNAi in controlling influenza virus replication. Initially, we reported that influenza virus (H9N2) infection increases the viral load and the expression of autophagy markers while inhibiting the RNAi pathway. Subsequent studies employing autophagy enhancer and inhibitor treatments confirmed that avian influenza virus (AIV, H9N2) promotes viral replication by enhancing autophagy pathways. Further analysis revealed that ATG7, an autophagy protein, can interact with dicer to affect its antiviral functions. Finally, we discovered that infection with other avian RNA viruses, including infectious bursal disease virus (IBDV) and infectious bronchitis virus (IBV), induced the upregulation of ATG7, which blocked the RNAi pathway to facilitate virus replication. Our findings suggested that virus infection might trigger the upregulation of autophagy and downregulation of the RNAi pathway, revealing a complex interaction between these two biological processes in the defence against viral replication.

Supplementary Information

The online version contains supplementary material available at 10.1007/s00018-025-05603-1.

Introduction

Autophagy is essential for maintaining cell homeostasis since it recycles damaged proteins or organelles and even removes intracellular infections. This process involves the engulfment of intracellular cargo in autophagosomes, which then combine with lysosomes to complete cargo destruction [1–3]. Autophagy-related protein 7 (ATG7) is an E1-like enzyme that participates in the synthesis of microtubule-associated protein 1 light chain 3 (LC3-II), which is required for the autophagy process [4, 5]. Notably, many studies have revealed that autophagy plays a role in the replication of viruses, such as influenza virus [6, 7], avian leukosis virus subgroup J [8], and infectious bursal disease virus [9, 10].

RNA interference (RNAi) contributes to cell-intrinsic antiviral immunity in a variety of eukaryotic organisms, including plants, birds, and mammals [11]. In the RNAi antiviral process, the Dicer protein cleaves the double-stranded RNA (dsRNA) formed during virus replication into 21–23 nt virus-derived siRNA (vsiRNA). Subsequently, vsiRNAs are loaded onto the Ago2 protein and target the virus’s homologous RNA to block virus replication [12]. There are currently many studies on the interaction between autophagy and RNAi, but there is no detailed explanation of how autophagy and RNAi interact during viral replication [13–15]. RNAi can affect autophagy-related genes in various ways, thereby regulating the autophagy process [16, 17]. Moreover, autophagy can potentially impact RNAi effectiveness by breaking down dsRNA to preserve cell homeostasis [18, 19]. Since both autophagy and RNAi are involved in regulating virus replication, we wondered how these two processes collaborate in defending against virus replication [20–22].

A recent study reported that autophagy is involved in influenza virus replication and that the inhibition of autophagy results in a significant decrease in the influenza virus titer [23, 24]. Several studies have reported that autophagy and RNAi are involved in the replication of influenza viruses, but the association between autophagy and RNAi has never been explored [6, 23]. In this study, we selected the avian influenza virus (AIV, H9N2) as a model to illustrate whether and how cell autophagy promotes viral self-replication or clears intracellular viruses, especially whether this process involves RNAi.

Results

AIV-H9N2 infection affects autophagy and the RNAi pathway

Autophagy and RNAi are important antiviral regulatory mechanisms in eukaryotes. Hence, we determined whether autophagy and RNAi are involved in AIV-H9N2-infected DF-1 cells. Initially, the immunofluorescence assay (IFA) and plaque assay results revealed that the viral load increased with increasing MOI from 0.01 to 1 and that H9N2 infection at an MOI of 1 caused severe cellular damage (Fig. 1A-B). Moreover, Western blotting revealed that AIV-H9N2 infection at different MOIs decreased RNAi-related protein (Dicer) levels and increased autophagy-related protein levels (Fig. 1C-D and S1A). Moreover, we selected 0.1 MOI AIV-H9N2 for the subsequent experiments. With the prolongation of 0.1 MOI AIV-H9N2 infection, the NS1 and HA levels of AIV-H9N2 and the LC3II/LC3I ratio increased, and the Dicer and P62 levels decreased (Fig. 1E-G and S1B). A decrease in P62 and an increase in the LC3II/LC3I ratio represent the occurrence of autophagy. These results indicate that AIV-H9N2 infection causes upregulation of the autophagy pathway and downregulation of the RNAi pathway.

Fig. 1.

Autophagy and RNAi are involved in AIV-H9N2 replication. A. DF-1 cells were infected with different MOIs (0.01, 0.1, and 1) of AIV-H9N2 for 48 h, and then, immunofluorescence was used to detect AIV-H9N2-HA protein expression. The blue fluorescence represents DAPI-stained nuclei, whereas the red fluorescence represents the AIV-H9N2-HA protein. The magnification is 10×. The right image shows the fluorescence intensity analysis. B. DF-1 cells were infected with AIV-H9N2 at different MOIs (0.01, 0.1, and 1) for 48 h, and the supernatant was collected and incubated with MDCK cells. The release of virus particles was detected through a plaque assay after 72 h, and statistical analysis of the number of speckles was performed. C-D. Western blotting detection of autophagy- and RNAi-related protein expression levels in DF-1 cells infected with different MOIs (0.01, 0.1, and 1) of AIV-H9N2 (C) and grayscale analysis of the data in Fig. 1C (D). GAPDH was used as an internal reference protein, and the expression level of the blank group was set to “1”, while the expression levels of the other groups were multiples of that of the blank group. E-G. RT-qPCR (E) and Western blotting (F) detection of RNAi- and autophagy-related protein expression levels in DF1 cells infected with 0.1 MOI H9N2 at different time points; grayscale analysis of the data in Fig. 1F (G). The data are shown as the means ± SDs, n = 3. The differences were not significant at P ≥ 0.05 (ns) or significant at P < 0.05 (*), P < 0.01 (**)

Autophagy and the RNAi pathway are involved in AIV-H9N2 replication

Previous studies have suggested that AIV-H9N2 infection affects both the autophagy and RNAi pathways. We then wondered whether autophagy and RNAi could influence virus replication. On the one hand, we constructed Dicer overexpression and silencing plasmids to investigate whether the increase in the AIV-H9N2 virus is related to RNAi (Fig. 2A-B and S1C). The Western blotting and RT-qPCR results revealed that the plasmids successfully overexpressed (PC-Dicer) and knocked down (PS-Dicer) Dicer. In addition, plaque and Western blotting analyses revealed that overexpression of the Dicer protein can reduce the number of virus particles and protein levels of AIV-H9N2, whereas silencing dicer modestly increased virus replication (Fig. 2C-F).

Fig. 2.

RNAi is an important antiviral pathway, and autophagy is beneficial for AIV-H9N2 replication. A. DF-1 cells were transfected with the Dicer overexpression plasmid PC-Dicer for 48 h, and then Western blotting was used to detect the expression of the Dicer protein. GAPDH was used as an internal reference protein, and the expression level of the blank group was set to “1”, while the expression levels of the other groups were multiples of that of the blank group. The PC group was used as the control. B. DF-1 cells were transfected with Dicer knockdown plasmids (PS-Dicer-01, PS-Dicer-02, and PS-Dicer-03), and Western blotting was used to detect Dicer protein expression to determine the silencing efficiency. The PS group was used as the control. C-F. DF-1 cells were transfected with the PC-Dicer plasmid or PS-Dicer plasmid for 24 h, and then 0.1 MOI AIV-H9N2 was used for infection for 48 h. The supernatant was collected and incubated with MDCK cells. The release of virus particles was detected through a plaque assay after 72 h, and statistical analysis of the number of speckles was performed (C-D). For 48 h, Western blotting (E) and gray value analysis (F) were performed to detect the expression of the AIV-NS1 protein. G-H. DF-1 cells were treated with RAPA (100 nM) and CQ (100 µM) for 4 h and then infected with 0.1 MOI AIV-H9N2 for 48 h. The protein expression of H9N2-NS1 and the autophagy-related proteins P62, LC3I, and LC3II was detected via Western blotting (G), and grayscale analysis was performed below the band (H). I-J. DF-1 cells were transfected with PC-Dicer for 48 h, and RT-qPCR (I) and Western blotting (J) were used to detect LC3II/LC3I expression. The PC group was used as the control. The data are shown as the means ± SDs, n = 3. The differences were not significant at P ≥ 0.05 (ns) or significant at P < 0.05 (*)

On the other hand, we tested the cytotoxicity of the autophagy enhancer rapamycin (RAPA) and the autophagy inhibitor chloroquine (CQ) on DF-1 cells and found that these two reagents did not exhibit significant cytotoxicity at working concentrations (Figure S1D). Further Western blotting and plaque assays revealed that the activation of autophagy by the activator rapamycin (RAPA, 100 nM) was beneficial for the replication of AIV-H9N2. In contrast, the use of the autophagy inhibitor chloroquine (CQ, 100 µM) effectively reduced the viral load of AIV-H9N2 (Fig. 2G-H and S1E-F). We also found that Dicer overexpression did not significantly influence the expression level of LC3II/LC3I according to the RT-qPCR and Western blotting results (Fig. 2I-J). For further investigation, DF-1 cells were transfected with PC-Dicer or PS-Dicer for 24 h and then infected with AIV-H9N2 at an MOI of 1 or treated with 100 nM RAPA for 24 h. RT-qPCR revealed that dicer did not influence the RAPA-mediated autophagy pathway but influenced the virus-mediated autophagy pathway (Figure S2A-B). The above phenomenon may be caused by dicer affecting virus replication, and more experiments are needed to confirm the specific verification of this phenomenon. Finally, the above results indicated that both autophagy and RNAi could influence AIV-H9N2 replication.

Autophagy suppresses dicer-dependent antiviral RNAi response via ATG7

As a highly conserved antiviral mechanism, RNAi is essential for thwarting RNA viruses. First, to explore whether the inhibition of RNAi in AIV-H9N2-infected cells is related to the upregulation of autophagy, we analyzed the expression of dicer after treatment with the autophagy activator rapamycin (RAPA) and the autophagy inhibitor chloroquine (CQ). The Western blotting and RT-qPCR results revealed that 100 nM RAPA inhibited Dicer expression, whereas 100 µM CQ promoted Dicer expression, indicating that the activation of the autophagy pathway inhibited dicer expression (Fig. 3A-B and S2C). Moreover, to investigate which autophagy-related protein affects Dicer expression, we tested 12 autophagy-related genes and found that AIV-H9N2 infection significantly increased the expression of HIF1A and ATG7 but decreased the expression of NDP52 and Beclin1 (Fig. 3C-D and S2C-D). We subsequently constructed NDP52, Beclin1, HIF1A, and ATG7 overexpression plasmids to test their effects on Dicer (Fig. 3E and S3A-B). The Western blotting results revealed a significant decrease in dicer in the ATG7 group and a modest increase in dicer in the Beclin1 group compared with the NDP52 and HIF1A groups (Fig. 3F-I).

Fig. 3.

AIV-H9N2 infection promotes the upregulation of autophagy and inhibits the Dicer-dependent antiviral RNA response via ATG7. A-B. DF-1 cells were treated with 100 nM RAPA or 100 µM CQ for the indicated time points (12 h, 24 h, and 48 h), after which the protein expression levels of Dicer, P62, LC3I, and LC3II were detected via Western blotting (A). The right panel shows the gray value analysis for Fig. 3A (B). C. DF-1 cells were treated with 100 nM RAPA or 100 µM CQ for 24 h, after which the mRNA expression levels of autophagy-related genes were detected via RT-qPCR. D. DF-1 cells were infected with 0.1 MOI AIV-H9N2 for the indicated time points (6 h, 12 h, 24 h, 36 h, and 48 h), after which the protein expression levels of NDP52, HIF1A, Beclin1, and ATG7 were detected via Western blotting. E. DF-1 cells were transfected with the overexpression plasmids PC-Beclin1, PC-HIF1A, PC-ATG7, and PC-NDP52 and then subjected to protein expression detection by immunofluorescence, with the blank group as a negative control and the PC group (containing the mRFP gene) as a positive control. The magnification is 10×. F-I. DF-1 cells were transfected with PC-NDP52 (F), PC-HIF1A (G), PC-Beclin1 (H), or PC-ATG7 (I) for 48 h, after which the protein expression of dicer was detected via Western blotting. The image on the right shows a grayscale analysis. The data are shown as the means ± SDs, n = 3. The differences were not significant at P ≥ 0.05 (ns) or significant at P < 0.05 (*), P < 0.01 (**)

ATG7 directly interacts with dicer via the D684 site to facilitate AIV-H9N2 replication

A previous study suggested that AIV-H9N2-induced autophagy might block cellular RNAi by influencing dicer. We then wondered whether the above-selected 4 autophagy-related genes might directly interact with dicer to promote viral replication. Initially, co-immunoprecipitation (co-IP) results revealed that the ATG7 protein can interact with Dicer (Fig. 4A). Both exogenous and endogenous interactions occur between ATG7 and Dicer (Fig. 4B). To confirm whether ATG7 can bind to dicer to inhibit its antiviral function and promote viral replication, we also designed ATG7-knockdown plasmids, and the Western blotting results revealed that the silencing effect of PS-ATG-03 was greatest (Fig. 4C). In addition, plaque assays confirmed that ATG7 overexpression strongly increased the virus titer, whereas ATG7 knockdown significantly decreased the virus load (Fig. 4D). Subsequent Western blotting revealed that ATG7 overexpression resulted in a modest decrease in Dicer protein expression and a slight increase in NS1 protein expression, which suggested that the ATG7 protein is important for inhibiting Dicer protein expression and promoting AIV-H9N2 replication (Fig. 4E-F). Moreover, molecular docking was performed to study further how ATG7 interacts with the Dicer protein.

Fig. 4.

ATG7 protein promotes AIV-H9N2 replication through interaction with dicer. A. 293T cells were transfected with PC-Dicer, PC-NDP52, PC-HIF1A, PC-Beclin1, or PC-ATG7 for 48 h and then pulled down with the Dicer antibody for the CO-IP experiment. The lysates were subjected to detection of NDP52, HIF1A, Beclin1, ATG7, and dicer protein expression via Western blotting. In addition, the internal control and IgG were used as inputs for the negative control. B. Supplement to Fig. 3A. DF-1 cells were transfected with PC-Flag or PC-ATG-Flag for 48 h and subjected to co-IP. In addition, the Flag antibody was used for pulldown on the left, and the Dicer antibody was used on the right. β-actin was added as the control. C. DF-1 cells were transfected with ATG7-knockdown plasmids (PS-ATG7-01, PS-ATG7-02, and PS-ATG7-03) for 48 h, and the silencing effect of ATG7 was detected via Western blotting. The image below the strip is used for gray value analysis. D-F. DF-1 cells were transfected with PC-ATG7 or PS-ATG7 for 24 h and infected with 0.1 MOI AIV-H9N2 for 48 h. The supernatant was collected and incubated with MDCK cells. The release of virus particles was detected through a plaque assay after 72 h. Statistical analysis of the number of speckles (D). In addition, the 48 h samples were subjected to Dicer and H9N2-NS1 expression Western blotting, and the results were subjected to gray analysis (E-F). The data are shown as the means ± SDs, n = 3. The differences were not significant at P ≥ 0.05 (ns) or significant at P < 0.05 (*), P < 0.01 (**)

The results not only revealed that the positively charged amino acids of ATG7 Lys673 might interact with the negatively charged amino acids of Dicer GLU1827 to form a salt bridge interaction but also revealed that SER399, ARG484, ASN451, and ASP684 of ATG7 form hydrogen bonds with GLU830, ARG1090, LYS1289, GLU1721, GLU1827, and CYS1640 of Dicer (Fig. 5A-B and S3C). Furthermore, we constructed the ATG7 site deletion plasmids ATG7ΔS399, ATG7ΔN484, ATG7ΔR451, ATG7ΔD673, and ATG7ΔD684 to validate which ATG7 site interacted with Dicer (Fig. 5C and S3D). Western blotting revealed that PC-ATG7ΔD684 significantly increased Dicer protein levels, indicating that ATG7 mainly interacts with dicer through the ATG7 D684 site (Fig. 5D-E). In addition, the D684 site of the dicer is a key point for AIV-H9N2 replication, as shown in Fig. 5D. We then detected virus-derived small RNAs (vsiRNAs) in the ATG7-overexpressing and ATG7-silenced groups. The results revealed that ATG7 overexpression reduced the number of AIV-H9N2-derived total siRNAs and 21–23 nt vsiRNAs, whereas ATG7 silencing increased the number of total siRNAs and AIV-H9N2 vsiRNAs. In addition, when D684 combines the above two results, we can reasonably conclude that the binding of Dicer to ATG7 would suppress its cleavage function and decrease vsiRNA synthesis, which would impact AIV-H9N2 replication. D684 is located in the RNase III region of the Dicer domain, which is primarily used for cleaving dsRNA or double-stranded RNA intermediates produced during viral replication. Among them, small RNAs of different sequence lengths and quantities have different effects on virus replication. It cannot be ruled out that D684 is a key site for dicer to cleave dsRNA, but more experimental evidence is still needed in the future.

Fig. 5.

D684 is the main site at which ATG7 interacts with dicer to facilitate AIV-H9N2 replication. A-B. Two-dimensional representations of the ATG7 protein and Dicer protein binding interface (A). Three-dimensional docking of ATG7 and Dicer (green ribbon represents the chain of ATG7; sky blue represents the chain of dicer) (B). C. The CDS region of the Gallus-ATG7 protein and the interaction sites with the Dicer protein, mainly S399, N484, R451, D673, and D684. D. DF-1 cells were transfected with PC, PC-ATG7, and ATG7 mutant plasmids (S399, N484, R451, K673, and D684) for 48 h, and Western blotting was used to detect ATG7 and Dicer protein expression. In addition, PC and PC-ATG7 were used as controls. E. Gray value analysis of Fig. 5D. The data are shown as the means ± SDs, n = 3. The differences were not significant at P ≥ 0.05 (ns) or significant at P < 0.05 (*), P < 0.01 (**)

ATG7 affects the ability of the dicer to produce small RNA

The siRNA generated by Dicer cleavage is primarily responsible for the antiviral activity of RNAi. First, we detected changes in microRNAs in the context of ATG7 overexpression, and groups in which microRNAs were silenced were also produced by Dicer cleavage. The small RNA sequencing results revealed 217 significantly changed microRNAs (118 downregulated and 99 upregulated) in the AIV-H9N2-infected ATG7 overexpression group compared with 105 altered microRNAs (53 downregulated and 52 upregulated) in the H9N2-treated control group. In contrast, we also detected 205 significantly changed microRNAs (59 downregulated and 146 upregulated) in the AIV-H9N2-infected ATG7-silenced group compared with those in the AIV-H9N2-infected control group (Fig. 6A-B). Second, Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses were performed on the differentially expressed miRNAs identified in this study. The KEGG analysis indicated that the differentially expressed miRNAs in all groups were involved in regulating autophagy components (Fig. 6C). Moreover, the differentially expressed miRNAs in both the PC-ATG7 + H9N2 vs. PC + H9N2 and PS-ATG7 + H9N2 vs. PS + H9N2 groups participated in the MAPK signaling pathway and metabolic pathways (Fig. 6C). We then analyzed whether ATG7 influenced the production of small RNA in the AIV-H9N2 infection group. Third, we detected total siRNAs and virus-derived small RNAs (vsiRNAs) in the ATG7-overexpressing and ATG7-silenced groups (Fig. 6A). On the one hand, the results revealed that ATG7 overexpression (PC-ATG7) reduced the number of AIV-H9N2-derived total siRNAs and 21–23 nt vsiRNAs compared with those in the AIV-H9N2-infected PC-control group (Fig. 6A). On the other hand, compared with the AIV-H9N2-infected PC-control group, the ATG7-silenced group (PS-ATG7) presented an increased number of total siRNAs and AIV-H9N2 vsiRNAs (Fig. 6D). This finding suggested that the binding of ATG7 to Dicer would suppress its cleavage function and lower vsiRNA synthesis, which would impact AIV-H9N2 replication.

Fig. 6.

The ATG7 protein affects the dicer’s ability to cleave and produce small RNAs. A. PC-ATG7 and PS-ATG7 plasmids were transfected into DF-1 cells, and RNA samples were collected for transcriptome sequencing analysis. Statistics of the quantity of 21–23 nt miRNA siRNAs and vsiRNAs. B-C. The volcano plot displays the differences in miRNAs among the different groups (B). KEGG analysis of the impact pathways in each group (C). D. After the DF-1 cells were transfected with PC-ATG7 or PS-ATG7 for 24 h, 0.1 MOI AIV-H9N2 was used for infection. Small RNAs were aligned with the AIV-H9N2 genome, and the abundance and distribution of 21–23 nt vsiRNAs were determined via IGV software

ATG7 also suppresses dicer expression during infection with other RNA viruses

Evaluating whether ATG can bind to dicer to influence viral replication is a common phenomenon in other poultry viruses. Initially, we selected the infectious bursal disease virus (IBDV) and infectious bronchitis virus (IBV) to repeat the above experiment. As expected, IBDV and IBV infection of DF-1 cells led to an increase in ATG7 protein expression and a decrease in Dicer protein expression (Fig. 7A-B). Moreover, we transfected an ATG7 overexpression plasmid (PC-ATG7) into DF-1 cells and infected them with IBDV and IBV. The Western blotting results revealed that the viral load in DF-1 cells significantly increased and that the expression level of dicer significantly decreased after transfection with PC-ATG7 (Fig. 7C-D).

Fig. 7.

ATG7 suppresses Dicer expression during infection with other RNA viruses. A-B. DF-1 cells were infected with 1 MOI IBDV (A) or 5 MOI IBV (B) for the indicated time points (12 h, 24 h, 48 h, and 72 h), after which Western blotting was used to detect the expression of autophagy-related proteins and viral proteins. The blank group was used as the control, and the gray value analysis was performed on the right. C-D. DF-1 cells were transfected with PC-ATG7 for 24 h and then infected with 1 MOI IBDV (C) or 5 MOI IBV (D) for 24 h. Western blotting was used to detect dicer and viral protein expression. The PC group was used as the control, and the images on the right were subjected to gray analysis. E. Single-cell analysis of the distribution of bursal B cells and FACS validation of the changes in the subtypes of bursary B cells after IBDV infection. t-SNE visualization of the major nine cell types characterized by different colors. The encircled light-green cells represent the total fibroblast population. F. t-SNE of Dicer and ATG7 in the fibroblast population. G. KEGG and GO analyses of the IBDV-infected 2-week-old SPF chickens and the blank control chickens. The data are shown as the means ± SDs, n = 3. The differences were not significant at P ≥ 0.05 (ns) or significant at P < 0.05 (*), P < 0.01 (**), P < 0.001 (***)

Furthermore, we analyzed our previous single-cell sequencing data from chickens infected with IBDV. In addition to the identification of nine cell subpopulations, the results revealed that IBDV infection repressed fibroblast dicer and activated ATG7 (Fig. 7E-F and S4A). Moreover, ATG4A, ATG5 and ATG9 were highly expressed in IBDV-infected fibroblasts (Figure S4B). Finally, KEGG and GO analysis results revealed that IBDV infection affects the processes of genetic information processing (translation and transcription), cellular component organization or biogenesis, cellular pro-cell, binding, and enzyme regulator activity (Fig. 7G and S4C-D).

Overall, our study demonstrated that autophagy activation, which is mediated by ATG7, inhibited the RNAi pathway by suppressing Dicer expression, which is crucial for antiviral defence.

Discussion

RNAi is a conserved posttranscriptional silencing mechanism in eukaryotes [25–27]. At present, studies have reported the importance of RNAi in antiviral therapy in poultry and mammals [28–30]. The collaboration between autophagy and RNAi in antiviral defence is multifaceted and can occur at several levels [31–34]. First, they might cooperate to degrade viral components. RNAi methods focus primarily on viral mRNA, while autophagy encapsulates and degrades entire viral particles or viral factories [35]. Second, autophagy can modulate antiviral signaling pathways, which then influence the activity of RNAi machinery components [36–38]. Third, the components of autophagy and RNAi might directly interact with each other [39–41]. For example, certain autophagy-related proteins may regulate the trafficking or processing of dsRNA, thereby influencing the efficiency of RNAi-mediated viral suppression [37, 42]. Hence, the collaborative efforts of autophagy and RNAi represent a robust antiviral defence mechanism that operates at multiple levels to inhibit viral replication and promote viral clearance [40, 43].

The intricate interplay between autophagy and RNAi has emerged as a pivotal research area in virology, particularly in understanding the mechanisms that regulate viral replication [37]. Previous studies have revealed that autophagy is an intracellular degradation process that prevents the accumulation of damaged proteins and organelles and the release of necessary nutrients in the event of nutrient deficiency [2, 9]. Autophagy acts as an antiviral defence mechanism by degrading viral particles and promoting viral clearance [44, 45]. Many RNA viruses have developed various strategies to disrupt autophagy and promote their replication [3, 46]. Some studies have shown that IBDV promotes self-replication by inhibiting autophagosome formation through the IBDV-VP3 protein [10]. In addition, other studies reported that inhibiting autophagy with 3-methyladenine or knocking down ATG5 significantly suppressed AIV-H9N2 replication, suggesting that these autophagy-related proteins might influence RNA virus replication [47, 48]. Our study also revealed that autophagy activation promoted AIV-H9N2 replication (Fig. 2G and S1E). This finding is consistent with previous studies reporting the involvement of autophagy in influenza virus replication [49, 50]. However, our study extends these observations by demonstrating that autophagy, specifically through ATG7, can inhibit the antiviral RNAi pathway to facilitate viral replication (Fig. 4D-F).

Studies have shown that TRIM29 regulates the PERK-mediated ER stress immune response to control picornavirus CVB3-induced viral myocarditis and that PERK activation is reported to be dependent on ATG7 [51, 52]. Given the critical role of TRIM29 in regulating antiviral innate immunity against RNA and DNA viruses [53–55], whether ATG7 can regulate TRIM29 to promote virus replication is a topic worth further exploration. Moreover, ATG7 acts as a novel player in Autophagy-RNAi cross talk, and it functions as a ubiquitin E1-like enzyme crucial for autophagosome formation [56, 57]. A previous study demonstrated that silencing the ATG7 gene in pepper plants enhances the replication levels of pepper mild mottle virus [58]. Interestingly, the barley stripe mosaic virus disrupts the ATG7-ATG8 interaction, thereby impairing autophagy-mediated antiviral defence mechanisms [59]. These findings suggest that the ATG7 protein influences virus replication within host cells. Team Gibbings reported increased expression levels of the key protein dicer in the RNAi pathway upon siRNA-mediated knockdown of the ATG7 protein in HeLa cells [36].

While the activity and expression levels of dicer are crucial for cellular antiviral responses, the interplay between the ATG7 protein and dicer during virus replication has not yet been explored. Our study first identified ATG7 as a direct interactor with dicer, the enzyme responsible for processing dsRNA into siRNAs and pre-miRNA to miRNA in the RNAi pathway. We confirmed that ATG7 interacts with dicer primarily through the D684 site (Fig. 5). Site-specific deletion of ATG7 at D684 abolished its inhibitory effect on Dicer expression, further confirming the critical role of this residue in the ATG7-Dicer interaction. Further functional assays demonstrated that ATG7 overexpression inhibits Dicer expression and promotes viral replication, whereas ATG7 silencing has the opposite effect, underscoring the importance of this interaction in regulating viral replication. These findings suggest that ATG7 could be a potential therapeutic target for modulating RNAi-mediated antiviral responses (Fig. 8). Our study lays the groundwork for future research in this area by identifying the ATG7-Dicer interaction as a potential target for antiviral therapies. The prospect of developing inhibitors that can selectively enhance antiviral RNAi without disrupting beneficial autophagy is an intriguing area for further exploration, with the potential to offer new strategies for combating viral infections.

Fig. 8.

Pattern diagram of RNA viruses that use the ATG7 protein to inhibit the Dicer protein. The RNA virus pattern diagram shows the inhibition of the RNAi pathway by the ATG7 protein. RNA viruses such as AIV-H9N2 can upregulate the autophagy process in DF-1 cells, hence facilitating the conversion of LC3I to LC3II and increasing the level of the ATG7 protein. Furthermore, the substantial overexpression of the ATG7 protein caused by the virus encouraged ATG7 to bind to the Dicer protein via the D684 site in the RNAi pathway, which prevented the Dicer protein from recognizing and cleaving the viral dsRNA

In addition to the interaction of ATG7 and Dicer, other studies have shown that ATG5 and ATG7 are considered important molecules for inducing autophagy [5]. Among them, the expression of a lncRNA (DICER1-AS1) and an NF-κB-related mRNA encoding gene (ATG5) is correlated. Compared with that in noncancer tissues, the relative expression of DICER1-AS1 in cancer tissues is often associated with histological grade [60]. Moreover, a study revealed that the downregulation of SQSTM1 (P62) leads to a decrease in the expression of the miRNA processing enzyme DICER and the miRNA effector AGO2 [61].

The decrease in miRNA, siRNA, and vsiRNAs resulted in a compromised antiviral RNAi response, allowing the virus to replicate unchecked. The finding that ATG7 overexpression reduced the number of vsiRNAs produced during AIV-H9N2 infection further highlights the importance of this interaction in regulating viral replication. Among them, the use of the autophagy promoter RAPA can cause high expression of the ATG7 protein and low expression of the Dicer protein, leading to a further increase in virus replication. The use of the autophagy inhibitor CQ can restore the expression level of the Dicer protein and effectively reduce the viral load.

In addition, our study suggested that ATG7-mediated inhibition of dicer and the RNAi pathway is not limited to the AIV-H9N2 influenza virus but may be a common strategy employed by other avian RNA viruses (Fig. 7). We observed similar trends in IBDV and IBV infections, where ATG7 expression was upregulated, and Dicer expression was downregulated. These observations have important implications for understanding the mechanisms of viral replication and suggest that targeting ATG7 could be a potential antiviral strategy for a broad range of RNA viruses. This study revealed for the first time that the critical role of the ATG7 protein in regulating autophagy and Dicer enzymes significantly affects RNA virus replication levels in avian cells. The ATG7-Dicer interaction is vital in viral replication. Potential therapeutic strategies include the development of small molecule inhibitors that mimic dicer’s binding site to disrupt the ATG7-Dicer complex or the use of peptides/antibodies to block key interaction sites, such as D684. Modulating autophagy levels can indirectly affect this interaction, and enhancing the RNAi pathway independently of ATG7 by delivering exogenous siRNAs/miRNAs are also viable. Targeting this interaction can impact various viral diseases. Moreover, RNA viruses such as AIV, IBDV, and IBV can reduce the viral load and severity. These findings also have implications for emerging viruses, allowing early identification of therapeutic targets. Combination therapies involving the use of antiviral drugs with agents disrupting the ATG7-Dicer complex could be more effective. Further research and targeted therapies may revolutionize viral disease treatment, improving antiviral defences and reducing the public health burden of viral infections.

Conclusion

The prevention and control of RNA viruses in poultry have always been a challenge in the poultry industry. Among them, infection of DF-1 cells with poultry RNA virus can cause high expression of autophagy-related genes and low expression of RNAi. The upregulation of the autophagy pathway also leads to high expression of the ATG7 protein. Notably, our study revealed that the ATG7 protein can interact with the Dicer protein, thereby affecting dicer’s ability to cleave viral dsRNA, leading to host cells not being highly resistant to viral replication.

Experimental procedures

Antibodies and reagents

Rabbit anti-AIV H9N2 HA (Sino Biological, 1:500), mouse monoclonal anti-AIV NS1 (GeneTex, 1:500), rabbit polyclonal anti-Dicer (Bioworld, 1:500), rabbit polyclonal anti-Ago2 (Bioworld, 1:1000), rabbit anti-SQSTM1/P62 (YEASEN, 1:500), rabbit anti-LC3B (Abways TECHNOLOGY, 1:500), mouse monoclonal anti-GAPDH HRP (Bioworld, 1:5,000), rabbit polyclonal anti-His (YEASEN, 1:500 dilution), rabbit anti-ATG7 (Abways, 1:1000), rabbit polyclonal anti-NDP52 (AiFang Biological, 1:500), rabbit anti-HIF1A (Abways, 1:500), mouse monoclonal anti-His (YEASEN, 1:2000), mouse monoclonal anti-Flag (YEASEN, 1:1000), rabbit polyclonal anti-Flag (YEASEN, 1:1000), mouse monoclonal anti-IBV (HyTest, 1:500), Goat anti-mouse/rabbit IgG (H&L)-HRP (Bioworld, 1: 10000) and goat anti-mouse/rabbit IgG H&L AF594 (Bioss, 1:1000) secondary antibodies were incubated with the samples for 1 h at room temperature.

The cells were incubated with different concentrations of rapamycin (RAPA, Solarbio), chloroquine (CQ, Solarbio), L-1-tosylamide-2-phenylethyl chloromethyl ketone (TPCK)-treated trypsin (Sigma, 2 µg/mL), and Fluoroshield™ with DAPI (Sigma, 1:1000) at the appropriate doses. The reagents, instruments, and software used in this study can be found in supplemental material Table S1.

Cells and viruses

DF-1 cells, Madin-Darby canine kidney (MDCK) cells, and human embryonic kidney (HEK-293T) cells were purchased from the American Type Culture Collection (ATCC, USA). DF-1 and MDCK cells were cultured in DMEM/high-glucose medium (Bio-Channel) supplemented with 10% heat-inactivated fetal bovine serum (FBS; ExCell Bio). HEK-293T cells were cultured in Roswell Park Memorial Institute (RPMI) 1640 medium (Gibco^™) supplemented with 10% FBS (Gibco^™). All cell media contained 2% penicillin-streptomycin solution (Bio-Channel, 100 U/mL penicillin, 0.1 mg/mL streptomycin) and were grown at 37 °C in a 5% CO₂ environment.

The AIV-H9N2 strain A/duck/Nanjing/01/1999 (EID₅₀ = 10^− 7/0.1 mL), IBV-M41 strain (EID₅₀ = 10^− 6.8/0.1 mL), and classical virulent IBDV BC6/85 strain (TCID₅₀ = 10^− 4.5/0.1 mL) were kindly provided by the Jiangsu Academy of Agricultural Sciences (JAAS, Nanjing, China) and stored in our laboratory [62, 63].

Virus infection experiment

For the different multiplicities of infection (MOIs) AIV infection experiments, DF-1 cells were infected with 0.01, 0.1, or 1 MOI AIV-H9N2 containing 2 µg/mL TPCK-treated trypsin. After 2 h of adsorption, the cells were washed three times with phosphate-buffered saline (PBS, Solarbio) and then incubated in serum-free DMEM with 2 µg/mL TPCK-treated trypsin for 48 h. For the 0.1 MOI AIV-H9N2 infection experiment, the indicated time was adjusted to 6 h, 12 h, 12 h, 24 h, 36 h, or 48 h. For the IBDV and IBV infection experiments, DF-1 cells were incubated with 1 MOI IBDV or 5 MOI IBV-M41 for 2 h at 37 °C and cultured in 2% FBS DMEM for the indicated time (12 h, 24 h, 48 h, or 72 h) after three washes with PBS. For the overexpression experiment, DF-1 cells were transfected with 2 µg of PC-Dicer, PC-ATG7, or the mutant plasmid for 24 h, after which the virus was used for infection (0.1 MOI AIV, 1 MOI IBDV, or 5 MOI IBV) for 24 h.

Plaque-forming assays

Virus replication was titrated via plaque-forming assays. Briefly, the supernatants of cultured DF-1 cells were obtained after 48 h of AIV-H9N2 infection. MDCK cells infected with the supernatants were washed with PBS and overlaid with DMEM 2 × High Glucose (Livning, lvn1001-8) containing low melting point agarose (Solarbio, A8350) and TPCK-treated trypsin. After incubation for 72 h, the plaques were stained and counted.

RNA extraction, RT-PCR, and RT-qPCR

Total RNA from DF-1 cells was isolated with TRIzol™ reagent (Invitrogen) according to the manufacturer’s instructions. One thousand nanograms of RNA was reverse transcribed via HiScript II Q Select RT SuperMix for qPCR (Vazyme). The ChamQ Universal SYBR qPCR Master Mix (Vazyme) was used for qPCR in the 7500 fast real-time PCR system. For quantification, qPCR data for gene expression levels were analyzed via the 2^−ΔΔCt method to calculate relative RNA levels, which were normalized to Gallus-GAPDH expression [64]. The sequences of the RT-qPCR and plasmid construction primers used in this study are listed in supplemental material Tables S2–S3 and were generated by Sangon Biotech Co., Ltd. (Shanghai, China).

Plasmid and ATG7 mutation construction

The full-length cDNAs encoding Gallus-ATG7, Gallus-NDP52, Gallus-Beclin1, and Gallus-HIF1A were amplified via RT-PCR from the total RNA of DF-1 cells and cloned and inserted into pcDNA3.1-3×Flag-C (YouBio, VT9221) to form PC-ATG7, PC-NDP52, PC-Beclin1, and PC-HIF1A. The pcDNA3.1-Dicer-myc-HisB (PC-Dicer) plasmid constructed from pcDNA3.1(+)/myc-HisB (YouBio, VT1024) was stored in our laboratory. The plasmid construction process was as follows: 25 µL of 2 × Taq Plus Master Mix II Dye Plus reagent (Vazyme, P211), 2 µL of Primer F, 2 µL of Primer R, 50 ng of cDNA, and ddH₂O to 20 µL. The program was as follows: 95 °C, 3 min; 95 °C, 15 s, 58 °C, 20 s, 72 °C, 30 s (35 cycles); 72 °C, 7 min. The linearization system was as follows: 2×Taq Plus Master Mix II (Vazyme, P213) 2 µL, BamH I 1.5 µL, Hind III 1.5 µL, PS plasmid 1000 ng, ddH₂O to 20 µL. The program was as follows: 30 °C, 30 min; 37 °C, 30 min; and 85 °C, 5 min. After electrophoresis and the use of a gel extraction kit (OMEGA, D2500), T4 DNA ligase (ATG Biotechnology Co., Ltd., M101–40000) was used for ligation and transformation. The Gallus-ATG7 shRNA (PS-ATG7-01, 5’-TGACAGTTGGAGTTTATGA-3’; PS-ATG7-02, 5’-ATGTGGCTACACTTGAAGA-3’; PS-ATG7-03, 5’-GGTGGAATTAAT-GGTTTCT-3’) and Gallus-Dicer shRNA (PS-Dicer-01, 5’-GTGATTCTGAAGATGAT-GA-3’; PS-Dicer-02, 5’-AGAAAGAGAAGCCAGAGAC-3’; PS-Dicer-03, 5’-GAAGCCAGAGACTAATTTC-3’) were designed on the Thermo Fisher BLOCK-iT™ RNAi Designer (https://www.thermofisher.cn/bogus.html). The PS-shRNA primers used for plasmid construction can be found in supplemental material Table S3, and the vector was linearized via QuickCut™ Hind III (TAKARA, 1615) and QuickCut™ BamH I (TAKARA, 1605). The PC-ATG7 mutants (ATG7ΔS399, ATG7ΔN484, ATG7ΔR451, ATG7ΔK673, and ATG7ΔD684) were generated via a ClonExpress II One Step Cloning Kit (Vazyme, C112) and overlap cloning techniques. Notably, QuickCut™ Hind III, QuickCut™ Xho I (TAKARA, 1635), and specific primers were used in plasmid construction. The primers used are listed in the supplemental material in Table S4.

Cell transfection and treatment

For the overexpression experiments, 2 µg of PC-ATG7, PC-NDP52, PC-HIF1A, PC-Beclin1, or PC-Dicer was transfected via Hieff Trans™ Liposomal Transfection Reagent (YEASEN) following the manufacturer’s protocols. At 48 h posttransfection, subsequent related experiments, such as Western blotting experiments, were performed on the cells. After the DF-1 cells were transfected with PC-ATG7 or PC-Dicer for 24 h, the cells were further infected with 0.1 MOI AIV-H9N2 for another 24 h and subjected to plaque assays, Western blotting, and co-immunoprecipitation.

For the knockdown experiments, the ATG7 knockdown plasmids PS-ATG7 (PS-ATG7-01, PS-ATG7-02, and PS-ATG7-03) and the Dicer knockdown plasmids PS-Dicer (PS-Dicer-01, PS-Dicer-02, and PS-Dicer-03) were transfected with 2 µg using Lipofectamine^® 2000 Reagent (Invitrogen) following the manufacturer’s protocols. At 48 h posttransfection, the silencing efficiency was detected through Western blotting experiments. After the DF-1 cells were transfected with PS-ATG7 or PS-Dicer for 24 h, the cells were further infected with 0.1 MOI AIV-H9N2 for another 24 h and subjected to plaque and Western blotting assays.

RAPA and CQ treatment

For the cytotoxicity assay, a Cell Counting Kit-8 (Sangon Biotech) was used for detection. DF-1 cells were treated with different concentrations of RAPA (1 nM, 10 nM, 100 nM, 1 µM, and 10 µM) and CQ (10 µM, 100 µM, 1 mM, 10 mM, and 100 mM) for 2 h. For other experiments, the concentrations of RAPA and CQ used were 100 nM and 100 µM, respectively. DF-1 cells were then collected at the indicated time points (4 h, 12 h, 24 h, and 48 h) for subsequent detection.

Western blotting and co-immunoprecipitation

DF-1 cells and 293T cells were lysed with RIPA lysis buffer (Epizyme Biotech) containing a PMSF protease inhibitor (Solario) for Western blotting and co-immunoprecipitation. The supernatants were collected, and the protein concentration was determined via a BCA protein assay (Epizyme Biotech). Later, the cell lysates were resolved and separated on SDS-polyacrylamide gels and transferred to polyvinylidene fluoride (PVDF) membranes (Epizyme Biotech). After being blocked with 5% bovine serum albumin (BSA, Sigma) in TBST (Solario), the membranes were incubated at 4 °C overnight with primary antibodies as indicated, followed by incubation with secondary antibodies. The signal was detected by using an Ultrasensitive ECL Chemiluminescence Kit (Sangon Biotech). In addition, when the density of the HEK-293T cells and the DF-1 cells reached 70–80%, the control plasmid (PC), PC-Beclin1, PC-ATG7, PC-HIF1A, and PC-NDP52 plasmids were cotransfected with the PC-Dicer plasmid. After transfection with Hieff Trans™ Liposomal Transfection Reagent (YEASEN) for 48 h, the cells were lysed with RIPA lysis buffer. Later, a portion of the supernatant was used as input, and the remaining lysates from the cells were incubated with anti-FLAG or anti-His antibodies and protein A/G beads for 4 h at 4 °C. After centrifugation, the supernatant was removed, and the pellets were suspended in a washing buffer. The final pellets were lysed in lysis buffer for Western blot analysis. For Fig. 4B, PC-Flag-C (YouBio, Flag-Mock) was used as the control.

Immunofluorescence assay

DF-1 cells were plated on coverslips in 6-well plates at a density of 80% and were infected with AIV-H9N2 (0.01 MOI, 0.1 MOI, or 1 MOI) or treated with PBS (Sango Biotech). After being incubated in virus-maintenance DMEM (2% FBS) for 24 h, the cells were washed with PBS three times and fixed with 4% paraformaldehyde (Sangon Biotech) at room temperature for 20 min. After being rinsed with PBS and incubated with 0.4% Triton X-100 (Sangon Biotech) for 15 min, the DF-1 cells were blocked with 5% bovine serum albumin (BSA, Sigma) for 1 h. After being washed three times, the cells were incubated with the rabbit anti-H9N2 HA antibody at 4 °C overnight, followed by incubation with the goat anti-rabbit IgG H&L AF594 antibody for 1 h. After being washed three times with PBST and stained with DAPI, the slides were observed under a confocal laser microscope (Zeiss, Germany).

Protein molecular docking

The protein molecular docking experiment was as follows. Gallus-ATG7 (Q5ZKY2) and Gallus-Dicer (Q25BN1) were downloaded from the UniProt website (https://www.uniprot.org/). Protein pretreatment, including removal of water molecules, hydrogenation, charge extraction, and extraction of the original ligand from the structure, was performed via Discovery software (http://www.discoverystudio.net/). The Zock website (https://zdock.umassmed.edu/) was used for protein-to-protein docking research, with Gallus-Dicer as the docking receptor and Gallus-ATG7 as the ligand. During the docking process, the angular step size was set to 6, and the angle was 15°. The docking result generated 2,000 poses, and further tools were used to list the top 100 well-rated poses. The zdock score of these poses was subsequently used as the X-axis, and the cluster was used as the Y-axis for visualization. A zdock score greater than 19 was selected as the baseline, and key poses were screened for Rdock docking analysis. In the Rdock docking analysis, a lower energy score indicates a more stable pose. The polar interactions between proteins were analyzed via PyMOL software (https://www.pymol.org/), and the two-dimensional interactions between proteins were analyzed via Ligplot.

Single-cell analysis

The accession number for Associate Abid’s article single-cell sequence data is GEO GSE167377 [65]. The groups were divided into S1 (control 2-week-old chicks), S2 (IBDV-infected 2-week-old chicks), S3 (control 3-week-old chicks), and S4 (IBDV-infected 3-week-old chicks). The Ranger software pipeline provided by 10 × Genomics was used to divide the cells into nine cell types: B cells, B cells, B cells, epithelial cells, epithelial cells, T cells, B cells, T cells, B cells, untitled cells, and fibroblasts. The gene names ATG7 and Dicer were input into the software, and the gene expression levels of ATG7 and Dicer in various cell subpopulations, especially in fibroblasts, were analyzed. Subsequently, KEGG and GO analyses were conducted.

Transcriptome sequencing and virus-derived small RNA analysis

The RNA samples were sent to LC-BioTechnology Co. Ltd. for transcriptome sequencing analysis, which specifically detected differences in miRNAs and virus-derived small RNAs (vsiRNAs) among different groups. Using the self-developed ACGT101 miR (v4.2) for miRNA data analysis, the analysis process was as follows: (1) the 3’ adapter and garbage sequence were removed to obtain clean data; (2) sRNA length screening: the plant retention length range was 18–25 nt, and the animal retention length range was 18–26 nt; (3) the remaining sequences were compared with the mRNA, RFam, and Repbase databases (excluding miRNAs) and filtered; (4) miRNA identification: reads filtered by databases such as length filtering and RFam were obtained, and precursors and genomes for miRNA identification were compared; (5) differential expression analysis of miRNAs was performed; and (6) prediction and enrichment analysis of target genes for differentially expressed miRNAs were performed. Please refer to the detailed analysis methods. Specifically, all reads are matched to the viral genome to identify 18-25-nt short RNAs corresponding to necessary vsiRNAs. In addition, emphasis is placed on counting the number of 21–23 nt vsiRNAs and their distribution in the viral genome.

Statistical analysis

Statistical analyses of RT-qPCR, fluorescence intensity, and Western blotting gray value data were carried out via ImageJ software (https://imagej.nih.gov/ij/), Microsoft Excel (https://www.microsoft.com/zh-cn/microsoft-365/excel), and GraphPad Prism 10 (https://www.graphpad.com/). Statistical significance was established via Student’s t-test for each group and one-way analysis of variance (ANOVA) for multiple groups. All the data are presented as the means ± SDs of at least three independent experiments. The figures’ statistical significance was as follows: ****P < 0.0001, ***P < 0.001, **P < 0.01, *P < 0.05; ns, insignificant.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Abbreviations

AIV

Avian influenza virus

IBV

Infectious bronchitis virus

IBDV

Infectious bursal disease virus

RNAi

RNA interference

LC3-II

Licensed microtubule-associated protein 1 light chain 3

ATG7

Autophagy-related 7

SQSTM/p62

Chromosome 1

BECN1

Beclin 1

RAPA

Rapamycin

CQ

Chloroquine

eGFP

Enhanced green fluorescent protein

mRFP

Monomeric red fluorescent protein

MOI

Multiplicity of infection

qPCR

Quantitative real-time PCR

CO-IP

Co-immunoprecipitation

IFA

Immunofluorescence assay

PC

PcDNA3.1

PS

Psilencer4.1

Author contributions

Yaotang Wu was responsible for designing the ideas for the manuscript, arranging the experiments, conducting the bioinformatics analysis, processing the Western blotting results, re-examining the single-cell data, and drawing summary graphs for the abstract. Yang Wu and Zhixin Li constructed the H9N2 infection experiment, which is related to Figs. 1, 2 and 5. Chenlu Wang performed the ATG7 knockdown experiments related to Figs. 4 and 6. Ningna Xiong, Wenxin Ji, Mei Fu, and Junpeng Zhu helped to construct the ATG7 plasmid model and autophagy experiment, related to Figs. 3 and 4. Jian Lin designed and carried out molecular genetic studies, bioinformatic analyses, and manuscript drafting; he was also the principal investigator of the project. Professor Qian Yang conceived the study and participated in the manuscript.

Funding

This work was supported by the Project of Sanya Yazhou Bay Science and Technology City (SCKJ - JYRC - 2022 - 31), the National Natural Science Foundation of China (32072835), and the Fundamental Research Funds for the Central Universities (KYT2023004) to J.L. Moreover, this work was also supported by the Ningxia Natural Science Foundation Project: Gene Evolution and Transmission Dynamics of Avian Influenza Virus in Ningxia (Project No.: 2022AAC02071) to ZX.L and the National Natural Science Foundation of China (31930109 and 31772777) and a project funded by the Priority Academic Program Development of Jiangsu Higher Education Institutions (PAPD) to Q.Y. and J.L.

Data availability

The data are now available at NAR online.

Declarations

Competing interests

The authors disclosed no possible conflicts of interest. All the authors have agreed to publish this manuscript.