The apple C2H2-type zinc finger transcription factor MdZAT10 positively regulates JA-induced leaf senescence by interacting with MdBT2

Jasmonic acid (JA) plays an important role in regulating leaf senescence. However, the molecular mechanisms of leaf senescence in apple (Malus domestica) remain elusive. In this study, we found that MdZAT10, a C2H2-type zinc finger transcription factor (TF) in apple, markedly accelerates leaf senescence and increases the expression of senescence-related genes. To explore how MdZAT10 promotes leaf senescence, we carried out liquid chromatography/mass spectrometry screening. We found that MdABI5 physically interacts with MdZAT10. MdABI5, an important positive regulator of leaf senescence, significantly accelerated leaf senescence in apple. MdZAT10 was found to enhance the transcriptional activity of MdABI5 for MdNYC1 and MdNYE1, thus accelerating leaf senescence. In addition, we found that MdZAT10 expression was induced by methyl jasmonate (MeJA), which accelerated JA-induced leaf senescence. We also found that the JA-responsive protein MdBT2 directly interacts with MdZAT10 and reduces its protein stability through ubiquitination and degradation, thereby delaying MdZAT10-mediated leaf senescence. Taken together, our results provide new insight into the mechanisms by which MdZAT10 positively regulates JA-induced leaf senescence in apple.


Introduction
Plant leaf senescence, the last stage of leaf development, is accompanied by a series of physiological and biochemical changes, including the degradation of intracellular organelles and hydrolysis of macromolecules for the relocation of nutrients and energy into newly developing tissues or storage organs 1 . It is important to understand how plants regulate the senescence process to prevent major yield losses in agriculture. The leaf senescence process can be triggered and promoted by unfavorable environmental cues, including extended darkness, drought, and pathogen attack [2][3][4] , and by endogenous factors such as age, developmental stage, and plant hormones 5,6 . Leaf senescence inhibits photosynthetic capacity and thus decreases crop quality and yield 7,8 . Therefore, delaying leaf senescence offers potential economic benefits 7 .
Plant hormones are known to affect the timing of leaf senescence. Hormones such as abscisic acid (ABA), jasmonic acid, ethylene (ET), and salicylic acid (SA) accelerate the leaf senescence process, whereas auxin, cytokinins (CKs), and gibberellic acid (GA) delay leaf senescence 9 . JA is a lipid-derived phytohormone that is ubiquitous in the plant kingdom and plays essential roles in the regulation of multiple physiological processes in plants, including root growth, leaf senescence, and the response to wounding and pathogens [10][11][12][13] . In Arabidopsis, the endogenous JA content is higher in senescent leaves than in nonsenescent leaves 14 . Consistent with this increased JA content, several genes involved in the JA biosynthesis pathway, such as LIPOXYGENASE 1/3/4 (LOX1/3/4) and ALLENE OXIDE CYCLASE 1 (AOC1), are also markedly upregulated during leaf senescence 14 . In response to JA, the JASMONATE ZIM-DOMAIN (JAZ) proteins interact with CORONATINE INSENSITIVE1 (COI1), a component of the SCF COI1 complex 15,16 . The JAZ proteins are then degraded by the 26S proteasome, thereby releasing downstream JA-responsive genes such as the bHLH transcription factor MYC2 16,17 . MYC2 positively regulates JA-induced leaf senescence by directly activating the expression of SENESCENCE-ASSOCIATED GENE 29 (SAG29), and MYC2 interacts with the bHLH subgroup IIId TFs bHLH03, 13, 14, and 17, which antagonistically regulate leaf senescence 18 .
A large number of genetic and transcriptome studies have shown that TFs regulate the leaf senescence process 10,19 . Some TFs play critical roles in leaf senescence regulatory networks; these TFs include members of the bHLH, NAC, MYB, WRKY, bZIP, C2H2-type zinc finger, and AP2⁄ EREBP families 19,20 . C2H2-type zincfinger proteins (ZFPs) are a large family of transcriptional regulators in plants 21 . Various C2H2-type zinc finger TFs are known to be involved in plant development and stress responses 22,23 . Most ZFPs contain one to four highly conserved QALGGH motifs 24 . In addition, a few ZFPs contain Cterminal ERF-associated amphiphilic repression (EAR) motifs, which function as transcriptional repressors 25,26 . Several members of the C2H2-type zinc finger TF family were found to be up-or downregulated during natural leaf senescence 19 , indicating that C2H2-type zinc finger TFs may participate in leaf senescence. Arabidopsis zinc-finger protein 2 (AZF2), a C2H2-type zinc finger TF, positively regulates age-triggered leaf senescence 27 .
ABSCISIC ACID-INSENSITIVE5 (ABI5), a basic leucine zipper (bZIP)-type TF, positively regulates ABA signaling and participates in seed germination, abiotic stress tolerance, and leaf senescence 4,28,29 . Previous studies have revealed that ABI5 modulates leaf senescence by transcriptional regulation. ABI5 positively regulates darkinduced leaf senescence by directly repressing the expression of ABA-response protein (ABR) 30 and activating the expression of the chlorophyll degradation genes NON-YELLOW COLORING1 (NYC1) and STAY-GREEN 1 (SGR1/NYE1) 4 . The bHLH TFs PHYTOCHROME-INTERACTING FACTORS 4/5 (PIF4/5) directly activate ABI5 during dark-induced senescence 4 . In rice (Oryza sativa), ONAC054 directly activates OsABI5 in response to leaf senescence 31 . A recent study found that MdABI5 is involved in ABA-induced leaf senescence 32 . These findings indicate that ABI5 plays a crucial role in the leaf senescence process.
Previous studies have identified several genes that promote or delay leaf senescence in apple [33][34][35] . In this study, we identified a C2H2-type zinc finger TF, MdZAT10, and demonstrated that it positively regulates leaf senescence in apple (Malus domestica). Further experiments showed that MdZAT10 interacts with MdABI5 and accelerates the MdABI5-mediated leaf senescence process. MdZAT10 was also found to accelerate JA-induced leaf senescence. We found a negative JA regulator, MdBT2, which interacts with MdZAT10 and modulates its stability, thereby repressing MdZAT10-mediated leaf senescence. In summary, we used protein-protein interactions to delineate the relationships among MdBT2, MdZAT10, and MdABI5 during leaf senescence.

MdZAT10 positively regulates leaf senescence
The expression of a large number of C2H2-type TFs is markedly induced during natural senescence 19 . The MdZAT10 gene, a C2H2-type zinc finger TF in subclass C1-2i, is homologous to Arabidopsis STZ/ZAT10 (SALT TOLERANCE ZINC FINGER). The MdZAT10 (MDP0000198015) gene was identified by a BLAST search against the apple genome database (Apple Gene Function & Gene Family DataBase version 1.0). To identify proteins homologous to ZAT10, a phylogenetic tree containing sequences from 21 different plant species was constructed. All proteins contained two conserved zinc finger domains and an EAR motif, and MdZAT10 was highly homologous to PbZAT10 from Pyrus bretschneideri ( Supplementary Fig. S1).
We measured the expression level of MdZAT10 in apple leaves at different developmental stages, including the nonsenescent (NS), early-senescent (ES), and latesenescent (LS) stages. MdZAT10 expression was higher in at ES and LS stages than at the NS stage (Fig. 1a). To confirm the function of MdZAT10 during leaf senescence, an MdZAT10 overexpression vector was transformed into Arabidopsis, generating three independent Arabidopsis lines (MdZAT10-L1, L2 and L3) (Supplementary Fig. S2). Detached leaves from transgenic Arabidopsis seedlings showed greater leaf yellowing than those from wild-type (Col) seedlings (Fig. 1b). Consistent with their differences in color, the leaves of the transgenic plants showed significant decreases in chlorophyll content and maximum quantum yield of photosystem II (Fv/Fm) (Fig. 1c, d). To further confirm these leaf senescence phenotypes, MdZAT10 overexpression and antisense suppression vectors were transformed into detached apple leaves using a transient expression system (Supplementary Fig. S2). Consistently, apple leaves overexpressing MdZAT10 exhibited early senescence, whereas apple leaves expressing MdZAT10 antisense suppression showed delayed senescence (Fig. 1e). Furthermore, apple leaves' MdZAT10 overexpression and antisense suppression vectors also showed corresponding differences in chlorophyll content (Fig. 1f). We found that MdZAT10 overexpression increased the expression of MdNYC1 and MdNYE1 in apple leaves (Fig. 1g, h). These results suggest that MdZAT10 positively regulates leaf senescence.

MdZAT10 interacts with the MdABI5 protein
To further explore the mechanism by which MdZAT10 promotes leaf senescence, a liquid chromatography/mass spectrometry (LC/MS) assay was carried out to screen proteins that interact with MdZAT10 using MdZAT10-GFP as bait. After screening, the MdABI5 protein (GenBank accession number: LOC103430245) was found to interact with MdZAT10, and a yeast two-hybrid (Y2H) assay was performed to confirm this interaction. The fulllength cDNA of MdZAT10 was fused to the pGAD424 vector as prey (pGAD-MdZAT10), and the full-length cDNA of MdABI5 was fused to the pGBT9 vector as bait (pGBD-MdABI5). The pGAD-MdZAT10 and pGBD-MdABI5 plasmids were cotransformed into yeast. The results showed an interaction between the MdZAT10 and MdABI5 proteins (Fig. 2a). To identify the regions in MdZAT10 that interact with MdAB15, MdZAT10 was divided into N-terminus (MdZAT10-N) and C-terminus (MdZAT10-C) fragments. These results indicated that the zinc finger domains and EAR motif of MdZAT10 are essential for the interaction between MdZAT10 and MdABI5 (Fig. 2a). MdZAT10 interacted with MdABI5, but not interacted with MdABI1, MdABI2, or MdABI4 ( Fig. 2a, b); MdABI5 also specifically interacted with MdZAT10 but not with other MdZATs (Fig. 2c). In addition, we carried out an in vitro pull-down assay and found that MdABI5-His could be pulled down by the MdZAT10-GST fusion protein (Fig. 2d). Finally, in a BiFC assay, a strong yellow fluorescent protein (YFP) fluorescence signal in the nuclei was observed only when MdZAT10-cYFP and MdABI5-nYFP were cotransformed into Nicotiana benthamiana leaves (Fig. 2e). These results indicated that MdZAT10 physically interacts with MdABI5.

MdABI5 promotes leaf senescence
Previous studies have reported that MdABI5 regulates ABA-induced leaf senescence 32 . We detected the expression level of MdABI5 in apple leaves at different developmental stages. The MdABI5 expression level was higher at the ES and LS stages than at the NS stage ( Supplementary Fig. S3). To elucidate the role of ABI5 during leaf senescence, an MdABI5 overexpression vector was transformed into Arabidopsis, generating three individual transgenic lines (MdABI5-L1, L2, and L3) (Supplementary Fig. S2). MdABI5 overexpression clearly promoted leaf yellowing and reduced the chlorophyll content and Fv/Fm after 5 days in the dark (Supplementary Fig. S3). Furthermore, the MdABI5 overexpression and MdABI5 antisense suppression vectors were transiently transformed into detached apple leaves. Apple leaves that overexpressed MdABI5 showed a more severe senescence phenotype and lower chlorophyll content, whereas MdABI5 antisense suppression showed a delayed senescence phenotype and higher chlorophyll content ( Supplementary Fig. S3). Previous studies have shown that ABI5 induces leaf senescence by directly regulating the expression of the chlorophyll degradation genes NYC1 and NYE1 4,31 . In this study, the expression of MdNYC1 and MdNYE1 was enhanced in apple leaves overexpressing MdABI5 (Supplementary Fig. S3). These results suggest that MdABI5 promotes leaf senescence.

MdZAT10 promotes MdABI5-regulated leaf senescence
Given the interaction between MdZAT10 and MdABI5, we suspected that MdZAT10 participates in MdABI5-mediated leaf senescence. An Arabidopsis line . a A yeast two-hybrid (Y2H) assay showed that MdZAT10 interacts with MdABI5. Full-length MdZAT10 and truncated MdZAT10 sequences were cloned into the pGAD424 vector. Full-length MdABI5 was cloned into the pGBT9 vector. The empty pGAD vector was used as a negative control. b A Y2H assay showed that MdABIs (MdABI1, MdABI2, and MdABI4) and MdZAT10 interact. c A Y2H assay showed that MdABI5 interacts with MdZAT proteins (MdAZF1, MdZAT5, MdZAT6, MdZAT10, MdZAT11, MdZAT14, MdZAT16, and MdZAT18). d In vitro MdZAT10 and MdABI5 pull-down assay. The MdABI5-His protein was incubated with MdZAT10-GST and GST. Proteins pulled down with GST beads were detected using anti-GST and anti-His antibodies. e BiFC assay. The MdZAT10-cYFP and MdABI5-nYFP constructs were transiently expressed in Nicotiana benthamiana leaves, and the fluorescence signal was observed by fluorescence microscopy. Nuclei are indicated by DAPI staining. Scale bars, 10 μm overexpressing MdZAT10 was crossed with an Arabidopsis line overexpressing MdABI5. The resultant Arabidopsis plants overexpressing MdZAT10/MdABI5 turned yellow much faster and had less chlorophyll content than the plants overexpressing MdABI5 alone (Fig. 3a, b). Consistent with the observed phenotype in Arabidopsis, the overexpression of MdZAT10 increased the MdABI5-mediated leaf senescence in apple leaves (Fig. 3c, d). These results revealed that MdZAT10 accelerates MdABI5-promoted leaf senescence.
Therefore, we hypothesized that MdZAT10 affects the transcriptional activation of MdNYC1 and  into the effector construct pGreenII 62-SK (MdZAT10-SK, MdABI5-SK). We found that MdABI5 activated the promoters of MdNYC1 and MdNYE1 and had a stronger effect when MdZAT10 and MdABI5 were cotransformed ( Fig. 3g-i). These results indicated that the interaction between MdZAT10 and MdABI5 enhances the transcriptional activity of MdABI5 for MdNYC1 and MdNYE1.

MdZAT10 activates JA-induced leaf senescence
We also found that MdZAT10 was induced by MeJA (Fig. 4a). To further confirm the function of MdZAT10 in response to JA, we produced transgenic apple calli that expressed β-glucuronidase (GUS) driven by a region 2 kb upstream of the MdZAT10 gene. Histochemical staining showed that MeJA treatment significantly increased GUS activity (Fig. 4b, c). Detached leaves from MdZAT10 transgenic Arabidopsis showed increased leaf yellowing with MeJA treatment (Fig. 4d). The chlorophyll content in these leaves was significantly reduced compared to the leaves of Col plants (Fig. 4e). Consistently, upon MeJA treatment, apple leaves overexpressing MdZAT10 also exhibited early senescence, whereas MdZAT10 antisense suppression showed delayed senescence (Fig. 4f). Furthermore, leaves MdZAT10 overexpression and MdZAT10 antisense suppression showed different chlorophyll  indicate significant differences by t-test ( * P < 0.05, ** P < 0.01) contents (Fig. 4g). These results indicated that MdZAT10 acts as a positive regulator of JA-induced leaf senescence in apple.

MdBT2 physically interacts with the MdZAT10 protein
In addition to its transcriptional regulation, we found that MdZAT10 was regulated at the posttranslational level in response to MeJA treatment. An in vitro protein degradation assay was performed to measure the protein level of MdZAT10 in response to MeJA treatment. The fusion protein MdZAT10-His was incubated with total protein from apple calli with or without MeJA treatment. The MdZAT10-His level dropped rapidly without MeJA treatment, but the drop in the MdZAT10 protein level was markedly abrogated by treatment with MeJA or the 26S proteasome inhibitor MG132 ( Supplementary  Fig. S4a). Furthermore, the MdZAT10 protein level in MdZAT10-overexpressing transgenic apple calli increased with MeJA treatment (Supplementary Fig.  S4b). These results indicated that the presence of MeJA reduced MdZAT10 protein degradation by the 26S proteasome pathway.
In addition to MdABI5, MdBT2 was screened as a potential interaction protein of MdZAT10. MdBT2 plays a key role in the regulation of JA-mediated leaf senescence 36 . We used a Y2H assay to determine whether MdBT2 and MdZAT10 interact. The full-length cDNA of MdBT2 was fused to the pGBT9 vector as bait (pGBD-MdBT2). Only yeast cells that contained pGAD-MdZAT10 and pGBD-MdBT2 grew well on -Trp/-Leu/-His/-Ade screening medium (Fig. 5a). To determine the regions of MdBT2 that interact with MdZAT10, MdBT2 was divided into Nterminal (MdBT2-N) and C-terminal (MdBT2-C) fragments. As shown in Fig. 5b, the BTB and TAZ domains of MdBT2 were indispensable for this interaction (Fig. 5b). To determine whether MdBT2 specifically interacts with MdZAT10, seven apple C2H2-type ZFPs (MdAZF1, MdZAT5, MdZAT6, MdZAT11, MdZAT14, MdZAT16, and MdZAT18) were fused to the prey vector pGAD424. However, all of them failed to interact with MdBT2 ( Supplementary Fig. S5). MdBT1, MdBT2, MdBT3.1, and MdBT4 are important members of the apple MdBT protein family, and MdBT1 and MdBT2 interact with MdZAT10 ( Supplementary Fig. S5).
The MdBT2-MdZAT10 interaction was further verified by pull-down and BiFC assays. For the pull-down assay, the fusion protein MdBT2-GST and GST as a control were incubated with the fusion protein MdZAT10-His. Only MdBT2-GST could pull down MdZAT10-His (Fig. 5c). Next, we conducted a BiFC assay to further confirm the interaction. MdZAT10 and MdBT2 were fused to the Cterminal (cYFP) and N-terminal (nYFP) regions of yellow fluorescent protein, respectively. As shown in Fig. 5d, a strong fluorescence signal was detected in the nucleus when MdZAT10-cYFP and MdBT2-nYFP were injected into N. benthamiana leaves, whereas no YFP fluorescence signal was detected in the negative controls. Taken together, these results suggested that MdBT2 physically interacts with MdZAT10.

MdBT2 promotes degradation of the MdZAT10 protein
Recent studies have shown that MdBT2 generally negatively regulates the stability of its target proteins by ubiquitination 37,38 . We incubated the MdZAT10-His fusion protein with total protein from apple calli treated with or without the proteasome inhibitor MG132. The protein stability of MdZAT10 was enhanced by MG132, indicating that the MdZAT10 protein is degraded by the 26S proteasome pathway (Supplementary Fig. S4).
Then, we speculated that MdBT2 mediates MdZAT10 protein stability. To confirm this speculation, we generated transgenic MdBT2-overexpressing and antisense apple calli (MdBT2-OX and MdBT2-Anti). A cell-free MdZAT10 degradation assay was performed by incubating MdZAT10-His with total protein extracted from wild-type, MdBT2-OX, and MdBT2-Anti calli. As shown in Fig. 6a, degradation of the MdZAT10-His protein was much more rapid in the MdBT2-OX calli than in the wild-type calli, whereas MdZAT10-His degraded more slowly in the MdBT2-Anti calli. However, the proteasome inhibitor MG132 noticeably repressed degradation of the MdZAT10-His protein (Fig. 6a). These results indicated that the MdZAT10 protein is degraded by the 26S proteasome pathway, and a ubiquitination assay was performed to further verify this finding. The total protein extracted from MdBT2-GFP calli and GFP was incubated with the MdZAT10-His protein. The ubiquitinated MdZAT10-His protein was assessed using anti-His and anti-Ubi antibodies. A larger amount of high-molecularmass MdZAT10-His was detected in the MdBT2-GFP + MdZAT10-His mixture (Fig. 6b). To confirm that MdBT2 promotes the degradation of MdZAT10 in vivo, we generated 35S::MdZAT10-GFP and 35S::MdZAT10-GFP + 35S:: MdBT2-OX transgenic apple calli. Western blotting was performed with an anti-GFP antibody, and the MdZAT10-GFP protein abundance was lower in the 35S::MdZAT10-GFP + 35S::MdBT2-OX transgenic calli (Fig. 6c). Taken together, these results demonstrated that MdBT2 mediates degradation of the MdZAT10 protein.

MdBT2 delayed the MdZAT10-mediated leaf senescence
Previous studies have revealed that MdBT2 delays JAinduced leaf senescence 36 . Given that MdBT2 promoted the ubiquitination of MdZAT10, we speculated that MdBT2 is involved in the regulation of MdZAT10promoted leaf senescence. To investigate the potential function of MdBT2 in leaf senescence, we obtained three transgenic Arabidopsis lines (MdBT2-L1, L2, and L3) and transgenic apple plants overexpressing BT2 (MdBT2-OE-L1 and MdBT2-OE-L5) and BT2 antisense (MdBT2-Anti-L13 and MdBT2-Anti-L23) (Supplementary Fig. S2). MdBT2overexpressing leaves from both Arabidopsis and apple showed a delayed senescence phenotype, whereas MdBT2 antisense plants showed accelerated JA-induced leaf senescence. The chlorophyll content was consistent with the phenotype (Fig. 7a, b and Supplementary Fig. S6). The Arabidopsis line overexpressing MdZAT10 was crossed with the Arabidopsis line overexpressing MdBT2. The resultant Arabidopsis plants overexpressing MdZAT10/ MdBT2 showed delayed leaf senescence and enhanced chlorophyll content compared to those of plants overexpressing MdZAT10 (Fig. 7c, d). Consistent with these results, the overexpression of MdBT2 decreased the MdZAT10-promoted leaf senescence phenotype in apple leaves and increased the chlorophyll content (Fig. 7e, f). Taken together, these results indicated that MdBT2 delayed the MdZAT10-promoted leaf senescence.

Discussion
Leaf senescence is a complex process that involves the degradation of cellular components such as chloroplasts 39 . Accompanied by the degradation of massive amounts of chlorophyll, the most visible feature of plant senescence is leaf yellowing. Leaf senescence also affects crop productivity and plant fitness 40 . Many TFs show expression changes during leaf senescence 41 . In this study, we identified a C2H2-type zinc finger TF, MdZAT10, which positively regulates dark-and JA-induced leaf senescence. STZ/ZAT10 is a member of the C2H2-type zinc finger TF family in subclass C1-2i that is involved in different abiotic stresses such as drought, salinity, cold, and osmotic stresses [42][43][44][45][46] . In addition, it plays central roles in plant growth and development 47 . Previous studies showed that several C2H2-type zinc finger TFs are induced during senescence in Arabidopsis 19,41 . AZF2 functions as a positive regulator of age-dependent leaf senescence, and the loss of AZF2 function delayed leaf senescence 27 . Our results showed that MdZAT10 expression is higher in senescent leaves than in young leaves (Fig. 1), and overexpressing MdZAT10 accelerated leaf senescence (Fig. 1). We further explored how MdZAT10 promotes leaf senescence and measured the expression levels of some senescence-related genes. MdZAT10-OX transgenic calli showed significantly enhanced the expression of senescence-related genes MdSAG29 and MdPAO and downregulated the expression of MdWRKY70 and MdAPX2 (Supplementary Fig. S7). Overexpression of SAG29 in Arabidopsis accelerated leaf senescence 48 . SAG29 acts as a target gene of MYC2 and is involved in JA-induced leaf senescence 18 . Pheophorbide a oxygenase (PAO) is induced by natural senescence, and the pao1 mutant exhibits a stay-green phenotype 49 . In Arabidopsis, WRKY70 negatively regulates age-dependent leaf senescence 50 . APX2 (ASCORBATE PEROXIDASE2) encodes cytosolic APX2, which plays a key role in removing H 2 O 2 51 . H 2 O 2 is the most commonly used inducer of leaf senescence 52 . Elevated reactive oxygen species (ROS) levels have been reported to accelerate leaf senescence [53][54][55] . It is possible that MdZAT10 promotes leaf senescence by reducing ROS scavenging or regulating the expression of senescence-related genes.
In addition, we found that MdZAT10 promotes JAinduced leaf senescence. MdZAT10 was induced with MeJA treatment, and MdZAT10-OX transgenic calli showed an enhanced expression of MdAOC1 and MdAOS (ALLENE OXIDE SYNTHASE) (Supplementary Fig. S7). AOC1 and AOS are associated with the JA biosynthesis pathway, and their expression is upregulated during leaf senescence 10 . In Arabidopsis, MYC2 binds the promoter of STZ/ZAT10 and regulates its expression 56 . STZ/ ZAT10 and AZF2 can bind the LOX3 promoter to regulate the early response to MeJA 57 . These results indicate that STZ/ZAT10 is involved in JA signaling.
In addition to regulating downstream genes, MdZAT10 may regulate leaf senescence by interacting with other proteins. Here, we demonstrated that MdZAT10 interacts with MdABI5 (Fig. 2). ABI5 regulates dark-and ABAinduced leaf senescence 4,30,32,58 . Recent studies have indicated that ABI5 negatively regulates photosynthesis and chloroplast development under dark treatment 4,58 . RNA-seq data showed that StABI5 negatively regulates the expression level of photosynthesis-related genes 58 . ABI5 positively regulates leaf senescence by directly repressing the expression of ABR 30 and activating the expression of NYC1 and NYE1 4,31,58 . Here, our data showed that MdABI5-overexpressing transgenic plants showed enhanced leaf senescence and that MdABI5 antisense suppression delayed leaf senescence (Supplementary Fig. S3). MdABI5 activated the expression of MdNYC1 and MdNYE1 in MdABI5-OX transgenic apple calli, further confirming the positive regulation of leaf senescence by MdAB15 (Fig. 3). We found that MdZAT10 accelerated MdABI5-mediated leaf senescence and increased the transcriptional activation of MdNYC1 and MdNYE1 by MdABI5 (Fig. 3). It is possible that MdZAT10 promotes leaf senescence by interacting with MdABI5 to affect the transcriptional activity of MdABI5 for target genes.
In addition to MdABI5, MdZAT10 also interacts with MdBT2. MdBT2 delayed JA-induced leaf senescence, and MdBT2 delayed the MdZAT10-promoted leaf senescence  (Fig. 7). These results indicated that MdBT2 plays an opposite role in MdZAT10-promoted leaf senescence. Ubiquitination is widely involved in plant biological processes 20 . Several studies have shown the involvement of some E3 ligases in leaf senescence via the 26S proteasome pathway 59 . A recent study revealed that MdBT2 interacts with MdMYC2 and MdJAZ2, thereby regulating JAinduced leaf senescence 36 . BT2 is a component of the CRL3 complex that promotes target protein ubiquitination 60 . Our results showed that MdBT2 promoted the degradation of MdZAT10 and delayed JA-induced leaf senescence (Figs. 5-7). In the presence of JA, the degradation of MdBT2 was promoted 36 , thus releasing MdZAT10 from MdBT2-mediated degradation. Our results provide insight into the molecular mechanisms by which BT2 mediates JA-induced leaf senescence. These results imply that MdBT2 dynamically regulates JAinduced leaf senescence by regulating different target proteins. Although MdZAT10 interacts with MdBT2 and MdABI5, it is unclear whether these two interactions are related. The same region of MdZAT10 interacts with MdBT2 and MdABI5, and both MdBT2 and MdABI5 interact with full-length MdZAT10 (Figs. 2 and 5). It is possible that MdBT2 and MdABI5 competitively interact with MdZAT10 to antagonistically regulate leaf senescence, but this hypothesis requires further verification.
In summary, a working model to summarize the role of MdZAT10 in leaf senescence is proposed (Fig. 8). On the one hand, MdZAT10 positively regulates leaf senescence through its interaction with MdABI5, enhancing the transcriptional activity of MdABI5 for the chlorophyll degradation genes MdNYC1 and MdNYE1. On the other hand, in the absence of MeJA, MdBT2 interacts with MdZAT10 and ubiquitinates MdZAT10 to degrade MdZAT10, thereby negatively regulating MdZAT10promoted leaf senescence. In contrast, MdZAT10 is induced by exogenous MeJA, which promotes leaf senescence. MeJA accelerates the degradation of MdBT2, releasing MdZAT10, which contributes to JA-induced leaf senescence. In this study, we have characterized the role of MdZAT10 in the leaf senescence regulatory network through its direct interaction with MdBT2 and MdABI5.

Vector construction and plant transformation
To construct overexpression vectors, the full-length coding sequences of MdBT2 and MdABI5 were cloned into the pCXCN-Myc vector, and the full-length coding sequence MdZAT10 was cloned into the pRI-101 vector (with a GFP tag). To generate antisense suppression vectors (MdBT2-Anti, MdABI5-Anti, and MdZAT10-Anti), fragments of the MdBT2, MdABI5, and MdZAT10 sequences were cloned into the pRI-101 vector. MdBT2-GFP transgenic apple calli were obtained as described in our previous study 38 . To generate the proMdZAT10:GUS construct, the promoter fragment of MdZAT10 was inserted into the pCAMBIA1300-GUS vector, which was then transformed into apple calli. Transgenic apple seedlings, apple calli, and Arabidopsis seedlings were obtained according to previously described methods 61,62 . The transient transformation apple leaves were obtained according to a previously described method 62 . All the primers used for gene cloning are listed in Supplementary Table 1.

RNA extraction and gene expression analysis
Total RNA was isolated from apple seedlings, apple calli, Arabidopsis seedlings, and treated seedlings using the RNAplant Plus kit (TIANGEN) according to the manufacturer's protocol. First-strand cDNA was synthesized using the PrimeScript™ RT Reagent Kit (TaKaRa) according to the manufacturer's instructions. qRT-PCR was performed on a StepOnePlus instrument (Applied Biosystems) using UltraSYBR Mixture (Takara). All primers are listed in Supplementary Table 2. Three biological replicates and three technical replicates were performed for each experiment.

MdZAT10-interacting protein screen
A LC/MS assay was performed with the MdZAT10-GFP protein to screen out MdZAT10-interacting proteins as previously described 63 . The MdZAT10-GFP protein was extracted from MdZAT10-overexpressing transgenic calli and purified using a Pierce Classic IP Kit (Thermo Fisher). The MdZAT10-GFP protein was incubated with total protein extracted from apple seedlings for 6 h. The mixed protein solutions were incubated with protein A/G agarose beads and anti-GFP antibody according to a standard co-immunoprecipitation protocol. Then, the resin was eluted, and the eluted proteins were separated on SDS-PAGE gels. Protein identification was carried out by LC-MS/MS (OE Biotech, Shanghai, China).

Analysis of leaf senescence phenotype
To examine the leaf senescence phenotype, detached leaves were placed on 3 mM MES buffer (pH 5.8) in the dark at 22°C. To examine phytohormone-induced leaf senescence, detached leaves were floated on 3 mM MES buffer (pH 5.8) containing 100 μM MeJA at 22°C and kept under dim light for the indicated time period.

Quantification of the chlorophyll content and Fv/Fm ratio
To measure the total chlorophyll concentration, total pigments were extracted from plant leaves with 95% (v/v) ethanol for 24 h. The absorbance at 649 and 665 nm was measured using an ultraviolet/visible spectrophotometer (SOPTOP UV2800S, Shanghai, China). To calculate the Fv/Fm ratios, leaves were analyzed with a closed chlorophyll fluorescence imaging system (Photon System Instruments, Brno, Czech Republic) according to the manufacturer's instructions as previously described 31 .

Protein degradation and ubiquitination assays
For the in vitro protein degradation assay, total protein was extracted from wild-type and transgenic apple calli using degradation buffer (25 mM Tris-HCl, pH 7.5, 10 mM NaCl, 10 mM MgCl 2 , 4 mM phenylmethylsulfonyl fluoride, 5 mM DTT, and 10 mM ATP). The protein extracts were incubated with MdZAT10-His at 22°C and assessed by western blotting with an anti-His antibody (Abmart).
We performed an in vivo ubiquitination assay as described previously 64 . In brief, MdBT2-GFP transgenic calli was extracted using a Pierce TM Classic IP Kit (Thermo Fisher), and the extracts were incubated with MdZAT10-His protein at 4°C overnight. The in vivo ubiquitination of MdZAT10 was detected by western blotting with anti-His (Abmart) and anti-Ubi (Sigma-Aldrich) antibodies.

Transient expression assay
To carry out the transient expression assay, the MdNYC1 and MdNYE1 promoter sequences were inserted into the pGreenII 0800-LUC vector. Full-length MdZAT10 and MdABI5 were cloned into the pGreen 62-SK vector. The recombinant plasmids were transformed into N. benthamiana leaves by Agrobacteriummediated transformation, and LUC/REN activity ratio was detected using a dual-luciferase reporter assay system (Promega) 64 .

Statistical analyses
Each experiment in this study was repeated at least three times. The data were analyzed by t-test using GraphPad Prism 6.02 software, and asterisks denote significant differences ( * P < 0.05, ** P < 0.01).