Kinases and protein motifs required for AZI1 plastid localization and trafficking during plant defense induction

Summary The proper subcellular localization of defense factors is an important part of the plant immune system. A key component for systemic resistance, lipid transfer protein (LTP)‐like AZI1, is needed for the systemic movement of the priming signal azelaic acid (AZA) and a pool of AZI1 exists at the site of AZA production, the plastid envelope. Moreover, after systemic defense‐triggering infections, the proportion of AZI1 localized to plastids increases. However, AZI1 does not possess a classical plastid transit peptide that can explain its localization. Instead, AZI1 uses a bipartite N‐terminal signature that allows for its plastid targeting. Furthermore, the kinases MPK3 and MPK6, associated with systemic immunity, promote the accumulation of AZI1 at plastids during priming induction. Our results indicate the existence of a mode of plastid targeting possibly related to defense responses.


INTRODUCTION
Plants have an innate, non-adaptive, immune system based on the ability to recognize non-self-molecules (Spoel and Dong, 2012). Pathogen recognition depends largely on two types of proteins: pattern recognition receptors that reside on the plasma membrane and perceive microbe-associated molecular patterns (MAMPs; e.g. flg22 peptide, derived from bacterial flagellin) (Macho and Zipfel, 2014), and intracellular resistance proteins (R proteins), receptors that recognize specific effector proteins, or the effector-induced alterations on their targets, that are injected by pathogenic microbes to promote virulence (Jones and Dangl, 2006;Cesari, 2018). After pathogen recognition, plants trigger local defense responses, such as reactive oxygen species accumulation, callose depositions and the activation of key signaling kinases, including MPK3 and MPK6 (MPK3/6) (G omez-G omez and Boller, 2000;Chinchilla et al., 2006;Boller and Felix, 2009;Schwessinger et al., 2011). Additionally, pathogen infections can also induce long-lasting and broad-spectrum systemic resistance (Fu and Dong, 2013;Pieterse et al., 2014). Depending on the plant tissue involved in the initial recognition, different types of systemic resistance programs are established. Aerial tissues induce systemic acquired resistance (SAR) (Fu and Dong, 2013), whereas roots trigger the so-called induced systemic resistance (ISR) that extends to the aerial tissue (Pieterse et al., 2014). In addition, when a MAMP(s) is/are perceived, it can also induce systemic resistance termed MAMP-triggered SAR (mSAR) (Mishina and Zeier, 2007;Cecchini et al., 2015b).
Systemic immunity programs are typically characterized by an alert or "primed" state, which allows the plant to efficiently reactivate its defenses (more rapidly and/or strongly) against a subsequent pathogen attack (Jung et al., 2009;Parker, 2009;Conrath et al., 2015;Martinez-Medina et al., 2016). For the establishment of systemic immunity after microbial recognition, the local production of one or more signal molecules capable of moving to distal tissues and priming defenses is needed. Several systemic signals have been described and many appear to be acting together during different systemic resistance programs (Park et al., 2007;Truman et al., 2007;Jung et al., 2009;Chanda et al., 2011;Chaturvedi et al., 2012;N avarov a et al., 2012;Wittek et al., 2014;Chen et al., 2018). One of those signals is the lipid-derived azelaic acid (AZA). AZA is locally generated in plastid envelopes and possibly thylakoid membranes and moves from local leaves to the systemic tissues (Ren et al., 2008;Jung et al., 2009;Zoeller et al., 2012;Yu et al., 2013;Gao et al., 2014;Cecchini et al., 2015b). AZA can induce a primed state when exogenously applied to aerial tissues (Jung et al., 2009;Cecchini et al., 2015b). A key component shared between SAR, ISR and mSAR programs and the associated priming, is the lipid transfer protein (LTP)-like AZI1 (AZELAIC ACID INDUCED 1), and its close paralog EARLI1 (EARLY ARABIDOPSIS ALUMINUM INDUCED 1) (Jung et al., 2009;Cecchini et al., 2015b). AZI1 is required for the systemic movement of AZA specifically affecting systemic, but not local disease resistance (Cecchini et al., 2015b). Remarkably, AZI1 is also needed for the action of other proposed systemic defense signals glycerol-3-phosphate, dehydroabietinal and pinene-monoterpenes (Chaturvedi et al., 2012;Yu et al., 2013;Riedlmeier et al., 2017). Proteins proposed to posttranslationally alter AZI1 are MPK3/6, which can phosphorylate AZI1 in vitro (Pitzschke et al., 2014); these kinases have prominent roles in SAR, ISR and defense priming induced by AZA (Beckers et al., 2009;Cecchini et al., 2019).
AZI1 is a membrane protein and a pool of it, together with EARLI1, exists near the site of AZA production, the plastid outer envelope membrane (Zoeller et al., 2012;Cecchini et al., 2015b). AZI1 also localizes to the endoplasmic reticulum (ER), plasma membrane (PM) and plasmodesmata (Cecchini et al., 2015b;Lim et al., 2016). It was proposed that AZI1 forms part of membrane contact site complexes between plastids and ER membranes, allowing the non-vesicular transport of AZA and possibly other nonpolar signals to systemic tissues (Cecchini et al., 2015b). After SAR-triggering infections, AZI1/EARLI1 becomes highly enriched at plastid envelopes, which suggests that plastid targeting is critical for AZI1/EARLI1's role in signaling (Cecchini et al., 2015b). However, AZI1 and EARLI1 do not possess a "classical" predicted or known signal sequence/motif that can explain their localization. AZI1 motifs can be divided into an amino terminal hydrophobic domain (HD, which is also a putative signal peptide (SP)), a central proline-rich region (PRR) with unknown function, and a C-terminal LTP domain (8 cysteine motif; 8CM) predicted to bind lipids ( Figure 1a). Proteins like AZI1 that have a PRR plus an 8CM domain are considered "Hybrid Proline Rich Proteins" (HyPRPs) (Dvor akov a et al., 2007). Previously, it was suggested that AZI1 employs an undescribed N-terminal bipartite signal (SP+PRR) that drives its plastid targeting (Cecchini et al., 2015b).
Although many proteins that are localized to plastids show no recognizable signals, plastid targeting mechanisms for nuclear-encoded proteins have been defined to some extent and can be divided into three groups: (1) targeted proteins with cleavable transit peptides or pre-sequences that are recognized in the plastid envelope and In AZI1 variant IV, both acylation sites were replaced by alanine (C28/30A). HD: hydrophobic domain/predicted signal peptide, similar to signal peptide, possible transmembrane domain; CxC: possible acylation sites; CPR: positively charged region; PRR: proline-rich region; 8CM: lipid transfer domain (8-cysteine motif). (b) Western blots of microsomal membrane (M) and soluble (S) protein fractions from N. benthamiana expressing GFP alone or fused to AZI1 variants (a) or DIR1:GFP. Bands were revealed using anti-GFP antibody. Similar results were observed in three independent experiments. (c) Western blots of microsomal membrane (M) and soluble (S) protein fractions samples used in (b). Bands were revealed using anti-H + ATPase and anti-FBP antibodies as integral membrane protein and cytosolic markers, respectively. The blots in (b, c) stained with Coomassie blue (CBB) are presented to show loading. The asterisks indicate possible cleavage products.
© 2020 The Authors. then imported to the stroma, thylakoids or move back to inner or outer envelopes (Lee et al., 2017); (2) proteins with no cleavable signal where a transmembrane domain (TMD) acts as a targeting signal and anchors them to the outer envelope membrane (Kim and Hwang, 2013); (3) and the (less understood) plastidic b-barrel proteins in which the secondary/tertiary structures constitute the organelle targeting region . In addition, TMD-driven plastid proteins can be separated into tail-anchored (C-terminal region TMD) or signal anchored (N-terminal region TMD) (Kim and Hwang, 2013). Signal-anchored proteins are flanked by a charged positive region (CPR), usually containing at least three basic residues, important for the targeting as an ER import evading signal (Waizenegger et al., 2003;Lee et al., 2011). It was proposed that the degree of hydrophobicity of the TMD determines whether the region will anchor to plastid, mitochondria or ER membranes. Although there is a degree of overlap, signal-anchored proteins in which the TMDs show a Wimley and White hydrophobicity score >0.4 mainly target the ER, whereas those with lower values mainly target plastids and/or mitochondria (Lee et al., 2011;Lee et al., 2014).
Here, we characterized AZI1 motifs in relation to their impact on subcellular targeting (with a focus on plastids) and identified pathogen defense components that modulate AZI1 localization. We report that specific features of the AZI1 amino terminus and the defense-associated kinases MPK3/6 mediate AZI1-plastid targeting and/or intracellular trafficking. Our results suggest the existence of a mechanism of plastid targeting and trafficking that is active during defense responses against pathogens.

AZI1 is a signal-anchored protein
The TargetP algorithm (Emanuelsson et al., 2007) predicts that AZI1's hydrophobic amino terminus is a signal peptide. However, since AZI1 strongly associates with membranes (Cecchini et al., 2015b), this domain might be a non-cleavable TMD anchor, as has been shown for many signal-anchored proteins (Figure 1a I) (Jayasinghe et al., 2001;Kim and Hwang, 2013). In support of this idea, amino acids 31-37 (KPSPKPK) in AZI1's amino terminus constitute a charged protein region (CPR) that is characteristic of signal-anchored proteins (Figure 1a I).
To further analyze the possibility that AZI1's N-terminal region (AZI1 D38-161 containing the HD+CPR) functions as a signal anchor, we tested if it was sufficient to confer membrane association. We generated a construct where AZI1 D38-161 was fused to a GFP construct (Figure 1a III) and analyzed its localization by fractionation. AZI1 fused to GFP was previously shown to retain function (Cecchini et al., 2015b). Microsomal and soluble fractions were obtained from agroinfiltrated Nicotiana benthamiana leaves and examined by immunoblot. A large pool of the AZI1 D38-161 : GFP fusion protein was found in the microsomal fraction (M), with only traces present in the soluble fraction (S) (Figure 1b III). Because AZI1 also possesses two possible acylation sites after the HD (C28/C30; CSS-Palm 2.0; (Ren et al., 2008)) that could relate to its microsomal localization, we repeated the analysis using a second construct with these residues mutated (cysteines to alanines, CxC? AxA) (Figure 1a IV). Figure 1 (b IV) shows that this variant also largely partitioned with the microsomal fraction, indicating that these sites do not significantly affect membrane localization. As previously shown, full length AZI1 or AZI1 without the LTP domain (8CM) also localized to microsomal fractions (Figure 1b I and II) (Cecchini et al., 2015b). As controls, the soluble GFP and the soluble/microsomal DIR1:GFP constructs showed the expected fractionation patterns indicating that we can detect transiently expressed proteins located to the soluble and membrane fractions of N. benthamiana (Figure 1b, right panel) (Cecchini et al., 2015b). In addition, the distribution of markers for microsomal and soluble fractions, H + ATPase and FBPase, respectively, indicated that the fractionation worked as anticipated ( Figure 1c).
Together, these results strongly suggest that the N-terminal 37 amino acids that include AZI1's putative SP (the HD region in Figure 1) are sufficient to confer membrane anchoring, indicating that AZI1 is a signal-anchored protein.
The PRR and TMD are required for the normal pattern of AZI1 plastid envelope targeting Most signal-anchored proteins targeted to plastid envelopes (or mitochondria) have an N-terminal TMD Wimley and White hydrophobicity score below 0.4 (Lee et al., 2011). In contrast, AZI1's HD displays a 0.58 score, indicative of an ER membrane resident signal-anchored protein (MPEx software 3.2, amino acids 8-26, http://blanco.b iomol.uci.edu/mpex/index.html; Jayasinghe et al., 2001;Kim and Hwang, 2013)). Thus, to determine whether other domain/motifs are required for AZI1 plastid outer envelope membrane targeting, we generated AZI1 variants for each one ( Figure 2a) and fused them to GFP or an HA tag.
To test the subcellular localization of AZI1-GFP variants, we imaged agrotransformed N. benthamiana leaves by confocal microscopy. The deletions of the PRR motif (AAs 40-76), or CPR+PRR (AAs 32-76) and CxC+CPR+PRR (AAs 28-76) regions, abolished AZI1's plastid targeting (Figure 2b,compare I and II to V,VI and VII). Quantitation of the GFP fluorescence showed a significant decrease in GFP signal localized to plastids upon deletion of these regions (Figure 2c I, II, V, VI, and VII). Consistent with our previous work (Cecchini et al, 2015b), fractionation confirmed the loss of plastid targeting for PRR deletion variants V and VI; quantitation of chlorophyll and immunoblot analysis of BiP2 indicates that chloroplast enrichment was successful without significant ER contamination ( Figure 2d). In contrast, plastid localization was largely retained in AZI1 variants upon the loss of the CPR region or putative acylation sites ( . Surprisingly, even though neither variant displayed AZI1's ring-like localization around plastids, fractionation revealed that both variants X and XI associate with plastids ( Figure 2e). Further partitioning of plastids into membrane and soluble fractions indicates that variant X is plastid-membrane associated ( Figure 2f). Close inspection of the confocal micrographs supports the fractionation results, as there were points of contact between GFP filaments and plastids (Figure 2b, X and XI yellow arrowheads, and X and XI insets at bottom-right panel). This was also observed in the GFP fluorescence quantitation when compared to control GFP (Figure 2c X and XI). We noted that variants X and XI were similar in apparent mass to AZI1 (Figure 2e). It is possible that the higher percentage of proline residues in AZI1 variants relative to AZI1 causes anomalous migration patterns during SDS-PAGE similar to what has been previously described for other proline-rich proteins (Hames, 1998). Alternatively, higher than expected migration may be due to extra posttranslational modifications related to the formation into filaments of variants X and XI. These results corroborate our earlier observations about the importance of the PRR for plastid targeting (Cecchini et al., 2015b). Furthermore, they show that the HD/TMD is required for the normal ring-like pattern of plastid membrane association. Together, they indicate that the motifs required for full AZI1 plastid outer envelope membrane targeting are the HD/TMD plus the PRR (Table S1), suggesting a bipartite signal-anchored protein targeting mechanism.

Loss of the TMD causes AZI1 to stably associate with microtubules
The fluorescence patterns of AZI1-GFP variants that lack the TMD showed strong similarities to microtubules (MTs) and/or actin filaments (AFs) (Figure 2b X and XI; Kang et al., 2014;Kumar et al., 2018). Therefore, we investigated the effects of Oryzalin (Ozn) and Latrunculin B (LatB), MT and AF inhibitors, respectively, on the AZI1 D2-30 :GFP filamentous pattern. As Figure 3(a) (left panel) shows, Ozn but not LatB treatment vastly reduced AZI1 D2-30 :GFP filaments compared with mock treatment. The effectiveness controls for the inhibitors, on MT-targeted (RFP-TUB6) and actin-targeted (LifeAct) markers, showed the expected results (Figure 3a, middle and right panels). Moreover, when we coexpressed AZI1 D2-30 :GFP together with RFP-TUB6, we observed a robust co-localization between GFP and RFP signals ( Figure 3b, upper panels). No obvious co-localization was observed between RFP-TUB6 and LifeAct markers ( Figure 3b, bottom panels). Importantly, live imaging of AZI1:GFP together with RFP-TUB6 showed that full length AZI1 in vesicle-like structures moved in close association with MT bundles (Figure 3c and Movie S1). The proportion of AZI1:GFP colocalizing with RFP-TUB6 represents~16% of the total signal compared with a~3% found in control GFP ( Figure S1a). Thus, loss of the TMD anchor induced AZI1 to (more) stably associate with MTs network, suggesting that there is a close relationship between AZI1 and cell cytoskeleton. We considered the possibility that the association of AZI1 D2-25 :GFP and AZI1 D2-30 :GFP with discrete contact points on plastids (Figure 2b,e) might be due to the accumulation of these AZI1 variants along the MTs. However, tubulin was absent from plastid fractions, suggesting that these variants associate with plastids without stable/ strong MT associations ( Figure S1b).
Next, to analyze which of the AZI1 motifs/regions are required for AZI1 D2-25/2-30 :GFP association with MTs, we  (version IXa:C28A, IXb: C30A, IXc: C28/30A). AZI1 variants X and XI inset panel (at bottom-right position) shows a close up of the filamentous signal overlapping ("touching") with plastid autofluorescence. White arrowheads indicate GFP fluorescence in plastid envelopes (note the rings of GFP signal in green that surround autofluorescent plastids in blue in the micrographs). Yellow arrowheads indicate filamentous GFP signal overlapping plastid autofluorescence. Bar = 10 µm. (c) Quantification of mean GFP signal intensity of AZI1 and its variant constructs within the plastids. The GFP signal intensity from 20 to 25 plastids was quantified for each construct, from 4 to 5 independent imaging experiments. The bars represent standard errors of means of these values. Individual data points are shown on the bar chart as scatter-dots. (d-f) Western blots of plastid (P) and total (T) fractions and total (Pt), envelope (Pe) and soluble (Ps) plastid subfractions from N. benthamiana expressing GFP-or HA-tagged AZI1 or AZI1 deletion and mutation variants. Bands were revealed using anti-GFP or anti-HA antibody as indicated. Asterisks indicate unspecific bands. The blot stained with Coomassie blue (CBB) is presented to show loading. 7-10 ug of protein were loaded on blots probed with anti-GFP antibody and 30 ug of protein were loaded on the blot probed with anti-HA antibody. Chlorophyll amount (µg) is shown for each fraction to indicate the plastid enrichment. Similar results were observed in two or more independent experiments. Western blots of the same total and plastid extracts were also probed with anti-BiP2 to assess the level of ER contamination in plastid fractions. For the blots containing variants VIII, IXb, and IXc, the lanes containing the size marker between variants were cropped from the images. For the panels displaying variants VIII, IXb, and IXc, the anti-GFP and anti-BiP2 blots were yielded from separate SDS-PAGE gels.
© 2020 The Authors.   cytoplasm and nuclei. Western blot analysis indicated that the GFP was not cleaved from the expressed fusion proteins ( Figure 3f). These results suggest that AZI1's PRR and 8CM regions are required together for the direct or indirect association with MTs.

Microtubules are dispensable for flg22-induced enrichment of AZI1 to plastids
To dissect which cellular components are needed for AZI1 targeting, we began by identifying a defined, strong defense-inducing stimulus other than pathogen infection Figure 4. Impact of MAP kinase mutations and cytoskeleton inhibitors on AZI1/EARLI plastid targeting during defense signaling. (a) Western blot of total extracts from WT Col-0 Arabidopsis leaves to test AZI1-EARLI1 protein levels at different times post infiltration with 1 µM flg22 or water (mock). (b) AZI1-EARLI1 levels in total and plastid fractions from Arabidopsis leaves of WT Col-0, mpk3, and mpk6 plants 12 h post treatment with 1 µM flg22 (+) or water (mock, À). Shown in the graph is quantitation of AZI1-EARLI1 levels relative to the total protein content in each Coomassie blue membrane lane, as quantified by densitometry. The highest value in total or plastid fractions was set to 100 in the relative units. (c) AZI1-EARLI1 protein levels in total and plastid fraction from Arabidopsis WT Col-0 leaves 6 h post treatment with 1 µM flg22 and 3 h post-infiltration with inhibitors of actin filaments (Latrunculine B, LatB), microtubules (oryzalin, Ozn) or mock. Right graph: ratio of AZI1/EARLI1 in plastids versus total in different treatments, as quantified by densitometry. Western blot panels separated with vertical line belong to the same blot. (d) Actin and microtubule cytoskeleton changes in Arabidopsis after inhibitor treatments. Col/Lifeact-GFP and Col-gl1/GFP-TAU6 transgenic plants expressing actin or microtubule markers, respectively, were imaged by laser-scanning confocal microscopy at 6 h post treatment with 1 µM flg22 and 3 h post-infiltration with inhibitors of actin filaments (Latrunculine B), microtubules (oryzalin) or mock. The representative micrographs shown are Z-series maximum intensity projections. Micrographs show GFP signal (greyscale). Bar = 20µm. (e) AZI1-EARLI1 protein levels in total and plastid fractions from untreated Arabidopsis leaves of acd6-1 and acd6-1mpk3-1 mutant plants. Right graph: plastid to total ratio of AZI1-EARLI1 levels relative to the total protein content in each Coomassie blue membrane lane, as quantified by densitometry. The averages AE SE from three (b,c) or four (e) independent experiments are shown. In (b,c), different letters (lower or uppercase) indicate statistically significant differences (P < 0.01, analysis of variance (ANOVA), SNK test (b) or Tukey test (c)). The dotted line indicates that the total and plastid fractions were analyzed independently in (b). In the graph in (e), the asterisk indicates statistically significant differences determined by t-test (*P < 0.05, n = 4). In (b-d) the individual data points are shown on the bar chart as scatter-dots. In (a-c,e) the blots stained with Coomassie blue (CBB) are presented to show loading. Bands were revealed using anti-AZI1/EARLI1 polyclonal serum. Red arrow indicates the AZI1/EARLI1 monomer band. Asterisks indicate unspecific bands.
© 2020 The Authors.  (Cecchini et al., 2015b) that might cause enrichment of AZI1 to plastids. Since AZI1 and its paralogue EARLI1 are also needed for mSAR induction, we measured AZI1/EARLI1 levels in total and plastid extracts after flg22 treatment.
We infiltrated Arabidopsis WT plants with flg22 or H 2 O (mock) and analyzed the native AZI1/EARLI1 levels in total extracts at different times post-treatment by Western blot. Figure 4(a) shows that between 3-6 h post treatment (hpt), the total amount of AZI1 greatly increased in response to flg22 (+) compared to mock (À). Local AZI1/EARLI1 induction by flg22 treatment was also pronounced at 12 hpt (Figure 4b, total -WT plants). To determine if this AZI1 increase translated to higher levels in plastids, we treated Arabidopsis with flg22 and analyzed the levels of AZI1/ EARLI1 in total and plastid fractions at 6 and 12 hpt (Figure 4b; Figure S2a, WT plants). The amount of plastid-localized AZI1/EARLI1 increased in flg22-treated samples compared to mock; at 12 hpt there was a greater fold increase in the amount of AZI1/EARLI1 targeted to plastids (~109) relative to the fold increase in the total extract (~39). Thus, flg22 MAMP treatment strongly induces AZI1/ EARLI1 protein levels and increases their relative enrichment in the plastid fraction.
Considering the above results and the role of cytoskeletal dynamics for protein trafficking, we next studied the importance of the cytoskeleton for AZI1 targeting to plastids during MAMP stimulation. We quantified AZI1/EARLI1 protein levels in Arabidopsis WT Col-0 total and plastid fractions 6 hpt with flg22 and 3 h post-infiltration with or without inhibitors for microtubules (Ozn) and actin filaments (LatB), respectively. This set up allowed us to make the inhibitor treatments before the earliest robust increase of AZI1 in plastids ( Figure S2a; Figure 4a). As shown in Figure 4(c), Ozn or LatB treatments did not affect the plastid AZI1 levels compared to mock. As efficacy controls for the inhibitor treatments, we used the same treatment conditions to study transgenic plants expressing Col/Lifeact-GFP and Col-gl1/GFP-TUA6, actin or microtubules markers, respectively. Both inhibitor treatments strongly disrupted the respective AF and MT cytoskeletons (Figure 4d).
These results suggest that although AZI1 can traffic in close association with MTs, disruption of the cytoskeleton does not significantly impact AZI1 targeting to plastids.
MPK3 and MPK6 enhance AZI1/EARLI1 plastid targeting during defense induction AZI1's PRR is a proposed phosphorylation target for MPK3 and possibly MPK6 (Pitzschke et al., 2014), two key signaling factors associated with biotic and abiotic stress (Rodriguez et al., 2010). Moreover, we recently showed that these kinases are also needed for underground AZA-induced priming (Cecchini et al., 2019). Since the PRR is required for plastid targeting (Figure 2), we analyzed if MPK3/6 affect plastid localization of AZI1/EARLI1 proteins.
We treated WT Col-0, mpk3 and mpk6 mutant plants with flg22 or H 2 O (mock) and analyzed the levels of native AZI1/EARLI1 in plastids fractions compared to total extracts (Figure 4b; Figure S2a). Compared to WT, the mpk3 and mpk6 plants displayed decreased levels of AZI1/EARLI1 specifically in plastid fractions 12 h after flg22 treatment, but not in mock-treated plastid fractions or any total extracts (Figure 4b). In addition, we also observed a reduction of AZI1/EARLI1 in mpk6 plastids fraction at 6 hpt (Figure S2a). Immunoblots with phospho-antibody show that MPK3/4/6 were still active at both 6 and 12 h post flg22 treatment compared with mock treated plants ( Figure S2b).
Together, these results strongly suggest that both MPK3 and MPK6 are required for robust AZI1 plastid enrichment during flg22-MAMP defense induction. Supporting this idea, acd6-1, a plant with constitutively active MPK3 and pattern receptor-mediated immunity , showed a reduced proportion of AZI1/EARLI1 in the plastid fraction relative to the total extract when an mpk3 mutation was present (Figure 4e). The lower total levels of AZI1 in acd6-1 versus acd6-1mpk3 may be due to an altered balance of signaling molecules/hormones including salicylic acid, which is affected by both mutations (Vanacker et al, 2001;Zhang et al., 2014).

MPK3 and MPK6 are required for mSAR induction in Arabidopsis
Considering the requirement of AZI1 and EARLI1 for mSAR (Cecchini et al., 2015b), we next analyzed if MPK3 and MPK6 are also needed for this systemic defense These data indicate MPK3/6 are required for mSAR induction, possibly related to their effect on AZI1/EARLI's plastid targeting. Because MPK3 and MPK6 are also required for SAR and ISR (Beckers et al., 2009;Cecchini et al., 2019), our data strongly indicate a key role for these kinases in plant systemic resistance programs.

AZI1, MPK3 and MPK6 co-localize at sites of plastid-ER contacts
It was previously suggested that MPK3 interacts with AZI1 and forms protein complexes near the cell boundary  (Pitzschke et al., 2014). To get deeper insight into the subcellular localization and possible complex formation between AZI1 and MPK3/6, we analyzed their co-localization by confocal microscopy.
We co-expressed MPK3:RFP or MPK6:RFP together with AZI1:GFP by agro-transformation of N. benthamiana leaves. As shown in Figure 6(a), both MPK3:RFP and MPK6:RFP localized to the nucleus in a spherical pattern and partially co-localized with AZI1:GFP at perinuclear ER and plastid/stromule-ER contact sites. Some co-localization signals are likely to be plastid-ER and/or stromule-ER contact sites (Figure 6a; bottom panels) (Cecchini et al., 2015b). No co-localization was found between control soluble GFP and MPK3:RFP or MPK6:RFP at those places (Figure 6b). As expected, control GFP and MPKs co-localize at nuclei and cytoplasm. Immunoblots to detect MPK3 and MPK6 fusion proteins established that free RFP was not detectable (Figure 6c,d).
Next, we assessed putative interactions between AZI1 and MPK3 during defense induction. Immunoprecipitation (IP) assays were implemented by using Arabidopsis that expressed functional HA-tagged AZI1 previously shown to complement azi1 (Cecchini et al., 2015b). AZI1:HA-expressing plants treated with flg22 or H 2 O (mock) for 6 hpt were used for immunoprecipitation with an anti-HA matrix. MPK3 did not co-precipitate with AZI1:HA in either mockor flg22-treated samples (Figure 6e).
Altogether, these results indicate that although they partially co-localize, AZI1 does not form stable complexes with MPK3 in Arabidopsis under our assay conditions.

DISCUSSION
Not much information exists about the subcellular targeting/trafficking regulation of immune components (Khaled et al., 2015;Wang et al., 2016;Gu et al., 2017). AZI1, a key factor for the plant systemic defenses, shows dynamic levels in plastid envelopes (Jung et al., 2009;Cecchini et al., 2015b). Interestingly, AZI1 does not have a recognizable "classic" plastid targeting signal. Here, we discovered and characterized a N-terminal bipartite signature that explains AZI1's subcellular location: a predicted signal peptide followed by a PRR (Cecchini et al., 2015b). AZI1's putative signal peptide is a non-cleavable TMD that anchors the protein to membranes, placing this LTP-like protein as a signal-anchored protein. In addition, unlike what was found for other signal-anchored proteins, the presence of a PRR after the TMD (and not the TMD hydrophobicity (Kim and Hwang, 2013)) is required for plastid targeting. Together, our data establishe AZI1 as belonging to a previously undescribed class of signal-anchored proteins. Remarkably, the defense-associated kinases MPK3 and MPK6 are required for the increased targeting of AZI1 to plastids in response to flg22. Moreover, our results suggest that AZI1 utilizes the MT network for intracellular trafficking. Therefore, we propose that after pathogen or MAMP recognition, MPK3/6 are locally stimulated and, by acting on its N-terminal bipartite signal, enhance AZI1 plastid targeting. This in turn could determine how much systemic defense priming ensues.

AZI1 is a variant of signal-anchored proteins
AZI1 is a signal-anchored protein targeted to plastids. This finding explains our previous work showing that AZI1 is present in microsomal fractions (Cecchini et al., 2015b). Supporting this, other reports showed no cleavage of the AZI1 putative signal peptide (Zhang and Schl€ appi, 2007;Pitzschke et al., 2016). Because traces of AZI1 TMD:GFP were detected in the soluble fraction (AZI1 D38-161 and AZI1 C28/30A D38-161 ), we cannot completely rule out that a very small amount of AZI1 is capable of entering the secretory pathway. In agreement with this possibility, using protoplasts and root exudates, the localization of a small proportion of AZI1 was consistent with its being secreted to plant cell walls (Zhang and Schl€ appi, 2007;Pitzschke et al., 2016). Alternatively, soluble traces of AZI1 TMD:GFP in our experiments might be contamination from the microsomal extraction fractionation.
Most signal-anchored proteins targeted to plastids have relatively low TMD hydrophobicity indices (Lee et al., 2011). In contrast, AZI1's TMD hydrophobicity index is higher and its targeting to/co-fractionation with plastids depends on the presence of the PRR (Cecchini et al., 2015b;this work). It seems possible that the PRR substitutes for the requirement for a low degree of hydrophobicity within the TMD as the factor for plastid targeting. Interestingly, the PRR has features that are shared with bona fide chloroplastic transit peptides (cTPs). These include a large number of serine, arginine and proline residues, and overall low peptide complexity (targetP, (Emanuelsson et al., 2000); (Bruce, 2000;Lee et al., 2018)). Thus, AZI1 may use a chimeric targeting mechanism with components shared by signal anchor-and cTP-containing proteins. In this scenario, proteins required for both signal anchor and cTP targeting may also be involved in AZI1 targeting.
Interestingly, an algorithm trained to find apicoplast (plastid analogue in apicomplexan organisms) protein targeting signals, a putative SP+cTP (Zuegge et al., 2001), correctly predicts AZI1's plastid localization (Cecchini et al., 2015b). In addition, because transit peptides/presequences and the AZI1 PRR exhibit a high content of positively charged residues (Mackenzie, 2005), the PRR may act as a CPR region (albeit the PRR is longer than the CPRs in signal-anchored proteins). This would explain our findings that deletion of the AZI1 CPR does not strongly affect AZI1's targeting.
AZI1's predicted S-acylation sites (Ren et al., 2008) are conserved in most HyPRP members, suggesting an important role for these residues. However, mutation of these sites did not affect AZI1's plastid localization. S-acylation often acts as a mechanism to fine-tune proteins (re) localization and/or stability (Hemsley, 2015;Daniotti et al., 2017). Thus, small differences in AZI1 trafficking dynamics may have been missed in our experiments. Alternatively, the predicted acylation sites (cysteine residues) may be involved in the formation of disulphide bonds and the formation of AZI1-complexes, as previously suggested (Zhang and Schl€ appi, 2007;Cecchini et al., 2015b).
MPK3/6 are needed for defense-related AZI1 plastid targeting and mSAR As observed after stimulation of R protein-mediated defenses, MAMP treatment greatly induces the fraction of AZI1 localized to plastids, probably reflecting shared signals and/or components between the SAR and mSAR programs (Mishina and Zeier, 2007;Cecchini et al., 2015b). The kinases MPK3 and MPK6 are known shared factors between R-and PRR-induced systemic responses and priming (Beckers et al., 2009;Rodriguez et al., 2010;Cecchini et al., 2019). Remarkably, MPK3 and MPK6 specifically impact AZI1 targeting and are required for the increased accumulation of AZI1 at plastids. Thus, it is possible that MPK3/6 act in systemic defense programs, at least in part, by regulating AZI1's increased levels in plastids, which in turn may affect the movement of systemic defense priming signals such as AZA or glycerol-3-phosphate (Yu et al., 2013;Cecchini et al., 2015b;Cecchini et al., 2019). Moreover, MPK3 and MPK6 can promote plastid-associated ROS production (Su et al., 2018), which play an important role in SAR and could facilitate the generation of AZA in this organelle (Zoeller et al., 2012;Wang et al., 2014). Considering that MPK3 and MPK6 are activated in response to diverse abiotic stresses (Rodriguez et al., 2010;Jalmi and Sinha, 2015), it is also possible that AZI1 is required in plastids during other environmental conditions. We speculate that in these situations, AZI1 could facilitate the movement of other signals, like other plastid oxylipins related to stress resistance (Bl ee, 2002;Prost et al., 2005;Breuers et al., 2011).
How might MPK3/6 regulate AZI1's abundance in plastids? Phosphorylation of cTP residues by STY kinases regulates chloroplast import rate (Martin et al., 2006;Lamberti et al., 2011). Thus, one idea is that MPK3/6 might impact AZI1 targeting by directly phosphorylating the PRR. In support of this, AZI1's PRR has several predicted MAPK-phosphorylation sites and MPK3 can phosphorylate the PRR in vitro (Pitzschke et al., 2014). Previous work indicated that AZI1 might interact with MPK3 in an in planta heterologous system (Pitzschke et al., 2014). However, although AZI1 and MPK3/6 partially co-localized, stable complexes containing AZI1 and MPK3 in Arabidopsis were not detectable by co-immunoprecipitation. Labile and/or short-lived interactions between AZI1-MPK3/MPK6 may not permit coimmunoprecipitation in our conditions. Since MPKs are thought to have many substrates, it is not surprising that stable complexes were not detectable Rayapuram et al., 2018).

MTs and AZI1 trafficking
Surprisingly, the loss of the N-terminal TMD causes AZI1 to stably co-localize with microtubules. AZI1's PRR and 8CM domain together may directly or indirectly interact with MT filaments, suggesting that AZI1 uses them for intracellular trafficking. Supporting this idea, full length AZI1-GFP moves along (or very close to) MT bundles. It is unlikely that AZI1-MT interactions guide AZI1 to plastids, since MT inhibitor treatment did not affect the abundance of AZI1 in plastids. MTs play a key role for the extension of stromules (Erickson et al., 2017;Erickson and Schattat, 2018;Kumar et al., 2018), plastid structures where AZI1 also resides (Cecchini et al., 2015b). Both stromules and AZI1 enrichment in plastids are highly induced with similar kinetics by flg22 ( (Caplan et al., 2015) and this work). One possibility is that plastid-anchored AZI1 in close proximity to MTs might help stromules extend to and/or make contact with other structures in cells. It is also possible that the MT network is required for AZI1 to target non-plastid subcellular sites like plasmodesmata (PD), which was suggested to be required for AZA transport and ultimately systemic resistance (Cecchini et al., 2015b). The localization of AZI1 to PD during defense signaling may indicate the importance of AZI1's trafficking among distinct subcellular compartments for SAR and priming establishment (Lim et al., 2016). Since AZI1 is also required for the action of other systemic signals generated in plastids like glycerol-3phosphate, dehydroabietinal and pinene-monoterpenes (Chaturvedi et al., 2012;Yu et al., 2013;Riedlmeier et al., 2017), it is possible that it also allows small metabolites to reach PDs for simplastic movement. Future work will address these possibilities.
In summary, AZI1 uses a previously undescribed variant of the signal anchor proteins mechanism to target plastids. This targeting and/or stability in the plastid pool of AZI1 is impacted by two key defense associated kinases, MPK3 and MPK6, probably for a tight modulation during defense induction against pathogens and possible other stresses. This could be especially important considering how fundamental plastids are for many types of defense responses (Caplan et al., 2008;Grant and Jones, 2009;Nomura et al., 2012;Zeier, 2013;de Torres Zabala et al., 2015;Cecchini et al., 2015a;Medina-Puche et al., 2020). In particular, a timely regulation of the AZI1 pool in plastids could grant an efficient movement of AZA for the establishment of systemic defenses. Additionally, since several HyPRPs contain an N-terminal bipartite signal, this targeting mechanism could also be functional for all of them. It is also © 2020 The Authors. conceivable that other non-HyPRP plant proteins also use a similar N-terminal targeting signal. Supporting this idea, several signal-anchored proteins with high TMD hydrophobicity and predicted to be retained in the ER, were reported to be plastid-localized (Lee et al., 2011).
Collectively, this work strengthens the idea that subcellular re-localization of defense components is a key aspect of plant immune signaling, especially with regard to the phenomena of priming, in which a well-placed ambush can be the difference between disease and death.

Vectors and constructs
All primers and vectors used in this study are described in Table S1. The vectors and constructs carrying full length AZI1 as well as the deletion AZI1-variants II. and VI. were previously described (Cecchini et al., 2015b). The AZI1 deletion and point mutation variants as well as the MPK3 and MPK6 full coding region were cloned from A. thaliana cDNA using PCR primers linked to specific sequences compatible with the TOPO pENTR â and GATEWAY â cloning procedure, and introduced into the plant expression vector pBAV150 (Vinatzer et al., 2006). MPK3 and MPK6 were also introduced into pSITE-4NA (CD3-1642/ABRC (Nelson et al., 2007)). To generate internal deletion and point mutation constructs the proper primers were used to amplified the fragments to be linked by PCR (Table S2). When required, a start codon was added in the forward primers. All these constructs permitted the expression of the transgenes with C-terminal GFP or HA-tag (pBAV150/pBAV154; (Vinatzer et al., 2006)) controlled by the Dex-inducible promoter or with C-terminal mRFP1 driven by the 35S promoter (pSITE-4NA). The vector used for the expression of the actin cytoskeleton marker Lifeact-GFP was previously described (Smertenko et al., 2010). The microtubule marker RFP-TUB6 construct was kindly provided by Dr. Wasteneys (Ambrose et al., 2011). All vectors, constructs and primers used in this study are listed in Table S2.

Subcellular localization
For localization studies, 4-week-old N. benthamiana leaves were infiltrated with Agrobacterium tumefaciens C58C1 or GV3101 strains carrying the different constructs. In order to express fusion proteins from the Dex-inducible vectors, 20 or 30 lM Dexamethasone (D4902; Sigma-Aldrich, St. Louis, MO, USA) was infiltrated into leaves 1 day after agroinfiltration. Fractionation and confocal microscopy studies were done 21 h after Dex. Agrobacterium harboring different constructs were infiltrated together for co-expression analysis. N. benthamiana leaves were prepared for microscopy as described (Littlejohn et al., 2010). A Zeiss LSM710 laser-scanning confocal microscope (Carl Zeiss Microscopy GmbH, Jena, Germany) was used to capture GFP fluorescence (excitation: 488 nm; emission: 505-530 nm), RFP fluorescence (excitation: 561 nm; emission: 570-620 nm) and plastid autofluorescence (excitation: 633, emission: 650-750 nm). Images were taken using a LD C-Apochromat 409/1.1 W Korr objective using a sequential acquisition mode. For time series acquisition, images were taken at low resolution scanning and maximum speed mode. Images were processed using ImageJ (http://rsb.inf o.nih.gov/ij), ZEN 2012 (Zeiss) and Adobe Photoshop software.

Quantification of fluorescence
For the analyses of GFP fluorescent intensity of AZI1 and variants in plastids, confocal images were captured using two channels: plastid autofluorescence (blue, 650-750 nm) and GFP fluorescence (green, 505-530 nm) as described above. Using the Fiji software (https://imagej.net/), individual plastids were identified in the blue channel and a spherical region of interest (ROI) was drawn around each plastid such that it incorporated the entire region of autofluorescence. The intensity of the GFP signal within the ROI was measured. Four or more images from independent experiments were analyzed for each construct and 4-5 plastids were quantified per image to calculate the mean plastid GFP fluorescence AE SE. To test the significance across values, one-way ANOVA followed by Tukey's HSD post hoc tests were performed. For the analyses of GFP signal that colocalized with RFP-TUB6, confocal image channels for GFP (green, 505-530 nm) and RFP fluorescence (red, 570-620 nm nm) were used. The ImageJ Colocalization plugin (http:// rsb.info.nih.gov/ij/plugins/colocalization.html) was used to generate ROIs from GFP/RFP colocalized signal regions (using thresholds = 50% and ratio = 50%). The intensity of the total GFP signal and within the ROI was measured for each image. Six or more images from independent experiments were analyzed to calculate the percentage of GFP fluorescence AE SE that colocalize with RFP-TUB6 with respect to total. To test the significance, one-way ANOVA followed by Tukey's HSD post hoc tests was performed.

Fractionation
Microsomal fractions were obtained as described in . 1 g of N. benthamiana leaf tissue expressing the different constructs was used for the fractionation. Plastids were isolated from 1 g of Arabidopsis or N. benthamiana leaves following the protocol described in (Cecchini et al., 2015b). Chlorophyll content was measured by spectrophotometric analysis as described in Lamppa (1995). For N. benthamiana transiently expressing AZI1 deletion variants V, VIII, IXc, and X, plastids were isolated from 1 g of leaves following the protocol described in Cecchini et al. (2015b) with a slight modification. The total/impure plastids were separated at 4°C on two percoll gradients (80-40%), initially in 2 ml microfuge tubes then again in 4 ml polycarbonate tubes. To enrich for plastid envelopes, purified plastids were frozen in liquid nitrogen to lyse membranes then thawed on ice. Thawed samples were centrifuged at 13 000g, the supernatant was collected, then the pellet was resuspended in ice-cold dH 2 O (with protease inhibitor cocktail) and frozen at À80°C o/n to further lyse membranes. The envelope fraction was again centrifuged at 13 000g, the supernatant fraction was collected, then the pellet was © 2020 The Authors.

flg22 treatment and systemic resistance induction
To analyze AZI1/EARLI1 levels and localization after flg22 peptide treatment, Arabidopsis WT and mpk3 and mpk6 mutant leaves were infiltrated with 1 µM flg22 or mock (H 2 O). Total or plastid fraction protein samples were obtained after different times postinfiltration as indicated in Figure legends. Plants were treated soon after lights were turned on in the morning.

Exogenous application of actin and microtubules inhibitors
Solutions of Latrunculin B (20 µM, Sigma-Aldrich) or oryzalin (40 µM, Sigma-Aldrich) inhibitors were directly infiltrated into leaves with a needleless syringe. In Arabidopsis, treated leaves were analyzed by microscopy or total and plastid fraction proteins isolated after 3 h. In N. benthamiana, discs were immediately cut after leaves' infiltration and floated in the inhibitor solutions for 16-18 h before the microscopy analysis.

Statistical analysis
Analyses in this study were done with the software SigmaPlot v11.0 (Systat Software, Inc.). ANOVA (log-transformed data for bacterial growth curves) followed by the Tukey or Newman-Keuls (SNK) post hoc test and one-tailed Student's t-test were used as indicated in Figure legends. Figure S1. Quantification of the AZI1:GFP that colocalize with RFP-TUB6 and anti-tubulin Western blot of total and plastid fractions from N. benthamiana expressing AZI1 D2-25 :GFP (X), AZI1 D2-30 :GFP (XI), or mock-treated. Figure S2. Flg22 treatment affects the levels of AZI1/EARLI1 and phosphorylation state of MPK3/MPK6 in mpk3 and mpk6 mutants. Table S1. A summary of the in vivo microscopy and fractionation plastid association data for AZI1 variants. Table S2. Vectors, constructs and primers list used in this study.