The hnRNP-Q Protein LIF2 Participates in the Plant Immune Response

Eukaryotes have evolved complex defense pathways to combat invading pathogens. Here, we investigated the role of the Arabidopsis thaliana heterogeneous nuclear ribonucleoprotein (hnRNP-Q) LIF2 in the plant innate immune response. We show that LIF2 loss-of-function in A. thaliana leads to changes in the basal expression of the salicylic acid (SA)- and jasmonic acid (JA)- dependent defense marker genes PR1 and PDF1.2, respectively. Whereas the expression of genes involved in SA and JA biosynthesis and signaling was also affected in the lif2-1 mutant, no change in SA and JA hormonal contents was detected. In addition, the composition of glucosinolates, a class of defense-related secondary metabolites, was altered in the lif2-1 mutant in the absence of pathogen challenge. The lif2-1 mutant exhibited reduced susceptibility to the hemi-biotrophic pathogen Pseudomonas syringae and the necrotrophic ascomycete Botrytis cinerea. Furthermore, the lif2-1 sid2-2 double mutant was less susceptible than the wild type to P. syringae infection, suggesting that the lif2 response to pathogens was independent of SA accumulation. Together, our data suggest that lif2-1 exhibits a basal primed defense state, resulting from complex deregulation of gene expression, which leads to increased resistance to pathogens with various infection strategies. Therefore, LIF2 may function as a suppressor of cell-autonomous immunity. Similar to its human homolog, NSAP1/SYNCRIP, a trans-acting factor involved in both cellular processes and the viral life cycle, LIF2 may regulate the conflicting aspects of development and defense programs, suggesting that a conserved evolutionary trade-off between growth and defense pathways exists in eukaryotes.


Introduction
Living organisms are exposed to various abiotic and biotic stresses and continuously integrate diverse signals to limit the spread of microbial pathogens. RNA regulatory processes, which modulate transcriptional programs and post-transcriptional events in response to various cues, are an important component of the tolerance and adaptation strategies of organisms. Thus, RNAbinding proteins (RBPs), which are involved in various aspects of RNA processing (i.e., mRNA maturation, editing, splicing, and mRNA trafficking), are thought to be key regulators of stress responses in both animals and plants. However, this possibility has been scarcely studied.
The heterogeneous nuclear ribonucleoprotein (hnRNP) group of RBPs forms a large family of ancestral modular proteins with a high degree of functional diversification [1]. In humans, this group contains 37 members, only a few of which are involved in the stress response. For instance, hnRNP-K participates in the response to genotoxic stress [2]. The human NS1-associated protein 1/ synaptotagmin-binding cytoplasmic RNA-interacting protein (NSAP1/SYNCRIP), another cellular hnRNP protein, participates in the translational activation of both cellular and viral RNAs and functions in the Hepatitis C virus (HCV) life cycle by interacting with HCV RNA [3,4,5]. Knock-down of NSAP1/ SYNCRIP significantly decreased the amount of HCV RNA in mammalian cells, suggesting that NSAP1/SYNCRIP is a negative regulator of viral defense responses. Recently, hnRNP-I was shown to interact with the long non-coding RNA RoR to modulate the expression of the tumor suppressor p53 in response to DNA damage and to participate in a surveillance network [6].
Plants contain more RBPs than do animals, but the number of plant hnRNPs is similar in both plants and animals. The functions of hnRNPs remain poorly described in plants [1,7]. Few studies have reported a role for plant hnRNPs in the stress response [8]. Until now, the glycine-rich RBP AtGRP7 was the only hnRNP known to play a role in the plant's response to pathogens [9]. AtGRP7 is the substrate of the type III effector HopU1, which is injected by the bacterial pathogen Pseudomonas syringae into plant host cells. HopU1 mono-ADP ribosylates a conserved arginine residue of AtGRP7, preventing binding to immunity-related RNA [10,11]. In a previous study, we identified LIF2, a novel hnRNP of the cruciferous plant Arabidopsis thaliana [12]. LIF2 is a nucleocytoplasmic RBP that contains three RNA Recognition Motifs (RRMs), the most frequent RNA-binding domain in the hnRNP family. LIF2 and its close homologs, AtLIL1 and AtLIL2, belong to the hnRNP-Q subfamily and LIF2 is structurally homologous to human NSAP1/SYNCRIP [12]. LIF2 interacts in vivo with LHP1, a Polycomb Repressive Complex1 (PRC1) subunit [13,14]. LIF2 is involved in the maintenance of plant cell identity and cell fate decision [12]. Indeed, loss-of-function of LIF2 affected various aspects of growth and development, such as flowering time and leaf size. More dramatically, lif2 mutations induced indeterminate growth of ovaries, resulting in the formation of ectopic inflorescences bearing severely affected flowers. Besides its developmental functions, LIF2 might be involved in the stress response. Indeed, a large set of stress-related genes (193) was found to be deregulated in lif2-1. The set of common genes deregulated in both lif2-1 and lhp1 mutants was even more enriched in stress-related genes than was that in lif2-1 [12]. Furthermore, loss-of-function of LHP1 (also named TFL2 or TU8) altered the glucosinolate profile and reduced symptoms in response to infection by the obligate biotrophic fungus Plasmodiophora brassicae, which causes clubroot disease, a damaging disease in Brassicaceae [15,16].
These data suggest that LIF2 and LHP1 have common functions in the stress response and prompted us to investigate the function of LIF2 in biotic stress responses. Here, we show that lif2 mutations conferred altered susceptibility to the necrotrophic fungal pathogens Sclerotinia sclerotiorum (S. sclerotiorum) and Botrytis cinerea (B. cinerea) and to the hemi-biotrophic bacterial pathogen P. syringae pv. tomato (Pst), three biological agents with substantial impacts on the agronomical production of various plant species [17,18,19]. To better understand the altered pathogen susceptibility of the lif2 mutants, we investigated the expression of defense marker genes regulated by jasmonic acid (JA) and salicylic acid (SA), two critical signaling hormones in the activation of plant defense, as well as the production of stress-associated metabolites and hormones. Finally, we showed that the transcriptomic profiles of stress-response factors in lif2 are specific and we identified key defense regulators (such as the WRKY18 and WRKY33 transcription factors, which are known to be involved in B. cinerea susceptibility), whose deregulation may contribute to the observed pathogen response. Together, our data suggest that the LIF2 hnRNP-Q may suppress the plant immune response by an unknown SA-independent pathway. Given that the human homolog of LIF2, the NSAP1/SYNCRIP protein, is a trans-acting factor involved in both cellular processes and the viral life cycle [3,4,5], we propose that the conserved hnRNP-Q proteins may have an evolutionary conserved function in regulating the trade-off between growth and defense in eukaryotes.

Results
Analysis of the lif2-1 transcriptome reveals a potential stress-related function for LIF2 In a previous transcriptome profiling experiment, we showed that genes that were deregulated in the A. thaliana lif2-1 null mutant were greatly enriched in Gene Ontology (GO) terms involved in responses to stress stimuli [12]. Here, we further examined the deregulated gene set (1008 genes) using Singular Enrichment Analysis (SEA) implemented in the agriGO toolkit [20]. This analysis revealed 293 deregulated genes associated with the GO term ''response to stimulus'' and 193 with ''response to stress'' (Table 1). We noticed that the normed frequency (NF) of the ''response to biotic stimulus'' (GO:000960, NF 4.22, p-value 7610 223 ) was higher than that of the ''response to abiotic stimulus'' (GO:0009628, NF 2.75, p-value 2610 216 ) ( Table 1). Furthermore, several genes were associated with GO terms related to the JA-defense signaling pathway, the glucosinolate metabolic pathway, and responses to fungal and bacterial pathogens ( Table 1). These data suggest that even if LIF2 is induced only weakly upon pathogen infection (Fig. S1), this gene might be involved in defense responses to various bioaggressors.
JA and SA pathways are altered in the lif2-1 mutant Based on the lif2 transcriptome data [12], we further analyzed the expression of a key marker gene of the JA-mediated defense, PDF1.2 (a JA-responsive gene), and the expression of genes of the JA biosynthesis and signaling pathways, LOX3, AOS, AOC, and OPR3 (genes encoding enzymes of the JA biosynthesis pathway), JAR1 (encoding an enzyme that converts JA to the bioactive JA-Ile molecule), and COI1 (encoding an F-box subunit of the JAreceptor complex) [21] (Fig. 1A). All of these genes were downregulated in lif2-1, suggesting that the JA defense pathway is globally repressed in the lif2-1 mutant.
Since the JA-dependent defense network cross-communicates with the SA-dependent signaling pathway to fine-tune the plant's defense response [22,23], we assessed the expression of ICS1/ SID2, which is involved in SA biosynthesis; ENHANCED DISEASE SUSCEPTIBILITY1 (EDS1) and PAD4, which are involved in the SA-signaling pathway; and NON-EXPRESSOR OF PR GENES 1 (NPR1) and PATHOGENESIS-RELATED 1 (PR1), which are important positive regulators of SA responses [24,25]. EDS1, PAD4, and ICS1 were upregulated in the lif2-1 mutant relative to wild-type plants (Fig. 1B). PR1 was also upregulated in lif2-1, whereas NPR1 expression was not significantly affected in lif2-1 (Fig. 1B). These data suggest that most genes involved in the SAmediated defense pathway are constitutively upregulated in the lif2 mutants, in the absence of pathogen. The upregulation of genes of the SA signaling pathway and the downregulation of genes of the JA signaling pathway in lif2 illustrates the antagonistic action of these two pathways.
Prompted by the finding that genes involved in JA-and SAdefense-related networks are differentially expressed in lif2-1, we quantified the endogenous levels of these hormones in lif2-1 using HPLC-electrospray-MS/MS (Fig. 1C). Surprisingly, whereas ICS1 was upregulated in lif2-1, free SA levels were similar in the mature rosette leaves of lif2-1 and wild-type plants (Fig. 1C). Similarly, the JA levels were not significantly affected in the lif2-1 mutant (Fig. 1C). Furthermore, we found that the levels of auxin (AIA) and abscissic acid (ABA), two other phytohormones recently shown to be involved in the stress response [26,27], were not significantly altered in lif2-1 in the absence of pathogen challenge (Fig. 1C).
The lif2 mutants are less susceptible to the bacterial pathogen Pseudomonas syringae As JA-and SA-defense-related gene expression were altered in the lif2 mutant ( Fig. 1A-B), we investigated the response of lif2 to inoculation with the hemi-biotrophic bacterial pathogen P. syringae [28]. The bacterial growth of the virulent DC3000 strain was 10 times lower in lif2-1 leaves than in wild-type leaves at 24 hours post-inoculation (hpi) ( Fig. 2A). Consistently, the disease symptoms were reduced on the leaves of two independent lines, lif2-1 and lif2-3, whereas severe symptoms were observed on the leaves of wild-type plants five days post-inoculation (dpi) (Fig. 2B). As expected, the lif2-c line, a lif2-1 mutant line complemented with LIF2 under the control of its own regulatory regions, was as susceptible as the wild-type plant (Fig. 2B). Similar results were obtained with the avirulent DC3000 avrRpm1 strain (Fig. 2C). These data are consistent with the observation that SA-related genes are constitutively expressed in the lif2 mutant.
In response to P. syringae infection, the JA-related biosynthesis genes were upregulated in wild-type plants and to a lesser extent in lif2-1 (Fig. 3), with the exception of JAR1, which was not induced. COI1, which encodes a component of the JA-receptor complex interacting with the bacterial phytotoxin coronatine, was upregulated in both genotypes (Fig. 3A). PDF1.2 was downregulated in wild-type plants and not activated upon infection in lif2. In contrast, genes involved in SA biosynthesis and signaling pathways were upregulated in both wild-type and lif2 plants upon DC3000 inoculation (Fig. 3B). PAD4 induction was weaker in lif2 than in wild-type plants, whereas PR1 expression was similar in the mutant and wild-type plants upon P. syringae infection. Therefore, The LIF2 RBP in Biotic Stress PLOS ONE | www.plosone.org despite a significant basal activation in the absence of pathogen, lif2 mutant plants were able to further activate SA-related defense genes in response to pathogen attack, and the response of the lif2 mutant to P. syringae infection differed from the wild type only in the level of defense gene expression. Together, these results indicate that LIF2 is not necessary for the activation of SA-and JA-related defense genes.
Response of the lif2 sid2 double mutant to the biotrophic P. syringae pathogen Since neither the high level of PR1 expression in the lif2 mutant nor the reduced susceptibility of lif2 to P. syringae infection were correlated with an increase in free SA content, we hypothesized that LIF2 may act downstream of SA production. To test this hypothesis, we crossed the lif2 mutant with the salicylic acid induction deficient2 (sid2/ics1) mutant.
The rosettes of sid2 lif2-1 double mutant plants were smaller than those of the two parental lines, sid2-2 and lif2-1, in both short-day (SD) and long-day (LD) conditions ( Fig. 4A and S2). Interestingly, in SD conditions, similarly to lif2-1 plants, the double mutant was early flowering (Fig. S2) and produced indeterminate ovaries (IDO) with an ectopic inflorescence, suggesting that the lif2-1 mutation is epistatic to sid2-2.
We then investigated the response of the lif2-1 sid2-2 double mutant to pathogens. Bacterial growth of the DC3000 strain was reduced at 24 hpi (by about 10-fold) in the lif2-1 sid2-2 rosette leaves compared with those of the wild type (Fig. 4B), similarly to lif2-1. These results suggest that the reduced susceptibility of lif2 is independent of SA biosynthesis. Interestingly, the level of PR1 expression in lif2-1 sid2-2 was similar to that of the wild type ( Fig. 4C), suggesting that the decreased susceptibility of lif2 may involve an SA-independent defense pathway.

Response of the lif2 mutant to a necrotrophic fungal pathogen
Since responses to pathogens are usually largely dependent on the pathogen lifestyle, we then assessed the defense response of lif2 to the necrotrophic fungal pathogen S. sclerotiorum (strain S55). We used two A. thaliana accessions as controls, Rubezhnoe-1 (Rbz-1) and Shahdara (Sha), which were previously shown to be resistant and susceptible to S. sclerotiorum, respectively [29]. After inoculation, the symptoms were stronger in lif2 than in wild-type plants ( Fig. 5A and B), revealing that lif2 is susceptible to S. sclerotiorum. This observation is consistent with the constitutive downregulation of JA-related genes in lif2 reported in this study, and the finding that the JA pathway is essential for resistance to this pathogen [29,30,31].
We then inoculated these plants with a second necrotrophic fungal pathogen, B. cinerea. Unexpectedly, we observed that lif2 was less susceptible to two virulent strains of B. cinerea, the wildtype virulent B0510 and Bd90 strains, the latter of which was less virulent (Fig. 6A-C). These data suggest that the immune response of lif2 is independent of the pathogen's lifestyle.
To decipher the molecular basis for the resistance of lif2 to B. cinerea, we assessed the expression of a number of regulators known to be involved in disease resistance to B. cinerea (Table 2). For instance, mutations in the MYB46 transcription factor gene [32], ATGXRS13 [33], or the basic helix-loop-helix MYC2 transcription factor [34] led to increased resistance to B. cinerea. In our transcriptome data, ATGXRS13 was upregulated in lif2, while MYB46 was not affected, suggesting that these two genes are probably not involved. However, we found that MYC2 was downregulated in lif2 (Fig. 6D), which is in agreement with the decreased susceptibility of lif2 to B. cinerea. We also evaluated the expression profiles of a few other candidate genes whose downregulation led to an increased susceptibility to B. cinerea (i.e., PAD2, CYP71B15/PAD3, BIK1, WRKY18, WRKY33, WRK40, and WRKY70) [35,36,37,38,39,40]. Interestingly, only the expression of WRKY18 and WRKY33 was significantly different in the two mutant alleles compared with wild-type or lif2-c plants.
The loss-of-function of CATALASE 2 (CAT2) triggers pathogen defense responses and resistance [41]. Furthermore, the inhibition of NADPH oxidase and of other flavoprotein enzymes involved in oxidative stress limits B. cinerea infection [42]. We found that CAT2 was downregulated in lif2, which is in agreement with the reduced susceptibility of this mutant to B. cinerea. Thus, CAT2 and the MYC2, WRKY18, and WRK33 transcription factors are interesting candidate proteins that participate in the lif2 response to B. cinerea.

Glucosinolate content is altered in the lif2 mutant
Due to the observed enrichment in the GO term ''glycosinolate metabolic process'' (GO:0019757) in the lif2-1 transcriptome (Table 1), and to the established role of glucosinolates (GLSs) and their breakdown products in plant defense [43], we extracted and quantified twenty-one of these secondary metabolites (b-thioglucoside-N-sulfated oximes), belonging to three main GLS families (aliphatic, indolic, and benzoate GLSs) in mutant and wild-type seedlings (Table S1). The GLSs were analyzed by negative electrospray ionization liquid chromatography coupled with mass spectrometry (ESI-HPLC-MS). Globally, the GLS content was lower in the lif2 mutant than in wild-type plantlets, with decreases observed in fourteen GLSs belonging to all three main GLS families (Fig. 7A). The levels of indol-3-ylmethyl glucosinolate (I3M), 4-methylsulfinylbutyl glucosinolate (4MSOB), 4methylthiobutyl glucosinolate (4MTB), and 5-methylthiopentyl glucosinolate (5MTP) were significantly decreased in the lif2 mutant. Interestingly, the indolic GLS family was less affected by the lif2 mutation, with only a decrease in indol-3-ylmethyl glucosinolate (glucobrassicin, I3M) being observed. As a control, we quantified the GLS content in the lhp1 mutant (Fig. 7B), which is known to have an altered glucosinolate profile [15]. In lhp1, significantly increased levels were observed for six GLSs (i.e., 8MSOO, 3MTP, 4MSOB, 5MSOP, 6MSOH, and 7MSOH), whereas only one GLS, the indolic 4MOI3M GLS, had decreased levels relative to the wild type. Similar results were obtained using lhp1 seeds [15]. Next, we quantified the levels of raphanusamic acid (RA), an important breakdown product of GLS that forms during the biotic stress response [44]. PEN2 myrosinase catalyzes the formation of RA from the I3M or 4MOI3M substrates. We observed a significant reduction in RA in both lif2 (0.91 ng/mg, n = 7-8, p,0.01) and lhp1 (0.89 ng/mg, n = 7-8, p,0.001) compared with wild-type plants (1.39 ng/mg).
Therefore, lif2 has a GLS profile that is distinct from that of wild-type and lhp1 seedlings. Thus, the control of GLS metabolism seems to be dependent on LIF2 and partially independent of the LHP1/LIF2 interaction [12]. Alterations to the GLS pathways may also contribute to the responses of the lif2 mutant to pathogens.

Molecular signature of the lif2 mutant
In an attempt to identify the key genes involved in the lif2 response to biotic stress, we further analyzed the expression of transcription factors (TFs), due to their prominent role in regulating gene expression. Among the 66 TFs deregulated in the lif2-1 transcriptome [12], we noticed that the APETALA2/ ethylene-responsive element-binding protein (AP2/EREBP), WRKY, MYB, and basic helix-loop-helix (bHLH) transcription factor families were overrepresented (Table S2) (Fig. 8A). We also noticed a strong bias towards downregulation of genes of these families (Fig. 1C). For example, 11 of the 13 deregulated genes of the AP2-EREBP family were downregulated in lif2 (Fig. 8). Out of the four subfamilies of AP2/EREBP (i.e., AP2, RAV (related to ABI3/VP1), dehydration-responsive element-binding protein (DREB), and ERF) [45,46], only the two largest ones, the DREB and ERF subfamilies, were deregulated in the lif2-1 mutant. The plant-specific WRKY transcription factors are key regulators of stress and plant immune responses [47], whereas the NAC TFs are involved in both development and the abiotic and/or biotic stress responses. Two stress-responsive NACs (SNACs) that were recently described [48,49] were deregulated in the lif2-1 transcriptome. In the MYB and bHLH families, MYC2, which encodes a key defense transcription factor involved in JA responses, was deregulated, (Fig. 6D) as well as the JA-ASSOCIATED MYC2 LIKE1 gene (JAM1).
Furthermore, we noticed that 6 genes associated with the GO term ''negative regulators of defense response'' (GO:0031348) were downregulated in the lif2-1 transcriptome. These genes might participate in the lif2 stress response (Fig. 8B). Among these genes, JAR1 represses the SA-dependent signaling pathway; PEN3, which is an ABC transporter, restricts pathogen proliferation in the hosts; and BAP1 is a negative regulator of programmed cell death and is involved in membrane trafficking in response to external conditions [50]. These data suggest that LIF2 affects a variety of defense-related pathways.
The 38 TFs that were both deregulated in the lif2-1 rosette transcriptome (Table S2) and belong to the AP2/EREBP, WRKY, NAC, MYB, or bHLH family, as well as the 6 negative regulators (Fig. 8B) represent a sort of molecular signature of the lif2 mutant phenotype. We thus wondered how these 44 genes were deregulated in wild-type genetic backgrounds in response to biotic stress. To address this question, we selected 62 transcriptome profiles from plants treated with pathogens compared with those that were not, in the Genevestigator database [51] and extracted the expression of these 44 genes. We then performed a hierarchical The LIF2 RBP in Biotic Stress PLOS ONE | www.plosone.org clustering analysis using of the MultiExperiment Viewer application [52], and identified three main gene clusters, which are characterized by specific co-expression profiles of MYC and MYB (Fig. 8C). The small Cluster I (containing MYC4/MYB28/MYB48) contained genes that are upregulated in lif2-1 and downregulated in response to most pathogen treatments. The large Cluster II (containing MYC2/MYB15) contained genes downregulated in lif2-1 and up-or downregulated in response to pathogen treatments, with no obvious link to the biology of the pathogens. Finally, Cluster III (containing MYC7E/MYB34) mainly contained genes repressed in lif2-1 and in response to most biotic stresses, except in response to P. syringae, and thus, which are probably associated with the JA pathway due to the presence of JAR1. In conclusion, our clustering analysis revealed that (i) the expression profiles of the 44 selected genes involved in the response to pathogens were diverse, (ii) the MYC and MYB genes were coexpressed in different combinations, and (iii) that lif2 exhibited a unique expression pattern, with opposite profiles to most of the 62 analyzed transcriptome profiles, which might be in line with its specific response to pathogens.

Discussion
Plants have evolved both constitutive and induced defense mechanisms to counteract pathogen attacks. Most of the induced mechanisms are based on the production of a complex repertoire of plant metabolic compounds and signaling hormones, which activate appropriate defense pathways, based on the nature of the pathogens [53,54]. Salicylic acid, a phenolic phytohormone, and the jasmonic acid phytohormone, a polyunsaturated fatty acid derived from a-linolenic acid, are two key molecules in plant defense mechanisms [55]. SA, which also possesses medicinal properties [24] and plays a central role in animal immunity [56], and its derivatives, have crucial roles in defense against biotrophic pathogens [57]. They also regulate various aspects of abiotic stress responses, plant growth, and development. The JA and SA defense signaling pathways have been shown to cross-communicate and to be mostly antagonistic, providing plants with a regulatory potential to fine-tune their defense reaction depending on the type of pathogen encountered [22,23]. SA and JA are mainly involved in the response to biotrophic and necrotrophic pathogens, respectively. However, complex crosstalk exists between the hormone signaling pathways involved in plant responses to pathogens [26]. Secondary metabolites, such as glucosinolates, are also involved in the biotic stress response in plants [43].
Only a few RBPs have emerged as regulators of the biotic stress response in plants [8,58,59,60]. For instance, AtRBP-DR1, a putative component of the resistance protein RPS2 complex, positively regulates defense responses mediated by SA in A. thaliana [61]. The conserved MOS2 protein, a putative RBP with both Gpatch and KWO domains, is required for resistance against P. syringae pv. maculicola ES4326 and Peronospora parasitica Emoy2 [62], whereas overexpression of the Gossypium hirsutum GhZFP1 protein, which contains a CCCH-type zinc finger present in some RBPs, has been shown to enhance both disease resistance to the Rhizoctonia solani fungus and salt stress tolerance [63]. Overexpression of the Capsicum annuum RNA-binding protein 1, CaRBP1, in A. thaliana conferred reduced susceptibility to infection by the biotrophic oomycete Hyaloperonospora arabidopsidis [9]. Except for the hnRNP AtGRP7 [9], the RBP-mediated defense mechanisms remain poorly understood. Here, we studied the role of A. thaliana RBP LIF2 in the plant's response to biotic stress. We demonstrated that LIF2 loss-offunction results in pathogen resistance, in a manner that is independent of the pathogen's lifestyle or infection strategy. Indeed, the lif2 mutant was less susceptible to both the hemibiotrophic bacterial P. syringae pv. tomato pathogen and the necrotrophic ascomycete B. cinerea, two pathogens with substantial impacts on agronomical production due to their wide range of host specificities [17,18]. However, lif2 was more susceptible to the necrotrophic ascomycete S. sclerotiorum, which attacks more than 400 plant species around the world.
The three pathogens used in this study induce different plant defense pathways and their associated defense molecules: the JA and ethylene pathways, and possibly the SA pathway, are activated in response to S. sclerotiorum infection [29,30]; the SA pathway is key in the plant's response to P. syringae infection; and JA and camalexin are involved in the response to B. cinerea infection [64]. Our analysis of the lif2 mutant revealed that SArelated genes (ICS1/SID2, PR1) were constitutively activated. The upregulation of PR1 was in agreement with the decreased susceptibility of lif2 to P. syringae, and this response was also observed in other mutants with reduced susceptibility to P. syringae, such as cpr1 and cpr5 [65]. However, the constitutive activation of the SA biosynthesis pathway did not lead to an overaccumulation of free SA in the lif2 mutant. SA undergoes various chemical modifications, which affect its activity, catabolism, transport, and storage [66]. The accumulation of free SA is fine-tuned by the transcriptional regulation of genes involved in SA biosynthesis, but also by the modulation of enzymes that modify SA [24]. Our data suggest that genes of the SA pathway may undergo posttranscriptional regulation or that a homeostasis mechanism may control free SA accumulation and maintain it at wild-type levels in lif2. Furthermore, the SA production pathway is not fully elucidated in A. thaliana, since no isochorismate pyruvate lyase, which is required to convert isochorismate into SA in bacteria, has been identified to date. A second SA biosynthesis pathway, which is regulated by phenylalanine ammonia lyase (PAL), has been proposed to exist in plants [57,67]. In lif2, the discrepancy between the constitutive upregulation of the SA/ICS1-dependent pathway and the lack of SA accumulation may also suggest some crosstalk between the ICS1 and PAL pathways to fine-tune SA production. Finally, the lif2 sid2 double mutant was less susceptible than wildtype plants to P. syringae, despite exhibiting a PR1 expression level similar to that of wild-type plants. Therefore, the reduced susceptibility of the lif2 mutant to P. syringae was independent of SA in non-challenged plants, suggesting that LIF2 participates in plant defense via a novel defense pathway that is independent of the SA-signaling defense pathway.
Furthermore, we showed that the GLS profile of the lif2 mutant was altered. These changes might indirectly participate in the increased resistance phenotype of lif2. Indeed, GLSs function in the plant's defense response to herbivores and fungal pathogens [44,68,69,70]. Ward et al. (2010) suggested that GLSs do not play a direct role in the response to bacteria, but possibly have an intermediary role by influencing some defense pathways. Indeed, after infection of A. thaliana with P. syringae DC3000, a significant reduction in I3M and 4MOI3M was observed [71]. In the lif2 mutant, a low level of I3M was observed, but no change was observed in 4MOI3M in the absence of bacterial challenge. The global decrease in GLS and RA contents observed in lif2 might thus contribute to the control of the basal defense pathways. Interestingly, JA is a regulator of GLS gene expression and of GLS accumulation via multiple pathways [72,73]. Several TFs have recently been shown to be involved in GLS gene regulation, including MYB28, MYB29, and MYB76, which are key regulators of the aliphatic-GLSs, and MYB34 and MYB51, which are regulators of the indole-GLSs. Furthermore, MYC2, MYC3, and MYC4 can form protein complexes with all known GS-related MYBs to regulate GLS biosynthesis [74]. In the lif2 mutant, MYB28 is downregulated, whereas MYB34 is upregulated, and the levels of members of the two GLS families were decreased. However, MYC2 is downregulated in the lif2 mutant. Furthermore, several TFs related to defense responses are deregulated in lif2. WRKY33 is an essential transcription factor in the defense against B. cinerea that acts by controlling the expression of genes involved in redox homeostasis, SA signaling, and camalexin biosynthesis and thus affecting the SA-JA balance [40]. The activation of WKRY33 observed in lif2 may have downstream effects, which may play a role in the resistance of lif2 to B. cinerea. Therefore, it is likely that multiple components of different defense pathways contribute to the primed state of lif2 in the absence of pathogen, and to its reduced pathogen susceptibility. Interestingly, LIF2 interacts with the chromatin-associated protein LHP1, a subunit of the Polycomb Repressive Complex, which interacts with numerous genomic sites and regulates their expression [12,14]. Some defense-related genes are present among the LHP1 targets. Whether LIF2 acts coordinately with LHP1 at these defense-related loci to control them constitutes an interesting research question, as the role of chromatin proteins in plant immunity is poorly documented [75,76,77,78,79]. For instance, PIE1, a member of the SWR1 subfamily, was shown to negatively regulate plant defense [80], whereas loss-of function of the SIRTUIN2 histone deacetylase, a homolog of the yeast Silent information repressor 2 (Sir2) protein [81] and of SDG8 histone methyltransferase [82] alter plant-pathogen responses. It was proposed that chromatin modifications may participate in defense priming in plants [83]. Thus, our study highlights an emerging role for the LIF2 chromatin-associated protein in biotic stress responses, with a putative suppressor function in plant immunity (Fig. 9).
Finally, LIF2 is also involved in the maintenance of growth and cell determination during floral development [12]. Therefore, our study illustrates the trade-off existing between plant development and plant defense against various enemies in a changing environment [84,85]. LIF2 may thus regulate the balance between development and plant immunity by limiting the cost associated with the plant defense response (Fig. 9). It was suggested that priming of SA-related defense responses greatly enhances disease resistance and plant fitness, but diminishes fitness in the absence of pathogens [85]. Furthermore, since both developmental and plant defense processes require adaptation to environmental conditions, sharing common elements with dual functions may allow for better fine-tuning [86]. Since the human hnRNP-Q homolog of LIF2, SNAP1/SYNCRIP, also acts as a cellular RBP and as a suppressor of human immunity against virus infection, our data highlight a conserved role for hnRNP-Q in eukaryote immunity. The dual function of hnRNP-Q proteins in development and defense is likely to involve key conserved molecular events in eukaryotic cells.   Table S1. doi:10.1371/journal.pone.0099343.g007 The LIF2 RBP in Biotic Stress PLOS ONE | www.plosone.org Identifying the associated RNA partners of LIF2 RBP may elucidate the underlying mechanism.

Plant materials
Arabidopsis thaliana lines used in this study were in the Columbia (Col-0) accession. The homozygous lif2-1 and lif2-3 mutants were described in [12]. The lif2-c line is a lif2-1 mutant complemented with the LIF2 genomic region under the control of a 3-kb promoter region. The sid2-2 [87] mutant was crossed with the lif2-1 allele. Double mutants were selected by PCR using specific primers (Table S3, [12]).
For P. syringae, B. cinerea, and S. sclerotiorum pathogen response analyses, plants were grown on soil in growth chambers, under controlled short-day conditions (8 hours light/16 hours dark), at 20uC and 70% hygrometry.

Primers
All primers are described in Table S3.

Bacterial inoculation
The virulent strain P. syringae pv. tomato DC3000 (Pst DC3000) [88] and the avirulent strain P. syringae pv. tomato DC3000, carrying the avirulent gene avrRpm1 (Pst DC3000 avrRpm1) [28], were grown at 28uC on King's B solid medium plates containing the appropriate antibiotics. Briefly, a fresh bacterial suspension was scraped off the growth plate, and resuspended in 10 mM MgCl 2 at an OD600 of 0.01 (at a concentration of 10 6 cfu ml 21 ). The bacterial suspension was infiltrated into the abaxial side of 24 to 30 rosette leaves of seven-week-old plants (6 to 10 plants per genotype) using a 1-ml syringe without a needle. Control inoculations (Mock) were performed using 10 mM MgCl 2 . Bacterial growth was assessed 24 hours after infiltration. Whole leaves were homogenized in 500 ml of 10 mM MgCl 2 . After serial dilutions, homogenates were plated on King's B medium and incubated at 28uC for 2 days before colonies were counted. The surface area of each leaf was measured using ImageJ software in order to calculate cfu/cm 2 . Day 0 titers ranged from log 10 = 3.6-3.9. Leaves were photographed 5 days post-infiltration (dpi). All experiments were repeated at least twice with similar results.

Fungal pathogen inoculation procedure
Four-week-old plants were inoculated with discs carrying S. sclerotiorum (S55 strain) mycelia, as previously described [29]. Plant inoculations were performed in a Helios 1200 Phytotron under a light period of 9 h at 22uC and 80% relative humidity during the first 3 days and 66% thereafter. At least 20 plants of each genotype were inoculated. The disease score of each line was evaluated seven days post-inoculation. Two independent experiments were performed with similar results.
Leaves of seven-week-old plants were inoculated with mycelial plugs of the wild-type strain B. cinerea B0510 and Bd90 (diameter of 3 mm), as previously described [89]. Infected leaves were placed in a Petri dish under high humidity and incubated at 21uC. Lesion surfaces were measured daily for three days using ImageJ software.

Quantification of hormonal contents
Rosettes of seven-week-old plants were frozen immediately after harvest and ground in liquid nitrogen. Four pools of 6 rosettes were collected. Frozen material (100 mg) was extracted with 3 ml of acetone/water/acetic acid (80/19/1, v:v:v) containing the following stable isotope labelled internal standards: 10 ng The instrumental parameters were set as follows: capillary, 2.70 kV (negative mode); extractor, 3 V; and source block and desolvation gas temperatures, 120uC and 350uC, respectively. Nitrogen was used for the nebulization and desolvation (77 L h 21 and 365 L h 21 , respectively), and argon was used as the collision gas at 2.83 10 23 mbar. For a 5-mL injection volume of sample prepared and reconstituted in 140 ml of 50/50 acetonitrile/H2O (v/v), the limit of detection (LOD) and limit of quantification (LOQ) were extrapolated for each hormone from a calibration curve and sample using the Quantify module of MassLynx (version 4.1 software). The parameters used for MRM quantification and the LOD and LOQ are listed in Tables S4 and S5, respectively. The amount of JA was expressed as a ratio of peak areas (209.62/214.62) per fresh weight, due to impurities contained in the D5-JA standard.

Glucosinolate extraction and quantification
Glucosinolate analyses were performed on 15-day-old in vitro seedlings grown under long-day conditions. The plant material was rapidly collected, frozen, and lyophilized. Lyophilized material (30-50 mg) was ground in liquid nitrogen. Then, 0.2 mg of sinigrin hydrate standard (Sigma Aldrich Ref. 85440) was added as an internal tracer for recovery and analytical The LIF2 RBP in Biotic Stress purposes. The samples were extracted using 2.5 ml of extraction solution (75% acetonitrile/25% water solution), vigorously homogenized for 5 min in a Polytron homogenizer (Fischer Scientific), and centrifuged (8000 g, 20uC, 15 min). The supernatants were collected and the pellets were extracted with 1 ml of the extraction solution and sonicated (15 min,25 Hz). Following centrifugation, the supernatant was pooled with the first supernatant. The extracts were then evaporated to dryness using a Thermo Savant Speedvac overnight, at ambient temperature. A water/acetonitrile solution (95/5) was added to the dried extracts, which were filtered and diluted to the third with water/acetonitrile (95/5), prior to negative electrospray LC-MS analysis on an Alliance 2695 system coupled to Quattro LC (Waters). Glucosinolates were identified by retention time, mass, isotopic pattern, and fragment ions [90,91]. The concentrations of the metabolites of interest were quantified using the sinigrin response, as previously described [92]. Three independent extraction analyses were carried out per biological experiment and two to three biological replicates were performed. Biochemical information about the different classes of GLSs and their biosynthesis pathways can be found in various databases (e.g., http://www.genome.jp/ kegg-bin/show_pathway?ko00966+C08417 or the AraCyc website).

Gene expression analysis
Total RNA was isolated from various tissues, using the RNeasy Plant Mini Kit (QIAGEN), and treated with RNase-free DNaseI (Invitrogen). Reverse transcription (RT) reactions were performed with Superscript II reverse transcriptase (Invitrogen), according to the manufacturer's instructions. Quantitative real-time PCR (qPCR) was performed on Eppendorf MastercyclerH ep realplex (Eppendorf) using MESA FAST qPCR MasterMix Plus for SYBRH Assay (Eurogentec), as per the manufacturer's instructions.