An oomycete effector protein induces shade avoidance in Arabidopsis and attenuates salicylate signaling by binding to host proteins of the RADICAL-INDUCED CELL DEATH1 family

The oomycete pathogen Hyaloperonospora arabidopsidis (Hpa) causes downy mildew disease on Arabidopsis. During infection, Hpa like other biotrophic pathogens, suppresses activation of plant innate immunity by translocating effector proteins into host cells. Some of these effectors localize to the host cell nucleus where they may manipulate transcriptional reprogramming of plant defense genes. Here we report that the nuclear-localized Hpa effector HaRxL106, when expressed in Arabidopsis, induces shade avoidance and attenuates the transcriptional response to the defense signaling molecule salicylic acid. HaRxL106 interacts with RADICAL-INDUCED CELL DEATH1 (RCD1) and loss of RCD1 function renders Arabidopsis resilient against HaRxL106-mediated suppression of immunity. To further characterize the molecular functions of RCD1 we solved a crystal structure of RCD1’s Poly-(ADP-ribose)-Polymerase (PARP) domain and, based on non-conservation of amino acids constituting the active site of canonical PARPs, conclude that RCD1 has no PARP activity. We report that RCD1-type proteins are phosphorylated and identified histone-modifying Mut9-like kinases (MLKs) as RCD1-interacting proteins. A mlk1,3,4 triple mutant exhibits stronger SA-induced defense marker gene expression compared to wild-type plants. Our data suggest that HaRxL106 suppresses Arabidopsis innate immunity by manipulating the function(s) of RCD1 in the host cell nucleus and point towards a role of RCD1 as a transcriptional co-regulator that integrates signals from light and pathogen sensors.


Introduction
Plants rely on their innate immune system to distinguish beneficial microbes from harmful pathogens or commensal bacteria. While plant innate immunity fends off the majority of attempted infections, specialized pathogens can subvert host defenses with effector proteins that are translocated into host cells. Many pathogen effectors interfere with cellular processes that are essential for innate immunity such as formation of cell wall appositions, secretion of antimicrobial compounds, production of reactive oxygen species (ROS) or transcriptional activation of defense genes (DebRoy et al., 2004;Nomura et al., 2006;Bozkurt et al., 2011;Anderson et al., 2012;Gangadharan et al., 2013;Asai et al., 2014). Bacterial pathogens evolved specialized secretion systems to deliver effectors into host cells (Deng et al., 2017). Likewise, the fungal rice blast pathogen Magnaporthe oryzae employs a specialized secretion pathway to deliver host cell-targeted effectors into a host-derived membrane-rich compartment named the biotrophic interfacial complex (Khang et al., 2010;Giraldo et al., 2013). How other filamentous plant pathogens, such as oomycetes, translocate effectors into plant cells remains poorly understood (Petre and Kamoun, 2014).
Plants respond to infection by biotrophic pathogens with elevated biosynthesis of the defense hormone salicylic acid (SA). Elevated SA levels induce fluctuations in the cellular redox status culminating in activation of the NONEXPRESSOR OF PATHOGENESIS-RELATED GENE 1 (NPR1) protein (Mou et al., 2003). Export of SA from chloroplasts to the cytoplasm leads to thioredoxin-catalyzed reduction of disulfidelinked oligomeric complexes of the NPR1 protein (Tada et al., 2008). Monomeric NPR1 translocates to the nucleus where it functions as a transcriptional co-activator and is 4 indispensable for SA-responsiveness for most SA-induced genes (Wang et al., 2006). Some biotrophic plant pathogen effectors actively suppress SA accumulation and/or SA signaling. The maize smut fungus Ustilago maydis produces a host cell-targeted chorismate dismutase that may suppress SA-mediated immunity by diverting the SAprecursor chorismate into the phenylpropanoid pathway (Djamei et al., 2011). The oomycete pathogen Hyaloperonospora arabidopsidis (Hpa) suppresses transcriptional upregulation of the SA marker gene PATHOGENESIS-RELATED GENE 1 (PR1) in infected cells of its host Arabidopsis thaliana (Caillaud et al., 2013). At least two Hpa effector proteins interfere with SA signaling when expressed as transgenes in Arabidopsis. Effector HaRxL44 appears to attenuate SA signal transduction by targeting the MEDIATOR subunit Med19 for proteasomal degradation (Caillaud et al., 2013), while effector HaRxL62 interferes with SA signaling by an unknown mechanism (Asai et al., 2014). Whether pathogen effectors manipulate proteins of the NPR class directly remains unknown. However, pathogens interfere with other processes that indirectly promote SA signaling. The Xanthomonas campestris effector protein XopJ is a Cys protease that cleaves the RPT6 subunit of the 19S regulatory particle of the proteasome, thereby interfering with targeted degradation of poly-ubiquitinated proteasome substrates. Via this mechanism, XopJ also attenuates proteasomal turnover of NPR1, which is required for full transcriptional activation of NPR1 target genes (Üstün and Börnke, 2015;Spoel et al., 2009).
Shade avoidance in Arabidopsis, activated by a low red/far-red light ratio, attenuates the transcriptional response to SA (Genoud et al., 2002;de Wit et al., 2013;Gangappa et al., 2016). Notably, simulated shade conditions also suppress transcript changes 5 induced by exogenous application of Methyl-Jasmonate (MeJA) and attenuate plant defense towards necrotrophic pathogens and herbivores (Izaguirre et al., 2006;Cerrudo et al., 2012;de Wit et al., 2013). This is remarkable given that SA-and JA-responsive gene networks are antagonistically regulated in response to infection by biotrophic and necrotrophic pathogens Caarls et al., 2015). As a consequence, plants become more susceptible to infection by both biotrophic and necrotrophic pathogens when shade avoidance is activated either environmentally, or genetically by mutations in PHYTOCHROME B (PHYB) (Genoud et al., 2002;Izaguirre et al., 2006;de Wit et al., 2013).
Arabidopsis RADICAL-INDUCED CELL DEATH1 (RCD1) has also been proposed to act as a positive regulator of SA signaling. Loss of RCD1 function does not alter SA levels but transcript levels of many NPR1 target genes are lower in rcd1 mutants when compared to wild-type plants (Ahlfors et al., 2004;Brosché et al., 2014). RCD1 is the founding member of a plant-specific protein family characterized by a central domain with sequence similarity to the catalytic domain of Poly-(ADP-ribose)-polymerases (PARPs) (Lamb et al., 2012). In contrast to canonical PARPs that covalently modify target proteins by ADP-ribosylation, Arabidopsis RCD1 does not show PARP activity in vitro when expressed as a GST fusion (Jaspers et al., 2010a). In addition to the central PARP domain, RCD1 and its paralog SIMILAR TO RCD ONE 1 (SRO1) have an Nterminal WWE domain and a C-terminal RST domain. Proteins with this domain architecture are conserved in all land plants and are referred to as type I proteins. In contrast, the presence of additional family members that lack the N-terminal WWE domain (type II) appears to be specific to the Brassicaceae (Jaspers et al., 2010a). Type 6 I proteins from Arabidopsis and rice localize to the plant cell nucleus and bind to many sequence-unrelated transcription factors via their RST domains (Katiyar-Agarwal et al., 2006;Jaspers et al., 2009;You et al., 2014). Therefore, RCD1 might influence SA signal transduction by interacting with transcription factors that mediate SA-induced transcriptome changes. Notably, an RCD1 homologue from wheat (TaSRO1) shows PARP activity when expressed in E. coli, suggesting that some RCD1-type proteins may be enzymatically active (Liu et al., 2014).
RCD1 interacts with the Hpa effector HaRxL106 in a yeast-two-hybrid (Y2H) assay and this effector renders Arabidopsis more susceptible to biotrophic pathogens when expressed as a transgene (Fabro et al., 2011;Mukhtar et al., 2011). In plant cells, HaRxL106 binds to nuclear transport receptors of the importin- class with affinity in the low micro-molar range and is actively transported into the nucleus, indicative of a virulence-promoting activity of the effector in the host cell nucleus (Wirthmueller et al., 2015). Here we report that HaRxL106, when expressed as a transgene, affects both SA signaling and light-dependent developmental processes. We identify RCD1 as a likely virulence target of HaRxL106 and report that RCD1 interacts with histone-modifying kinases that impinge on SA signaling.

HaRxL106-expressing Arabidopsis plants exhibit attenuated light and defense signaling
To characterize HaRxL106-interacting proteins from Arabidopsis we generated transgenic lines expressing HaRxL106 with an N-terminal YFP or 3xHA-StrepII (HS) 7 epitope tag under control of the 35S promoter. As previously reported for transgenic plants expressing untagged HaRxL106 (Fabro et al., 2011), these lines are hypersusceptible to infection by the compatible Hpa isolate Noco2 ( Fig. 1A; one-way ANOVA, F8,171=3.14, p=5.73x10 -16 ; Tukey-Kramer post-hoc test, p<0.05). While analyzing lines expressing HaRxL106 we noticed that they show signs of constitutive shade avoidance, specifically longer hypocotyls and elongated petioles under white light conditions (Fig.   1B). Differences in hypocotyl length between wild type plants and transgenic lines were more pronounced when we grew seedlings under a lower fluence rate of white light (12 mol m -2 s -1 ) (Fig. 1,C and D;F5,174=215.48, ; Tukey-Kramer post-hoc test, p<0.05). Under these conditions HaRxL106-expressing seedlings were indistinguishable from the phyB-9 mutant that shows constitutive shade avoidance (Reed et al., 1993). Lines expressing control constructs YFP and HS did not differ from wild type plants in hypocotyl length (Fig. 1,C and D). In contrast, differences in hypocotyl length between HaRxL106-expressing transgenic lines and wild type plants were much smaller when we grew seedlings in darkness ( Fig. 1D; one-way ANOVA, F5,174=3.14, p=9.67x10 -3 ; Tukey-Kramer post-hoc test, p<0.05). This suggests that in addition to defense signaling, either light perception or light signal transduction is attenuated in lines expressing HaRxL106.

Effector HaRxL106 suppresses SA signal transduction but not SA levels
Plants that undergo shade avoidance, either induced by supplementary FR light or by mutations in PHYB, show an attenuated transcriptional response to SA and MeJA (Genoud et al., 2002;de Wit et al., 2013). As suppression of SA signal transduction would be a conceivable virulence mechanism for an effector of a biotrophic pathogen, 8 we tested SA-induced up-regulation of the SA marker gene PR1 by qPCR in Col-0 plants and two transgenic lines expressing YFP:HaRxL106 and HS:HaRxL106, respectively ( Fig. 2A). As expected, SA treatment induced PR1 mRNA levels in Col-0 plants but not in the npr1-1 mutant (Cao et al., 1994) (one-way ANOVA, F7,16=78.60, p=3.39x10 -11 ; Tukey-Kramer post-hoc test, p<0.05). In contrast, PR1 expression levels in SA-treated HaRxL106 transgenic lines were comparable to those in mock treated Col-0 plants suggesting that HaRxL106 affects either endogenous SA levels or SA signal transduction ( Fig. 2A). To distinguish between these two possibilities we quantified levels of SA in wild type plants, the sid2-1 mutant that is impaired in pathogen-triggered SA biosynthesis (Wildermuth et al., 2001), and HaRxL106 transgenics. There was a trend for lower SA levels in the sid2-1 mutant compared to Col-0 in plants infiltrated with Pseudomonas syringae pv. tomato (Pst) DC3000 (1x10 8 cfu/ml) 24h earlier (Fig. 2B). In contrast, SA levels in HaRxL106-expressing lines were comparable to  or intermediate between Col-0 and sid2-1 (experiment C). When we analyzed the data using a linear mixed effects model we found no statistical differences between the genotypes (p<0.05). Despite some variability between SA measurements in the three independent biological experiments, these results suggest that HaRxL106 does not substantially alter SA levels but nevertheless strongly attenuates SA-induced transcriptional regulation of the SA marker gene PR1.

Effector HaRxL106 attenuates NPR1-dependent defense activation
HaRxL106 has a nuclear localization sequence (NLS) and is actively transported into nuclei of plant cells via karyopherins of the importin- group (Wirthmueller et al., 2015).
Given that NPR1 is an important nuclear signal integrator of the SA pathway, we tested whether HaRxL106 affects NPR1 localization or NPR1 protein levels. The npr1-1 mutant has been complemented by a 35SPro:NPR1:GFP transgene under long day conditions (Kinkema et al., 2000). HaRxL106 suppresses constitutive PR1 gene expression induced by the 35SPro:NPR1:GFP transgene, the effector must either act on a step downstream of nuclear NPR1 signaling or disrupt the nuclear transactivator function of NPR1 itself.

HaRxL106 over-expressing lines show a partial transcription profile overlap with the radical-induced cell death1-1 mutant
Several Arabidopsis proteins that interact with HaRxL106 in the yeast-two-hybrid (Y2H) system have been reported (Mukhtar et al., 2011). HaRxL106 interactors identified included several importin--type karyopherins, the tri-helix transcription factor ARABIDOPSIS 6b-INTERACTING PROTEIN 1-LIKE1 (ASIL1), transcription factor TEOSINTE BRANCHED CYCLOIDEA AND PCF 14 (TCP14) and RCD1. We reasoned that if one or several of these proteins constitute virulence targets of HaRxL106, the transcriptome profile of the corresponding mutants might show similarities to the transcriptome profile of HaRxL106 over-expressors. The transcriptome of tcp14 knockout mutants has recently been analyzed and revealed a set of 18 genes that are differentially expressed in two independent tcp14 T-DNA lines (Yang et al., 2017). We performed transcriptome profiling using the EXPRSS RNAseq pipeline Sohn et al., 2014) to compare the transcriptome profiles of Col-0, the HS:HaRxL106 over-expressor line #2, rcd1-1, asil1-1 and the importin- mutant mos6-1 (Overmyer et al., 2000;Gao et al., 2009;Palma et al., 2005). To compare the transcriptional response of all lines to a biotrophic pathogen, we performed transcriptome profiling of non-treated plants as well as plants infiltrated with 5 x 10 5 cfu/ml of Pst DC3000. Plants infiltrated with 10 mM MgCl2 served as mock control and we harvested all samples 24 h after infiltration. We applied a false discovery rate (FDR) of less than 0.001 and a two-fold change in expression as criteria to identify differentially expressed genes (DEGs) (Supplementary Table S1). As shown in mock-infiltrated and Pst-infiltrated plants, respectively. In rcd1-1, between 181 and 429 genes differed in expression compared to Col-0. In contrast, analysis of the asil1-1 and mos6-1 mutants, as well as the published tcp14 dataset, revealed a comparatively small number of DEGs (<48).
We noticed a partial overlap in DEGs between HaRxL106 and rcd1-1, particularly in the repressed genes ( Fig. 4 A, B, see Fig. S1 for induced genes). In non-treated plants, 63% of genes repressed in rcd1-1 also exhibited lower transcript abundance in the HaRxL106 transgenic line. In mock-treated and Pst-infiltrated tissue this overlap was substantially lower, with 31% and 37% of shared repressed genes, respectively (Fig. 4B). Vice versa, out of all genes repressed in the HaRxL106 transgenic line only 25%, 7% and 5% were also expressed at lower levels in rcd1-1 in non-treated, mock-treated and Pst-infiltrated tissue, respectively (Fig. 4B). Among the genes repressed in non-treated rcd1-1 and the HaRxL106-expressing line there was an over-representation of defense-related transcripts as determined by analysis of Gene Ontology (GO) terms (Vandepoele et al., 2009) and cis-regulatory promoter elements (O'Connor et al., 2005) (Table S1). Lower transcript abundance of SA defense genes in rcd1-1 is consistent with a previously reported transcriptome profiling of rcd1 mutants without pathogen challenge (Brosché et al., 2014). Figure 4C shows expression profiles of two selected SA marker genes, PR1 and WRKY38. Transcript levels of both genes were induced by Pst infection and to lesser extent by MgCl2 infiltration in wild type plants. In contrast, in the HaRxL106expressing line and in rcd1-1, transcriptional up-regulation of both SA marker genes by Pst infection or MgCl2 infiltration was attenuated ( Fig. 4C; PR1 one-way ANOVA, F8,18=5.94, p=9.5 x 10 -5 ; WRKY38 one-way ANOVA, F8,18=7.64, p=1.81 x 10 -4 ; Tukey-Kramer post-hoc tests, p<0.05). Therefore, both loss of RCD1 function and ectopic expression of HaRxL106, lead to repression of SA defense genes in naïve and challenged plants. The partial overlap between genes that were repressed in the HS:HaRxL106 line and rcd1-1 prompted us to characterize the interaction between the two proteins in more detail.

HaRxL106 interacts with RCD1 and SRO1 proteins and RCD1 quantitatively contributes to SA signal transduction
To test for interaction between HaRxL106 and RCD1 in Arabidopsis, we made use of a transgenic line in which the rcd1-3 mutation is complemented by expression of an RCD1Pro:RCD1:HA construct (Jaspers et al., 2009). We transformed the RCD1:HA line with YFP:HaRxL106 and selected double transgenic lines. When we immunoprecipitated YFP:HaRxL106 from these lines, RCD1:HA co-purified with HaRxL106 whilst a crossreacting band detected by the -HA antibody did not (Fig. 5A). Therefore YFP:HaRxL106 interacts with functional epitope-tagged RCD1:HA protein in

Arabidopsis.
RCD1 and its paralog SRO1 show unequal genetic redundancy with respect to plant development and responses to abiotic stress with RCD1 making a stronger contribution (Jaspers et al., 2009;Teotia and Lamb, 2009). RCD1 and SRO1 share the same domain structure comprising an N-terminal WWE domain, a central PARP domain and a Cterminal RST domain (Jaspers et al., 2010b). To test whether HaRxL106 also interacts with SRO1 we used Agrobacterium tumefaciens-mediated transient protein expression in Nicotiana benthamiana. We co-expressed HS:HaRxL106 and C-terminally GFP-13 tagged versions of RCD1 and SRO1. Although all constructs were expressed from the strong 35S promoter we did not detect RCD1:GFP by Western blot in total protein extracts (Fig. 5B). In contrast, SRO1:GFP and HS:HaRxL106 were detectable with the respective antibodies. As shown in Fig. 5B HS:HaRxL106 co-immunoprecipitated with both RCD1:GFP and SRO1:GFP but not with free YFP that we used as control. This suggests that HaRxL106 interacts with both RCD1 and SRO1 in plant cells.
The partial redundancy between RCD1 and SRO1 prompted us to test if SRO1 also contributes to transcriptional regulation of NPR1 target genes. We compared SAinduced transcriptional upregulation of PR1 in wild type plants, the YFP:HaRxL106 line, rcd1-1 and sro1-1. PR1 levels 8h after SA spraying were approximately 3-fold lower in rcd1-1 and sro1-1 but the difference was only statistically significant in the case of rcd1-1 (one-way ANOVA, F9,36=7.7, p=3.32 x 10 -6 ; Tukey-Kramer post-hoc test, p<0.05). The rcd1-1 and sro1-1 mutations had a weaker effect on PR1 expression than the YFP:HaRxL106 transgene (Fig. 5C). Nevertheless, our data suggest that RCD1 is required for full SA-induced PR1 expression.

The C-terminal 58 amino acids of HaRxL106 are required for RCD1-binding and attenuation of light and defense signaling
To test whether HaRxL106 binding to RCD1 correlates with its defense-suppressing activities we generated a mutant variant of HaRxL106 that does not bind to RCD1. The protein sequence C-terminal to the HaRxL106's signal peptide and RxLR motif can be divided into two regions based on predicted protein secondary structure: a larger domain with a predicted -helical WY-fold (Win et al., 2012) and a 58 amino acid C-terminal region that mediates binding to importin- (Wirthmueller et al., 2015). To narrow down 14 the RCD1-binding site in HaRxL106, we used the Y2H system to compare RCD1 binding to full-length HaRxL106, an HaRxL106 variant lacking the 56 C-terminal amino acids (HaRxL106C) and to the C-terminal 58 amino acids alone (HaRxL106-Cterm58).
Unexpectedly, we did not detect interaction of the two full-length proteins in yeast suggesting that the sensitivity of the Y2H reporters under our conditions was lower than in Mukhtar et al., 2011. However, we found the HaRxL106 C-terminus interacts with RCD1 ( Fig. 6A). Next, we tested which domain(s) of RCD1 are required for binding to the HaRxL106 C-terminus. The WWE, PARP and RST domains of RCD1 are separated by regions of 50-100 amino acids predicted to be disordered (Kragelund et al., 2012). As shown in Fig. 6A the HaRxL106 C-terminus did not interact with the isolated WWE, PARP or RST domains, nor did we detect binding to a construct encompassing PARP and RST domain. However, a fragment spanning the WWE and PARP domains activated the HIS3 reporter when co-expressed with the HaRxL106 C-terminus. Notably, this region of RCD1 also showed a weaker interaction with full-length HaRxL106 protein ( Fig. 6A). This suggests that HaRxL106 specifically binds to the RCD1 WWE-PARP domains via its C-terminal 58 amino acid peptide, and that RCD1's RST domain might negatively affect binding of the effector to the N-terminal domains of RCD1. We next tested whether the HaRxL106 C-terminal 58 amino acids are necessary for altered light and SA signaling in Arabidopsis. To this end we transformed an RFP:NLS:HaRxL106C construct lacking the C-terminus of the effector into Col-0. Because the HaRxL106C construct also lacks the effector's NLS this fusion protein carries a SV40 NLS to ensure efficient nuclear import (Wirthmueller et al., 2015). As controls we also generated transgenic RFP:HaRxL106 lines and lines expressing a fusion of RFP and the 58 Cterminal amino acids of HaRxL106 (RFP:HaRxL106-Cterm58). All constructs were under 15 control of the 35S promoter and we confirmed expression of the RFP fusion proteins by Western blot (Fig. 6B).
As expected, transgenic lines expressing RFP-tagged HaRxL106 developed signs of constitutive shade avoidance (Fig. 6C). In contrast, RFP:NLS:HaRxL106C lines were indistinguishable from wild type plants. Notably, RFP-HaRxL106-Cterm58 lines resembled transgenics expressing full-length HaRxL106 suggesting that the C-terminal 58 amino acids of the effector are required and sufficient for attenuation of light signaling ( Fig. 6C). We then tested resistance to Hpa Noco2 in these lines. While expression of RFP:HaRxL106 led to enhanced disease susceptibility in two independent transgenic lines (one-way ANOVA, F2,342=102.26, p=1.55 x 10 -35 ; Tukey-Kramer post-hoc test, p<0.05), the RFP:NLS:HaRxL106C fusion protein failed to suppress defense. In contrast, transgenic lines expressing RFP:HaRxL106-Cterm58 were more susceptible to Hpa Noco2 than Col-0 (one-way ANOVA, F4,590=578.15, p=2.28 x 10 -25 ; Tukey-Kramer post-hoc test, p<0.05), but less so than lines expressing the full-length effector.
Therefore, the C-terminal 58 amino acids of HaRxL106 are required to alter light and defense signaling. Similar to RFP:HaRxL106 lines, transgenics expressing the HaRxL106 C-terminus responded with lower PR1 transcript levels than wild type plants to SA spraying (Fig. 6E). In contrast, PR1 transcript in lines expressing RFP:NLS:HaRxL106C responded like wild type plants to SA spraying (one-way ANOVA, F15,32=27.54, p=4.42 x 10 -14 ; Tukey-Kramer post-hoc test, p<0.05).

RCD1 is dispensable for basal resistance to Hpa but required for HaRxL106mediated suppression of defense
To test if RCD1 contributes to basal resistance to Hpa, we infected the rcd1-1 mutant with Hpa Noco2. Despite the observed lower transcript abundance of NPR1 target genes in rcd1-1 (Fig. 4), the mutant was not more susceptible than wild type plants (Fig.   7A). Consistent with a previous large-scale Hpa phenotyping report (Weßling et al., 2014), rcd1-1 showed enhanced disease resistance compared to Col-0 (one-way ANOVA, F5,1149=79.22, p=1.70 x 10 -71 ; Tukey-Kramer post-hoc test, p<0.05). This suggests that, if RCD1 is a relevant virulence target of HaRxL106, inhibition of RCD1's function(s) or signaling is unlikely to be responsible for the enhanced disease susceptibility induced by HaRxL106. We considered that HaRxL106 may manipulate RCD1 in a way that is not mimicked by complete loss of RCD1 function, for example by altering RCD1's interaction with other proteins or ligands. We compared susceptibility to Tukey-Kramer post-hoc test, p<0.05). Therefore, functional RCD1 protein is essential for suppression of basal defense by HaRxL106. Loss of RCD1 function also attenuated the extent of petiole elongation in the YFP-HaRxL106 background ( Fig. 7C; line #5) However, this effect was only partial with YFP:HaRxL106 rcd1-1 #5 transgenics exhibiting an intermediate phenotype between rcd1-1 and YFP:HaRxL106 in Col-0 background, perhaps because SRO1 partially compensates for loss of RCD1 function in such lines (Jaspers et al., 2009), or because HaRxL106 might target other host proteins in addition to RCD1.

A crystal structure of RCD1's PARP domain suggests that RCD1-type proteins do not function as canonical ADP-ribosyl transferases
Our findings that suppression of basal defense by HaRxL106 is largely dependent on RCD1 and that the effector binds to RCD1's WWE-PARP domains prompted us to further investigate the molecular function(s) of RCD1. We reasoned that if RCD1 had PARP or a related transferase activity, HaRxL106 might manipulate this enzymatic function. To generate a structural framework for testing putative enzymatic functions of RCD1's PARP domain we solved its structure by x-ray crystallography at 2.5 Å resolution (for data collection and refinement statistics see Table S2). As shown in Fig. 8A the RCD1 PARP domain adopts a fold that is overall similar to mammalian PARP domains. The closest structural homologue of the RCD1 PARP domain in the protein data bank was human PARP14 [PDB identifier 3SE2, in complex with inhibitor 6(5H)phenanthridinone] with a root mean square deviation of 1.45 Å over 176 amino acids at a sequence identity of 17%. However, our structure of the RCD1 PARP domain also confirmed that the amino acid triad H-Y-E constituting the active site of mammalian and plant PARPs is not conserved in RCD1 (Kleine et al., 2008;Jaspers et al., 2010a) (Fig.   8B). The conserved Histidine (H1600) of HsPARP14 corresponds to RCD1 L333 whilst 18 the position of Y1633 in HsPARP14 is filled by H365 of RCD1. These two first residues of the H-Y-E triad make direct contact to NAD + and are essential for ADP-ribosyl transferase activity in mammalian enzymes (Kleine et al., 2008). The Glu residue of the H-Y-E triad that is critical for elongation of ADP-ribose chains, and therefore distinguishes mono-ADP-ribosyl-transferases from PARPs, is not conserved in HsPARP14. Consequently the enzymatic activity of HsPARP14 is limited to mono-ADPribosylation (Kleine et al., 2008). In the RCD1 PARP domain residue N428 takes the equivalent position. Non-conservation of the His and Tyr residues critical for NAD + binding in canonical ADP-ribosyl-transferases suggests that the RCD1 PARP domain does not bind NAD + . Furthermore, the RCD1 PARP domain lacks an acidic residue in the third position of the triad. Overall, non-conservation of all of the residues critical for NAD + binding and PARP activity strongly indicate that RCD1 does not have canonical PARP activity.
Many inhibitors of mammalian PARPs mimic the nicotinamide moiety of NAD + and stabilize canonical PARP domains in thermal shift assays (Wahlberg et al., 2012). The PARP inhibitor 6(5H)-phenanthridinone binds to several human PARPs with affinity in the low micro-molar range (Wahlberg et al., 2012) and has also been proposed to bind to the corresponding cleft of RCD1 based on homology modelling (Rissel et al., 2017).
To test whether 6(5H)-phenanthridinone also stabilizes RCD1 we compared thermal shifts induced by the inhibitor for the PARP domains of HsPARP1 and RCD1 (Fig. 8C).
As previously shown, 6(5H)-phenanthridinone stabilized the catalytic domain of HsPARP1 (Wahlberg et al., 2012). In contrast the RCD1 PARP domain did not show any changes in stability in thermal shift assays, even in presence of 2 mM 6(5H)-19 phenanthridinone (Fig. 8C). This suggests that 6(5H)-phenanthridinone either does not bind to RCD1 or that the inhibitor binds in a way that is different from binding to human PARPs and therefore does not stabilize the RCD1 PARP domain.
Conceivably, the cleft of RCD1 that corresponds to the catalytic center of active PARPs has evolved to bind other small compounds. To unequivocally test if the cleft corresponding to the catalytic center of active PARPs is required for RCD1 function we designed structure-based site-directed mutants of RCD1 ( Fig 8B). We mutated RCD1 residue H365 to Gln to abolish possible ring-stacking of small aromatic compounds, which is an important function of the corresponding Tyr (H-Y-E) residue in canonical PARPs. Furthermore, we mutated RCD1 S375 to a Trp. This mutation will abolish possible hydrogen bonding of the Ser residue with Nicotinamide-related compounds. In addition, we hypothesized that inserting a bulky hydrophobic residue at this position will occupy the cleft and block binding of potential ligands. Finally, we mutated RCD1 D421 to Ala. D421 hydrogen bonds to N428 and is the only acidic residue in close proximity that might possibly substitute for the lack of a Glu residue in RCD1 (Fig. 8B).
We transformed the RCD1 site-directed mutant constructs with a C-terminal triple HA epitope tag alongside with an RCD1-3HA wild-type control into the rcd1-1 background.
All constructs were under transcriptional control of 2.5 kb of the native RCD1 promoter.
As shown in Fig. 8D Fig. 8D). However, an independent transgenic line (A) expressing the same construct fully complemented rcd1-1, suggesting that differences in expression levels might account for partial complementation in line B. As a quantitative read-out for RCD1 function under stress conditions we determined the tolerance to oxidative stress induced by 1 M Paraquat (Fig. 8E). The rcd1-1 mutant was more resistant to Paraquat treatment than wild type plants, as previously reported (Ahlfors et al., 2004). The enhanced Paraquat tolerance of rcd1-1 was reverted by expression of RCD1Pro:RCD1:3HA. Likewise, all mutant versions of RCD1 complemented the Paraquat tolerance phenotype of rcd1-1 and we did not detect strong quantitative differences between RCD1 and its mutant variants (Fig. 8E). For the RCD1 D412A construct line A fully complemented the enhanced Paraquat tolerance of rcd1-1.
In contrast, line B that only partially complemented the developmental phenotype of rcd1-1 did not complementation the Paraquat tolerance. Nevertheless, for each RCD1 mutant variant we identified at least one transgenic line that fully complemented both, rcd1-1's growth phenotype and the elevated tolerance to Paraquat. This demonstrates that mutations in the putative NAD + binding site of RCD1 do not affect RCD1's function in plant development and oxidative stress signaling. Taken together our results strongly suggest that RCD1 has no canonical PARP activity, does not bind to a broad-spectrum PARP inhibitor in thermal shift assays, and that the integrity of the presumed NAD + binding cleft is not essential for RCD1's functions in plant development and signal transduction under oxidative stress conditions.

HaRxL106 binds to the N-terminal domains of RCD1 and SRO1 that mediate homoand hetero-dimerization
Although RCD1 does not carry a canonical PARP domain, HaRxL106 appears to manipulate RCD1 function by binding to the WWE-PARP region of RCD1 (Fig. 6A).
Using the Y2H system we further narrowed down the HaRxL106 binding site of RCD1 to an N-terminal fragment encompassing the WWE domain and the linker region up to the beginning of the PARP domain (Fig. 9A). Deletion of the linker region resulted in loss of interaction with HaRxL106 suggesting that the WWE domain on its own is not sufficient for binding to the effector. The isolated PARP domain did not bind to HaRxL106, irrespective of whether or not we included the linker region ( Fig. 9A). HaRxL106 showed binding to a construct containing WWE, linker and PARP domain but to a lower extent than binding to the WWE-linker region (Fig. 9A). In contrast, the HaRxL106C control did not interact with any of the RCD1 deletion constructs tested. This suggests that the WWE-linker region is required and sufficient for binding to HaRxL106. WWE domains are thought to mediate protein-protein interactions (Aravind, 2001). We therefore tested if the WWE-linker regions of RCD1 and SRO1 can mediate formation of homo-or hetero-oligomers. We found that the RCD1 WWE-linker region interacts with itself and the corresponding region of SRO1 in Y2H assays indicative of the formation of homo-and hetero-oligomers. We obtained comparable results for the corresponding part of the SRO1 protein (Fig. 9C). These data suggest that the RCD1 and SRO1 WWElinker regions could mediate formation of RCD1/SRO1 oligomers.

RCD1's WWE domain forms protein complexes with histone modifying kinases
Given that RCD1 does not have PARP activity, we further characterized RCD1 protein function(s) by screening for in planta interactors of RCD1. Attempts to immuno-purify epitope-tagged RCD1 protein in amounts sufficient for LC-MS/MS analysis of copurifying proteins from transient expression assays in N. benthamiana or stable Arabidopsis transgenics were not successful. We therefore resorted to screening for interactors of RCD1's WWE-linker region following transient expression in N.
benthamiana as this part of the protein binds to HaRxL106 and is more stable (Fig. 9D).
The predominant interactors were several importin- isoforms, full-length RCD1-type proteins and protein kinases with sequence homology to Casein II kinases ( Fig. 9E; Table S3). RCD1 and related proteins carry conserved NLS motifs at their N-termini (Katiyar-Agarwal et al., 2006) and co-purification of importin-proteins confirms that these NLS bind to nuclear transport adapters in plant cells. Given that we identified several peptides from the PARP and RST domains (Fig. S2) of RCD1-type proteins by LC-MS/MS, the WWE-linker fragment appears to form homo-and hetero-oligomers with endogenous RCD1-type proteins in N. benthamiana, which is consistent with oligomer formation of the WWE-linker regions in Y2H ( Fig. 9B and 9C). Our pull-down strategy therefore provides an indirect method to immuno-purify full-length RCD1-type proteins.
Another prominent group of proteins co-precipitating with the WWE-linker fragment were Casein II-like kinases. A blastp search against the Arabidopsis protein database (TAIR11) identified MUT9-like kinases (MLKs) as likely orthologs (Fig. 9E). To our knowledge MLKs have not been characterized in N. benthamiana. MLKs are nuclearlocalized Ser/Thr kinases that phosphorylate histones. In Arabidopsis and Chlamydomonas phosphorylation of histone H3 Thr3 (H3T3ph) is the best characterized 23 MLK phosphorylation site (Casas-Mollano et al., 2008;Wang et al., 2015). We found several residues in the RCD1 WWE linker region to be phosphorylated ( Fig. S3; Table   S4). While most of these phosphorylated peptides were located in the GFP:WWE-linker bait protein from Arabidopsis, we also detected two phospho-peptides from the WWE-PARP linker region of a co-purifying N. benthamiana RCD1 ortholog (Fig. S3) Table S5). In accordance with the results from N. benthamiana experiments we also identified several peptides from the PARP and RST domains of RCD1 and SRO1 (Fig. S2).
MLKs have been previously reported to affect H3T3ph levels in response to osmotic and salt stress and the mlk1,2 double mutant is hypersensitive to sub-lethal concentrations of PEG and NaCl (Wang et al., 2015). As MLKs and RCD1 form protein complexes in plant cells, and given that SA marker genes are expressed at lower levels in rcd1 mutants, we asked if MLKs also affect the transcriptional response to SA. We sprayed mlk1,2,3 and mlk1,3,4 triple mutants with SA and determined transcriptional upregulation of PR1 8 h later (Fig. 9F). The mlk1,3,4 triple mutant consistently showed elevated PR1 transcript levels in response to SA (one-way ANOVA, F3,8=79.22, p=0.01; Tukey-Kramer post-hoc test, p<0.05), whereas basal PR1 expression levels in the mock control were not strongly altered. In contrast the mlk1,2,3 triple mutant responded like wild type, suggesting that MLK4 might be particularly relevant for SA-induced transcript changes.

Discussion
Several biotrophic pathogens evolved virulence mechanisms to counteract activation of SA-dependent defense genes (Asai et al., 2014;Lewis et al., 2015), but how pathogen effectors suppress SA-mediated immunity remains only partially understood. Apart from active conversion of SA (Djamei et al., 2011), effector-mediated activation of JA signaling appears to be the main strategy of biotrophic pathogens to attenuate SAdependent defense (Zheng et al., 2012;Caillaud et al., 2013;Yang et al., 2017). Here we show that Hpa effector HaRxL106 induces shade avoidance in Arabidopsis and suppresses SA signaling while JA marker genes are either not affected or slightly repressed (Table S1).
We mapped both, the light signaling-and defense-manipulating activities of HaRxL106 to a short C-terminal part of the effector. Intriguingly, loss of RCD1 function renders Arabidopsis resilient to HaRxL106-mediated suppression of defense and diminishes HaRxL106-induced petiole elongation (Fig. 7). These results suggest that RCD1 integrates signals downstream of both, pathogen-and photoreceptors and that Hpa is 25 exploiting this function of RCD1 to attenuate plant immunity. Notably, the morphological and defense phenotypes of rcd1 null mutants are opposite to those induced by ectopic expression of HaRxL106 (Fig. 7). Therefore, it is conceivable that the effector manipulates RCD1 to boost its cellular activities although this hypothesis cannot easily be tested without a better understanding of the molecular function of both proteins.
HaRxL106 is predicted to have an -helical WY structure, a fold which likely evolved as a versatile building module of oomycete effectors and can mediate different molecular functions in fusion with small peptides or other domains (Boutemy et al., 2011;Maqbool et al., 2016). The WY domain in HaRxL106 might function as a scaffold that stabilizes and/or presents the C-terminal peptide that is essential for suppression of plant immunity. In accordance with this model, expressing a fusion of the C-terminal 58 amino acids of HaRxL106 to RFP is sufficient to trigger shade avoidance and attenuate basal defense (Fig. 6). Notably, manipulation of selective autophagy by the host-targeted Phytophthora infestans effector PexRD54 is also based on a disordered C-terminal peptide that is stabilized by five tandem WY domains . RCD1, and sequence-related proteins from Arabidopsis and rice, bind transcription factors via their C-terminal RST domains. In contrast the function(s) of RCD1's Nterminal WWE and central PARP domains have not been characterized. Although an RCD ortholog from wheat shows PARP activity (Liu et al., 2014), our structural analysis suggests that Arabidopsis RCD1 is unlikely to be enzymatically active. Our crystal structure of the RCD1 PARP domain provides first insights into plant PARP domains and we identified several molecular differences between RCD1's PARP domain and the 26 catalytic domain of mammalian PARPs. None of the amino acids constituting the H-Y-E triad that are essential for PARP activity in previously characterized enzymes are conserved in RCD1 (Fig. 8B). Our complementation analysis of RCD1 site-directed mutant variants shows that RCD1 tolerates relatively severe amino acid exchanges in the cleft region of its PARP domain without losing its functions in plant development and oxidative stress signaling ( Fig. 8D and E). Taken together, our structural and molecular analyses of RCD1 suggest that the protein does not have PARP activity and can therefore, in analogy to pseudo-kinases, be classified as a pseudo-PARP. Given that the N-and C-terminal domains of RCD1 mediate protein-protein interactions, the PARP domain of RCD1 might act as a scaffold bridging and/or coordinating the action of the terminal protein interaction domains. RCD1's WWE domain and the linker region up to the PARP domain are essential for binding to HaRxL106 (Fig. 9A). We also found that the WWE domains of RCD1 and its paralog SRO1 can form homo-and hetero-oligomers in Y2H (Fig. 9). It will be interesting to clarify if oligomer formation has a biological function in signal transduction of RCD1type proteins. The WWE domain is a conserved iso-ADP-ribose binding domain but it is not known if plant WWE domains bind PAR chains Wang et al., 2012 Table S3). The interaction between MLKs and the RCD1 Nterminus implies that phosphorylation of the linker region might be mediated by MLKs, but our data do not rule out alternative kinases. The ~90 amino acid linker region between the RCD1 WWE and PARP domains is predicted to be disordered (Ishida and Kinoshita, 2007;Kragelund et al., 2012) but it is conceivable that phosphorylation or binding of interacting proteins induce a specific fold in this region (Wright and Dyson, 2009;Bah et al., 2015). As the RCD1 PARP domain appears to act as a scaffold, reversible phosphorylation of residues in the linker region between WWE and PARP domains could regulate the cooperation of these two domains.
MLKs are recruited to the evening complex in a PHYB-dependent manner and have previously reported functions in light signaling, circadian rhythm and abiotic stress responses (Casas-Mollano et al., 2008;Wang et al., 2015;Huang et al., 2016). Here we show that an mlk1,3,4 triple mutant responds with higher expression of the SA marker gene PR1 to exogenous application of SA while basal PR1 levels are unaffected (Fig.   9F). This suggests that MLK phosphorylation sites on histones or other proteins also influence transcriptional mechanisms required for fine-tuning the amplitude of SAinduced PR1 expression. H3T3 is the best-characterized phosphorylation site of MLKs in Chlamydomonas and Arabidopsis. Notably, in mammalian cells repressive H3T3ph and activating tri-methylation of the adjacent K4 form a molecular switch that directly affects TFIID binding thereby regulating gene expression in different phases of the cell cycle (Varier et al., 2010). In mammals the TAF3 subunit anchors the TFIID complex onto H3K4me3-marked nucleosomes (Vermeulen et al., 2007;van Ingen et al., 2008). In contrast, the TAF3 ortholog of plants, if it exists, has not been identified (Lawit et al.,28 2007; Fromm and Avramova, 2014). Consistent with H3T3ph being a repressive chromatin mark, we find that simultaneously knocking out MLK1, 3, and 4 results in higher SA-induced PR1 expression.
Although attenuated defense responses under shade conditions have been reported for multiple plant-pathogen or plant-herbivore interactions, the underlying molecular mechanisms are only partially understood (Ballaré, 2014). Notably, constitutive SA defense mutants still undergo shade avoidance, arguing against simple resource partitioning scenarios to explain the 'growth-defense tradeoff' (de Wit et al., 2013).
Reciprocally, the sav3-2 mutant that is impaired in the morphological responses to shade still shows attenuated activation of JA-dependent pathogen resistance under low R/FR but it remains unknown if the same holds true for SA-dependent immunity (Cerrudo et al., 2012). Of note, unlike HaRxL106 over-expressors, plants infected by (hemi-)biotrophic pathogens do not show morphological signs of shade avoidance. We hypothesize that this discrepancy is explained by the tissue-unspecific expression of the HaRxL106 transgene by the 35S promoter. In contrast, Hpa primarily infects epidermal and mesophyll leaf cells in natural infections (Koch and Slusarenko, 1990).
Overall, these observations indicate that activation of at least one signaling pathway under shade, rather than the morphological responses triggered by it, are responsible for suppression of innate immunity. PIF transcription factors integrate light, temperature and other environmental signals (Wigge, 2013). Recently, PIF4 was shown to act as a negative regulator of plant immunity at elevated temperature (Gangappa et al., 2016).
Constitutive defense activation and the dwarf stature of the Nod-like receptor mutant 29 snc1-1 is suppressed when plants are grown at elevated temperature, and this depends on PIF4. A similar effect was observed for overexpression of PHYB in the snc1-1 mutant background (Gangappa et al., 2016). Interestingly, also mutations in RCD1 render snc1-1 insensitive to elevated temperature although this effect was only observed for plant size and not for disease resistance to the bacterial strain Pst DC3000 (Zhu et al., 2013).
Nevertheless, the similar phenotypes observed in snc1-1 pif4, snc1-1 35S:PHYB and snc1-1 rcd1 double mutants indicate that RCD1 might be a component of a regulatory node integrating light, temperature and defense signals. RCD1 interacts with PIF transcription factors and rcd1 mutants show reduced hypocotyl elongation under red and blue light (Jaspers et al., 2009;Salazar, Felipe Sarmiento and Neuhaus, Gunther, 2010).
Although the molecular function(s) of RCD1 remain poorly characterized, its localization to the nucleus and interaction with transcription factors point to a role as a transcriptional co-regulator. Consistent with such a role of RCD1 is our finding that HaRxL106 interferes with SA signaling at the level of transcription (Fig. 3).

Generation of transgenic Arabidopsis lines
Transgenic Arabidopsis plants expressing YFP-tagged HaRxL106 have been described  (Wirthmueller et al., 2015). Plasmids pXCSG-HS and pH7WGR2 were recombined with the above pENTR plasmids using Gateway® LR clonase II to generate HS-and RFP-tagged versions of the HaRxL106 constructs, respectively.
Transgenic HS-tagged HaRxL106 lines were generated by transforming Col-0 or 35SPro:NPR1:GFP plants with A. tumefaciens strain GV3101 pMP90 RK carrying pXCSG-HS:HaRxL106 constructs. Transgenic lines expressing RFP-tagged HaRxL106 constructs were generated by transforming Col-0 with A. tumefaciens strain GV3101 pMP90 carrying the corresponding pH7WGR2 plasmids. We transformed plants using floral dip (Logemann et al., 2006). Col-0 lines expressing RCD1 amino acids 1-265 as Cterminal fusion to GFP were generated by recombining a corresponding pENTR clone with pK7WGF2 (Karimi et al., 2002), transforming the construct into A. tumefaciens strain GV3101 pMP90 and subsequently transforming Arabidopsis Col-0 plants by floral dip. Site-directed mutants of RCD1 were generated using the QuikChange method (Agilent). To generate site-directed mutants of RCD1 a 428 bp BamHI/NcoI fragment was subcloned from the RCD1 coding sequence into the BamHI/NcoI sites of pET-DUET. Following successful mutagenesis using the QuikChange method, BamHI/NcoI fragments were cloned into a pENTR/D-TOPO plasmid carrying 2499 bp of the RCD1 promoter sequence fused to the RCD1 coding sequence. The pENTR/D-TOPO plasmids carrying wild-type or mutated RCD1 were then recombined with pGWB13 (Nakagawa et al., 2007) to generate a translational fusion with a triple HA-tag at the C-terminus of the protein. The constructs were transformed into the rcd1-1 mutant by floral dip as described above.

Hpa infection and quantification
Basal resistance to Hpa isolate Noco2 was either tested on adult leaves of 6-week-old plants (Fig. 1A)  For the seedling assay 35-40 seedlings per genotype were incubated in a 0.02% (w/v) solution of Uvitex 2B, then destained in water for 2 min, mounted on a Styrofoam rack and imaged through a Leica UV filter (Leica #10447415) using a Leica M165 FC fluorescent stereo microscope connected to a EL6000 laser source. Only conidiophores on the upper side of the cotyledons were counted.

Quantification of SA
Six-weeks-old Arabidopsis plants were syringe-infiltrated with 1 x 10 8 cfu/ml Pst DC3000 in 10 mM MgCl2 or a 10 mM MgCl2 mock solution. Samples were taken 24 h after infiltration including control samples from non-treated plants. SA was extracted and quantified as previously described (Aboul-Soud et al., 2004). Briefly, leaf tissues (0.2 g) were extracted in 1 ml 90% methanol following homogenization in liquid nitrogen. 3hydroxybenzoic acid (HBA, Sigma) was used as an internal standard. The level of SA was determined by a fluorescence detector (Shimadzu RF-20AXS, excitation at 305 nm and emission at 405 nm) in reverse-phase HPLC using a Nexera UHPLC (Shimadzu) system with a C18 (Kinetex 2.6µm XB-C18) column.

SA treatment
For SA treatment Arabidopsis plants were grown on soil under short day conditions. One h after dawn (09:00) we sprayed 4-week-old plants with a solution containing 0.1 mM SA and 0.01% Silwet L-77 until run-off and took samples for RNA preparations 8 h later.

Paraquat treatment
Arabidopsis seedlings were sown on GM medium supplemented with 1 M Paraquat and placed in long-day conditions. After 14 days the number of seedlings with expanded true leaves was counted.

RNA extraction, cDNA synthesis and RNA-Seq transcriptome profiling
Five-week-old Arabidopsis plants were syringe-infiltrated with 5 x 10 5 cfu/ml of P. syringae DC3000 at 12:00 (4 h after lights on). Rosette leaf samples from non-treated, mock-and bacteria-infiltrated plants were harvested 24 h later. Total RNA was extracted using the TRI reagent (Sigma) and 1-Bromo-3-chloropropane (Sigma), as per manufacturer's guidelines. RNA was precipitated with half volume of isopropanol and half volume of high salt precipitation buffer (0.8 M sodium citrate and 1.2 M sodium chloride). RNA samples were treated with DNaseI (Roche) according to the manufacturer's recommendation and phenol/chloroform extracted and ethanol precipitated. For qRT-PCR assays, mRNA was reverse transcribed using SuperScript II Reverse Transcriptase (Thermo Fisher) according the manufacturer's conditions. cDNA 34 samples were diluted fivefold and 1 µl was used as template in a 20 µl qRT-PCR reaction. For RNA-seq, 3 µg of total RNA was used to generate first strand cDNA using a oligo(dT) primer comprising the P7 sequence of Illumina flow-cell. Double strand cDNA was synthesized as described previously (Okayama and Berg, 1982). Purified cDNA was subjected to Covaris shearing to a target size of 200bp (parameters: Intensity -5, Duty cycle -20%, Cycles/Burst -200, Duration -90seconds). End repairing and Atailing of sheared cDNA was carried out as described by Illumina. Y-shaped adapters were ligated to A-tailed DNA and subjected to size selection on 1x TAE agarose gels.
The gel-extracted library was PCR enriched and quantified using qPCR with previously sequenced similar size range Illumina library. Transcriptome data has been deposited at NCBI's Gene Expression Omnibus (https://www.ncbi.nlm.nih.gov/geo/) with identifier GSE89402.

Yeast-two-Hybrid
All constructs for the Y2H assay were cloned into pDEST32 (bait) or pDEST22 (prey) vectors (Invitrogen) using Gateway® recombination. The RCD1, WWE-PARP (amino acids 1-471) and PARP-RST (amino acids 241-589) deletion constructs used for Fig. 6A have been published (Jaspers et al., 2009). Two additional constructs, WWE (amino acids 1-155) and PARP (amino acids 247-472), were cloned into pDEST22 using the same strategy. For the Y2H assays in Fig. 9 A-C the following RCD1 and SRO1 deletion For HaRxL106 constructs the pENTR plasmids carrying HaRxL106, HaRxL106C or HaRxL106-Cterm58 were recombined into pDEST32. Yeast strain Mav203 was cotransformed with pDEST32 and pDEST22 plasmids and double-transformed yeast cells were selected on SD -Leu -Trp plates. Selected clones were grown in liquid SD -Leu -Trp medium at 30 °C for 48 h until cultures had reached saturation. The OD600 was adjusted to 0.1, and the yeast strains were plated in 10-fold serial dilutions onto SD -Leu -Trp -His medium containing 0, 1, 5, 10 or 20 mM 3-AT. Serial dilutions were also plated on SD -Leu -Trp medium to compare growth rate of the yeasts under nonselective conditions. The plates were photographed 3-5 days after plating.

qRT-PCR
qRT-PCR was performed using 10 l SYBR Green qPCR premix (Sigma) and the The data were analyzed using CFX Manager™ software (Bio-Rad). The average relative transcript levels from three independent biological replicates were analyzed using oneway ANOVA followed by a Kramer-Tukey post hoc test.

Transient expression
RCD1:GFP and SRO1:GFP constructs were generated by cloning their coding sequences into pENTR/D-TOPO and recombining these plasmids with pK7FWG2 36 (Karimi et al., 2002). A. tumefaciens GV3101 strains were grown on selective plates, resuspended in 10 mM MgCl2 10 mM MES pH 5.6 and incubated with 100 M acetosyringone for 2 h at RT. Prior to infiltration each strain was mixed with A. tumefaciens strain GV3101 expressing the silencing suppressor 19K at a ratio of 1:2 [19K]. For co-expression, strains were mixed in a 1:1:2[19K] ratio. We infiltrated leaves of 4-week-old N. benthamiana plants with a needleless syringe and harvested the leaves 48-72 h later.

Protein extraction from Arabidopsis, N. benthamiana, immunoprecipitation and western blot
Protein extracts were prepared by grinding N. benthamiana or Arabidopsis leaf material in liquid nitrogen to a fine powder followed by resuspension in extraction buffer [50 mM Tris, 150 mM NaCl, 10% glycerol, 1 mM EDTA, 5 mM DTT, 1× protease inhibitor cocktail (Sigma #P9599), 0.2% NP-40, pH 7.5] at a ratio of 2 ml buffer per 1 g leaf material. For experiments that included RCD1 (domains) the proteasome inhibitor MG132 (10 M) and phosphatase inhibitors NaF (10 mM) and Na3VO4 (1 mM) were added to all buffers.
Crude protein extracts were centrifuged at 20.000 x g 4°C 20 min and the supernatant was either boiled in sodium dodecyl sulphate (SDS) sample buffer for western blots or used for immunoprecipitation. For western blots proteins were separated by SDS-PAGE and electro-blotted onto polyvinylidene difluoride membrane (Millipore). Antibodies α-HA 3F10 (Roche), α-GFP 210-PS-1GFP (Amsbio), α-RFP-biotin ab34771 (Abcam) were used for detection. For immunoprecipitation a fraction of the supernatant was saved as 'input' sample and 15 μl GFP-beads (GFP-Trap_A; Chromotek) were added to the remaining supernatant. The volume for immunoprecipitation from N. benthamiana for western blots was 1.4 ml. The volume for immunoprecipitation from Arabidopsis was 4 ml. The volume for immunoprecipitation experiments coupled to mass spectrometry was 15-25 ml. Following incubation of the samples on a rotating wheel at 4°C for 2 h the beads were collected by centrifugation at 1200 x g and 4°C for 1.5 min. The beads were washed 3 times with 1 ml extraction buffer and then boiled in SDS sample buffer to elute proteins from the beads.

Thermal shift assays
Thermal shift assays were performed in a CFX96 Touch™ Real-Time PCR Detection System (Bio-Rad) as described by Vivoli et al. (2014)  6(5H)-phenanthridinone (Sigma) was added from a 25 mM stock solution in DMSO to final concentrations of 2 nM to 2 mM or DMSO was used as a control. The total reaction volume was 25 l. Unfolding of the proteins was determined by running a thermal denaturation program with 0.5 °C temperature increase per cycle followed by measuring SYPRO orange fluorescence in FRET mode of the instrument. The first derivative of the fluorescence over temperature was plotted using CFX Manager™ software (Bio-Rad).

Confocal microscopy
Arabidopsis leaf discs were mounted onto microscopy slides in 50% glycerol or water and analyzed on a Leica DM6000B/TCS SP5 confocal microscope with the 488 nm as excitation wavelengths for GFP.

Recombinant expression and purification of PARP domains from E. coli
The expression construct for the human PARP1 PARP domain (amino acids 662-1014; L713F mutant) has been described (Langelier et al., 2012). The PARP domain of RCD1 (amino acids 269-460) was cloned into the pOPINF vector (Berrow et al., 2007) linearized by KpnI/HindIII digest using Gibson assembly. The expression construct was HEPES, 150 mM NaCl, pH 7.5]. PARP domains eluting from the column were concentrated to 0.5-1 mg/ml (HsPARP1) and used for thermal shift assays. For crystallization of the RCD1 PARP domain, the His6-tag was cleaved using 3C protease.

39
The protein was run through a 5 ml HisTrap HP IMAC column in buffer A1 to remove the His6 tag and residual uncleaved fusion protein, followed by injection onto the Superdex 75 26/60 PG column and eluted as above. The protein was concentrated to 35 mg/ml using Vivaspin 20 and 2 columns (Sartorius) with a molecular weight cut-off of 5 kDa and flash-frozen in liquid nitrogen. Seleno-Met-labelled RCD1 PARP protein was produced using feedback-inhibition and purified as described above for native RCD1 PARP.

Crystallization and data collection
The RCD1 PARP domain was used for crystal screens at a concentration of 35 mg/ml. Crystals of native and Seleno-Met-labelled protein grew in 0.1 M HEPES pH7.5, 1.6 M Ammoniumsulfate, 2% PEG1000 in hanging drops at 20 °C. The crystals were harvested in Paratone-N oil (Hampton) and frozen in liquid nitrogen. Data collection was performed on beamlines i04 (native) and i24 (Seleno-Met) at Diamond Light Source, Oxford, UK. X-ray data were processed with iMosflm (Battye et al., 2011) and scaled with Aimless (Evans and Murshudov, 2013) from the CCP4 suite (Collaborative Computational Project, Number 4, 1994). For X-ray data collection statistics see Table   S2. The RCD1 PARP structure was solved by single wavelength anomalous dispersion using Phaser (McCoy, 2007). Iterative building and refinement cycles with Coot (Emsley et al., 2010), Refmac5 (Murshudov et al., 2011) and Phenix (Adams et al., 2010) were used to obtain the final model with statistics given in Table S2. Validation tools in Molprobity  and Coot were used to analyze the final structure. 3D visualizations of protein structures were prepared using PyMOL software v1.7.2 (http://sourceforge.net/projects/pymol/). Reflection data and the RCD1 PARP domain structure have been deposited at the Protein Data Bank with identifier 5NGO.

Protein mass spectrometry
Samples for LC-MS analysis were prepared by excising bands from one dimensional SDS-PAGE gels stained with colloid Comassie Brilliant Blue (Instant Blue, Expedeon).
The gel slices were destained with 50% Acetonitrile and cysteine residues modified by 30 min reduction in 10 mM DTT followed by 20 min alkylation with 50 mM chloroacetamide. After extensive washing with 30% Acetonitrile and dehydration with 100% acetonintrile the pieces were incubated with 100 ng of trypsin (Promega) in 100 mM ammonium bicarbonate, 10% acetonitile at 37 ºC overnight.
LC-MS/MS analysis was performed using a hybrid mass spectrometer Orbitrap Fusion and a nanoflow UHPLC system U3000 (Thermo Scientific). The generated peptides were applied to a reverse phase trap column (Acclaim Pepmap 100, 5 μm, 100 μm x 20 mm) connected to an analytical column (Acclaim Pepmap 100, 3μm, 75μm x 500mm; Thermo Scientific). Peptides were eluted in a gradient of 9-50% acetonitrile in 0.1% formic acid (solvent B) over 50 min followed by a gradient of 50-60% B over 3 min at a flow rate of 300 nL min -1 . The mass spectrometer was operated in positive ion mode with nano-electrospray ion source with an ID 0.01 mm fussed silica PicoTip emitter (New Objective). Voltage +2.2 kV was applied via conductive T-shaped coupling union.
Transfer capillary temperature was set to 320 ºC, no sheath gas, and the focusing voltages were in factory default setting. Method MS events consisted from full scan in Orbitrap analyzer followed by two collisions of "softer" CID and more "energetic" HCD to maximize the chances to acquire spectra with structurally important information. Orbitrap            Table S4.