PARylation of the forkhead‐associated domain protein DAWDLE regulates plant immunity

Abstract

Protein poly(ADP‐ribosyl)ation (PARylation) primarily catalyzed by poly(ADP‐ribose) polymerases (PARPs) plays a crucial role in controlling various cellular responses. However, PARylation targets and their functions remain largely elusive. Here, we deployed an Arabidopsis protein microarray coupled with in vitro PARylation assays to globally identify PARylation targets in plants. Consistent with the essential role of PARylation in plant immunity, the forkhead‐associated (FHA) domain protein DAWDLE (DDL), one of PARP2 targets, positively regulates plant defense to both adapted and non‐adapted pathogens. Arabidopsis PARP2 interacts with and PARylates DDL, which was enhanced upon treatment of bacterial flagellin. Mass spectrometry and mutagenesis analysis identified multiple PARylation sites of DDL by PARP2. Genetic complementation assays indicate that DDL PARylation is required for its function in plant immunity. In contrast, DDL PARylation appears to be dispensable for its previously reported function in plant development partially mediated by the regulation of microRNA biogenesis. Our study uncovers many previously unknown PARylation targets and points to the distinct functions of DDL in plant immunity and development mediated by protein PARylation and small RNA biogenesis, respectively.

Introduction

In addition to the preformed physical barriers, the innate immune system is essential for sessile plants to ward off potential infections 1. Perception of microbe‐associated molecular patterns (MAMPs) by cell surface‐resident pattern recognition receptors (PRRs) activates the first line of plant immunity, termed as pattern‐triggered immunity (PTI) 2, 3. Bacterial flagellin, lipopolysaccharide (LPS), peptidoglycan (PGN), elongation factor Tu (EF‐Tu), and fungal chitin are among well‐characterized MAMPs that elicit robust defense responses in plants. Recognition of flg22, a 22 amino acid peptide derived from bacterial flagellin, by the cognate PRR flagellin‐sensing 2 (FLS2) initiates immune signaling via heterodimerization with brassinosteroid (BR)‐insensitive 1 (BRI1)‐associated kinase 1 (BAK1) in Arabidopsis 4, 5. Members of Botrytis‐induced kinase 1 (BIK1) family receptor‐like cytoplasmic kinases (RLCKs) are rapidly phosphorylated by the FLS2/BAK1 complex upon flg22 perception to transduce intracellular signaling 6, 7. BIK1 directly phosphorylates NADPH oxidase respiratory burst oxidase homolog D (RBOHD), thereby contributing to the production of reactive oxygen species (ROS) 8, 9. In addition, MAMP perception triggers rapid activation of mitogen‐activated protein kinases (MAPKs) and calcium‐dependent protein kinases (CDPKs), calcium influx, stomatal closure, callose deposition, and production of antimicrobial compounds and defense hormones 2, 10. The robust defense responses are intertwined with MAMP‐induced massive transcriptional reprogramming which are orchestrated by the coordinated action of gene‐specific transcription factors and the general transcriptional machinery 11, 12, 13.

Protein poly(ADP‐ribosyl)ation (PARylation) is an important post‐translational modification process that plays a crucial role in a broad array of cellular responses such as DNA damage, cell death, and inflammation 14, 15. PARylation is primarily catalyzed by poly(ADP‐ribose) polymerases (PARPs), which transfer ADP‐ribose moieties from donor nicotinamide adenine dinucleotide (NAD⁺) molecules to target proteins often at glutamate (Glu), aspartate (Asp), or in some cases lysine (Lys) residues, resulting in the formation of linear or branched poly(ADP‐ribose) (PAR) polymers on target proteins 14, 15. PARylation is a reversible reaction and the covalently attached PAR on the target proteins can be hydrolyzed to free PAR or mono‐(ADP‐ribose) by poly(ADP‐ribose) glycohydrolases (PARGs). PAR activities and PARPs have been found in a wide range of organisms from archaebacteria to mammals and plants 16. Similar with their animal counterparts, plant PARPs and PARGs are important regulators in plant growth, genotoxic and abiotic stress responses, circadian clock, and immunity to pathogen infections 16, 17. In contrast to animals, which often contain more than a dozen of PARP family members in the genome, the Arabidopsis genome encodes three PARPs, PARP1, PARP2, and PARP3. Although AtPARP1 bears the highest sequence homology to human hsPARP‐1, which accounts for more than 90% of PARP activities in humans, PARP2 appears to be the most active and important PARP enzyme in Arabidopsis 18, 19. Protein PARylation regulates multiple PTI responses, such as immune gene activation and callose deposition. Importantly, MAMP treatment enhances PARP2 auto‐PARylation activity, suggesting that PARylation is an integral part of plant innate immunity 18. However, protein PARylation targets remain largely unknown in plants.

To identify PARylation substrates, we deployed a high‐density Arabidopsis protein microarray coupled with in vitro PARylation assays and identified 54 proteins as the putative substrates of PARP2. Several of these candidates were confirmed with in vivo PARylation assays. DAWDLE (DDL), a forkhead‐associated (FHA) domain protein, was found to play an important role in plant defense responses. DDL was previously reported to control plant growth and development partially via regulating small RNA biogenesis 20, 21. We showed that the ddl mutants exhibited enhanced susceptibility to the infections by several bacterial pathogens accompanied with the compromised MAMP‐induced callose deposition and defense gene activation at late time points. PARP2 directly interacts with DDL in a PARylation‐dependent manner. Mass spectrometry (MS), mutagenesis, and complementation analyses indicate that protein PARylation of DDL is essential for its function in plant immunity, but dispensable for its function in plant development. Thus, we identified DDL as a physiological target of protein PARylation in plant immunity, which is uncoupled from its function in plant development.

Results

Identification of PARylated proteins using Arabidopsis protein microarray

We have previously established in vitro and in vivo PARylation assays of Arabidopsis PARPs and PARGs, and demonstrated that PARP2 possesses a robust polymerase activity 18. To identify PARylated Arabidopsis proteins, we employed a high‐density Arabidopsis protein microarray containing 10,048 proteins and performed in vitro PARylation assays using recombinant proteins of PARP2 fused with maltose‐binding protein (MBP). The protein microarray chips were incubated with MBP‐PARP2 proteins in the presence of NAD⁺, the donor of ADP‐ribose groups. The chips incubated without NAD⁺ served as a negative control. After extensive washing, the protein chips were incubated with α‐PAR antibody, which detects the PAR polymer of PARylated proteins, and then Cy3 fluorophore‐conjugated secondary antibody. The fluorescence signals were scanned and imaged by GenePix4100A scanner (Fig 1A). As the same proteins were duplicated as two individual spots on the protein chips, the proteins with both spots consistently exhibiting significantly higher signals than those on the negative control chips were identified as PARylated proteins by PARP2 (Fig 1B, Appendix Fig S1 and Dataset EV1). We have detected 54 candidates that are likely PARylated by PARP2 using this approach (Appendix Table S1). Notably, based on the prediction of protein subcellular localization by TAIR, 56% of candidates are predicated as nuclear‐localized proteins (Fig 1C) compared to 34% of total proteins on the chip to be nuclear‐localized (Dataset EV1), consistent with the nuclear localization of PARP2 18. Gene Ontology (GO) analysis indicates that proteins associated with DNA/RNA metabolism, response to stresses, response to biotic/abiotic stimuli, and transcription are enriched among candidates compared to the total proteins on the chip (Fig 1D and Dataset EV1). This is consistent with the notion that protein PARylation is often involved in stress responses, DNA damage repair, chromatin modification, and gene transcriptional regulation 22.

Figure 1. Identification of PARylation targets using Arabidopsis protein microarray.

1. Scheme of identification of Arabidopsis proteins PARylated by PARP2 using protein microarray. Same proteins were spotted as duplicates on the chips. Protein chips were first incubated with MBP‐PARP2 proteins in the PARylation reaction buffer with NAD⁺ (left) or without NAD⁺ (right). The PARylated proteins indicated as the orange dotted chain were detected with α‐PAR antibody hybridization followed by Cy3 fluorophore‐conjugated secondary antibody. The fluorescence signals were scanned and imaged by GenePix4100A scanner.
2. Scanned images from a pair of representative protein chips after in vitro PARylation reaction.
3. Subcellular localization of identified candidates based on TAIR prediction. The percentage of each subgroup among the total candidates was presented.
4. GO analysis of identified candidates (dark gray bars) compared to the total proteins on the chip (light gray bars). The proteins were categorized based on the GO annotation by TAIR. The x‐axis shows the percentage of genes in a specific functional category; the y‐axis represents different categories.
5. Confirmation of PARylated proteins with an in vivo PARylation assay. FLAG‐tagged candidate proteins were expressed in protoplasts and immunoprecipitated by α‐FLAG antibody after feeding protoplasts with ³²P‐NAD⁺. The immunoprecipitated proteins were separated in 10% SDS–PAGE and detected by autoradiography (top panel). The input of FLAG‐tagged proteins is shown in an α‐FLAG immunoblot (middle panel), and the protein loading control is shown by Ponceau S staining for RuBisCo (RBC) (bottom panel). The experiments were repeated three times with similar results.

To validate whether some of the candidates are PARylated in vivo, we transiently expressed the candidate genes with a FLAG‐epitope tag in Arabidopsis protoplasts fed with radiolabeled ³²P‐NAD⁺ as the ADP‐ribose donor. The candidate proteins were immunoprecipitated with α‐FLAG antibody and subjected to SDS–PAGE autoradiograph. The PARylated proteins were detected as a ladder‐like smear due to the incorporation of radiolabeled ADP‐ribose with different length of PAR polymer formation. As shown in Fig 1E, DAWDLE (DDL, AT3G20550), PLANT TUDOR‐LIKE PROTEIN (AT1G06340), HYALURONAN/mRNA‐BINDING PROTEIN (AT5G47210), and METHYL‐CPG‐BINDING DOMAIN 11 (MBD11, AT3G15790) were PARylated in vivo compared with the transfection of a control plasmid (Fig 1E) or a FLAG‐tagged nuclear protein, AT5G03660 (Fig 1E, left lane).

The compromised disease resistance to bacterial pathogens in the ddl mutants

Protein PARylation has been implicated to play an important role in plant defense 18, 19, 23. To determine whether any identified PARylation target candidates are involved in plant defense responses, we characterized homozygous T‐DNA insertional mutants of several candidates in response to the infection by a virulent bacterial pathogen, Pseudomonas syringae pv. tomato (Pst) DC3000. Among five homozygous T‐DNA insertional mutants examined (ddl, mbd11, ubc13b, at1 g10520, and at4 g22150) (Appendix Fig S2A), the ddl‐6 (SALK_045025) mutant exhibited high susceptibility to Pst DC3000 infection compared to Col‐0 WT and other mutant plants (Appendix Fig S2B). We failed to identify the homozygous mutants for AT5G47210 and AT1G06340 that were PARylated in vivo (Fig 1E). The transcript of MBD11 and UBC13B was not altered in the corresponding T‐DNA insertion plants (Appendix Fig S2A). Quantitative RT–PCR (qRT–PCR) analysis indicated that DDL transcripts were substantially reduced in ddl‐6 (Fig 2A and B). The bacterial multiplication of Pst DC3000 was more than tenfold higher in the ddl‐6 mutant than that in WT plants 3 days post‐inoculation (dpi) (Fig 2C and Appendix Fig S2B). The disease symptom development with yellowing and necrosis leaves was more pronounced in the ddl‐6 mutant than that in WT plants (Fig 2C). We also isolated another allele of ddl T‐DNA insertional mutants, ddl‐7 (SAIL_1281_F08), which had about fourfold reduction in DDL transcripts compared to WT Col‐0 plants (Fig 2A and B). The ddl‐7 mutant also exhibited enhanced susceptibility to Pst DC3000 infections compared to WT plants although to a less extend than ddl‐6 (Fig 2C). The previously reported ddl1‐1 and ddl1‐2 mutants in the Ws background were dwarf with severely stunted growth under the 12‐h light/12‐h dark growth condition 20 (Appendix Fig S3A). However, the severity of growth defects of ddl‐6 and ddl‐7 mutants in the Col‐0 background was less pronounced than that of ddl‐1 and ddl‐2 in the Ws background (Appendix Fig S3A). Clearly, ddl‐6 and ddl‐7 mutants were much bigger than ddl‐1 and ddl‐2 at 4 weeks after germination when we performed bacterial infection assays (Appendix Fig S3A). Similar with ddl‐1 and ddl‐2, ddl‐6 displayed delayed flowering and reduced apical dominance compared with Col‐0 WT plants (Appendix Fig S3B). The ddl‐1 mutant was also more susceptible to Pst DC3000 infections than Ws plants (Fig 2D). In addition, ddl‐6 was more susceptible to another virulent bacterial pathogen P. syringae pv. maculicola (Psm) ES4326 (Fig 2E), non‐pathogenic bacterium Pst DC3000 hrcC, a type III secretion mutant of Pst DC3000 (Figs 2F and EV1A), and non‐adaptive pathogen P. syringae pv. phaseolicola (Psp) (Fig 2G). Expression of DDL under the control of the native promoter in ddl‐6 restored the disease susceptibility to Pst DC3000 infection to the WT level in two independent transgenic lines W7 and W18 (Fig 2H). The data indicate that DDL plays an important role in plant defense to bacterial infections.

Figure 2. DDL is involved in plant defense.

1. Scheme of DDL protein domains and gene structures with molecular lesions of five mutants. N‐terminal NLS (nuclear localization signal) and C‐terminal FHA domains are labeled. The dark gray bars indicate exons, and lines in between indicate introns. The light gray bars indicate 5′ and 3′ UTRs. The ddl‐1, ddl‐2, and ddl‐3 mutants are in the Ws background, whereas ddl‐6 (SALK_045025) and ddl‐7 (SAIL_1281_F08) are in the Col‐0 background. ddl‐1, ddl‐2, ddl‐6, and ddl‐7 are T‐DNA insertional lines, whereas ddl‐3 is a TILLING line with a point mutation.
2. The DDL transcripts and growth phenotype of ddl‐6 and ddl‐7. DDL transcripts were detected in fully expanded leaves of 4‐week‐old plants by qRT–PCR (left panel). The expression of DDL in WT Col‐0 was set as 1. Plants grown under 12‐h light/12‐h dark condition for 4 weeks are shown (right panel). Scale bar = 1 cm.
3. ddl‐6 and ddl‐7 are more susceptible to Pst DC3000 infections. Leaves of 4‐week‐old plants were inoculated with Pst DC3000 at OD₆₀₀ = 5 × 10⁻⁴. Bacterial numbers were counted at 0 and 3 dpi. Leaf pictures were taken at 3 dpi. Scale bar = 1 cm.
4. The ddl‐1 mutant in the Ws background is more susceptible to Pst DC3000 infection.
5. ddl‐6 is more susceptible to Psm infection. Leaves of 4‐week‐old plants were inoculated with Psm at OD₆₀₀ = 5 × 10⁻⁴.
6. Increased bacterial growth of Pst DC3000 hrcC in ddl‐6. Leaves of 4‐week‐old plants were inoculated with Pst DC3000 hrcC at OD₆₀₀ = 5 × 10⁻⁴.
7. Increased bacterial growth of Psp in ddl‐6. Leaves of 4‐week‐old plants were inoculated with Psp at OD₆₀₀ = 5 × 10⁻⁴.
8. Complementation of DDL in disease resistance. Two independent T3 homozygous lines carrying pDDL::DDL‐FLAG (W7 and W18) in the ddl‐6 background were assayed for the susceptibility to Pst DC3000.

Data information: The data are shown as mean ± SD (n = 3) with one‐way ANOVA analysis and Tukey test (P < 0.05). Different letters, a, b, or c above the bars, indicate significant differences. One representative result from three independent repeats is shown.

Figure EV1. Defense responses in ddl‐6 .

1. Increased bacterial growth of Pst DC3000 hrcC in ddl‐6. Leaves of 4‐week‐old plants were inoculated with Pst DC3000 hrcC at OD₆₀₀ = 5 × 10⁻⁴. The data are shown as mean ± SD (n = 3) from three independent repeats with one‐way ANOVA analysis and Tukey test (P < 0.05). Different letters, a or b, indicate significant differences.
2. flg22‐induced PTI marker gene expression in Col‐0 and ddl‐6. Ten‐day‐old seedlings were treated with 100 nM flg22 for 60 min for qRT–PCR analysis. Data are shown as mean ± SD from three independent repeats.
3. Pst DC3000‐mediated induction of some late responsive genes is compromised in the ddl‐6 and parp1,2 mutants. Fully expanded leaves of 4‐week‐old plants were inoculated with Pst DC3000 (OD₆₀₀ = 0.01) and collected at 0, 8, and 24 hpi for qRT–PCR. Data are shown as mean ± SD from three independent repeats.
4. BIK1 phosphorylation upon flg22 treatment in Col‐0 and ddl‐6. BIK1‐HA was transiently expressed in Col‐0 or ddl‐6 protoplasts, and its phosphorylation upon flg22 treatment was detected as the band shift by an immunoblot with the α‐HA antibody.

To investigate the potential mechanisms underlying the compromised disease resistance in ddl mutants, we systematically examined various PTI responses in the ddl‐6 mutant. Callose deposition is a relatively late response upon MAMP perception. The ddl‐6 mutants showed reduced callose deposits at 24 h post‐inoculation (hpi) with flg22 or Pst DC3000 hrcC treatment compared to WT plants (Fig 3A). Consistent with an important role of protein PARylation in MAMP‐induced callose deposition, the parp1,2 double mutant also showed reduced callose deposits upon flg22 or Pst DC3000 hrcC treatment (Fig 3A). The parp1,2 mutant showed the reduced induction of some early MAMP responsive genes 18. It appears that the flg22‐mediated induction of FRK1, WRKY30, AT1G07160, and AT2G17740 by 1 h did not change significantly in the ddl‐6 mutant (Fig EV1B). We also investigated the induction of some MAMP responsive genes at late time points, such as 3, 8, or 24 h after treatment 24. Interestingly, the induction of PPDK (AT4G15530), VSR7 (AT4G20110), AT3G08870, and PEN3 (AT1G59870) at 24 hpi after Pst DC3000 hrcC treatment was reduced in the ddl‐6 mutant compared to that in WT plants. The induction of these genes by Pst DC3000 hrcC treatment was also reduced in the parp1,2 mutant (Fig 3B). Similarly, Pst DC3000‐mediated induction of these genes was reduced in both ddl‐6 and parp1,2 mutants (Fig EV1C). The PR1 and PR5 induction by Pst DC3000 was also diminished dramatically in the ddl‐6 mutant at 24 hpi (Fig 3C), suggesting that DDL is required for regulating some late responsive or secondary defense genes. The ddl‐6 mutant did not affect flg22‐induced BIK1 phosphorylation (Fig EV1D), MAPK activation (Fig 3D), or ROS burst (Fig 3E), which is consistent with the dispensable role of protein PARylation in these early responses 18. The data suggested that the compromised plant immunity in the ddl mutant might be partly due to altered PTI responses, especially those occurring at the late time points.

Figure 3. Differential PTI responses in ddl‐6 .

1. flg22‐induced callose deposition is reduced in ddl‐6 and parp1,2. Leaves of 4‐week‐old plants were inoculated with 0.5 μM flg22 for 20 h or hrcC at OD₆₀₀ = 0.2 for 24 h and stained with aniline blue. Callose deposits were visualized under UV light and quantified by ImageJ. The data are shown as mean ± SE (n = 12). Scale bar = 100 μm.
2. The hrcC‐mediated induction of some late responsive genes is compromised in the ddl‐6 and parp1,2 mutants. Fully expanded leaves of 4‐week‐old plants were hand‐inoculated with hrcC (OD₆₀₀ = 0.5) and collected at 0, 3, and 24 hpi for qRT–PCR. The data are shown as mean ± SE from three independent repeats.
3. Pst DC3000‐induced PR1 and PR5 transcripts are blocked in ddl‐6 and parp1,2. Fully expanded leaves of 4‐week‐old plants were inoculated with Pst DC3000 (OD₆₀₀ = 0.01) and collected at 0, 8, and 24 hpi for qRT–PCR. The data are shown as mean ± SE from three independent repeats.
4. flg22‐induced MAPK activation in Col‐0 and ddl‐6. Ten‐day‐old seedlings were treated with 100 nM flg22 and collected at the indicted time points. MAPK activation was analyzed by an immunoblot (IB) with α‐pERK antibody (top panel), and the protein loading is shown by Ponceau S staining for RuBisCo (RBC) (bottom panel).
5. flg22‐induced ROS in Col‐0 and ddl‐6. Leaf disks from 4‐week‐old plants were assayed for ROS production upon 100 nM flg22 treatment over 30 min. The data are show as mean ± SE (n = 24).

Data information: Experiments were repeated three times with similar results.

Flg22 treatment potentiates DDL–PARP2 interaction

Since DDL is PARylated by PARP2, we further investigated whether DDL directly interacted with PARP2 and whether this interaction is modulated upon MAMP elicitation. We expressed FLAG‐epitope tagged DDL and HA‐epitope tagged PAPR2 in Arabidopsis protoplasts. DDL‐FLAG could immunoprecipitate PARP2‐HA (Fig 4A). Apparently, the association was enhanced upon flg22 treatment (Fig 4A). The flg22‐induced DDL‐HA and PARP2‐FLAG association was also observed when they were transiently expressed in Nicotiana benthamiana (Fig 4B). Consistently, bimolecular fluorescence complementation (BiFC) assay also indicated the association of DDL and PARP2 in nuclei with co‐expression of PARP2 fused to amino‐terminal half of YFP (yellow fluorescence protein) (nYFP‐PARP2) and DDL fused to the carboxyl‐terminal half of YFP (cYFP‐DDL) in Col‐0 protoplasts (Fig 4C). JAM3, another nuclear protein 25, did not interact with PARP2 or DDL with BiFC assay (Fig EV2A). To test whether DDL directly interacts with PARP2, we performed an in vitro pull‐down assay with PARP2 fused to glutathione‐S‐transferase (GST) immobilized on Sepharose beads as a bait against DDL fused to 6×HIS‐SUMO at N‐terminus and an HA tag at C‐terminus. As shown in Fig 4D, HIS‐SUMO‐DDL‐HA could be pulled down by GST‐PARP2, but not GST itself. Interestingly, when GST‐PARP2 and HIS‐SUMO‐DDL‐HA were first subjected to a PARylation reaction before the pull‐down assay (Fig 4D, right panels), the interaction between DDL and PARP2 was markedly enhanced (Fig 4D, left panels), suggesting that PARylation may enhance the association of DDL and PARP2. The data are in line with the flg22‐induced DDL–PARP2 association (Fig 4A and B). We have shown that flg22 treatment enhanced PARylation activity of PARP2 18. The enhanced PARP2 activity may promote PARP2 and DDL interaction. Consistent with the direct interaction between DDL and PARP2, PARP2 PARylated DDL in vitro (Fig 4E). DDL‐FLAG immunoprecipitated from protoplasts could be PARylated by GST‐PARP2 in the presence of biotin‐NAD⁺ as detected by a Western blot using the streptavidin‐HRP antibody (Fig 4E). The PARylation of DDL, as well as PARP2, could be suppressed by the PARP inhibitor, 3‐aminobenzamide (3‐AB), or removed by GST‐PARG1 (Fig 4E).

Figure 4. DDL directly interacts with PARP2.

1. PARP2 and DDL association in Arabidopsis protoplasts. PARP2‐HA and DDL‐FLAG were transiently expressed in Col‐0 protoplasts, immunoprecipitated with α‐FLAG antibody (IP: α‐FLAG), and immunoblotted with α‐HA (IB: α‐HA) or α‐FLAG antibody (IB: α‐FLAG) (top two panels). The protein inputs are shown with immunoblotting before immunoprecipitation (bottom two panels). Protoplasts were treated without or with 100 nM flg22 for 30 min.
2. PARP2 and DDL association in N. benthamiana. PARP2‐FLAG and DDL‐HA were transiently co‐expressed in leaves of N. benthamiana. The samples were collected at 2 dpi for co‐immunoprecipitation (Co‐IP) assay with 100 nM flg22 treatment for 1 h.
3. PARP2 and DDL association with BiFC assay. Different combinations of constructs were transiently expressed in Arabidopsis Col‐0 protoplasts, and the fluorescence signals were observed with a confocal microscope. An NLS‐RFP construct was co‐transfected to show the nuclear signals. Scale bar = 10 μm.
4. PARP2 and DDL interaction in vitro. GST‐PARP2 protein immobilized on glutathione beads was used to pull down HIS‐SUMO‐DDL‐HA proteins (left panels). A PARylation reaction was performed prior to the pull‐down assay (the third lane of left panels). The PARylated PARP2 and DDL were detected with α‐PAR antibody (right panels).
5. In vitro PARylation of DDL by PARP2. Immunoprecipitated DDL‐FLAG proteins from Arabidopsis protoplasts were incubated with GST‐PARP2 in a PARylation reaction containing biotin‐NAD⁺. The reactions in lane 4 and 5 contained 1 mM 3‐AB or GST‐PARG1, respectively. PARylated proteins were detected by streptavidin‐HRP antibody (Strep), which recognizes biotin‐labeled PAR.

Data information: All experiments were repeated three times with similar results.

Figure EV2. Interaction of PARP2 with DDL‐N or DDL‐C in BiFC assays.

1. nYFP‐JAM3 did not interact with cYFP‐PARP2 or cYFP‐DDL with BiFC assay. Scale bar = 10 μm.
2. nYFP‐PARP2 interacts with cYFP‐DDL‐N. Different BiFC constructs were transfected in Arabidopsis protoplasts, and the fluorescence signals were observed with a confocal microscope. The NLS‐RFP construct was co‐transfected to show the nuclear localization. Scale bar = 10 μm.
3. Localization of GFP‐tagged DDL, DDL^12E1, DDL‐N, DDL‐C, and JAM3. The NLS‐RFP construct and indicated GFP constructs were co‐transfected in Arabidopsis protoplasts. The fluorescence signals were observed with a fluorescence microscope. Scale bar = 10 μm.

The flg22‐induced DDL PARylation by PARP2

We examined DDL PARylation dynamics upon flg22 treatment with an in vivo PARylation assay in which protoplasts expressing DDL‐FLAG were fed with radiolabeled ³²P‐NAD⁺. As shown in Fig 5A, PARylated DDL was detected as smear bands with higher molecular weight on the autoradiograph. The band intensity was enhanced upon flg22 treatment (Fig 5A), suggesting that flg22 perception induces DDL PARylation.

Figure 5. flg22 stimulates PARP1/2‐dependent DDL PARylation.

• A
  flg22 stimulates DDL PARylation. DDL‐FLAG was transiently expressed in Arabidopsis protoplasts fed with ³²P‐NAD⁺. Protoplasts were treated without or with 100 nM flg22 for 30 min. PARylated DDL was immunoprecipitated with α‐FLAG affinity beads, separated on 10% SDS–PAGE, and visualized by autoradiography. PAR(DDL) indicates PARylated DDL proteins. A representative result from three repeats is shown.
• B, C
  DDL PARylation is reduced in parp1, parp2, and parp1,2 mutants. Protoplasts isolated from parp1 or parp2 single mutants (B), or parp1,2 double mutants (C) were used for in vivo PARylation assay with ³²P‐NAD⁺. Protoplasts were treated with 100 nM flg22 for 30 min.

Data information: The experiments were repeated twice with similar results.

We determined whether endogenous PARP1 and PARP2, two enzymatic active PARPs primarily expressing in shoots 18, were required for DDL PARylation in vivo. DDL PARylation was slightly reduced in parp1, but more substantially suppressed in parp2 compared to WT plants (Fig 5B). Importantly, DDL PARylation appeared to be completely blocked in the parp1,2 double mutant (Fig 5C). The data support that PARP1 and PARP2 are two major enzymes for protein PARylation of DDL in Arabidopsis.

Identification of DDL PARylation sites

DAWDLE consists of 314 amino acids (AA) with two domains, a divergent N‐terminal domain, DDL‐N (1‐183 AA), and a highly conserved C‐terminal FHA domain, DDL‐C (184‐314 AA) (Fig 2A). We determined which domain of DDL mediates interaction and PARylation by PARP2. As shown in Fig 6A, PARP2 co‐immunoprecipitated with DDL‐N, but not DDL‐C. Similar with the full‐length DDL, flg22 treatment enhanced the interaction of PARP2 and DDL‐N (Fig 6A). Consistently, BiFC assay revealed that cYFP‐DDL‐N, but not cYFP‐DDL‐C, interacts with nYFP‐PARP2 (Fig EV2B). DDL‐GFP and DDL‐N‐GFP seem to exclusively localize in the nucleus, and DDL‐C‐GFP also largely localizes in the nucleus with some weak signal in the cytosol (Fig EV2C). In addition, DDL‐N but not DDL‐C could be PARylated in vivo (Fig 6B). The PARylation of DDL‐N but not DDL‐C by PARP2 was further confirmed by in vitro PARylation assay with biotin‐NAD⁺ (Fig 6C). Thus, it is likely that the PARylation residues of DDL by PARP2 mainly reside in its N‐terminus.

Figure 6. Identification of DDL PARylation sites.

1. DDL‐N but not DDL‐C associates with PARP2. FLAG‐tagged DDL‐N or DDL‐C was co‐expressed with PARP2‐HA in protoplasts for Co‐IP assay. Protoplasts were treated without or with 100 nM flg22 for 30 min.
2. In vivo PARylation of DDL‐N or DDL‐C in protoplasts fed with ³²P‐NAD⁺. PARylation was visualized with autoradiography.
3. In vitro PARylation of DDL‐N or DDL‐C. FLAG‐tagged DDL‐N or DDL‐C was expressed and immunoprecipitated from Arabidopsis protoplasts and subjected to in vitro PARylation by GST‐PARP2 with biotin‐NAD⁺. PARylated proteins were detected by streptavidin‐HRP antibody.
4. List of the mass/charge values of b⁺ ions and y⁺ ions of the peptide (₁₄₈AVeEALAAK₁₅₆). Red mass values correspond to the identified peak in the spectrum. The modified “e” is in bold and lowercase. The mass difference between b2⁺ and b3⁺ or y6⁺ and y7⁺ indicates that Glu150 is PARylated.
5. MS/MS spectrum of a doubly charged peptide with Glu (the third AA “e” in the peptide, Glu150) PARylation. 2⁺ indicates doubly charged, and ⁰ indicates loss of H₂O. The mass difference between y7⁺ and y6⁺ is 144.1, which is about 15 Da bigger than the Glu residue mass of 129.04.
6. DDL^12E mutants substantially diminish its PARylation in vivo. In vivo PARylation of DDL^12E1 or DDL^12E2 was performed using Arabidopsis protoplasts fed with ³²P‐NAD⁺.
7. Reduced association of DDL^12E1 with PARP2 in protoplasts. Protoplasts were treated with 100 nM flg22 for 30 min before Co‐IP assay.

Data information: The above experiments except MS analysis were repeated twice with similar results.

We next deployed a site‐specific proteomics approach, which has been recently developed to characterize the human PARylated proteome 26, to identify potential PARylation residues of DDL by PARP2. PARylation mainly occurs on the Glu (E) or Asp (D) residues of target proteins. In this approach, NH₂OH treatment of PARylated proteins resulted in the formation of hydroxamic acid derivative at ADP‐ribosylated Glu or Asp with an addition of 15 Da, an increment that can be readily distinguished by MS analysis 26. We first validated the effectiveness of NH₂OH treatment by performing an in vitro PARylation assay of DDL (HIS‐SUMO‐DDL‐HA) by GST‐PARP2. As shown in Appendix Fig S4A, after 1 M NH₂OH treatment, the smear bands corresponding to self‐modified PARP2 and PARylated DDL were largely diminished, suggesting that the poly(ADP‐ribose) covalently attached to PARP2 and DDL was effectively removed by NH₂OH treatment. We then sliced the gel corresponding to the treated HIS‐SUMO‐DDL‐HA proteins (arrow head 2) for MS analysis (Appendix Fig S4A). In total, we identified eight putative PARylated sites on six peptides (Appendix Fig S4B–D). Six residues were located on the N‐terminus of DDL and two on the C‐terminus (Appendix Fig S4B). We did not detect any PARylated sites on the gel slice with unmodified HIS‐SUMO‐DDL‐HA proteins (arrow head 1 in Appendix Fig S4A). Among eight identified PARylated sites, E150 is the most confident one, and D105 and E116 are not confident (Appendix Fig S4C). The residue mass between y7 and y6 in the peptide of ₁₄₈‐AVeEALAAK‐₁₅₆ is 144.1, which is about 15 Da bigger than the Glu residue mass of 129.04, suggesting that E150 was PARylated (Fig 6D and E). In addition, the b ion series also support such a modification (Fig 6D and E). Notably, the amino acid following E150 is also a Glu (E151). Two consecutive Glu residues are conserved sites that are often required for protein PARylation 27. We mutated E150/E151 to Alanine (2E mutant) and compared PARylation ability of WT and mutated DDL (Fig EV3A). Apparently, DDL^2E did not affect DDL PARylation with ³²P‐NAD⁺‐based in vivo PARylation assay (Fig EV3B). Mutation of some other sides identified by MS analysis, including D32 (DDL^1D), D105, and E116 (DDL^DE), also did not affect DDL PARylation (Fig EV3A and C). It is likely that multiple sites are required for DDL PARylation. We generated higher order DDL mutants with mutations in the potential PARylation sites, especially EEs in the N‐terminus (Fig EV3A). Mutations of two EEs (DDL^4E1 and DDL^4E2) and four EEs (DDL^8E) did not significantly affect DDL PARylation (Fig EV3B). However, mutations of six EEs (DDL^12E1 and DDL^12E2, mutation of 12 Glu to Ala) displayed markedly reduced protein PARylation compared to WT DDL (Fig 6F). PARylation of the DDL^12E1 mutant after flg22 treatment was almost completely diminished compared to WT DDL (Fig 6F). The protein expression of DDL^12E1 or DDL^12E2 was comparable with that of WT DDL (Fig 6F). DDL^12E1‐GFP still localizes in nucleus (Fig EV2C). Consistent with the requirement of PARylation for DDL and PARP2 interaction, DDL^12E1 showed the reduced interaction with PARP2 compared to WT DDL (Fig 6G).

Figure EV3. In vivo PARylation of DDL mutants.

1. A list of DDL mutants generated by site‐directed mutagenesis for in vivo PARylation assay.
2. In vivo PARylation assay of DDL mutants with different EE mutations listed in (A). The Glu residues were replaced with Ala. PARylation of DDL mutants was assayed with ³²P‐NAD⁺‐fed protoplasts.
3. In vivo PARylation of DDL mutants with residues (D32, D105, E116) identified by MS mutated to A.

DDL PARylation is essential for its function in plant immunity

To determine the biological functions of DDL PARylation, we transformed the ddl‐6 mutant with either WT DDL or DDL ^12E1 mutant under the control of its native promoter. Multiple transgenic plants with positive protein expression were obtained, and four representative transgenic lines with similar DDL protein levels were selected for further analysis (Fig 7A). WT DDL transgenic plants (lines W7 and W18) completely restored the disease susceptibility of ddl‐6 to Pst DC3000 infections to the WT level (Fig 2H and 7B). However, DDL ^12E1 mutant transgenic plants (lines E11 and E15) were still more susceptible to Pst DC3000 infections than WT plants (Fig 7B). The susceptibility of line E15 was similar with that of ddl‐6 (Fig 7B). The Pst DC3000‐mediated PR1 and PR5 induction was also restored to the WT level in WT DDL transgenic plants (lines W7 and W18) but remained substantially compromised in DDL ^12E1 mutant transgenic plants (lines E11 and E15) (Fig 7C). Similarly, flg22‐induced callose deposition was comparable between WT and DDL transgenic plants. However, DDL ^12E1 transgenic plants (lines E11 and E15) exhibited largely compromised callose deposits upon flg22 treatment (Fig 7D). Taken together, the data indicate that DDL PARylation is essential for its function in plant defense.

Figure 7. DDL PARylation is required for plant innate immunity.

1. Phenotype of plants complemented with WT DDL or DDL ^12E1. Four‐week‐old plants of two independent transgenic lines of pDDL::DDL‐FLAG (W7 and W18) or pDDL::DDL ^12E1 –FLAG (E11 and E15) were photographed. Scale bar = 1 cm. The expression of transgene was determined by an α‐FLAG immunoblot (bottom). RBC, Rubisco.
2. Susceptibility of complementation lines to Pst DC3000 infections. Four‐week‐old transgenic plants were inoculated with Pst DC3000 (OD₆₀₀ = 5 × 10⁻⁴), and the bacterial number was determined at 0 and 3 dpi. Data are shown as mean ± SD (n = 3) from three independent repeats with one‐way ANOVA analysis and Tukey test (P < 0.05). Different letters, a, b or c, indicate significant differences.
3. PR1 and PR5 induction in transgenic lines. Leaves of 4‐week‐old plants were inoculated with Pst DC3000 (OD₆₀₀ = 0.01) and collected at 0 and 24 hpi for qRT–PCR analysis. Data are shown as mean ± SD (n = 3) from three independent repeats with one‐way ANOVA analysis and Tukey test (P < 0.05). Different letters, a, b or c, indicate significant differences.
4. flg22‐induced callose deposition in transgenic lines. Leaves of 4‐week‐old plants were inoculated with 500 nM flg22, and callose deposits were stained with aniline blue at 20 hpi. Number of callose deposits was quantified with ImageJ (mean ± SE, n = 10). Scale bar = 100 μm.

Data information: All experiments were repeated three times with similar results.

Interestingly, both WT DDL and the DDL ^12E1 mutants were able to complement the growth defects of ddl‐6 to the same level (Fig 7A). About 80% of WT DDL and DDL ^12E1 mutant transgenic plants in the ddl‐6 background displayed the growth phenotype as WT plants (Fig EV4A). As shown in Fig EV4B, the transgenic plants with WT DDL (lines W7 and W18) or DDL ^12E1 mutant (lines E11 and E15) were no longer dwarf compared to ddl‐6, and reached a similar size of WT Col‐0 plants. The individual leaves of transgenic plants were also similar with WT and bigger than ddl‐6 leaves (Fig EV4C). The W7, W18, E11, and E15 transgenic plants also restored the deficiency of flowering time in ddl‐6 to the WT level (Fig EV4B). The ddl‐6 mutant also displayed defects in flower and seed development 20. The ddl‐6 mutant exhibited abnormal petal number and aborted ovules. The W7, W18, E11, and E15 transgenic plants restored these deficiencies to the WT level (Fig EV4D and E). Apparently, DDL PARylation is not required for its function in plant development.

Figure EV4. Phenotype of pDDL::DDL ^12E1 ‐FLAG transgenic plants.

1. Summary of the transgenic lines of pDDL::DDL‐FLAG and pDDL::DDL ^12E1 ‐FLAG. Number and percentage of transgenic lines that are ddl‐6‐like or wild type‐like are listed.
2. The pDDL::DDL ^12E1 ‐FLAG gene complements the delayed flowering phenotype of ddl‐6. Plants were grown for 4 weeks under the 12‐h light/12‐h dark condition and 2 weeks under the long day condition (16‐h light/8‐h dark) and photographed at the 6‐week‐old stage. W7 and W18 are two representative lines of pDDL::DDL‐FLAG in the ddl‐6 background; E11 and E15 are two representative lines of pDDL::DDL ^12E1 ‐FLAG in the ddl‐6 background. Scale bar = 1 cm.
3. The pDDL::DDL ^12E1 ‐FLAG gene complements the leaf development defects of ddl‐6. The plants were grown under the 12‐h light/12‐h dark condition, and all leaves from a single 4‐week‐old plant are shown. Scale bar = 1 cm.
4. The pDDL::DDL ^12E1 ‐FLAG gene complements flower development defects of ddl‐6. Plants were grown 4 weeks under the 12‐h light/12‐h dark condition and 3 weeks under the long day condition and the flowers from individual 7‐week‐old plants are shown. The ddl‐6 mutant exhibits abnormal petal numbers. Scale bar = 1 mm.
5. The pDDL::DDL ^12E1 ‐FLAG gene complements the ovule development defects of ddl‐6. The plants were grown 4 weeks under the 12‐h light/12‐h dark condition and 4 weeks under the long day condition, and the ovules from individual 8‐week‐old plants are shown. The ddl‐6 mutant has aborted ovules with withered tissues as indicated by arrows. Scale bar = 1 mm.

The growth defects of ddl mutants in the Ws background are in part due to the reduced biogenesis of small RNA, especially miRNA 21. Interestingly, the abundance of several miRNAs, including miR160, miR172*, miRNA166, miR167, miR393, and miRNA398, did not differ significantly between WT and the ddl‐6 mutant by Northern blot (Fig EV5A). This is consistent with the observation that the growth defects of ddl‐6 in the Col‐0 background were weaker than ddl‐1 in the Ws background. Similarly, the abundance of these several miRNAs was also not changed in the parp1,2 and parg1 mutants (Fig EV5A). We also performed small RNA‐sequencing of Col‐0 WT, parp1,2, and parg1. The abundance of these several and some other miRNAs did not change dramatically in the parp1,2 mutant (Fig EV5B) or parg1 mutant (Fig EV5C) compared to WT plants. The data suggest that protein PARylation may not be involved in the biogenesis of these several miRNAs.

Figure EV5. Expression of miRNAs in ddl‐6, parg1, and parp1,2 mutants.

• A
  The expression of different miRNAs in ddl‐6, parg1, and parp1,2 was detected with Northern blots. U6 is a control for equal loading. The radioactive signals were detected with a phosphorimager and quantified with ImageQuant (V5.2). The amount of miRNAs in each genotype was normalized to U6 RNA, compared with that in Col‐0, and indicated below each image. Value of miRNAs in Col‐0 was set as 1. Ten‐day‐old seedlings grown on ½MS plates were used for RNA isolation.
• B, C
  The expression of different miRNAs in WT vs parp1,2 (B) and WT vs parg1 (C) by sRNA‐seq. miRNA abundances were normalized relative to the total mapped reads in each library (TPM). miRNAs validated by Northern blots in (A) are indicated with a red label.

Discussion

Although protein PARylation is an essential post‐translational modification in higher eukaryotes, little is known about its targets 26, 28. Our protein microarray analysis coupled with in vitro PARylation assays identified 54 proteins as PARP2 targets in Arabidopsis (Fig 1A, Appendix Table S1). Consistent with a wide range of PARP functions, those candidates are implicated in different cellular responses. For example, DNA polymerase lambda (AT1G10520) and UBC13B (AT1G16890) play a role in DNA damage responses; MBD11 and a plant TUDOR‐LIKE PROTEIN (AT1G06340) are involved in chromatin modification; TIE2 (AT2G20080) and AtOFP14 (AT1G79960) belong to the families of transcription repressors; and several candidates (AT5G47210, AT4G09040, AT3G62310, and AT2G04420) are predicted to have RNA binding or helicase/nuclease activities (Appendix Table S1). Importantly, our protein microarray studies may shed light on the potential novel functions of PARylation in plants. Notably, five candidates are involved in protein ubiquitination process with two as ubiquitin‐conjugating enzymes (AT2G32790 and AT1G16890) and two as RING domain‐containing E3 ubiquitin ligases (AT2G24480 and AT3G11110). It is plausible that PARylation regulates the ubiquitination enzymatic activities. In addition, some of targets identified from protein microarray may provide important insights into mechanistic understandings of PARP and PARG functions. Arabidopsis PARG1 is implicated in the regulation of circadian rhythms 29. It has been speculated that certain circadian oscillator components are subjected to PARylation for the control of clock gene transcription. One of the potential candidates might be a CCT (CONSTANS, CO‐like, TOC1) domain‐containing protein (AT5G53420) identified as a PARylation target in our study. TOC1, a CCT domain‐containing protein, is a key regulator in establishing Arabidopsis circadian rhythms 30.

DAWDLE contains a C‐terminal FHA domain which is a phosphor‐binding domain found in all kingdoms from bacteria, to yeasts, plants, and animals 31. Interestingly, many FHA domain proteins are involved in DNA damage responses and DNA repair, similar with the primary functions that protein PARylation involves. The ddl mutants in the Ws background have severe growth defects with pleiotropic phenotypes 20. This growth defects appear to be partially due to the mis‐regulation of small RNA biogenesis 21. In the ddl‐1 mutant, the expression of several miRNA and endogenous small interfering RNA (siRNA) was reduced. DDL binds to DICER‐LIKE 1 (DCL1) and likely facilitates DCL1 to access or recognize pri‐miRNA 21. Notably, we observed that ddl‐6 and ddl‐7, the T‐DNA knockout lines in the Col‐0 background, exhibit relatively weaker developmental defects than ddl‐1 in the Ws background (Appendix Fig S3). When grown under the 12‐h light/12‐h dark condition, although ddl‐6 grows slower than Col‐0, the difference between ddl‐6 and Col‐0 is not dramatic at the 4‐week‐old stage (Appendix Fig S3B). Since the DDL protein sequence from Col‐0 is identical with that from Ws 20, 21, it is possible that a suppressor in Col‐0 weakens ddl defects or an enhancer in Ws potentiates ddl defects. Interestingly, deletion analysis indicates that DDL‐N‐terminus (DDL‐N) interacts with PARP2 and is PARylated by PARP2 (Fig 6). Consistently, the PARylation sites are mainly located on DDL‐N (Fig 6). DDL‐N bears certain homology with proteins in RNA metabolism, and DDL possesses the RNA binding activity in vitro 21, 32. It remains unknown whether DDL PARylation mutants affect RNA binding activity. Apparently, DDL PARylation is not essential for its function in plant development (Fig EV4). This is consistent with that although the parp1,2 mutant almost completely blocks protein PARylation, the mutant is phenotypically normal (Fig 5) 18, 19. We also did not detect the defect of miRNA expression in the parp1,2, parg1, and ddl‐6 mutants (Fig EV5). The data suggest that PARylation‐mediated DDL function in immunity is uncoupled from its function in small RNA biogenesis.

We have shown that MAMP treatment stimulates PARP2 activity in vivo 18. Consistently, flg22 treatment induces DDL PARylation (Fig 5A), which might be the consequence of enhanced PARP2 activity. Apparently, PARylation is coupled with PARP2 and DDL interaction. Flg22 treatment strongly induces DDL–PARP2 interaction (Fig 4B), and PARylated DDL and PARP2 bind much more tightly than un‐PARylated proteins (Fig 4D). The DDL deletions or mutations that are comprised in PARylation also show the reduced interaction with PARP2. It is likely that PARP2 is activated upon MAMP perception, leading to DLL PARylation, which in turn stabilizes PARP2 and DDL interaction. It will be interesting to determine how MAMP signaling regulates PARP activity. It appears that the PARP transcripts are not significantly induced upon PRR activation. However, human HsPARP‐1 is regulated by multiple post‐translational modification processes, such as phosphorylation, ubiquitination, SUMOylation, and cleavage 22. It remains unknown whether a similar activation mechanism exists for plant PARPs.

A major function of protein PARylation is to regulate gene transcription through modulating chromatin, functioning as transcriptional co‐regulators or mediating DNA methylation 33. In the Arabidopsis parp1,2 mutant, MAMP‐induced both early responsive genes (FRK1 and WRKY30) and late responsive genes (PPDK, VSR7, AT3G08870 and PEN3) are compromised (Fig 3B) 18. However, the Arabidopsis ddl mutant is only compromised in some of MAMP‐induced late responsive genes (Fig 3B). It is common that some primary immune response genes are activated within minutes, whereas others are activated rather slowly due to different signaling pathways that activate RNA polymerase II (RNAPII) for transcription. The genes that do not require nucleosome remodeling for transcription are usually activated more rapidly than remodeling‐dependent genes 34. Smad nuclear interacting protein 1 (SNIP1), the human homolog of DDL, regulates gene transcription by binding to the transcriptional co‐activator CREB‐binding protein (CBP)/p300 proteins, a group of histone acetyltransferases, that acetylate histones to relax the chromatin structure for transcriptional activation 35. DDL is also predicted to interact with several Arabidopsis histone acetyltransferases, such as HAC1 and HAC12 36. Modulation of histone modification is an important mechanism for PARylation‐regulated gene expression in mammals. PARylation of human histone lysine demethylase KDM5B inhibits its demethylase activity, thereby maintaining histone H3 lysine 4 trimethyl (H3K4me3), a histone mark associated with active promoters. In addition, HsPARP‐1 promotes exclusion of histone H1 and opening of promoter chromatin, collectively leading to a permissive chromatin environment for the RNAPII machinery loading 37. It remains possible that DDL is involved in PARylation‐mediated chromatin remodeling to create a permissive chromatin environment that allows loading of the RNAPII machinery on the target gene promoters in plant immunity.

Materials and Methods

Plant and pathogen materials

Arabidopsis accessions of Col‐0 and Ws, mutants of ddl‐1, ddl‐2, ddl‐6 (SALK_025250), ddl‐7 (SAIL_1281_F08), parp1 (GABI‐Kat 692A05), parp2 (SALK_140400) and parp1,2, and transgenic plants of pDDL::DDL‐FLAG/ddl‐6 or pDDL::DDL ^12E1 ‐FLAG/ddl‐6 were germinated on soil (Metro Mix 366) and grown under the condition of 23°C, 45% humidity, 75 μE m⁻² s⁻¹ light, and 12‐h light/12‐h dark photoperiod. The ddl‐1, ddl‐2, ddl‐6, ddl‐7 mutants, and the SALK lines for candidate targets of PARP2 listed in Appendix Fig S2 were obtained from the Arabidopsis Biological Resource Center (ABRC), and parp1, parp2, and parp1,2 were reported previously 18. The primers for T‐DNA knockout identification are listed in the Appendix Table S5. Four‐week‐old plants were used for protoplast isolation and pathogen infection assays. Ten‐day‐old seedlings grown on ½ Murashige and Skoog (MS) plates with 0.8% agar were used for flg22‐mediated MAPK activation and RT–PCR assays.

Bacterial strains, Pseudomonas syringae pv. tomato (Pst) DC3000, Pst DC3000 hrcC, P. syringae pv. maculicola ES4326 (Psm), and P. syringae pv. phaseolicola NPS3121 (Psp), were cultured in King's B (KB) liquid medium containing 50 μg/ml rifampicin or streptomycin. After overnight culture at 28°C, bacteria were collected by centrifugation at 1,200 g for 2 min at room temperature, washed twice with H₂O, and re‐suspended in 10 mM MgCl₂. For disease assay, bacterial suspension was diluted to the indicated titer and infiltrated into leaves of 4‐week‐old plants by using a 1 ml needleless syringe. To monitor the bacterial growth, leaves from different plants were taken for bacterial counting at 0 and 3 days post‐inoculation (dpi). For each sample, bacteria from two leaf disks were released into 100 μl H₂O by grinding, diluted into gradient titers, and spread on tryptone soya agar (TSA) plates (1% Bacto tryptone, 1% sucrose, 0.1% glutamic acid, and 1.5% agar) containing appropriate antibiotics. After 2 days incubation at 28°C, bacteria colonies were counted to determine colony forming units (cfu). All data points represent triplicates.

PARylation assays and protein microarrays

The MBP‐PARP2 proteins were purified from E. coli strain BL21 with amylose resins (New England Biolabs, USA, E8022L) according to the standard procedure. Proteins on microarray chips were incubated with 100 μl of 100 ng/μl MBP‐PARP2 for 1 h at room temperature in 1× PAR reaction buffer [50 mM Tris–HCl, pH 8.0, 50 mM NaCl, 10 mM MgCl₂, and 1× activated DNA (Trevigen, USA, 4671‐096‐06)] with or without 0.2 mM NAD⁺ (Sigma, N0632). The chips were washed with 1× Tris‐buffered saline with 0.1% Tween (1× TBST) for three times and blocked by SuperBlock Buffer in TBS (Thermo Scientific, USA, #37535) for 1 h at room temperature. The α‐PAR antibody (Trevigen, USA, 4335‐MC‐100‐AC) and Cy3‐conjugated goat α‐rabbit secondary antibody (Jackson Immuno Research, USA, 111‐166‐003) were sequentially incubated with the chips. The chips were imaged by the GenePix4100A scanner (Molecular Devices, USA) using the 532‐nm channel. A TIFF file of scanned image was generated for each probing. The image files and a grid file with spot information (protein identity and spot coordination on chips) were then used to generate the signal intensity readout for each protein spot via the GenePix Pro 6.0 software (Molecular Devices, USA). The original protein microarray results are listed in Dataset EV1.

Plasmid construction

The open reading frames (ORFs) of PARP2 target genes, including DDL, HYALURONAN/mRNA‐BINDING PROTEIN (AT5G47210), METHYL‐CPG‐BINDING DOMAIN 11 (MBD11, AT3G15790), PLANT TUDOR‐LIKE PROTEIN (AT1G06340), AT5G03660, and JAM3 (AT4G16430), were amplified from Col‐0 cDNA with primers listed in the Appendix Table S2. The PCR products were digested with BamHI and StuI or SmaI and ligated into the pHBT vector with a FLAG tag at the C‐terminus. The sequences of ORFs inserted in all constructs were confirmed by DNA sequencing. The pHBT‐PARP2‐HA and pMAL‐PARP2 constructs were reported previously 18. The DDL gene was further subcloned into a modified pET28a vector with a SUMO tag via BamHI and StuI restriction enzyme sites. pGST‐PARP2 was constructed by subcloning the PARP2 fragment from pHBT‐PARP2‐FLAG into a modified pGST vector via BamHI and StuI restriction enzyme sites. For the BiFC and GFP localization assay, the fragments containing DDL or PARP2 were subcloned into pHBT‐35S::nYFP, pHBT‐35S::cYFP, or pHBT‐35S::GFP, respectively, via BamHI and StuI restriction enzyme sites. The pCAMBIA2300‐35S::PARP2‐FLAG was generated by subcloning the PARP2 fragment into pCAMBIA2300 vector via BamHI and StuI restriction enzyme sites. The DDL promoter was amplified by PCR from Col‐0 genomic DNA and digested with SalI and BamHI. The DDL‐FLAG::NOS terminator was released from pHBT‐DDL‐FLAG vector via BamHI and EcoRI sites. The above two fragments were mixed and cloned into pCAMBIA1300 vector via SalI and EcoRI sites to generate pDDL::DDL‐FLAG construct. The DDL mutants were generated by the site‐directed mutagenesis using the primers listed in the Appendix Tables S2 and S3.

In vivo and in vitro PARylation assays

For in vivo PARylation assay, 1 ml of Arabidopsis protoplasts (2–3 × 10⁵/ml) was transfected with 200 μg of plasmid DNA to express FLAG‐tagged proteins for about 12 h. The protoplasts were stimulated with 100 nM flg22 for 30 min, pelleted and re‐suspended in 100 μl of WI solution (0.5 M mannitol, 20 mM KCl, 4 mM MES pH 5.7), and fed with 1 μCi of ³²P‐NAD⁺ for 1 h. The proteins were extracted by IP buffer (20 mM HEPES, pH 7.5, 100 mM NaCl, 5 mM EDTA, 1% Triton X‐100, 1× protease inhibitor, 1 mM DTT, 2 mM NaF, and 2 mM Na₃VO₄), and the FLAG‐tagged proteins were immunoprecipitated with α‐FLAG affinity beads (Sigma, USA, A2220) by incubating at 4°C for 3 h with gentle shaking. The proteins were separated in 10% SDS–PAGE for autoradiography to detect in vivo PARylated proteins. The inputs of proteins were detected by an immunoblot with the α‐FLAG antibody (Sigma, USA, A8592).

For in vitro PARylation assay, GST‐PARP2 and GST‐PARG1 recombinant proteins were purified from the E. coli BL21 strain with Pierce glutathione agarose (Thermo Scientific, USA, 16101) according to the standard manufacturer manual. FLAG‐tagged DDL, DDL‐N, and DDL‐C proteins were transiently expressed in Arabidopsis protoplasts and immunoprecipitated with α‐FLAG affinity beads, and incubated with 50 μl of PARylation reaction mix (100 ng of GST‐PARP2, 1× PAR reaction buffer, 20 μM biotin‐NAD⁺, 1× active DNA) at room temperature for 2 h with gentle shaking. Either 1 mM 3‐aminobenzamide (3‐AB, Sigma, A0788) (to inhibit PARylation) or 0.5 μg of GST‐PARG1 (to remove PAR) was included in the reaction. PARylated proteins were detected by an immunoblot using streptavidin‐HRP (Pierce, USA, 21126).

Identification of DDL PARylation sites by mass spectrometry

For in vitro PARylation of DDL and mass spectrometry analysis, ~100 ng of GST‐PARP2 was mixed with 1 μg of HIS‐SUMO‐DDL‐HA in the PARylation reaction buffer (50 mM Tris–HCl, pH 8.0, 50 mM NaCl, 10 mM MgCl₂, and 1× activated DNA with or without 0.2 mM NAD⁺) at room temperature for 6 h. Then, 1 M of NH₂OH was added in the reactions to remove PAR from the PARylated residues overnight. This treatment left on the PARylated residues a hydroxamic acid derivative, which has a signature mass increase of +15.0109 Da and could be distinguished by mass spectrometry. The proteins were separated in 10% SDS–PAGE by electrophoresis and stained with GelCode blue. The gel slices containing unmodified and modified HIS‐SUMO‐DDL‐HA were subject to mass spectrometry analysis as previously reported 38. Briefly, gel bands were in‐gel digested with trypsin overnight, and peptides were enriched for liquid chromatography–MS/MS analysis with a Q‐Exactive Plus Orbitrap mass spectrometer (Thermo Scientific). High energy collision dissociation (HCD) was used, and the normalized collision energy was 28. In each cycle, five most abundant ions were selected. The instrument was run in data‐dependent mode with a full MS (400–2,000 m/z) resolution of 70,000, and MS2 (200–2,000 m/z) was triggered at a minimal signal of 1,6e5. Dynamic exclusion for 10 s was used to prevent repeated analysis of the same peptides, and a lock mass of m/z 445.12003 (polysiloxane ion) was used for real‐time internal calibration. The MS system was interfaced with an automated Easy‐nLC 1000 system (Thermo Scientific). The peptides were loaded onto an Acclaim PepMap 100 pre‐column (20 mm × 75 μm; 3 μm‐C18) and separated on a PepMap RSLC analytical column (500 mm × 75 μm; 2 μm‐C18) at a flow rate at 300 nl/min during a linear gradient from the solvent A (0.1% formic acid (v/v)) to 25% solvent B (0.1% formic acid and 99.9% acetonitrile) in 45 min, ramp to 98% solvent B in 5 min, and then hold at 98% solvent B for additional 20 min.

The MS/MS spectra were analyzed using Mascot software (Matrix Science; version 2.4) with the following parameters: peptide tolerance at 10 ppm, tandem MS tolerance at ± 0.01 Da, peptide charge of 2+, 3+, or 4+, trypsin as the enzyme, iodoacetamide at cysteine as fixed modifications, and oxidation at methionine and PARylation at glutamic acid or aspartic acid residues with an addition of 15.0109 Da. The identified PARylated peptides were manually inspected to ensure confidence in PARylation site assignment. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository (http://www.ebi.ac.uk/pride/archive/) with the project PXD004267.

Co‐IP and pull‐down assays

Arabidopsis protoplasts (1 ml at 2 × 10⁵/ml) were transfected with different plasmids and lysed in Co‐IP buffer (20 mM HEPES, pH 7.5, 100 mM NaCl, 10% glycerol, 0.5% Triton X‐100, 1× protease inhibitor, 2 mM NaF, 2 mM Na₃VO₄, 1 mM DTT) by vortexing. The lysate was cleared by centrifugation at the speed of 14,500 g for 10 min at 4°C and incubated with α‐FLAG affinity beads at 4°C for 3 h with gentle shaking. The immunoprecipitated proteins were analyzed by an immunoblot with either α‐FLAG or α‐HA (Roche, USA, 3F10) antibodies. For Co‐IP in N. benthamiana, leaves of 4‐ to 5‐week‐old plants were hand‐inoculated with the mixture of Agrobacterium GV3010 strain (OD₆₀₀ = 0.3 for each plasmid) carrying different plasmids and the samples were collected at 2 days post‐inoculation.

For the pull‐down assay, recombinant GST‐PARP2 or GST proteins immobilized on glutathione‐Sepharose beads were incubated with purified HIS‐SUMO‐DDL‐HA proteins in TEN100 (20 mM Tris–HCl, pH 7.4, 0.1 mM EDTA, and 100 mM NaCl) by gentle shaking at 4°C for 1 h. To test the effect of PARylation on protein interaction, HIS‐SUMO‐DDL‐HA proteins and immobilized GST‐PARP2 protein beads were incubated in 50 μl of 1× PAR reaction buffer for 3 h at room temperature with gentle shaking prior to the pull‐down assay. The proteins were analyzed by immunoblotting with the α‐GST (Santa Cruz, USA, SC‐93909), α‐PAR, or α‐HA antibody.

BiFC assay

Different combinations of BiFC constructs were transfected into 200 μl of Arabidopsis protoplasts. Fluorescence images were taken using a Zeiss LSM 780 NLO multiphoton microscope. The YFP and RFP signals were excited with lasers at 516 and 561 nm, respectively, and analyzed with the ZEN lite software (Zeiss).

RNA isolation, RT–PCR, qRT–PCR, miRNA Northern blot and small RNA library construction, sequencing, and data analysis

Ten‐day‐old seedlings grown on ½ MS plates were transferred to 2 ml of H₂O in a 6‐well plate 10 days before 100 nM flg22 treatment. Fully expanded leaves from 4‐week‐old plants were hand‐inoculated with Pst DC3000 or hrcC for RNA extraction. RNA was extracted with TRIzol reagent (Life Technologies, USA, 15596018), and reverse transcription was performed with M‐MuLV reverse transcriptase (New England Biolabs, USA, M0253L) according to manufacturer's manual. Quantitative RT–PCR was performed with iTaq Universal SYBR Green Supermix (Bio‐Rad, USA, 172‐5124) in 7900HT Fast Real‐Time PCR System (Applied Biosystems, USA, 4329001). The primers for qRT–PCR and RT–PCR are listed in the Appendix Tables S4 and S6. The expression of each gene was normalized to the expression of UBQ10 (for qRT–PCR) or UBQ1 (for RT–PCR).

miRNA Northern blot was performed as described 20, 21. Basically, RNAs extracted from 10‐day‐old seedlings were resolved on a 15% PAGE gel containing 8% urea and transferred to a nitrocellulose membrane. Then, miRNAs and U6 RNA were probed with 5′‐end‐labeled ³²P antisense DNA oligonucleotides or LNA oligonucleotides. The radioactive signals were detected with a phosphorimager and quantified by ImageQuant version 5.2.

Small RNA library construction and sequencing were performed as described 39. Briefly, small RNA fragments (~15–30 nucleotides) were selected and recovered from 10 μg of total RNAs by 15% urea–polyacrylamide gel electrophoresis. Libraries were made using the NEBNext Small RNA Library Prep kit (NEB, USA) according to manufacturer's instructions. Briefly, 3′ and 5′ adapters were ligated sequentially to the small RNAs. Ligated small RNAs were converted to cDNA by reverse transcription followed by PCR amplification for 15 cycles. The barcoded libraries were sequenced in one lane on the Illumina HiSeq 2000. Raw reads were processed by first trimming 3′ adapter sequence using custom Perl script. Reads without 3′ adapter sequence or < 18 nucleotides after trimming were discarded from further analysis. Trimmed reads were aligned against a custom database containing rRNA, tRNA, snoRNA, and snRNAs using Bowtie 0.12.8 to filter out rRNA reads. Non‐aligned reads were subsequently aligned to the Arabidopsis genome (TAIR10) using Bowtie 0.12.8. Genome‐aligned reads were used for all downstream analysis. The small RNA‐sequencing data could be retrieved from the Gene Expression Omnibus (GEO) data repository at the National Center for Biotechnology Information (NCBI) with the accession number GSE81366.

Callose deposition assay

To induce callose deposits, 0.5 μM flg22 or Pst DC3000 hrcC (OD₆₀₀ = 0.2) was hand‐infiltrated into fully expanded leaves of 4‐week‐old plants. The leaves were collected at 20 hpi, cleared with 95% ethanol, washed 1 time with H₂O, and stained with 0.01% aniline blue (150 mM KH₂PO₄, pH 9.5) for 1 h. The callose deposits were visualized with UV light under a fluorescence microscope, and the number of callose deposits was counted by using ImageJ software.

MAPK activation assay

To determine MAPK activation in response to flg22, 10‐day‐old seedlings grown on ½ MS plates were transferred into a six‐well tissue culture plate containing 2 ml of H₂O for 1 day and then treated with 100 nM flg22 for 0, 5, 15, or 45 min. The phosphorylated MAPKs were detected by the α‐pERK antibody (Cell Signaling, USA, 9101) with an immunoblot.

ROS assay

For ROS assay, at least 20 leaf disks (0.25 cm² each) were cut from 4‐week‐old plants of each genotype (1 disk/leaf) and incubated overnight in 100 μl of H₂O in a 96‐well tissue culture plate. 100 μl of ROS reaction solution (50 μM luminol, 10 μg/ml horseradish peroxidase, with or without 100 nM flg22) was added to each leaf disk after removing H₂O. Luminescence was measured immediately with a luminometer (Perkin Elmer, 2030 Multilabel Reader, Victor X3) over a 40‐min time course with a 2‐min interval. The values of ROS production (RLU, relative light units) represent mean ± SE from 20 leaf disks.

Statistics

For statistical analysis, one‐way ANOVA and Tukey test were performed. The P‐value < 0.05 was considered as significant. Data were presented as mean ± SD or mean ± SE as indicated in the figures. The number of samples and experimental repeats was also indicated in the figure.

Author contributions

BF, SM, SC, NZ, SZ, and YY performed experiments; BF, SM, SC, BY, BL, XC, SPD‐K, LS, and PH analyzed data; and BF, LS, and PH conceived experiments and wrote the manuscript.

Conflict of interest

The authors declare that they have no conflict of interest.

Supporting information

Appendix

Expanded View Figures PDF

Dataset EV1

Review Process File

EMBO Reports (2016) 17: 1799–1813

See also: FLH Menke (December 2016)