Specific Missense Alleles of the Arabidopsis Jasmonic Acid Co-Receptor COI1 Regulate Innate Immune Receptor Accumulation and Function

Plants utilize proteins containing nucleotide binding site (NB) and leucine-rich repeat (LRR) domains as intracellular innate immune receptors to recognize pathogens and initiate defense responses. Since mis-activation of defense responses can lead to tissue damage and even developmental arrest, proper regulation of NB–LRR protein signaling is critical. RAR1, SGT1, and HSP90 act as regulatory chaperones of pre-activation NB–LRR steady-state proteins. We extended our analysis of mutants derived from a rar1 suppressor screen and present two allelic rar1 suppressor (rsp) mutations of Arabidopsis COI1. Like all other coi1 mutations, coi1rsp missense mutations impair Jasmonic Acid (JA) signaling resulting in JA–insensitivity. However, unlike previously identified coi1 alleles, both coi1rsp alleles lack a male sterile phenotype. The coi1rsp mutants express two sets of disease resistance phenotypes. The first, also observed in coi1-1 null allele, includes enhanced basal defense against the virulent bacterial pathogen Pto DC3000 and enhanced effector-triggered immunity (ETI) mediated by the NB–LRR RPM1 protein in both rar1 and wild-type backgrounds. These enhanced disease resistance phenotypes depend on the JA signaling function of COI1. Additionally, the coi1rsp mutants showed a unique inability to properly regulate RPM1 accumulation and HR, exhibited increased RPM1 levels in rar1, and weakened RPM1-mediated HR in RAR1. Importantly, there was no change in the steady-state levels or HR function of RPM1 in coi1-1. These results suggest that the coi1rsp proteins regulate NB–LRR protein accumulation independent of JA signaling. Based on the phenotypic similarities and genetic interactions among coi1rsp, sgt1b, and hsp90.2rsp mutants, our data suggest that COI1 affects NB–LRR accumulation via two NB–LRR co-chaperones, SGT1b and HSP90. Together, our data demonstrate a role for COI1 in disease resistance independent of JA signaling and provide a molecular link between the JA and NB–LRR signaling pathways.


Introduction
During their life cycle, plants have to fend off microbial pathogens including fungi, bacteria, viruses, and nematodes. To protect themselves, plants rely on the innate immune system of each plant cell to detect pathogen attack and subsequently activate disease resistance responses. The plant immune system relies on two inter-related branches. The first branch utilizes pattern recognition receptors (PRRs) to identify conserved pathogen associated molecular patterns (PAMPs). This recognition then initiates PAMP-triggered immunity (PTI) [1][2][3]. Although PTI can restrict further colonization in some cases, successful pathogens are still able to evade or suppress PTI with their effectors [4]. These proteins contribute to pathogen virulence by interfering with various plant defense-related cellular processes. However, effectors can also be recognized by the intracellular NB-LRR receptor proteins of the plant innate immune system [5]. Recognition of effectors results in effector-triggered immunity (ETI) and is the second branch of the plant immune system [1][2][3]. NB-LRR proteins contain a centrally located nucleotide binding site (NB) domain and a C-terminal leucine-rich repeat (LRR) domain. Mammalian NB-LRR containing (NLR) proteins mediate analogous processes in mammalian innate immunity [6].
As highly conserved proteins, SGT1 and HSP90 also interact with each other in mammalian cells, and play essential roles in mammalian immune responses mediated by NLR proteins. By coimmunoprecipitation experiments, both SGT1 and HSP90 were found to associate with many NLR proteins including NOD1 (Nucleotide-binding Oligomerization Domain 1), NOD2 (Nucleotide-binding Oligomerization Domain 2), and NALP3 (NACHT, LRR and PYD domains-containing Protein 3) [23,24]. In mammalian cells, treatment with geldanamycin (GDA), a chemical inhibitor of HSP90, impaired NOD2-induced NF-kB activity and NALP3-mediated inflammatory responses [24]. Knockdown of HSP90 by RNAi or GDA treatment also reduced the accumulation levels of NOD1 and NOD2 [23]. These results demonstrated that mammalian HSP90 is required for both NLR stability and function. In contrast, mammalian SGT1 is only required for NLR functions such as NOD1-mediated cytokine production, NOD1-mediated cell death, and NALP3-mediated inflammatory responses, but not for NLR stability [23,24]. Plant SGT1, however, functions in both NB-LRR activity and stability [25]. Moreover, mammalian SGT1 knockdown reduced the association between HSP90 and the NALP3 LRR domain, indicating that mammalian SGT1 functions as a co-chaperone of mammalian HSP90 to regulate client NLR protein [24]. Unlike plant RAR1, CHP1 (CHORD-containing Protein 1), a homolog of RAR1 in mammals, is not involved in regulating NLR protein accumulation or function [24]. Taken together, the SGT1-HSP90 chaperone complex has functions for mammalian NLR protein stability and activity, analogous to its functions for plant NB-LRR biology [19,20].
COI1 encodes an F-box protein that is a component of the SCF COI1 (Skp1/Cullin/F-box COI1 ) E3 ubiquitin ligase complex [31,32,42]. The function of COI1 is to specifically bind target proteins to promote ubiquitination and degradation by the 26S proteasome [31]. It is therefore assumed that COI1 regulates JA signaling and disease resistance via degradation of specific proteins. The connection between JA signaling and SCF COI1mediated protein degradation has been confirmed. The JASMO-NATE ZIM DOMAIN (JAZ) family proteins act as repressors of MYC2, a key transcriptional activator of JA responses, by directly interacting with MYC2. JA-Ile, a bioactive JA conjugate, induces the degradation of JAZ proteins by enhancing the protein interaction between JAZs and COI1, and thus de-represses JArelated transcription activation [28,29,31,43]. The JAZ and MYC proteins also play a role in disease resistance. Overexpression of JAZ1D3A, a C-terminal deletion form of JAZ1, led to enhanced disease resistance against Pto DC3000 in Arabidopsis [29]. The triple mutant for transcription factor genes MYC2, MYC3, and MYC4, which are all repressed by JAZ proteins, was as resistant against Pto DC3000 as the coi1 mutant [44].
In this study, we extend our previously described suppressor screen for new mutants that recover impaired RPS5 function in rar1 [21]. We introduce two novel missense alleles of COI1 that suppress the disease resistance phenotypes associated with rar1 mutation. Surprisingly, these two coi1 rar1 suppressor (rsp) alleles are completely fertility, in contrast to the male sterility associated with all other coi1 mutant alleles [32,43,45]. Like sgt1b and the hsp90.2 rsp alleles [21], these two coi1 rsp alleles interact with rar1 to restore the disease resistance responses mediated by some NB-LRRs and the accumulation of at least RPM1. Moreover, we demonstrate that overexpression of SGT1b can partially inhibit the coi1 rsp -enhanced accumulation of RPM1 and RPM1-mediated disease resistance in rar1. We also observe non-allelic non-complementation, a rare genetic interaction, between coi1 rsp mutants and hsp90.2-7 rsp mutant. These results support the hypothesis that coi1 rsp proteins regulate NB-LRR levels via SGT1b and HSP90.

Identification of new alleles of COI1 and of the rsp3 mutant
To identify new genes that act with RAR1 to regulate NB-LRR accumulation and activation, we performed a suppressor screen for new mutants which can suppress the disease susceptibility observed in rar1-21 (a stop mutation in Q52) [21]. Five rar1 suppressor (rsp) mutants were identified from approximately 200,000 M2 plants from 50 M2 pools that recover resistance responses to both Pto DC3000(avrPphB) and Pto DC3000(avrRpm1) [21]. Based on map-based cloning and subsequent allele sequencing, two of the five mutants were found to have mutations in COI1 (At2g39940). To follow accepted nomenclature conventions, we designated these two mutant alleles, coi1-21 rsp and coi1-22 rsp ,

Author Summary
To detect pathogen attack and subsequently trigger defense responses, plants utilize immune receptors composed of a nucleotide binding site (NB) domain and a Cterminal leucine-rich repeat (LRR) domain that function inside the cell. To identify regulators of NB-LRR protein accumulation and activity, we performed a genetic screen in the model plant Arabidopsis thaliana to isolate mutants that affect NB-LRR protein accumulation levels and NB-LRR triggered disease resistance. Here, we introduce two mutant alleles of COI1, a gene which encodes a wellcharacterized receptor for the phytohormone Jasmonic Acid (JA). It is widely accepted that COI1 is involved in JA signaling-dependent disease resistance. However, our new coi1 mutants affected NB-LRR accumulation in a manner independent of the JA signaling pathway. This indicated that not all disease resistance effects of COI1 require JA signaling. We also observed a link between COI1 and the RAR1-SGT1b-HSP90 co-chaperone complex, which plays a critical role in regulation of NB-LRR protein accumulations. respectively ( Figure 1). Based on disease symptoms after inoculation of Pto DC3000(avrRpm1) on backcross F1 and F2 populations, both of the coi1 rsp mutants were completely recessive (Table S1). This conclusion was also confirmed by growth assays of Pto DC3000(avrRpm1) in backcross F1 plants ( Figure 2). The coi1-21 rsp mutation is a G/A transition which leads to a G330E missense change in the COI1 protein. The coi1-22 rsp mutation is a G/A transition resulting in a G434E missense change in the protein.
Both mutations are within conserved LRR domains ( Figure 1).
Using the crystal structure of the Arabidopsis COI1 protein, we observed that neither coi1 rsp mutation is localized in the interfaces of COI1 that make up the ASK1-binding region and the ligandbinding pocket [31].
In addition, another rar1 suppressor (rsp) mutant called rsp3 was isolated from this screen. rsp3 suppressed all known rar1 phenotypes, and was localized in a 7 Mbp region on chromosome I ( Figure S1). A single allele, dominant mutation was identified in rsp3; its detailed characterization is beyond the scope of this work.  COI1 and HSP90 interact genetically to regulate disease resistance The disease resistance restoration phenotypes of hsp90.2-7 rsp and either coi1 rsp alleles in rar1 are fully recessive with respect to their respective wild type phenotypes ( [21], Figure 2). We monitored in planta growth of Pto DC3000(avrRpm1) to measure RPM1-mediated disease resistance in F1 plants of hsp90.2-7 rsp 6coi1 rsp crosses ( Figure 2). The resulting F1 plants were as resistant to Pto DC3000(avrRpm1) as their parental coi1 rsp plants. We also tested F1 plants of crosses between hsp90.2-7 rsp and either coi1 rsp allele for disease symptoms after inoculation of Pto DC3000(avrRpm1). The F1 plants displayed resistance against Pto DC3000(avrRpm1) ( Table  S2). In addition, we observed that a part of the F2 progenies from each F1 were susceptible to Pto DC3000(avrRpm1) ( Table S2). These results clearly demonstrate non-allelic non-complementation between hsp90.2-7 rsp and coi1 rsp mutants, suggesting that the two proteins function in the same process, and likely do so in physical proximity [14,46,47]. coi1 rsp alleles and coi1-16 partially suppress known rar1 phenotypes hsp90.2 rsp alleles isolated from our rar1 suppressor screen recover all known defective NB-LRR functions in a rar1 mutant background [21]. However, a previously published rar1 suppressor mutant, sgt1b, only affected a limited number of NB-LRR protein functions [25]. We therefore tested both coi1 rsp alleles to determine whether they have any NB-LRR specificity in their suppression of rar1. The coi1 rsp alleles partially suppress rar1 for RPS5 and RPM1 functions, and fully suppress rar1 for RPS2 function (Figure 2, Figure S2A and S2B, Figure 3A). rar1 exhibits enhanced disease susceptibility to the virulent bacterial strain Pto DC3000(EV) [18,21,25]. This phenotype might be due to a RAR1 function in basal defense, for example an additive effect of globally lowered accumulation of multiple NB-LRR proteins [1][2][3]. As measured by inhibition of bacterial growth, both coi1 rsp alleles completely suppressed the enhanced disease susceptibility phenotype in rar1 ( Figure 3B).
NB-LRR activation can trigger the hypersensitive response (HR) as well as disease resistance responses. RAR1 is required for HR mediated by many NB-LRR proteins. sgt1b is able to suppress the loss of RPS5-mediated disease resistance in a rar1 mutant, but not the loss of RPS5-mediated HR [25]. To test if NB-LRRdependent HR is also recovered in coi1 rsp rar1 double mutants, we measured ion leakage as a proxy for HR to quantify RPM1mediated HR in plants. Notably, the coi1 rsp alleles did not suppress rar1 for impaired RPM1-triggered HR ( Figure S2C). However, the coi1 rsp rar1 plants did recover RPM1-mediated disease resistance, measured via pathogen growth restriction ( Figure 3A).
RPS5, RPM1 and RPS2 all belong to the CC-NB-LRR subclass. The functions of some TIR-NB-LRR proteins also require RAR1. The effect of coi1 rsp on TIR-NB-LRR function was tested using the pathogenic oomycete Hyaloperonospora arabidopsidis (Hpa) isolate Emwa1 to trigger RAR1-dependent RPP4-mediated disease resistance [48]. Neither of the two coi1 rsp rar1 double mutants inhibited the growth of Emwa1 ( Figure S2D). This indicates that RPP4 function is not recovered in rar1 in the presence of either coi1 rsp allele. Thus, the coi1 rsp alleles possibly suppress rar1 only for CC-NB-LRR functions.
The accumulation of all tested NB-LRR proteins is reduced in rar1 plants, implying that the biochemical function of RAR1 is to maintain the stability of NB-LRR proteins [7,18,21,25,49]. We wondered whether coi1 rsp alleles could suppress the decrease of NB-LRR protein accumulation in rar1. We introduced our transgenic, myc-tagged RPM1 [50] into the coi1 rsp rar1 mutants by crossing and marker-assisted selection. The coi1 rsp alleles suppressed the lowered RPM1-myc accumulation in rar1 ( Figure 3C). Hence, the coi1 rsp alleles suppress the biochemical phenotype of rar1.
The coi1 rsp alleles are phenotypically different from two reference alleles, coi1-1 (a protein null (encoding W467STOP [32]; Figure 1) and coi1-16 (encoding L245F [45]; Figure 1), which are also completely or conditionally male sterile. We therefore tested whether either coi1-1 or coi1-16 could suppress rar1. Similar to the coi1 rsp alleles, coi1-1 and coi1-16 enhanced disease resistance responses against both Pto DC3000(avrRpm1) and Pto DC3000(EV) in a rar1 background ( Figure 3A, 3B). The increase in disease resistance against Pto DC3000(EV) was even higher than that caused by the coi1 rsp alleles ( Figure 3B). To our surprise, coi1-16 resulted in the recovery of RPM1-myc accumulation in rar1, but coi1-1 did not ( Figure 3C). However, coi1-16 and coi1-1 express equivalent enhanced disease resistance in rar1. Thus, the ''restoration'' of disease resistance responses against Pto DC3000(avrRpm1) that we observed in coi1-1 rar1 is not due to restoration of NB-LRR protein levels, but rather to bypass suppression of rar1 disease susceptibility. This is likely caused by enhanced basal defense possibly related to the antagonistic relationship between JA-and SA-dependent signaling ( Figure 3). The growth of Pto DC3000(avrRpm1) and Pto DC3000(EV) at 3 dpi was about the same in coi1-16 rar1 plants ( Figure 3). Thus, the restored disease resistance in coi1-16 rar1 is likely due to enhanced basal defense, not RPM1 function, although there is a restoration of RPM1-myc accumulation in coi1-16 rar1.
Since the coi1-1 null allele cannot suppress rar1, we suggest that the coi1 rsp alleles and coi1-16 are recessive gain-of-function alleles for the rar1 suppression phenotypes. They are also loss-of-function alleles for the JA response phenotypes as detailed below.
The coi1 rsp mutations negatively regulate RPM1dependent HR in otherwise wild-type plants We introduced the coi1 rsp alleles into an isogenic RAR1 background using marker-assisted breeding (see Methods). To further study the role of COI1 in regulating RPM1 function, we inoculated both coi1 rsp alleles, coi1-1 and coi1-16 plants with Pto DC3000(avrRpm1) and measured bacterial growth ( Figure 4B). The coi1 rsp and coi1-16 mutants were as resistant as wild type. The coi1-1 mutant displayed slightly enhanced resistance compared with wild type. We also measured RPM1-mediated HR in these coi1 single mutants using the ion leakage assay ( Figure 4C). Surprisingly, both coi1 rsp alleles weakly suppressed RPM1-mediated HR. We crossed RPM1-myc into these coi1 rsp , coi1-1 and coi1-16 single mutants and measured RPM1-myc protein levels ( Figure 3C). We observed no obvious changes in RPM1-myc levels in any of the single coi1 mutant. We conclude from these data that coi1 rsp mutations differentially regulate RPM1 function in rar1 or RAR1 backgrounds.
Increased RPM1 accumulation in coi1 rsp and coi1-16 is post-transcriptionally regulated Loss of COI1 leads to elevated levels of salicylic acid (SA) in plants [37], and elevated SA levels can induce the expression of some NB-LRR-encoding genes [51][52][53]. NB-LRR expression is not changed in rar1 ( Figure S4, [49]). We measured RPM1 mRNA levels in the coi1 rsp , coi1-1, and coi1-16 mutant plants in the context of wild-type RAR1 by RT-qPCR in order to determine whether the increased RPM1-myc protein levels noted in coi1 rsp and coi1-16 were due to enhanced transcription. Wild type and rar1 plants were used as controls. We detected no enhancement of RPM1 mRNA levels among the tested coi1 mutants ( Figure S4), indicating that the coi1 rsp and coi1-16 alleles restore RPM1 protein levels by a post-transcriptional mechanism in rar1.
The coi1 rsp alleles are JA-insensitive COI1 has an essential role in JA signaling; all previously isolated COI1 mutations caused insensitivity to JA-mediated inhibition of seedling growth [32,43,45]. We compared JAinsensitivity phenotypes of the coi1 rsp alleles to coi1-1 using a growth inhibition assay where plants were grown in the presence of MeJA, a functional JA derivative ( Figure 5). Like coi1-1, the MeJA-treated coi1 rsp seedlings grew on MeJA-containing media, while the growth of wild type seedlings was severely inhibited ( Figure 5A). MeJA treated coi1 rsp seedlings were clearly smaller than the untreated seedlings, suggesting that the coi1 rsp alleles are not as insensitive to JA as coi1-1. We quantified these phenotypes with a root elongation assay ( Figure 5B). The null allele coi1-1 displayed root growth inhibition of only about 14% in the presence of 50 mM MeJA. Compared with coi1-1, coi1-16 and both coi1 rsp alleles displayed intermediate insensitivity to MeJA treatment. Their root growth was inhibited about 27%, 30% and 42% respectively, while the root growth inhibition was more than 60% in wild type seedlings. Thus, the coi1 rsp alleles are JAinsensitive.
JA signaling is important in disease resistance responses. coi1 and other JA insensitive mutants exhibit enhanced resistance to the virulent bacterial strain Pto DC3000(EV) [29,37,38]. We measured the growth of Pto DC3000(EV) in our coi1 rsp alleles, coi1-1, and coi1-16 ( Figure 4A). The coi1 rsp alleles also displayed enhanced resistance to Pto DC3000(EV), although the increase in the coi1 rsp alleles was slightly lower than in the reference alleles coi1-1 and coi1-16.

coi1 rsp alleles are not null alleles
The coi1 rsp alleles are quantitatively different than the coi1-1 null allele with respect to JA responses ( Figure 5B) and enhanced resistance to Pto DC3000(EV) ( Figure 4A). We noted decreased COI1 protein accumulation levels in coi1-21 rsp , coi1-22 rsp and coi1-16 plants compared to wild type and rar1 plants ( Figure 3C). As expected, no detectable amount of COI1 protein was observed in coi1-1. The residual accumulations of COI1 protein confirmed that the coi1 rsp alleles and coi1-16 are not COI1 null alleles.

The sgt1b mutant is insensitive to JA responses
To determine whether other NB-LRR regulators function in regulating JA responses, we tested the JA response in the mutants of three NB-LRR co-chaperones, RAR1, SGT1b and HSP90.2 by the root elongation assay ( Figure S3). All rar1 and hsp90.2 mutants were as sensitive to MeJA treatment as wild type, suggesting that neither RAR1 nor HSP90.2, plays a role in JA responses. As expected, the sgt1b mutant displayed an obvious insensitivity to MeJA [54]. We also noted MeJA insensitivity in the rar1 sgt1b double mutant ( Figure S3). These results suggest that SGT1b is the only member of RAR1-SGT1-HSP90 NB-LRR co-chaperone complex required for JA signaling.
COI1 mutations do not affect the levels of RAR1, SGT1b, or HSP90 accumulation coi1 mutations restored the disease resistance responses mediated by three NB-LRR proteins in rar1 ( Figure 3A, Figure S2A and S2B) and thus possibly suppressed rar1 via effects upon NB-LRR regulators that control the accumulation, and hence the function, of multiple NB-LRR proteins. To examine this possibility, we determined the accumulation levels of three NB-LRR regulators, RAR1 ( Figure S5A), SGT1b ( Figure S5B), and HSP90 ( Figure S5C), in the coi1 rsp , coi1-1 and coi1-16 mutants in either RAR1 or rar1 backgrounds. These coi1 mutants did not exhibit any dramatic change of RAR1, SGT1b or HSP90 protein levels. Therefore, the coi1 rsp and coi1-16 alleles do not suppress rar1 influencing by regulating the steady state levels of RAR1, SGT1b and/or HSP90. SGT1b antagonizes coi1 rsp -mediated RPM1 accumulation and RPM1-dependent disease resistance in rar1 The coi1 rsp mutants displayed opposite phenotypes: increased NB-LRR accumulation and function in rar1 and decreased NB-LRR HR function in RAR1. A similar combination of phenotypes was previously observed in sgt1b as an rar1 suppressor [25]. The sgt1b mutation enhanced RPS5 accumulation and consequent restoration of RPS5-mediated disease resistance in rar1, but did not restore RPS5triggered HR in RAR1 [25]. This similarity implies that coi1 rsp mutants might regulate NB-LRR proteins by inhibiting the function of SGT1b and hence mimic sgt1b phenotypes.
Based on this hypothesis, we expected that a high dose of SGT1b would attenuate the rar1 suppression phenotypes of the coi1 rsp mutants. To test this, we introduced a 35S:SGT1b-HA construct into coi1-21 rsp rar1 plants containing RPM1-myc. Compared with parental coi1-21 rsp rar1 plants, four independent T3 lines that expressed relatively high levels of SGT1b::HA exhibited both reduced RPM1-myc levels ( Figure 6A) and RPM1-mediated disease resistance ( Figure 6B). However, the RPM1 accumulation and RPM1-mediated disease resistance observed in these T3 plants were still much higher than rar1 plants ( Figure 6A, 6B). These results demonstrated that modest over-expression of SGT1b can partially inhibit the rar1 suppression phenotypes of coi1 rsp alleles. As a control, we measured the growth of Pto DC3000(EV) in the plants used in the Pto DC3000(avrRpm1) growth assay. No enhanced growth of Pto DC3000(EV) was observed in these T3 lines ( Figure 6C), demonstrating that the reduction of RPM1mediated disease resistance in 35S:SGT1b-HA transgenic plants are not due to a decrease in basal defense.
In addition, we measured the HSP90 protein levels and RPM1myc mRNA levels in the transgenic plants used in the western blot analysis. No obvious decrease of HSP90 protein level ( Figure S6A) or RPM1-myc mRNA level was detected ( Figure S6B), indicating that the reductions of RPM1-myc accumulation in 35S:SGT1b-HA transgenic plants are not due to the decrease of HSP90 accumulation or the silencing of RPM1-myc gene.

Discussion
We initially performed a suppressor screen for mutants that could recover the diminished NB-LRR RPS5-mediated disease resistance phenotype of rar1 [21]. These suppressors were isolated in the null rar1-21 background (Figure S5A), and thus likely represent mutations that either bypass or counteract rar1. We reported two novel HSP90 alleles derived from this screen that function to mimic the effects of RAR1 on the HSP90 lid open/ close cycle required to stabilize NB-LRR clients [21]. Here, we detail the characterization of two coi1 alleles, coi1-21 rsp and coi1-22 rsp also identified in this screen (Figure 1), and we note that a third single allele locus defined by rsp3 has characteristics that suggest it might encode another new player in the regulation of NB-LRR accumulation ( Figure S1). Because rsp3 is a single, dominant allele, its description beyond the mutant phenotype was not pursued as part of this study.
The F-box protein COI1 is a core component of the receptor complex for jasmonate (JA) [28,29,31]. In plants, mutations in COI1 impair all known JA responses and thus result in insensitivity to JA or functional JA derivatives [32,33,37,43,45]. As expected, Bacteria were hand-infiltrated into leaves of each indicated genotype and counted at day 0 and day 3. Error bars represent 26 SE. Pairwise comparisons for all means for bacterial growth on day 3 were performed with One-Way ANOVA test followed by Tukey-Kramer HSD at 95% confidence limits. (C) Conductivity measurements after inoculation with high concentration Pto DC3000(avrRpm1) (5610 7 cfu/ml). Error bars represent 26 SE. The pathogen growth and HR assays were performed independently a minimum of three times with similar results. doi:10.1371/journal.pgen.1003018.g004 both of the coi1 rsp alleles were JA insensitive ( Figure 5). However, the MeJA insensitivity in coi1 rsp alleles is obviously weaker than in the null allele, coi1-1 ( Figure 5B).
In addition to being insensitive to JA, all coi1 alleles identified previously are, at least partially, male-sterile [32,33,37,43,45]. To our surprise, the two coi1 rsp alleles are completely fertile. Among the previously described alleles, only coi1-8 (encoding a missense change of E543K) exhibits partial fertility in regular growth conditions [43]. The other partially fertile allele, coi1-16 (encoding L245F), is fertile only at low temperature (16 degrees C) [45]. Similar to the coi1 rsp alleles, coi1-8 exhibited drastically reduced but still detectable COI1 protein levels [43]. A pull-down assay demonstrated that the COI1-8 E543K protein retains interaction with JAZ1, a substrate of COI1 in SCF COI1 -mediated protein degradation [43]. These results indicate that the weak MeJA insensitivity and intact fertility of the coi1 rsp mutants are likely due to lower accumulation of functional COI1 rsp proteins in these mutants. In other words, the coi1 rsp mutations, G330E and G434E, cause relatively weaker impairments of the COI1 protein stability and activity than the other reported coi1 missense alleles.
In Arabidopsis, resistance against the virulent hemi-biotrophic pathogen Pto DC3000 is a measure of basal defense [56]. In our study, all four tested coi1 alleles, coi1-21 rsp , coi1-22 rsp , coi1-1, and coi1-16 displayed enhanced disease resistance against Pto DC3000(EV) in both rar1 and RAR1 backgrounds ( Figure 3B, Figure 4A). These results correspond to previously published data [29,37,38], and confirm that COI1 represses basal defense, likely via JA-SA antagonism. Besides enhanced basal defense, the coi1 alleles also displayed enhanced ETI against Pto DC3000(avrRpm1) ( Figure 3A, Figure 4B). Hence, COI1 also inhibits ETI. Since the enhancement of ETI was found in rar1 mutant plants, RAR1, which is necessary for NB-LRR-mediated ETI in this and many other cases, is not required by COI1 to repress ETI.
A plausible mechanism explaining COI1 effects on NB-LRR accumulation in rar1 and RAR1 Although all four coi1 alleles we analyzed restored resistance against Pto DC3000(avrRpm1) in rar1 ( Figure 3A), we could classify them into three classes based on how they influence RPM1 accumulation and RPM1-mediated immune response ( Figure 3C, Figure 4C). Class I, represented by the null allele coi1-1, does not alter RPM1 levels. Class II, represented by coi1-16, enhances RPM1 levels in rar1 and has no effect on RPM1-mediated HR in RAR1. Class III, represented by coi1-21 rsp and coi1-22 rsp , enhance RPM1 levels in rar1, but reduce RPM1-mediated HR in RAR1. Since the null coi1-1 does not exhibit any detectable effect on RPM1 accumulation, the enhancement of RPM1 levels in rar1 is a gainof-function phenotype conferred by the COI1 mutant proteins accumulating in coi1-16 and the two coi1 rsp alleles. However, these alleles are all recessive for JA response phenotypes. The coexistence of these distinct genetic characteristics demonstrates that coi1-16 and coi1 rsp alleles are recessive gain-of-function alleles which have lost the JA signaling function of COI1, but gained new function, likely via interfering with the activity of other protein(s). RPM1 is associated with, and activated at, the plasma membrane; there is no current evidence suggesting that it shuttles into the nucleus [50,57]. COI1 is expected to be localized in the nucleus, because it binds to the nucleus-localized JAZ proteins [58]. A biochemical mechanism to explain our genetic results would require a reconciliation of these findings. There may be sufficient coi1 rsp protein at the plasma membrane to mediate the effects on RPM1 that we describe. Further, our inference that COI1 has a wild type function in mediating NB-LRR protein accumulation is consistent with suggestions that nucleo-cytoplasmic shuttling is required for the function of at least a subset of NB-LRR proteins [3].
Since the coi1 rsp alleles did not affect steady state SGT1b levels ( Figure S5B), coi1 rsp alleles might inhibit SGT1b activity to suppress the rar1 phenotype of reduced NB-LRR accumulation. To test this hypothesis, we overexpressed SGT1b in a coi1-21 rsp rar1 background. The rar1 suppression phenotypes of coi1-21 rsp , restored RPM1-myc accumulation and RPM1-mediated disease resistance, were partially complemented by SGT1b overexpression (Figure 6A, 6B). This result supports our hypothesis, and suggests that SGT1 functions with COI1 to regulate NB-LRR accumulation. On the other hand, the incomplete complementation could mean that we need higher levels of SGT1b over-expression, or that coi1 rsp proteins also down-regulate the activity of other targets, such as HSP90. Our speculation is supported by the nonallelic non-complementation observed between coi1 rsp mutants and hsp90.2-7 rsp mutant ( Figure 2, Table S2). This specific genetic relationship suggests that COI1 and HSP90 physically interact with each other or belong to the same protein complex.
The RAR1-SGT1-HSP90 chaperone complex has been related to the SCF complex by two sorts of evidence: 1) SGT1b and HSP90 associate and function with various SCF complexes [13,54,63,64]. RAR1 associates with the COP9 signalosome (CSN) which can inactivate the SCF complex [13,64,65]; 2) The SCF CPR1 complex negatively regulates the pre-activation steady state stability of two NB-LRR proteins, SNC1 and RPS2, via the F-box protein CPR1 [66]. The SCF component SKP1 is required for NB-LRR N protein-mediated resistance response against tobacco mosaic virus (TMV) [64]. This relationship suggests that RAR1-SGT1-HSP90 chaperone complexes function with an SCFmediated protein degradation pathway to control the accumulation levels of NB-LRR protein and thus avoid inappropriate NB-LRR activation [19]. The phenotypes observed in our recessive gain-of-function coi1 rsp mutants support this hypothesis. The coi1 rsp mutants suppressed the rar1 mutant for reduced NB-LRR RPM1 accumulation, and showed non-allelic non-complementation with hsp90.2. Moreover, overexpression of SGT1b partially inhibited the phenotypes of the coi1 rsp mutants. Similar to sgt1b and hsp90.2 lra mutants, coi1 rsp mutants caused impaired HR function when moved to a wild type background. The sum of these results is consistent the idea that the F-box protein COI1 functions with RAR1-SGT1-HSP90 chaperone complex and consequently affects NB-LRR protein accumulation and function.
To make the 35S:SGT1b-HA construct, the coding sequence of SGT1b without its stop-codon was amplified by PCR, and then moved into pGWB14 vector [70]. The final destination vector, pGWB14/35S:SGT1b-HA was electropolated into the Agrobacterium strain GV3101 for transformation of appropriate genotypes. Transformed plants were selected on MS medium plate (PhytoTechnology Laboratories, KS, U.S.) containing Hygromycin B (SIGMA, St. Louis, MO, U.S.).
Hyaloperonospora arabidopsidis (Hpa) isolate Emwa1 was used to inoculated ten-day-old cotyledons of plants as described [21]. Asexual sporangiophores were counted 7 days post-inoculation on at least 30 cotyledons for each genotype.

Identification and map-based cloning of mutations in COI1
The rar1 suppressor screen was previously described [21]. Standard genetic analyses and map-based cloning were performed as described [21]. We used 892 disease resistant F 2 individuals to define a 60 Kb interval on the chromosome II containing COI1. By sequencing COI1 in the originally isolated rar1 suppressor mutant, a G/A transition at position 1849 (nucleotide positions relative to the translation start site of the published sequence of COI1; AT2G39940) was identified in coi1-21 rsp . The other mutant, coi1-22 rsp , also contains a G/A mutation at position 2161 in COI1. To obtain coi1-21 rsp and coi1-22 rsp single mutants, we backcrossed the coi1 rsp alleles into an isogenic RAR1 background. PCR-based dCAP markers were designed for selecting these two coi1 rsp mutations.

MeJA treatment
For growth inhibition assays, seedlings were grown on MS medium with different concentrations of Methyl Jasmonate (MeJA) (SIGMA) at 22uC under 16 h light/8 h dark photoperiod. 10-day-old seedlings were taken picture to show the inhibition effects.
For root elongation assays, seedlings were horizontally grown on MS medium at 22uC under 24 h light for 4 d. Then seedlings were transferred to new MS medium with or without 50 mM MeJA, and grown for additional 4 d. Root elongations during these four days were measured.

Western blots
For detection of RPM1-myc in the genotypes mentioned in this study, we introduced by crossing and segregation the mutants into plants expressing RPM1-myc from the native RPM1 promoter as described [14]. The protein extraction and western blot were performed as described [14]. For detection of SGT1b-HA in plants, the protein extraction and western blot were carried out based on the protocol that was previously used for RPS5-HA [25]. The anti-COI1 antiserum was kindly provided by Daoxin Xie (Tsinghua University, Beijing, China). The protein extraction and western blot were performed as described [42]. anti-SGT1 and anti-RAR1 polyclonal antibodies against the full length SGT1b and full length RAR1 with C-terminus GST tag were generated in rabbits (custom products of Cocalico Biologicals, Inc.). anti-HSP90-2 was the product of Agrisera company (Swedish). The detailed protocols for detection of SGT1a, SGT1b, RAR1, and HSP90 proteins are provided as Text S1.

RT-qPCR
Plant RNA was extracted with RNeasy Plant Mini Kit (Qiagen). To eliminate DNA contamination, RNA was purified by Turbo DNA Free Kit (Ambion) and RNeasy Mini Kit (Qiagen). 2 mg RNA was reverse transcribed with Random Decamers and RETROscript kit (Ambion).
RT-qPCR was performed in a total volume of 25 ml (12.5 ml SYBR Green PCR Master Mix (Applied Biosystems), 0.5 ml cDNA, 1 ml Primer 1 (10 mM), 1 ml Primer 2 (10 mM) and 10 ml H 2 O) with MJ White 96-well plate and a DNA Engine OPTICON 2 system (MJ Research). The reaction was run at 95uC for 5 min, followed by 40 cycles at 95uC for 15 sec, 55uC for 30 sec and 72uC for 30 sec. Dissociation analysis was performed after each reaction to confirm the specificity. The relative expression of RPM1/RPM1-myc gene in different genotypes was calculated by DDCt method (User Bulletin #2, Manual of Applied Biosystems). The primers were newly designed or obtained from previous publication [73], and are provided as Text S1. Figure S1 rsp3 can suppress all known rar1 phenotypes. In addition to two hsp90.2 rsp and two coi1 rsp mutants, we isolated a fifth mutant, named rsp3 (rar1 suppressor 3), from the rar1 suppressor screen. Based on disease symptoms after inoculation with Pto DC3000(avrRpm1), we determined that rsp3 is dominant. The results of bacterial growth experiments showed that the rsp3 mutant suppressed rar1 for disease resistance functions of RPS5 ( Figure S1A), RPM1 ( Figure S1B) and RPS2 ( Figure S1C). rsp3 also fully suppressed the decreased basal disease resistance phenotype of rar1 ( Figure S1D). In addition to disease resistance, RPM1-mediated HR in rar1 was also suppressed by rsp3 ( Figure S1E). We also noted that rsp3 partially suppressed the loss of RPP4 function in rar1 following infection with Hpa Emwa1 ( Figure  S1F). Finally, we also found that rsp3 mutant suppressed the lowered accumulation of RPM1 in rar1 ( Figure S1G). The combined phenotypes of rsp3 rar1 mutant demonstrated that rsp3 suppressed, fully or partially, all known rar1 phenotypes. Using map-based cloning, we localized rsp3 mutation in a 7 Mbp (from 15.9 Mbp to 22.9 Mbp) mapping interval on chromosome I ( Figure S1H). No NB-LRR regulator has been found in this interval. However, because it is a single dominant allele, further characterization of rsp3 is beyond the scope of this paper. (A-B) Bacterial growth assays of (A) Pto DC3000(avrPphB), (B) Pto DC3000(avrRpm1), (C) Pto DC3000(avrRpt2) and (D) Pto DC3000(EV). Leaves of each indicated genotype were dip-inoculated [1].Bacteria were counted at day 0 and day 3. Error bars represent 26 SE; (E) Conductivity measurements after inoculation with high concentration of Pto DC3000(avrRpm1) (5610 7 cfu/ml). Error bars represent 26SE; (F) 10-day-old cotyledons were inoculated with Hpa isolate Emwa1. Asexual sprangiophores were quantified 7 days after inoculation on cotyledons for each of the indicated genotypes [2]. (Sp: sprangiophore); (G) Western blot analysis of RPM1-myc protein levels in indicated genotypes. RuBisCo levels stained by Ponceau S serve as loading control; (H) The positions of the Simple Sequence Length Polymorphisms (SSLP) markers used for rough map-based cloning on chromosome I. rsp3 mutation was localized in the mapping interval between marker T27K12(15.9 Mbp) and F19K23 (22.9 Mbp). The pathogen growth and HR assays were performed independently a minimum of three times with similar results. The RPM1-myc blot displayed is one of three independent blots giving similar results. (TIF) Figure S2 coi1 rsp alleles suppress some, but not all rar1 phenotypes for NB-LRR function. (A-B) Bacterial growth analysis of Pto DC3000(avrPphB) (A) and Pto DC3000(avrRpt2) (B) Leaves of indicated genotypes were dip-inoculated [1]. Bacteria were counted for day 0 and day 3. Error bars represent 26 SE. Pairwise comparisons for all means for bacterial growth on day 3 were performed with One-Way ANOVA test followed by Tukey-Kramer HSD at 95% confidence limits; (C) Conductivity measurements after inoculation with high concentration Pto DC3000(avrRpm1) (5610 7 cfu/ml). Error bars represent 26 SE; (D) 10-day-old cotyledons were inoculated with Hpa isolate Emwa1. Asexual sprangiophores were quantified 7 days after inoculation on cotyledons for each of the indicated genotypes [2]. (Sp: sprangiophore). The pathogen growth and HR assays were performed independently a minimum of three times with similar results. (TIF) Figure S3 coi1 rsp alleles and sgt1b are insensitive to MeJA. Inhibition of root elongation by 50 mM MeJA for the indicated genotypes. The root elongation assay was performed three times with similar results. At least fifteen seedlings per genotype were measured in each repeat. Error bar represents 26SE. (TIF) Figure S4 coi1 rsp and coi1-16 mutations do not enhance RPM1 transcript levels. RT-qPCR analysis of the expression of RPM1 for indicated genotypes. The result displayed is one of three independent RT-qPCRs giving similar results. (TIF) Figure S5 COI1 mutants studied express wild type levels of RAR1, SGT1 and HSP90 proteins. Western blot analysis of SGT1b, SGT1a, RAR1 and HSP90 protein levels for the indicated genotypes. RuBisCo levels stained by Ponceau S serve as loading control. The western blots were performed twice independently with similar results. (TIF) Figure S6 The reductions of RPM1-myc levels in the 35S:SGT1b-HA transgenic plants are not due to the decrease of HSP90 protein level or the silencing of RPM1-myc gene. (A) Western blot analysis of RPM1-myc, SGT1b, SGT1a, and HSP90 protein levels for the indicated genotypes. RuBisCo levels stained by Coomassie Brilliant Blue serve as loading control; (B) RT-qPCR analysis of the expression of RPM1 and RPM1-myc for the indicated genotypes. The western blot and RT-qPCR assay were performed independently a minimum of two times with similar results. (TIF)

Supporting Information
Table S1 Both of the coi1 rsp mutants are completely recessive. (DOC)

Table S2
Non-allelic non-complementation between coi1 rsp and hsp90 rsp mutants. RPM1-mediated resistance was tested by sprayinoculation with Pto DC3000(avrRpm1). Disease symptoms were evaluated 5 days after inoculation. (DOC) Text S1 Supporting information including the primers and corresponding enzymes for selecting of mutations, the primers for making 35S::SGT1b-HA construct, the primers used for RT-qPCR, and additional information of Western blots. (DOC)