NPR1 has been characterized as a result of observing the response to SA of the great number of alleles described for it 24 . NPR1 has also been described as essential for RIM, but there are differences between alleles, since npr1-1 and npr1-3 have different RIM ( 20 25 ). npr1-1 and npr1-3 have other differences in phenotypes related to MeJA. Thus, the SA-JA antagonism is not present in npr1-1, but it is active in npr1-3 16 . Other difference is the gene expression, whereas npr1-1 was affected in SA, JA, and ET dependent genes upon Pto inoculation, npr1-3 was only affected in SA dependent genes 17 . These different phenotypes have been attributed to the lack of nuclear localization in npr1-3, since the truncated cytosolic protein would be functional to modulated JA-dependent defense response 19 .

In order to determine the precise role of NPR1 in RIM, and to asses the role of the cytosolic vs. nuclear localization, we tested an allelic series of 43 npr1 alleles, defined by their inability to perceive SA 11 . The npr1 alleles produced a mixture of phenotypes in RIM (Figure 1), and out of the 43 alleles tested, 11 had a significant RIM (RIM+, Figure 1a), meaning that the growth of Pto in MeJA treated plants was significantly lower than in mock treated plants (Student’s test of one tail, with P<0.05). The rest of alleles showed no response (RIM-, 11 out of 32 are shown in Figure 1b). These two categories of alleles did not share any obvious feature and had the same proportion of stops and point mutations in each category. In fact, in the RIM+ alleles tested, six point mutations were widely scattered along the protein (Figure 1c, npr1-3 is included as reference).

These alleles could somehow affect the localization of the protein inside the cell, even if the mutation was not in the NLS. To check this possibility, cDNAs of three RIM+ alleles (npr1-22, -35 and −44, Additional file 1), and three RIM- alleles (npr1-1, -40, and −56, Additional file 1), chosen among the point mutations, were cloned along with the wild type NPR1. Then, these seven cDNAs were transiently expressed in N. benthamiana under the control of the 35S promoter and with the GFP marker. The free GFP was detectable in the nucleus and in the cytoplasm. But when the wild type and the six alleles of npr1 were expressed in the same conditions, GFP was detected mainly in the nucleus, with no difference existing between the two classes of alleles (Figure 2a).

As a complementary approach, we took advantage of the transgenic line that overexpresses NPR1 fused to the steroid hormone binding domain of the rat glucocorticoid receptor (HBD, and the transgenic plants are known as NPR1-HBD, 15 ). NPR1-HBD remains exclusively in the cytosol in mock conditions and should be functional in RIM. The original line is in an npr1-3 background (RIM+), and therefore the transgene was transferred to an npr1-1 background (RIM-) to check for complementation. Treatments with BTH and with and without glucocorticoid dexamethasone (DEX) showed that NPR1-HBD was functional (Additional file 1). NPR1-HBD, even under the control of the 35S promoter, did not complement the lack of RIM in npr1-1 (Figure 2b). When DEX was applied, NPR1-HBD moved to the nucleus and npr1-1 was complemented in the RIM phenotype. Note that the presence of cytosolic NPR1-HBD in an npr1-3 background did not enhance RIM in comparison to npr1-3 alone.

The npr1 alleles RIM+ or RIM- did not share any obvious feature, so it would be difficult to assign a precise role to the wild type gene. A critical genetic resource to discern the role of a gene is the null allele. Therefore, the response of two null alleles of npr1 to RIM was measured (Figure 3a). Both npr1-70 and npr1-71 are in Laer-0 background, so an introgression of npr1-1 in Laer-0 was used as control 11 . These two null alleles responded to RIM in all experiments like the wild type. npr1-70 introgressed in Col-0 responded again like the wild type, which ruled out any ecotype effect (Figure 3b). Since the direct role of NPR1 in RIM was in question as a consequence of the aforementioned results, we included an independent RIM- control, coi1-40 (see Methods).

The role of NPR1 in this response might be indirect. Thus, one scenario would be a reinforcement of the negative crosstalking between SA and MeJA. npr1 alleles produced more SA when infected with Pto 26 and seemed unable to metabolize it 27 . RIM- alleles -defective in terms of SA perception- might have left intact the negative crosstalk between SA and MeJA, and an excess of SA repressed the response to MeJA beyond the wild type levels. Therefore, the RIM+ alleles would be defective in terms of both SA perception and SA-MeJA crosstalking, an explanation that would also be in agreement with the behavior of the null alleles.

To test this hypothesis, the double mutants between npr1-1 and NahG (a transgenic plant that degrades SA, 28 ), eds5 (a mutant in SA transport, 29 ), and sid2 (a mutant in SA biosynthesis, 30 ), were constructed and tested for RIM. npr1-1 did not respond to MeJA even if the levels of SA were low (Figure 3c), so the hypothesis of a reinforcement of the negative crosstalk was not supported.

The experiments with the null alleles showed that NPR1 was not necessary for RIM. Perhaps NPR1 and other gene(s) would be redundantly responsible of RIM, and while null npr1 alleles would have a RIM+ phenotype, some npr1 alleles could be RIM- by interacting with other protein(s) negatively. The most likely candidates for these interactions were the NPR1 paralogs, since their proteins shared the same overall structure. There are five additional paralogs of NPR1 in the Arabidopsis genome 7 , and we analyzed double mutants of npr1-70 with different paralogs (Figure 4). There was no proof of NPR1 having a redundant role in RIM. NPR2 did not play a significant role in RIM (Figure 4a) and the same was true for the double npr3 npr4 (Figure 4b,c and 7 ). Strikingly, the double bop1 bop2 21 was RIM- (Figure 4d).

Single bop1 and bop2 were also tested for RIM and showed to be wild type (Figure 5a). BOP1 and BOP2 exert their function in part through transcriptional regulation of KNAT6 and physical interaction with PAN (TGA8, 31 ), but T-DNA insertions predicted to disrupt KNAT6 or PAN activity did not have an effect on RIM (Figure 5a). To rule out the possibility that other mutations besides bop1 and bop2 were producing this RIM- phenotype, we constructed an artificial microRNA (amiRNA, 32 ) to deplete the levels of BOP1 and BOP2 at the same time. Eight independent homozygous transgenic lines for amiRNA (BOP1 - BOP2) were analyzed for RIM (Figure 5b). Five out of eight lines were RIM-, and the remaining three responded less than the wild type control. The levels of both genes were partially depleted in the eight lines (Figure 5c); five of the lines had both genes significantly reduced, and all had BOP2 significantly reduced. None of these lines had the characteristic blade-on-petiole macroscopic phenotype, not even as the subtle phenotype of bop1 alone.

The previous experiments had shown that BOP1 and BOP2 were acting redundantly in RIM (Figure 4d). Therefore, increasing the amount of any of them should have an effect on RIM, especially since normal levels of BOP1 and BOP2 are quite low (Additional file 1, 21 ). The overexpression lines of BOP1 and BOP2 described 33 were analyzed for RIM. 35S:BOP1 had a stronger RIM than Col-0, and 35S:BOP2 had a strong variation in the MeJA treated plants (Figure 6a). At the time of the experiments, each population looked homogeneous, but when these plants were grown to set seeds, two phenotypes could be observed in each transgenic line. Approximately half of the plants showed a wild type phenotype, and the other half reproduced the dwarf plants described 33 . Seeds from both lines and from both phenotypes reproduced the two phenotypes. It seems that it was an issue of silencing, since RNA taken from plants classified by their mentioned phenotype diverged widely in the transgene expression (Figure 6b).

The bop1 bop2 double mutant lacked a RIM response, but this failure in MeJA signaling might occur at different points of the signal transduction. For example, the defect could target a general signaling component affecting all MeJA responses (e.g. coi1, 34 ), or a specialized part of the pathway, affecting a subset of MeJA responses (e.g. jin1, 35 ). When bop1 bop2 plants were grown in plates containing MeJA, the growth of the roots was similar to the wild type controls (Figure 7a). Other phenotypes of bop1 bop2 plants growing in MeJA plates were similar to the wild type controls (carotenoids production, size of aerial part, number of trichomes, etc.; data not shown). Another effect of MeJA is the increase of senescence in detached leaves, measured as chlorophyll production 36 . bop1 bop2 responded as the wild type in this particular system (Figure 7b). Coronatine is a virulence factor of several Pseudomonas isolates with structural and functional similarities to JA-Ile (the functional form of Jasmonate in planta, 5 ), therefore a mutant Pto that lacks coronatine 37 grows less in Col-0 than the wild type Pto. bop1 bop2 was also wild type in response to Pto with and without coronatine (Figure 7c). Inoculations with Plectospharaella cucumerina, a fungus that causes more disease in MeJA mutants than in wild type plants 38 , did not cause any more disease in bop1 bop2 than in Col-0 (data not shown). If bop1 bop2 was not a MeJA signaling mutant, but specifically in RIM, it could be defective in the other signaling required for ISR; ET. It was not; when 1 mM of 1-aminocyclopropane-1-carboxylic acid (ACC, an ET precursor) was sprayed to bop1 bop2, the resistance triggered was similar to that triggered in the wild type controls (Figure 7d). etr1 (ETHYLENE RESPONSE 1, 39 ) was included as a negative control of resistance induced by ethylene.

Once it was clear that both BOP genes are required for RIM, we tested the model that npr1 RIM- alleles could have a dominant negative effect on BOP activity, either directly or indirectly. To first test whether NPR1 had an effect in the interaction between BOP proteins, we used a yeast two-hybrid assay. As reported, BOP1 and BOP2 interacted with each other 40 . Next, we introduced in a third plasmid containing wt NPR1 or various mutant npr1 alleles presented in Figure 2a. If the effect of npr1 on RIM were a direct interaction between NPR1 and the two BOPs, the alleles that diverge in their RIM phenotype would diverge in their ability to interfere in the interaction of BOP1 and BOP2. The two classes of npr1 alleles did not have a distinct behavior (Figure 8a, the first three npr1 alleles are RIM-, and the last three are RIM+), therefore the dominant negative effect did not seem to be direct.

All the NPR1 paralogs tested interact with members of the TGA family in a different degree 7 21 . Therefore, the TGAs would be a reasonable candidate for being the third component, and their interaction with RIM- alleles would indirectly affect the function of BOP1 and BOP2. As a control, single mutants in TGA1 and TGA7 produced a significant RIM (Figure 8b), but when three specific tgas are knocked out at the same time (a triple which phenocopies an npr1 mutant in SA response, 41 ), there is no RIM (Figure 8b).

We reasoned that one or several of these three TGAs (TGA2, 5, and 6) might have a functional interaction with the BOPs, which might be affected by the RIM- alleles. To test this hypothesis BOP1 and each of the mentioned TGAs were introduced in the yeast two-hybrid system with the npr1 alleles mentioned above in a third plasmid. TGA2 and TGA6 interact differentially with BOP1 depending on the npr1 protein present (Figure 8c). There was an enhancement of the interaction in two out of the three RIM- alleles, and no interference in two out of the three RIM+ alleles. The interaction TGA5-BOP1 was not affected by the presence of npr1 proteins (data not shown). The experiments were repeated with BOP2 producing similar results (data not shown). In sum, the data indicated that BOPs and TGA2, TGA5 and TGA6 are required for RIM, that BOPs interact with these (and other) TGAs, and that NPR1 may modulate the affinity or stability of the interactions.

NPR1 is an essential gene for SAR and SA perception 6 . npr1-1, the most widely used allele, is also impaired in RIM 13 . We speculated that since npr1-3 is wild type for RIM ( 20 25 ), and it has been reported that the difference of some phenotypes between npr1-1 and npr1-3 was due to the lack of NLS in npr1-3 ( 17 19 ), the same could be true for RIM. However, we show here that the nuclear localization of the alleles makes no difference. This conclusion is supported by multiple lines of evidence. First, the npr1 alleles with RIM+ are not structurally similar to npr1-3, since not all of them are affected in the NLS (Figure 1c). Even an allele with a point mutation in the NLS (npr1-22, Additional file 1) should be partially localized in the nucleus 15 . Second, three RIM- and three RIM+ alleles do not differ in their nuclear localization or stability when transiently expressed in N. benthamiana (Figure 2a, Additional file 1). While these proteins are no longer functional, they respond to the signals of a wild type background by localizing in the nucleus. Third, when a functional NPR1 is anchored in the cytoplasm there is no complementation of the RIM- phenotype in an npr1-1 background (Figure 2b), nor there is an increase in RIM phenotype in an npr1-3 background. In fact, the application of DEX triggered an increased RIM in both backgrounds (discussed below).

But, most importantly, NPR1 is not required for RIM, since the null npr1 alleles are RIM+ regardless of the background (Figure 3a,b). We also discarded that NPR1 could be a part of RIM in a redundant fashion with its paralogs (Figure 4).

An interesting alternative for the role of NPR1 in RIM would be an effect on the crosstalking between SA and MeJA. NPR1 has been described as a key point in the negative regulation between SA and MeJA. Thus, the RIM+ alleles could be defective in both SA perception and in SA-MeJA crosstalk, while the RIM- alleles would be defective only in SA perception but not in SA-MeJA crosstalk. The inoculation with Pto triggers an increase in the levels of SA, and in the case of the npr1 alleles, there is more SA than in the wild type 42 . Although this hypothesis would explain the phenotype of the null alleles, it was rejected after the experiment of Figure 3c, where a severe reduction of SA levels in a RIM- allele did not have any effect on the phenotype.

The redundant functions of BOP1 and BOP2 are essential for normal development. Previous work has shown that the double mutant has numerous defects in plant architecture including altered leaf morphology 43 , changes in floral patterning 21 , defects in the conversion of shoots to flowers 44 and loss of floral-organ abscission 45 . The double mutant was tested for basal defense 21 and SA perception (Additional file 1) but no difference from wild type was found. We show here that both genes are also redundantly required in defense against pathogens triggered by MeJA. Interestingly, whereas significant loss of BOP activity is required to exert changes in development 21 , RIM is abolished in plants that are only partially silenced for the BOP genes (Figure 5b,c). Thus, the levels of gene expression required for RIM are higher than those required for normal development. Compatible with this idea, BOPs expression in plants is highly localized, restricted to young organ primordia, leaf petioles, and lateral organ boundaries, which may make systemic responses to MeJA sensitive relatively minor changes in BOP transcript abundance. Both NPR1 and the BOPs localize to the cytoplasm as well as nucleus and interact with members of the TGA family of bZIP transcription factors, albeit with different affinities (e.g., 12 ). In development, BOP1 and BOP2 form a nuclear complex with TGA8/PERIANTHIA (PAN) to regulate number of sepals and petals in flowers and potentially to promote floral meristem fate 21 . Given that pan loss-of-function did not reproduce the RIM- phenotype (Figure 5a) other genes, perhaps TGAs, are involved in this phenotype, as shown for SA perception 23 . Given that BOPs play both positive and negative roles in transcriptional regulation of the KNOX (Knotted1-like homeobox) gene KNAT6 46 , we also tested if RIM was affected by knat6 loss-of-function, but again, no difference was observed (Figure 5a). This may reflect redundancy with other KNOX genes, or more likely, that BOP regulation of RIM is independent of KNAT6.

Whether bop1 bop2 recapitulates or not all the phenotypes of the RIM- npr1 alleles (e.g. ISR, 13 ; Verticilium resistance, 18 ; resistance induced by Piriformospora indica, 47 ; etc.) remains to be assessed. We did check that there were similar phenotypes in the specificity of response to MeJA as well as the fact that bop1 bop2 was wild type for the rest of MeJA phenotypes (Figure 7). But there were strong differences, since npr1-3 is affected in basal defense and SA perception while bop1 bop2 is wild type for both phenotypes 21 . Regarding ET, the other hormone relevant for ISR, it has been proposed that applications of this hormone could render the crosstalk between SA and MeJA independent of NPR1 19 . It seems plausible that the ET induced resistance works as the MeJA induced resistance and other proteins -perhaps NPR1 paralogs, but not the BOPs (Figure 7d) - might also be affected by some alleles of npr1.

The RIM- npr1 alleles were the majority of the alleles found (32 RIM- vs. 11 RIM+). How is this compatible with the fact that the null npr1 alleles are RIM+? A possible explanation was the selection used in the screening. Since the selection was made for complete loss of SA perception, perhaps most of the RIM+ alleles had a phenotype of partial SA perception, as the null alleles. Then, the prediction would be that a good number of random alleles of npr1 would be RIM+ and partially receptive to SA. We previously showed that for SA perception, there are genetic interactions between the npr1 alleles and the NPR1 paralogs 11 . The work reported herein points to a genetic interaction too, this being between npr1 alleles on one side and the BOPs on the other. Thus, the RIM- alleles were a phenocopy of the bop1 bop2 mutant in defense but not in development. This discrimination was a consequence of the different thresholds for the phenotype in development and defense (Figure 5b,c).

Mechanistically, the levels of expression of the BOPs were low in comparison to NPR1 (Additional file 1), so a direct or indirect negative interference of NPR1 with the BOPs would be favored stoichiometrically. Once the pathogen was inoculated, the levels of SA would rise and in a wild type plant NPR1 is degraded as part of the signaling process 48 . In an npr1 background, this signaling would not be transmitted and perhaps the npr1 proteins would be able to interfere longer in RIM. This would explain the behavior of NPR1-HBD in npr1-1 (v 2b); NPR1-HBD in the cytoplasm did not complement npr1-1 in the RIM phenotype, but when DEX was applied there was complementation of the phenotype. Likely, when no DEX was present npr1-1 would somehow interfere with the function of the BOPs. When DEX was present, the presence of NPR1-HBD in the nucleus would trigger the degradation of both NPR1-HBD and npr1-1. If npr1-1 was degraded, the BOPs would function normally.

There was no evidence for a direct interaction in yeast, since the presence of NPR1 or mutated versions of this protein did not interfere in the interaction between BOP1 and BOP2 in a consistent manner with the phenotype (Figure 8a). A first alternative was that the interference of the RIM- alleles would occur with the BOPs without affecting the interaction between the BOPs. A second alternative would be that the RIM- alleles would interfere with other proteins that normally interact with the BOPs. In both cases there is a family of proteins that interacts with both NPR1 and the BOPs, the transcription factors TGAs 49 , with -again- functional redundancy (Figure 8b). Two out of three RIM- alleles enhanced or stabilized the BOPs-TGAs interaction, while two out of three RIM+ alleles did not (Figure 8c and data not shown). It was clear that the npr1 mutated proteins had an unpredicted effect on the BOPs-TGAs interaction, but the yeast experiments did not produce an absolute answer about the role of npr1 proteins in RIM. We speculate that in planta, all the RIM- alleles enhance the interaction between the BOPs and the TGAs, titering out the TGAs and thus rendering them unable to fulfill their function of triggering defense. On the other hand, the RIM+ alleles (including the null alleles), and NPR1 would not affect the interaction either way. Since in the yeast assays two out of three alleles worked as proposed in either way, it may be possible that a factor(s) is present in the plant that is not in yeast, or perhaps the fact that there are ten TGAs 49 , and that NPR1 is expressed between 3 and 18 times more than BOP1 + BOP2 (Additional file 1) could explain this difference. If this hypothesis were to be true, it will definitively explain the role of npr1 in RIM.

Arabidopsis thaliana (L.) Heynh. was sown and grown as described 23 in controlled environment rooms with days of 8 h at 21°C, 150 μmol m^-2 s^-1, and nights of 16 h at 19°C. Treatments, inoculations, and sampling started 30 minutes after the initiation of the artificial day to ensure reproducibility. The following genotypes were used: npr1-1 and npr1-3 27 ; npr2, npr1-20 to npr1-71, and combinations of npr1-70 with other genotypes 11 ; 35SCaMVp:NPR1HBD 15 ; sid2 30 ; eds5 29 ; NahG 28 ; npr3 and npr4 7 ; bop1-3 and bop2-1 21 ; coi1-40 (Dobón, Wulff, Canet and Tornero, to be published elsewhere); kant6, pan1-1 to pan1-3, tga1, and tga7 50 ; 35S:BOP1 and 35S:BOP2 33 ; etr1-3 39 ; tga2 tga5 tga6 41 . Pseudomonas syringae pv. tomato DC3000 (Pto) was grown, inoculated and measured as described 51 . Briefly, plants of 14 days were inoculated by spray with Pto at OD₆₀₀=0.1 with 0.02% Silwet L-77 (Crompton Europe Ltd, Evesham, UK). Three days later, the amount of colony forming units (cfu) per plant was quantified and represented in a logarithmic scale. When indicated, a strain of Pto lacking coronatine was used (Pto(cfa ^– ), 37 ). For all the experiments, at least three independent treatments were performed (three independent sets of plants sown and treated on different dates).

NPR1 and six alleles of this gene were cloned in pDONR222 or pDONR221 (Invitrogen, Barcelona, Spain) and then transferred to pMDC43 52 for expression in planta with GFP and to pARC352 53 for expression in yeast. Similarly, BOP1, BOP2, TGA2, TGA5, and TGA6, were cloned and then transferred to pDEST22 and pDEST32 (Invitrogen) for expression in yeast. Yeast n-hybrid analyses were done as described 54 , and the interactions were quantified as described 55 . N. benthamiana leaf tissue was mounted in water under a coverslip 4 days after infiltration with Agrobacterium tumefaciens containing the constructs. All imaging was conducted with a Leica TCS SL confocal laser scanning microscope (Leica, Barcelona, Spain) using an HCX PL APO CS 40X/1.25 oil objective to study the subcellular localization of the fluorescence-tagged proteins. Green fluorescent protein was visualized by 488-nm excitation with an Ar laser, and its emissions were examined with a band-pass filter for 500 to 530 nm. The primers used are included as Additional file 1. Primers and chemical products were purchased from SIGMA (St. Louis, MO, USA) unless otherwise is stated. For the construction of amiRNA(BOP1-BOP2), the plasmid pRS300 was modified 32 , cloned in pGW14 56 , and plants were transformed as described 57 .

To measure the effect in Pto growth 100 μM methyl jasmonate (MeJA) in 0.1% DMSO and 0.02% Silwet L-77 (Crompton Europe Ltd) was applied by spray one day before the pathogen inoculation 25 . Dexamethasone was applied at 2 μM diluted in water from a stock of 20 mM in EtOH. 1-Aminocyclo- propane-1-carboxylic acid (ACC) was sprayed at 1 mM in water with 0.02% Silwet L-77.

For in vitro culture, plants were grown in Johnson’s media 58 with 1 mM KH₂PO₄. When indicated, the plates were supplemented with 50 μM MeJA. The length of the roots was measured with ImageJ software 59 . Senescence induced by MeJA was measured as described 36 .

Total RNA from 3-week-old (Figure 5c) or 6-week-old plants (Figure 6b) was extracted with Trizol (Invitrogen), following the manufacturer’s instructions. cDNA was synthesized with RevertAid™ First Strand cDNA Synthesis Kit (Fermentas, Madrid, Spain), and the quantitative PCR performed with LuminoCt Sybr Green qPCR Ready Mix (SIGMA) in a 7000 RT-PCR Systems machine (Applied Biosystems, Madrid, Spain), following the manufacturer’s instructions. For each measurement three biological replicates were done. The obtained values were referred to the geometric average of three reference genes (At3G18780, At1G49240, and At5G60390), as described 60 , and normalized, being the value of Col-0 in mock equal to one. The list of primers used is provided in Additional file 1.