MAP kinases associate with high molecular weight multiprotein complexes

Depending on their phosphorylation status, MAP kinases (MAPKs) dynamically associate with stable high molecular weight multiprotein complexes.


Introduction
Some mitogen-activated protein kinases (MAPKs) respond to almost any disturbance of cellular homoeostasis, ranging from the seemingly innocuous touch stimulus to severe abiotic and biotic stress factors such as ultraviolet light, extreme temperatures, pathogen infection, and herbivory (Ichimura et al., 2000;Stratmann, 2008;Rodriguez et al., 2010;Sinha (Meng and Zhang, 2013). Two prominent MAPKs are MPK3 and MPK6 from Arabidopsis thaliana and their orthologs in other plants. In addition to regulating responses to environmental factors, these MAPKs are also involved in the regulation of developmental processes (Lampard et al., 2008;Wang et al., 2008;Zhou et al., 2009;Beck et al., 2010;Xu and Zhang, 2015). This astonishing pleiotropy is tied to signaling mechanisms that faithfully activate input-specific responses. Several mechanisms of MAPK signaling fidelity are known from animal systems, but they are poorly understood in plants.
MAPKs are components of the three-tiered MAPK module. MAPK kinase kinases (MAPKKKs) are activated by receptors and sensor proteins, and in turn activate MAPK kinases (MAPKKs) via phosphorylation. MAPKKs then phosphorylate and activate MAPKs. The number of putative MAPKKKs is large, with 60-80 in Arabidopsis (MAPK Group, 2002;Champion et al., 2004), 75 in rice (Rao et al., 2010), and 89 in tomato , whereas the number of MAPKKs is confined to 10 in Arabidopsis, 8 in rice, 11 in poplar (MAPK Group, 2002;Hamel et al., 2006), and only 5 in tomato . In contrast, the number of MAPKs is generally higher than the number of MAPKKs, with 20 members in Arabidopsis, 15 in rice, 21 in poplar (MAPK Group, 2002;Hamel et al., 2006), and 16 in tomato (Stulemeijer et al., 2007;Kong et al., 2012). Predictions strictly based on numbers seem to indicate that MAPKKs may function as convergence points for MAPKKKs, but the number of thoroughly studied MAPKKKs is too low to draw firm conclusions. Some MAPKs also represent convergence points for a number of MAPKKs. Arabidopsis MPK3 can interact with six and MPK6 with eight of the 10 MAPKKs (Andreasson and Ellis, 2010). The theoretical number of MAPK modules as a consequence of potential permutations of the three constituents is very high. Therefore, it is predicted that the same MAPK may engage with different substrate proteins depending on the specific MAPKKK/ MAPKK module it associates with. Indeed, a number of different in vivo substrates of MPK3 and MPK6 were identified (Meng and Zhang, 2013) and hundreds of putative MPK3/6 substrates were found in in vitro proteomic analyses (Popescu et al., 2009;Hoehenwarter et al., 2013;Lassowskat et al., 2014). In addition, MPK6 interacts with at least eight, and MPK3 with at least seven MAPK phosphatases that can dephosphorylate and inactivate them (Andreasson and Ellis, 2010;Bartels et al., 2010;Umbrasaite et al., 2010).
A fundamental question in signal transduction is how a signal relay such as a MAPK, which is shared among diverse input signals, can generate specific output responses.
The critical factor appears to be signaling dynamics, namely how MAPKs dynamically interact with other proteins in a spatiotemporal manner. A number of hypotheses exist to explain how MAPKs can accomplish signaling fidelity. The above-mentioned combinatorial diversity of module constituents can potentially result in the assembly of a large number of alternative MAPK modules with differential functions. These modules may combine with additional proteins such as scaffold proteins or MAPK-inactivating phosphatases to form larger multiprotein complexes. Assembly of distinct modules and complexes may result in different MAPK activation kinetics (speed of activation, and amplitude and duration of activity) with consequences for the output responses. Mechanisms that interpret and process MAPK kinetics include interactions of MAPKs with different substrates and other proteins. These interactions may occur in various subcellular compartments and tissues. In addition, MAPKs and/or their interacting partners may undergo subcellular translocations.
In non-plant systems, scaffold proteins play an important role in organizing specific MAPK modules to achieve signaling specificity. A MAPK scaffold protein is generally a nonenzymatic protein that interacts with at least two components of a MAPK module and often also with additional regulatory proteins. In some cases scaffold proteins themselves have catalytic activity; for example, PBS2 from yeast functions as a scaffold and as a MAPK kinase (Posas and Saito, 1997). A scaffold protein may regulate the subcellular localization of MAPKs, participate in positive and negative feedback loops, alter accessibility of MAPKs for MAPK-regulating proteins such as phosphatases or MAPKKs, and insulate specific MAPK modules from interactions with other MAPK modules (Whitmarsh, 2006;Shaw and Filbert, 2009;Zeke et al., 2009). Through these functions, a scaffold protein may have a profound influence on the outcome of MAPK signaling. The animal scaffold protein KSR binds all three components of the MAPK cascade and additional proteins such as 14-3-3 proteins, a phosphatase, and some heat shock proteins in a tissue-specific manner. In the absence of stimuli, it is present in the cytosol, but can translocate to the plasma membrane upon stimulation. Also, binding of its protein clients can occur in an input-specific manner (Kolch, 2005;Shaw and Filbert, 2009). These features highlight the dynamics of complex assembly and disassembly that are critical to achieve signaling fidelity. Recently, RACK1 was discovered as the first MAPK scaffold protein in plants (Cheng et al., 2015;Su et al., 2015). Arabidopsis RACK1 binds to the MAPKKK MEKK1, the MAPKKs MKK4 and MKK5, and the MAPKs MPK6 and MPK3.
We used gel filtration (GF) (size-exclusion chromatography) to separate high molecular weight (HMW) protein complexes from monomeric proteins in plant extracts. Probing GF fractions with antibodies against MAPKs, we identified various MAPKs as well as a MAPKK in HMW GF fractions (~250-550 kDa), indicating that MAPKs associate with multiprotein complexes. The presence of MAPK-containing multiprotein complexes was confirmed in five plant species from two different families (Solanaceae and Brassicaceae), showing that this is a general phenomenon. Only non-phosphorylated (i.e. inactive) MAPKs are present in the HMW GF fractions, unless dephosphorylation is blocked by the phosphatase inhibitor cantharidin. The predicted size of the complex suggests that MAPKs and MAPKKs can also interact with additional unknown proteins.

Immunoblot analysis
Proteins were boiled for 5 min before being separated via electrophoresis on a 10% polyacrylamide gel. Separated proteins were transferred to Immobilon-P PVDF membranes (Millipore, Billerica, MA, USA) using a mini transblot electrophoretic transfer cell (Biorad) for 1 h at 100 V in cold transfer buffer (48 mM Tris, 38 mM glycine). The membranes were blocked for 1 h in TBS-Tween-20 [10 mM Tris-HCl pH 7.5, 150 mM NaCl, 0.1% (v/v) Tween-20] containing 5% BSA (Fraction V, Fisher, Waltham, MA, USA). Membranes were washed twice with TBS-Tween-20 and then incubated with the primary antibody overnight at 4 °C with gentle rocking. Blots were washed five times in TBS-Tween-20 and then incubated with an alkaline phosphatase-conjugated secondary antibody (Sigma-Aldrich). After five washes in TBS-Tween-20, blots were analyzed by a chemiluminescence assay by incubating the blots for 5 min in LumiPhos (Pierce, Rockford, IL, USA) or Immuno-Star AP (Biorad) and visualized with HyBlot CL autoradiography film (Denville Scientific Inc., Metuchen, NJ, USA).
To generate SlMKK4-specific antibodies, the tomato (S. lycopersicum) MKK4 cDNA (GenBank accession no. AY691333) was expressed in Escherichia coli as an N-terminally His-tagged recombinant protein (Novagen pET28 system) and purified on a nickel column according to standard procedures (Novagen-EMD Millipore). Purified Coomassie G250-stained SlMKK4 protein in polyacrylamide gel was used to raise antibodies in chicken eggs (Cocalico Biologicals, Inc., Reamstown, PA, USA). Pre-bleeds did not contain cross-reacting bands at the predicted molecular weight of SlMKK4 (~37.5 kDa). For immunoblotting of plant leaf extracts, anti-SlMKK4 antibodies (IgY) were detected by incubation with anti-IgY secondary antibodies conjugated to alkaline phosphatase (Sigma-Aldrich) and a subsequent chemiluminescent assay as described above. To demonstrate antibody specificity, anti-SlMKK4 antibodies were pre-incubated with the recombinant MKK4 protein for 45 min prior to being used for immunoblotting.

Generation of vectors for virus-induced gene silencing
The modified Tobacco rattle virus (TRV) vectors, pTRV1 and pTRV2 (Liu et al., 2002), used for virus-induced gene silencing (VIGS) were obtained from S.P. Dinesh-Kumar (UC Davis, CA, USA). A portion of each target gene was cloned into the multiple cloning site (MCS) of the pTRV2 vector. A fragment for each target gene, which was predicted in silico to have minimal off-target gene silencing, was amplified from a cDNA library generated from S. lycopersicum cv. Rio Grande. For the VIGS-mpk1/2/3 construct, SlMPK1 and SlMPK2 (accessions Solyc12g019460 and Solyc08g014420; GenBank accessions nos AY261512 and AY261513) were co-silenced using a 577 bp fragment that was PCR amplified from the MPK1 ORF (primers: MPK1/2F 5'-GCGCGAGCTCCATGGTGGCAGGTTCATTC-3' and MPK1/2R 5'-CGGCGCTCGAGGCTCAGGTCCACGATA CCAT-3'). Tomato MPK1 and MPK2 are 89% identical at the nucleotide level. MPK3 (accession Solyc06g005170; GenBank accession no. AY261514) was specifically silenced using a 406 bp fragment amplified from the 3'-untranslated region (Primers: MPK3F 5'-GGCCGTCTAGAGCATAA GAGAAATCAGTTCTTC-3' and MPK3R 5'-CGCGCGGATCCACACCC AAAACTTCAAAAT GAC-3'). Forward and reverse primers were designed with the flanking restriction site pairs SacI/XhoI and XbaI/BamHI for ligation into the MCS of pTRV2 for MPK1/2 and MPK3, respectively. The resulting pTRV contains the 577 bp target sequence for MPK1/2 fused to the 406 bp sequence of MPK3 with the restriction site KpnI between the fragments. For the control vector, a 350 bp fragment of green fluorescent protein (GFP) was ligated into pTRV2 as previously described (Kandoth et al., 2007). Ligation products were expressed in One Shot DH5α competent cells (Invitrogen/Life Technologies, Grand Island, NY, USA). The sequence of the pTRV vectors was confirmed by DNA sequencing (Eton Biosciences Inc., Durham, NC, USA). The pTRV1 and modified pTRV2 vectors were transfected into Agrobacterium tumefaciens strain GV3101. A single colony was selected and inoculated into a liquid culture supplemented with kanamycin (50 μg ml -1 ), rifampicin (50 μg ml -1 ), and gentamycin (30 μg ml -1 ). The culture was incubated overnight at 30 o C with shaking at 250 rpm. To prepare the infiltration media, the overnight culture was centrifuged at 5000 g for 10 min at 4 o C. The pellet was washed twice with 10 mM MgCl 2 and resuspended in infiltration medium [10 mM MgCl 2 , 10 mM MES, and 150 μM acetosyringone (3', 5'-dimethoxy-4'-hydroxyacetophenone, Acros Organics, New Jersey, USA)] to a total OD 600 of 0.2 and 0.4 for pTRV2 and pTRV1, respectively. Infiltration medium was introduced into the apoplastic space in the cotyledon leaves of 10-to 14-day-old seedlings using a needleless syringe. Plants were sampled 4 weeks after infiltration.

Vector construction and transient transformation of N. benthamiana
Full-length tomato MPK1 and MPK2 (MPK1F 5'-ATGGATGGTTCCGTTCCGC-3', MPK1R 5'-TCACATGCG CTGGTATTCAGGAT-3', MPK2F 5'-ATGATGGTTCAGCTCCG CA-3', and MPK2R 5'-TCACATGTGCTGGTATTCGGGAT-3') were amplified from a Rio Grande cDNA library and cloned into the pCR™/GW/TOPO ® vector (Life Technologies). TOPO vector was transformed into competent DH5α E. coli cells, and plasmid DNA was isolated from positive transformants. Isolated plasmid was recombined into the Gateway ® vectors pEarleyGate 202 (N-terminal FLAG tag) or pEarleyGate 203 (N-terminal c-Myc tag) (Earley et al., 2006) via an LR recombination reaction using LR Clonase ® II enzyme mix and transformed by electroporation into One Shot ® Mach1™-T1R chemically competent E. coli (Life Technologies). Positive transformants were determined by restriction digest and confirmed by sequencing. Confirmed plasmids were transformed into A. tumefaciens strain GV3101. One colony was selected and used to start an overnight culture in 3 ml of LB medium. Cells were harvested by centrifugation, washed twice with 10 mM MgCl 2 , and resuspended in infiltration medium (10 mM MgCl 2 , 10 mM MES, and 150 μM acetosyringone) to a final OD 600 of 0.2-0.4. Using a needleless syringe, infiltration medium was forced into the abaxial side of N. benthamiana leaves, which were sampled 3-5 d later.

FLAG immunoprecipitation
For anti-FLAG immunoprecipitation (IP), GF fractions were collected and samples were pooled from fractions corresponding to HMW or low molecular weight (LMW) fractions (four fractions equaling 2 ml total for each). The IP was performed with an anti-FLAG M2 affinity gel using conditions recommended by the manufacturer (Sigma-Aldrich) with minor variations. Briefly, 1% (v/v) Triton X-100 (Sigma-Aldrich) and 0.5% (v/v) Igepal CA-630 were added to each pooled sample following GF. The affinity gel was prepared as described by the manufacturer using at least 40 μl of resin for each sample. Pooled fraction samples were added to the prepared resin and incubated with tumbling for 2 h at 4 o C. Samples were centrifuged and supernatant removed, then washed three times with TBS buffer (50 mM Tris-HCl,150 mM NaCl, pH 7.4). Samples were eluted with 3× SDS gel loading buffer. Samples were boiled for 5 min then subjected to SDS-PAGE and immunoblot analysis as described above or to Coomassie G250 or R250 staining.

Tomato suspension culture treatments
For treatments of samples used for immunoblotting (IB), 1.5 ml of S. peruvianum suspension-cultured cells were added to each well of 6-or 12-well tissue culture plates (BD Biosciences, San Jose, CA, USA) and shaken on an orbital shaker at 150 rpm under ambient room light and temperature. Cell suspensions were equilibrated for 1 h prior to treatment. For GF experiments, cells (~45 ml) in 125 ml Erlenmeyer flasks were directly treated and placed on an orbital shaker at 200 rpm in ambient room temperature and light. Then 5 ml of cells were collected at each time point for further analysis. Cantharidin (Enzo Life Sciences, Inc., Farmingdale, NY, USA) was dissolved in DMSO and added to the cells for a final concentration of 500 μM. Systemin and flg22 (GenScript, Piscataway, NJ, USA) were dissolved in water and added to the cells at a final concentration of 10 nM. Cells in multiwell plates were irradiated with UV-C (two 15 W UV-C lamps; GE Germicidal Lamp G15T8) at a distance of 10 cm for 5 min while shaking. Cells were removed from the suspension medium using a Büchner filter funnel lined with miracloth, and the cells were immediately flash-frozen in liquid nitrogen.
In-gel kinase assay Systemin-treated cell suspension extracts were analyzed by in-gel kinases assays (IGKAs) using myelin basic protein (MBP) as an artificial MAPK substrate as previously described (Stratmann and Ryan, 1997).

MAPKs are present in high molecular weight multiprotein complexes
Association of plant MAPK modules with other proteins such as scaffold proteins should lead to the formation of large multiprotein complexes. The molecular weight of the complex consisting of the human KSR scaffold (~100 kDa), the MAPKKK Raf-1 (~73 kDa), the MAPKK MEK1/2 (~43 kDa), and the MAPK ERK1/2 (~43 kDa) is ~260 kDa.
The size further increases with additional proteins that interact with the core components. To determine whether plant MAPKs can also be found in HMW multiprotein complexes, we separated plant protein complexes and monomeric proteins by GF. The column used here (Superose 6 10/300 GL) is designed to separate HMW molecules and complexes according to size, but has a poor resolution for LMW proteins, which tend to elute over a range of fractions. To better compare GF elution profiles in the figures shown, the elution volumes in increments of 0.5 ml fractions eluting from the GF column are indicated below the profiles, and the peak elution of molecular weight marker proteins, which provide a size estimate of the eluted protein complexes, are shown above the profiles.
We probed concentrated proteins from GF fractions by IB with an anti-AtMPK6 antibody, which was generated against A. thaliana MPK6 (AtMPK6). Using VIGS, we showed that this antibody also recognizes orthologs in tomato (S. lycopersicum) (SlMPK2 and/or SlMPK1) (Fig. 1B). Based on our IB data (Figs 1C,5) and the phylogenetic relationship among AtMPK6 orthologs (see Supplementary Fig. S1 at JXB online), this antibody probably recognizes AtMPK6 orthologs in the wild tomato species S. peruvianum (SpMPK1 and/or SpMPK2), and in tobacco (SIPK and/or NTF4). These MAPKs all have an apparent molecular weight of ~48 kDa. No bands corresponding to the lower apparent molecular weight of AtMPK3 and AtMPK4 orthologs were Fig. 1. Identification of a MAP kinase-containing multiprotein complex in tomato and tobacco. Total protein was extracted from leaf tissue and separated by GF. Eluate was collected in 0.5 ml fractions, concentrated, and analyzed by immunoblotting (IB). Only fractions representing 13-18.5 ml are shown (numbers underneath the immunoblots). The numbers above the panels indicate the peak elution of molecular weight standards in kDa. Input samples represent 30 μg of total protein extracted from leaf tissue and represent the same protein extracts that were used for GF. detected in these species, indicating that the anti-AtMPK6 antibody specifically recognizes MPK6 orthologs.
The GF elution profile from S. lycopersicum leaf extracts is biphasic. An ~48 kDa band was prominent in <100 kDa GF fractions (fractions ≥17 ml) and represents monomeric MPK1/2 (Fig. 1A). Based on the resolution of the Superose column, it cannot be excluded that these fractions contain MPK6 associated with another relatively small protein, but, as a working hypothesis, we will refer to the MPKs in LMW fractions (<100 kDa) as monomeric MPKs. In addition, we detected an ~48 kDa band in HMW GF fractions (Fig. 1A, fractions 13.5-15), corresponding to protein complex sizes between 300 kDa and 550 kDa.
In tomato plants, we used VIGS to reduce the expression of three highly homologous MAPK genes, SlMPK1, SlMPK2, and SlMPK3 (Holley et al., 2003;Kandoth et al., 2007). Sequences corresponding to the MAPKs were inserted into the genome of the TRV encoded on a binary vector, followed by transformation of the manipulated TRV vector into A. tumefaciens and delivery into plants by cotyledon infiltration. Four weeks later, plants were analyzed for silencing of MAPK genes. In leaf extracts of negative control plants that were infiltrated with TRV containing a fragment of GFP, SlMPK1/2 was detected by the anti-AtMPK6 antibody in HMW and LMW fractions (Fig. 1B, upper panel), similarly to uninfected plants (Fig. 1A). In contrast, in VIGS-mpk1/2/3 plants, only low amounts of SlMPK1/2 protein were detected (Fig. 1B, lower  panel). Additionally, an antibody against SlMPK2 (Holley et al., 2003) also recognized a band in HMW and LMW GF fractions from control plants, and this band was reduced in extracts from VIGS-mpk1/2/3 plants (see Supplementary Fig. S2). An anti-SlMPK3 antibody suitable for IB was not available, and a frequently used anti-AtMPK3 antibody (Sigma-Aldrich) does not recognize the tomato ortholog SlMPK3.
GF/IB analysis with anti-AtMPK6 antibody also detected ~48 kDa proteins in HMW GF fractions of leaf extracts from tobacco (N. tabacum) (~200-450 kDa; Fig. 1C) and N. benthamiana (Supplementary Fig. S3A; 300-500 kDa), as well as in extracts from S. peruvianum suspension cell cultures (~300-450 kDa) (Fig. 4). The double band in N. tabacum HMW fractions (Fig. 1C) may represent SIPK and NTF4, which are ordinarily not separated by SDS-PAGE (Ren et al., 2006). Figure 4 shows that the two bands do not represent the phosphorylated and unphosphorylated forms of a SIPK/ NTF4; however, they may have additional post-translational modifications that change their electrophoretic mobility. This could also be the case for the double band (MPK1/2) in LMW fractions of S. lycopersicum (Fig. 1B), although this pattern was not reproducible.
To demonstrate further the presence of MPKs in HMW GF fractions, N. benthamiana plants were transiently transformed with A. tumefaciens containing a c-myc-tagged SlMPK1 or SlMPK2. Both c-myc-tagged SlMPK1 and SlMPK2 proteins were incorporated into a HMW protein complex (Fig. 1D). The GF elution profiles for both the c-myc-tagged and the FLAG-tagged MAPKs was somewhat different as compared with GF profiles of untagged MAPKs, perhaps as an artifact of overexpression. Tagged MAPKs were also present in the fractions between the LMW and HMW fractions (100-300 kDa). In addition, transient expression of a FLAG-tagged SlMPK1 in N. benthamiana leaves and subsequent GF/IB of leaf extracts with an antibody against the FLAG tag resulted in detection of an ~48 kDa band in HMW and LMW fractions. However, there was also a cross-reacting band of a similar molecular weight present in untransformed control plants ( Supplementary Fig. S3B), although the signal was much weaker than in extracts from leaves expressing SlMPK1-FLAG. In addition, FLAG-tagged proteins in HMW and LMW fractions from leaf extracts were first immunoprecipitated with an anti-FLAG antibody to enrich the proteins, then analyzed by IB with the same anti-FLAG antibody. Transformed plants contained an ~48 kDa FLAG-tagged protein, which was more abundant in LMW GF fractions, but also clearly present in HMW GF fractions ( Supplementary Fig. S3C).
In A. thaliana, many MAPK mutant lines are available, which allowed us to ascertain specifically which MAPK components were present in the HMW versus the LMW fractions. In leaves of Arabidopsis ( Fig. 2A) a band with a size range Fig. 2. Identification of a MAP kinase-containing multiprotein complex in Arabidopsis thaliana. (A) GF/IB analysis of leaf extracts from A. thaliana ecotype Columbia-0 (Col-0, 1.5 mg of total protein), and the null mutants mpk6-2 (1.5 mg), mpk10-1 (0.5 mg), and mpk10-2 (0.5 mg). AtMPK6 and AtMPK10 were detected using anti-AtMPK6 antibody. (B) GF/IB analysis of leaf extracts from A. thaliana ecotype Col-0 (1.5 mg of total protein) and the mpk3-1 null mutant (0.75 mg). AtMPK3 was detected using anti-AtMPK3 antibody. Note: the Col-0 column on the left shows extracts (30 µg of total protein) from Col-0 as a reference for MPK6 (A) or MPK3 (B). These samples were analyzed on the same blot as the corresponding GF fractions, but additional bands were removed to adjust the elution profiles among panels. This is indicated by the white line that separates reference samples and GF fractions. For the two mpk-10 rows, shorter exposures for the Col-0 reference lane are shown as compared with the GF fractions. Numbers above and below the immunoblots are as described in Fig. 1. Immunoblots for Col-0 and mpk6-2 represent four independent experiments; and those for the two independent mpk10 mutant lines were performed once for each line.
between 300 kDa and 450 kDa was detected in HMW GF fractions after probing immunoblots of GF fractions with the anti-AtMPK6 antibody. However, the apparent molecular weight of the protein in HMW fractions is lower (indicating a size of ~45 kDa) than the molecular weight of the proteins in LMW fractions (~48 kDa). Importantly, this band was still present in the mpk6-2 null mutant, whereas the 48 kDa band in the LMW GF fractions was absent. A very weak band at ~45 kDa persisted in LMW fractions 16.5 and 17 ( Fig. 2A), indicating that the anti-AtMPK6 antibody recognizes an additional protein, which is present as a monomer but more abundant in a HMW protein complex. The C-terminal 12 amino acid peptide sequence of AtMPK6 against which the anti-AtMPK6 antibody was raised (REALAFNPEYQQ) shares eight amino acids with a sequence close to the C-terminus of the highly homologous AtMPK10 (EALAFNPE) (Supplementary Fig. S1). In two different mpk10 T-DNA insertional mutants (for characterization, see Supplementary Fig. S4), the 45 kDa bands detected by the anti-AtMPK6 antibody in the HMW GF fractions were absent, while the 48 kDa band (MPK6) in the LMW fractions persisted ( Fig. 2A). Together, this indicates that in Arabidopsis, MPK10 rather than MPK6 associates with a HMW protein complex and that the majority of the monomeric MPK protein in the LMW fractions is MPK6 protein. MPK10 is a closely related paralog of MPK6 ( Supplementary Fig. S1) and plays a role in polar auxin transport (Stanko et al., 2014). A role in stress responses has not been investigated. In Col-0, the elution profile of MPK6 in LMW fractions ranges from fraction 16 (150 kDa) to fraction 18 (<66 kDa). In the mpk10 mutants, a small fraction of MPK6 protein elutes in fraction 16.5 and the majority in fractions representing molecular weights of <66 kDa, which is more in line with elution profiles of extracts from solanaceous plants shown in Fig. 1. The differences in the elution profiles of Col-0 and mpk10 mutants are most probably due to different amounts of total protein analyzed confounded by the low resolution of the column for LMW proteins, and do not necessarily indicate an association of Arabidopsis MPK6 with other proteins.
Besides MPK10, MPK3 also associated with a HMW protein complex in Arabidopsis leaves (Fig. 2B), as shown by GF followed by IB with an anti-AtMPK3 antibody. Like MPK10, MPK3 protein was detected in HMW GF fractions ranging from 300 kDa to 450 kDa. All MPK3 bands were absent in the mpk3-1 null mutant.

MKK4 is part of a HMW multiprotein complex
If MAPKs associate with a protein complex to facilitate interactions with specific activating proteins such as MAPKKs, then these kinases may also be present in HMW protein complexes. We generated a specific antibody against tomato MKK4 (D-group of MAPKKs) (Fig. 3A). In tomato, MKK4 and MKK2 are the upstream MAPKKs that activate the MAPKs MPK1 and MPK2 (Pedley and Martin, 2004). We probed GF fractions of tomato leaf tissue with this antibody. MKK4 showed a biphasic elution pattern from the GF column with a major peak in the LMW fractions at <100 kDa and a minor peak in HMW fractions at >440 kDa (Fig. 3B). The MKK4 band was also detectable in fractions between the two peaks, unlike the MAPK bands. This demonstrates that MKK4 is also associated with HMW multiprotein complexes.

Phosphorylated MAPKs do not normally associate with a HMW multiprotein complex
The antibodies described so far do not distinguish between active and inactive MAPKs that are or are not phosphorylated on the threonine and tyrosine of the TEY MAPK phosphorylation motif. To test whether TEY-phosphorylated MAPKs associate with a HMW multiprotein complex, we probed GF fractions with anti-pERK antibody. This antibody was raised against the dually phosphorylated TEY motif of human p44 MAPK. Anti-pERK is widely used to demonstrate the presence of phosphorylated MAPKs across kingdoms, including plants (Hann et al., 2014;Buscà et al., 2015).
Probing with anti-AtMPK6 shows that SpMPK1/2 is present in both LMW and HMW fractions from extracts of both Fig. 3. Solanum lycopersicum MKK4 associates with a multiprotein complex. (A) Extracts from N. tabacum (Nt) and S. lycopersicum (Sl) leaves as well as affinity-purified recombinant SlMKK4 protein (RP) were analyzed by IB using either anti-SlMKK4 antiserum (left panel) or anti-SlMKK4 antiserum pre-incubated with recombinant MKK4 protein (right panel). Note that the lane showing purified MKK4 protein contains MKK4 degradation products. The binding of free MKK4 by the antibody prevented it from recognizing MKK4 on the immunoblot membrane. The MKK4 ortholog from N. tabacum (MKK9; accession no. NP_001311802) shares 88% identity with SlMKK4 at the amino acid level. Without a stretch of 16 amino acids that is absent in NtMKK9, the identity is 93%. This explains why the anti-SlMKK4 antibody recognizes a protein in N. tabacum, presumably NtMKK9. (B) GF/IB analysis of leaf extracts (1.5 mg) from S. lycopersicum plants. SlMKK4 was detected using anti-SlMKK4 antibody. Numbers above and below the immunoblots are as described in Fig. 1. All immunoblots represent a minimum of two independent experiments. treated and untreated suspension-cultured cells (Fig. 4A, B). This indicates that only unphosphorylated (i.e. inactive) SpMPK1/2 is present in the HMW multiprotein complex.
The situation was similar in leaf tissue of tobacco, where wound-induced phosphorylation of SIPK was only detected in LMW GF fractions (Fig. 4C).
According to the manufacturer, anti-pERK may also recognize MAPKs singly phosphorylated at the T of the TEY motif. Singly phosphorylated MAPKs are not enzymatically active (Robbins et al., 1993). In order to test whether enzymatically active MAPKs are also confined to LMW GF fractions, IGKAs were performed on extracts from S. peruvianum suspension cells treated with systemin (Fig. 4B) and from tomato plants 10 min after wounding (Supplementary Fig.  S5C). These results confirm the IB phosphorylation assays and together demonstrate that phosphorylated and enzymatically active MAPKs are only present in LMW GF fractions.

In the presence of a phosphatase inhibitor, phosphorylated MAPKs do associate with a HMW multiprotein complex
Using IGKAs and IB with anti-pERK antibodies, we were unable to detect phosphorylated MAPKs in HMW GF fractions in untreated and in stimulated S. peruvianum cells (Fig. 4). Notably, the ratio of complex-associated MAPKs to monomeric MAPKs remained unchanged up to 180 min after treatment with systemin or flg22 ( Fig. 4; Supplementary  Fig. S5D). To explore further the phosphorylation state of complex-associated MAPKs, we treated S. peruvianum cells with the phosphatase inhibitor cantharidin.
Cantharidin is a moderate (nM to high µM range) inhibitor of several serine/threonine protein phosphatases, notably PP2A (Pereira et al., 2011). Treatment of suspension-cultured cells with cantharidin induced a biphasic time course of MPK6 activity with a rapid early peak at 10 min and return to background levels, followed by a second slower increase starting 60-120 min after treatment and lasting for at least 6 h. During this second phase, two additional bands (~46 kDa and ~44 kDa) were detected by anti-pERK, most probably representing increased activity of SpMPK3 and the tomato ortholog of AtMPK4 (Fig. 5A). Extracts from cantharidin-treated cells were analyzed by GF/IB with anti-pERK and anti-AtMPK6 antibodies. Anti-pERK detected a phosphorylated 48 kDa MAPK in HMW and LMW fractions at 120-360 min after cantharidin treatment (Fig. 5B, C), but not in untreated cells (Fig. 4) and not during the first phase of cantharidin-induced MAPK activity at 10 min (Fig. 5A, C). Reprobing the immunoblots with anti-AtMPK6 antibody confirmed the presence of SpMPK1/2 in the same fractions that contained the phosphorylated 48 kDa MAPKs. No phosphorylated 46 kDa and 44 kDa MAPKs were detected in HMW fractions (Fig. 5B, C). These results indicate that cantharidin inhibition of MAPK dephosphorylation by phosphatases results in increased SpMPK1/2 activity over time and in association of phosphorylated SpMPK1/2 with a HMW protein complex.

Discussion
A number of more or less complete MAPK signaling pathways and networks have been described in plants, and the Fig. 4. Active phosphorylated MPK1/SIPK and MPK2/Ntf4 do not normally associate with a multiprotein complex in tomato and tobacco. Solanum peruvianum suspension-cultured cells were treated with 10 nM flg22 or 10 nM systemin. Total protein was extracted from suspension cells 0 and 10 min after elicitor treatment. Inputs represent 30 μg of total protein extracted from cells collected at the times indicated after treatment. Numbers above and below immunoblots are as described in Fig. 1. (A) GF/ IB analysis of extracts (0.9 mg of total protein) from untreated suspension cells or cells treated with flg22, probed with anti-AtMPK6 antibody for the detection of total MPK1/2 and anti-pERK antibody for the detection of phosphorylated MPK1/2 (p-MPK1/2). (B) The same as (A) but after treatment with systemin (0.75 mg of total protein). The additional panel shows enzymatic activity of MPK1/2 as determined by in-gel kinase assays (IGKAs). Total protein (1.5 mg) was separated by GF and each fraction was analyzed by IGKA. Signals represent 32 P-phosphorylated myelin basic protein. (C) GF/IB analysis of extracts (1.2 mg of total protein) from tobacco leaves left untreated or wounded and sampled 10 min later, probed with anti-AtMPK6 antibody for the detection of SIPK/Ntf4 and anti-pERK antibody for the detection of phosphorylated SIPK/Ntf4 (p-SIPK/ Ntf4). (A) and immunoblots in (B) are representative of four independent experiments; IGKAs in (B) and (C) of two independent experiments. For tobacco, similar results were seen for 180 min after wounding ( Supplementary Fig. S5B). *Unrelated protein that cross-reacts with anti-AtMPK6 antibodies. function of specific MAPK pathway components has been revealed. However, it is still poorly understood how MAPK modules are organized in response to an input signal. Several proteins have been identified in plants that interact with components of the MAPK pathways including phosphatases, 14-3-3 proteins, and MAPK substrate proteins (Andreasson and Ellis, 2010;Oh et al., 2010;Rodriguez et al., 2010), but it is largely unknown how, when, and where they interact with a MAPK module. In animal cells, MAPK-containing multiprotein complexes have been well characterized. With the exception of the recently discovered RACK1 complex, no such large complexes have been discovered in plant cells thus far. Examples of additional but smaller plant MAPK complexes include a nuclear complex consisting of MPK4, the MPK4 substrate MKS1, and the transcription factor WRKY33 (Qiu et al., 2008), and a potentially larger complex which contains OMTK1, an alfalfa MAPKKK. OMTK1 interacts directly with the MAPK MMK3, combining in one protein the kinase and the scaffolding function (Nakagami et al., 2004), and potentially giving rise to a larger complex. Here we provide evidence that certain plant MAPKs and MAPKKs associate with stable HMW multiprotein complexes. While the nature of the other constituents of these complexes remains to be determined, it is likely that they include the proteins known to interact with the members of a MAPK module as mentioned above.
The apparent size of the MAPK-containing complexes we detected by GF/IB varies depending on the plant species, the amount of protein loaded onto the GF column, and whether proteins from leaf material or suspension-cultured cells were analyzed. On average, complex-associated MAPKs eluted in fractions corresponding to molecular weights of 300-550 kDa. The peak elution profile for the tomato MAPKK MKK4 was similar to tomato MPK1/2 elution profiles, but additional fractions contained small amounts of MKK4 that did not contain MPK1/2 protein. There is precedent that different protein) from suspension cells sampled at 10, 120, and 180 min after treatment with 100 μM cantharidin, probed with anti-pERK antibody for the detection of phosphorylated MPK1/2 (upper panel). Immunoblots were stripped and reprobed with anti-AtMPK6 antibody for the detection of total MPK1/2 (lower panel). Only HMW fractions were analyzed (13-15 ml). Numbers above and below the immunoblots are as described in Fig. 1. Similar results were obtained in at least four independent experiments.
components of a multiprotein complex may show slightly different elution profiles (Dohmann et al., 2005;Gusmaroli et al., 2007;Hind et al., 2011). A biphasic elution profile was also obtained for c-myc-tagged MPK1 and MPK2 proteins, but the specific profiles were slightly different for each tagged MAPK and they were detected in all fractions tested. The elution patterns for MPK1-c-myc and MPK2-c-myc are similar, but the highest levels of MPK1-c-myc elute in fraction 13.5 and of MPK2-c-myc in fractions 14.5 and 15. This is probably experimental variation, but it cannot be excluded that both very similar MAPKs associate with different multiprotein complexes. Overexpression resulted in strongly increased amounts of MAPK protein in LMW fractions, but not in HMW fractions, as compared with proteins from wild-type plants and cells. We hypothesize that the amount of MAPKs that associate into a multiprotein complex is limited by other proteins (e.g. scaffolding proteins). Since such proteins are not up-regulated in MPK1/2-overexpressing plants, the amount of MPK1/2 in HMW fractions is not expected to increase, and the majority of the overexpressed proteins elute as monomeric proteins in LMW fractions. Since tagged MPK1/2 are present in HMW fractions, they probably occupy the same binding sites within the multiprotein complex as the native MPK1/2. The shift towards higher molecular weight (up to 150 kDa) in the LMW fractions is likely to be an artifact caused by the high amounts of tagged MPK1/2 protein due to overexpression.
Although the sizes of MAPKKK proteins can be highly variable (Lafleur et al., 2015), a complex consisting of the three components of the MAPK module is predicted to be ~200 kDa on average, indicating that additional proteins may be present in the HMW complexes we detected. The ratio of LMW (monomeric) to HMW MAPKs in GF fractions was somewhat variable between systems, but in general a significant amount of total MAPK protein eluted in the HMW fractions, with the exception of Arabidopsis MPK10. The KSR/ERK/MEK/B-RAF-containing complex from mouse brain tissue showed elution profiles similar to those shown here for plants (Nguyen et al., 2002). ERK was detected in GF fractions corresponding to molecular weights ranging from ~50 kDa to 500 kDa. The highest levels of ERK protein were detected in fractions corresponding to a molecular weight of ~ 200-400 kDa, but, in contrast to the GF profiles in plants, the relative amount of ERK in LMW fractions was lower (Nguyen et al., 2002). In Arabidopsis, we observed that the majority of MPK10 is associated with a HMW complex, similar to ERK in animals.
It was surprising to find that Arabidopsis MPK6 did not associate with a HMW complex, in contrast to its orthologs in tomato and tobacco. Instead, we demonstrated that the closely related Arabidopsis MPK10 associates with a HMW complex. MPK10 is the closest paralog of MPK6 and MPK3 in Arabidopsis (72% and 65% identity at the amino acid level, respectively). Thus it is not as similar to MPK6 as is tomato MPK1 to MPK2 (95.4% identity) or tobacco SIPK to Ntf4 (93.6% identity) ( Supplementary Fig. S1) (Holley et al., 2003;Ren et al., 2006;Kong et al., 2012). Therefore, MPK6 and MPK10 may be a product of an earlier gene duplication event. A detailed study demonstrated that MPK10 has a highly localized expression in leaves which supports its function in the development of leaf venation, and a role in polar auxin transport (Stanko et al., 2014). MPK10 was also shown to interact with MKK2 (Stanko et al., 2014) while MPK6 interacts with other MAPKKs. Since MPK6 functions in other developmental processes such as stomatal differentiation (Wang et al., 2007;Lampard et al., 2008) and stress responses, and a function of MPK10 in plant stress responses has not been reported, MPK10 and MPK6 seem to have undergone substantial neofunctionalization although they are close paralogs. However, too little is known about MPK10 to explain why MPK10 but not MPK6 associates with HMW complexes. On the other hand, AtMPK3 was clearly shown to associate with a HMW complex as no MPK3 protein was detectable in the mpk3 null mutant.
It has been demonstrated previously that tomato MKK4 functions upstream of MPK1/2 (Pedley and Martin, 2004), and we confirmed this activity using a protoplast transient transformation system in which MKK4 phosphorylates MPK1/2, but not MPK3 (data not shown). Our data are consistent with MKK4 and MPK1/2 being present in the same multiprotein complex. However, it cannot be excluded that they associate with separate complexes of a similar molecular weight.
In animal cells, some phosphatases associate with multiprotein complexes organized by scaffolds such as paxillin, KSR, and JIP1 (Willoughby et al., 2003;Deakin and Turner, 2008;Shaw and Filbert, 2009). In plants, different types of phosphatases can dephosphorylate and interact in vivo with MPK6, including: the dual-specificity MAPK phosphatases MKP1, MKP2, DsPTP1, PHS1, and IBR5; the PP2C-type Ser/Thr phosphatases AP2C1/2/3/4 and ABI1; and the Tyr-specific phosphatase AtPTP1 (Leung et al., 2006;Schweighofer et al., 2007;Bartels et al., 2009;Lumbreras et al., 2010;Umbrasaite et al., 2010). In the presence of cantharidin, tomato MPK1/2 activity gradually increased after 60 min, presumably due to inhibition of one or more MAPK phosphatases. Cantharidin is a PP2A-type phosphatase inhibitor, and thus far no such phosphatase has been characterized as a MAPK-inactivating phosphatase in plants. Therefore, it cannot be excluded that cantharidin activates MAPKs indirectly via inhibition of unknown phosphatases, which dephosphorylate other components of the MAPK pathways such as MAPKKs or MAPKKKs. In animal cells, PP2A can inhibit the activity of the MAPK ERK without dephosphorylation of the TEY motif (Letourneux et al., 2006). Regardless, cantharidin-activated MPK1/2 does associate with a HMW protein complex, suggesting that phosphatases regulate association of MAPKs with the complex.
The MAPK-containing multiprotein complex in S. peruvianum cells may operate in a similar manner to the RACK1 complex in Arabidopsis. Arabidopsis RACK1 binds to the MAPKKK MEKK1, the MAPKKs MKK4 and MKK5, and the two MAPKs MPK6 and MPK3. This complex is fully assembled in the absence of a stimulus, and, in response to treatment of Arabidopsis with a bacterial protease IV elicitor, the complex disassembles, thus releasing the active phosphorylated forms of MKK4/5 and MPK3/6, which can then activate defense gene expression (Cheng et al., 2015;Meng et al., 2015;Su et al., 2015). In addition, RACK1 appears constitutively to interact with the G-protein Gβ. Gβ and RACK1 did not play a role in the activation of the same MAPK module by the bacterial elicitor flg22, thus demonstrating how a scaffold protein confers signaling specificity (Cheng et al., 2015;Su et al., 2015). This result also indicated that only a portion of the MPK3/6 proteins resides in the RACK1 complex, while the remaining proteins are presumably either unbound or associated with other proteins. The latter is also true for the MAPKs we found associated with a HMW multiprotein complex (MPK1/2 and SIPK/NTF4), with the exception of MPK10, which is mainly associated with a HMW complex.
In tomato cells, our results suggest that a phosphatase inactivates MAPKs before or while they associate with the complex. This would result in a pool of MAPKs which can be re-activated by incoming signals. Blocking this phosphatase with a phosphatase inhibitor such as cantharidin results in aberrant association of active MAPKs with the complex. Interaction between inactive MAPKs and other complexassociated proteins may be destabilized upon MAPK phosphorylation by either a complex-associated or an unbound MAPKK. It is unclear if the phosphatase, while controlling the association/dissociation of MAPKs with/from the complex, is a constituent of the complex itself. This working model is speculative but can be tested, pending the identification of the components of the multiprotein complex. Our data suggest that the complex remains relatively stable, which should facilitate the identification of additional components.