Microtubules are polymers of α- and β-tubulin heterodimers that help form the eukaryotic cytoskeleton. They are involved in many physiological processes such as development, morphogenesis, intracellular transport, cell division, and motility. Plant microtubule arrays differ from those in other eukaryotes in that they are dispersed and lack conspicuous organization (Dixit et al., 2006; Hamada, 2014). In addition, plant microtubules are organized into several distinct structures during the cell cycle (Ehrhardt and Shaw, 2006; Yamada and Goshima, 2017). During mitosis and after chromosomal segregation, plant cells build the bipolar phragmoplast array composed mainly of aligned microfilaments and highly dynamic microtubules. These microtubules form parallel arrays in the cell cortex, where they are organized into highly diverse patterns, relying on a variety of mechanisms for orientation, assembly, and function. In interphase cells, microtubules form parallel arrays in the cortex. The orientation of these arrays varies according to the cell type and stage of development. Cortical microtubules, for instance, are perpendicularly oriented during cell elongation, yet form parallel arrays during interphase (Lloyd and Chan, 2008; Chen et al., 2016).

Mitogen-activated protein kinases (MAPKs) were originally shown to phosphorylate the neuronal microtubule-associated protein, MAP2, after stimulation by insulin (Ray and Sturgill, 1987, 1988). Since then, MAPKs have been recognized as one of the most important cellular regulators of biological processes such as differentiation, proliferation, cell death, hormone signaling, osmotic stress, etc. (Colcombet and Hirt, 2008; Lu et al., 2015). Canonical MAPK cascades operate via successive phosphorylation steps starting with an upstream MAPK kinase kinase (MAPKKK or MKKK) that phosphorylates and activates a MAPK kinase (MAPKK or MKK) on one of two serine/threonine residues in a conserved (S/T)-X₅-(S/T) motif in plants and (S/T)-X₃-(S/T) in yeast and animals (Garrington and Johnson, 1999; Chang and Karin, 2001). The MKK then phosphorylates a MAPK (or MPK) by dual phosphorylation of the conserved—TXY—motif located in its activation loop. Typical MAPK substrates include transcription factors in addition to other cytoplasmic or nuclear proteins (Rodriguez et al., 2010). Activation of the upstream MAPKKK is more complex, and poorly understood in plants. Studies in yeast and animals show that MAPKKK activation can be mediated by MAPKKK kinases (Elion, 2000), G proteins (Fanger et al., 1997), and G protein-coupled receptors (Sugden and Clerk, 1997) or less frequently by the receiver domain of a two-component histidine kinase complex (Posas and Saito, 1997). The Arabidopsis thaliana genome encodes around 80 putative MAPKKKs (including 21 from the MEKK subfamily), 10 MAPKKs (MKKs), and 20 MAPKs (MPKs) (Ichimura and Group, 2002). The plant MAPKKKs can be further classified into two major groups: the MEKKs subgroup containing 21 kinases similar to animal MEKKs and yeast MAPKKKs, and the Raf-like kinases. Phylogenetically close sister clades such as the ZIK and MLK-like kinases are also considered as MAPKKKs although, so far, only the MEKKs are involved in MAPK cascades (Rodriguez et al., 2010).These kinases have been implicated in a wide array of plant biological processes ranging from stress responses to development (Rodriguez et al., 2010). To date, however, only a limited number of complete MAPK cascades have been reported in plants, mainly in abiotic stress responses, plant–pathogen interactions and in plant development. For example, the first completely identified MAPK cascade was MEKK1-MKK4/MKK5-MPK3/MPK6, which acts upstream of two WRKY transcription factors, WRKY22/WRKY29. This cascade functions downstream from the FLS2 flagellin receptor kinase that ultimately enhance resistance and protects against multiple bacterial and fungal pathogens (Asai et al., 2002). In plant development, the A. thaliana MAPKKK YDA acts upstream of the MKK4/MKK5-MPK3/MPK6 signaling module, which functions in stomatal patterning and inflorescence development (Wang et al., 2007; Meng et al., 2012). In biotic and abiotic stress responses, the Nicotiana benthamiana MAPK cascade NPK1-MEK1-Ntf6 is involved in tobacco mosaic virus (TMV) resistance (Jin et al., 2002; Liu et al., 2004), while in A. thaliana, the MEKK1-MKK2-MPK4/MPK6 cascade mediates cold and salt stress responses (Teige et al., 2004), and the MEKK1-MKK1/MKK2-MPK4 cascade functions as a negative regulator in innate immune responses (Gao et al., 2008). More recently, some ABA-regulated abiotic stress genes where also shown to depend on the MAPKKK17/18-MKK3-MPK1/2/7/14 cascade (Danquah et al., 2015).

The role of MAPK signaling in regulating microtubule functions has been widely documented in animals (Hoshi et al., 1992; Mandelkow et al., 1995; Verlhac et al., 1996; Terret et al., 2003; Zhao and Chen, 2006; Liu et al., 2007; Sun et al., 2008). Meanwhile, the involvement of MAPK-signaling in the regulation of plant microtubule functions has begun to emerge. In Arabidopsis, inhibition of protein tyrosine kinases and tyrosine phosphatases was reported to affect the dynamics and organization of cortical microtubules (Yemets et al., 2008). AtMPK4 and AtMPK6 were also shown to be necessary for cell division and localize to the phragmoplast during cytokinesis (Beck et al., 2011; Komis et al., 2011). Phosphorylation of the tobacco microtubule-associated protein MAP65-1 by the MAPK NRK1/NTF6 down-regulates the microtubule-bundling activity of NtMAP65-1 and positively regulates expansion of the phragmoplast (Sasabe et al., 2006). In addition, the NtMAP65-1 Arabidopsis ortholog, AtMAP65-1, has been shown to bind microtubule bundles specifically during interphase, anaphase, and telophase. When ectopically expressed in tobacco cells, a non-phosphorylatable AtMAP65-1 version affects mitotic division (Smertenko et al., 2006).

Arabidopsis MKKK20 participates in the osmotic stress response through its regulation of MPK6 (Kim et al., 2012), and was shown to interact with calmodulins (CaM) and CaM-like proteins (CML) in protein array assays (Popescu et al., 2007). Recently, insertional mutants of MKKK20 showed insensitivity to ABA in terms of root growth and stomatal response. In this context, MKKK20 was demonstrated to modulates abscisic acid responses acting upstream of MKK5 and MPK6 (Li et al., 2017). Arabidopsis MAPKKs have been classified into four groups, A–D (Soyano et al., 2003; Teige et al., 2004; Dai et al., 2006), with AtMKK3 belonging to group B, a single-member clade characterized by the presence of a C-terminal NTF2 domain known to promote nuclear import of cargo proteins (Steggerda and Paschal, 2002). This feature could confer to MKK3 an important and singular rule among all other AtMKKs (Colcombet et al., 2016). MKK3 participates in pathogen signaling and is required for regulation during dark–light transition (Doczi et al., 2007; Lee, 2015). Arabidopsis root growth inhibition by jasmonate is negatively regulated by the MKK3/MPK6 cascade (Takahashi et al., 2007). This cascade, along with MYC2, is involved in blue light-mediated seedling development in Arabidopsis (Sethi et al., 2014). MKK3 also participates in an abscisic acid-activated MAPK cascade (Danquah et al., 2015). Recently, two MKK3 orthologs have been found to regulate seed dormancy in barley (Hordeum vulgare L.) and bread wheat (Triticum aestivum L.) (Nakamura et al., 2016; Torada et al., 2016). MAPKs are also divided into four groups (Ichimura and Group, 2002). AtMPK18 belongs to group D, whose members harbor a—TDY—motif in their activation loop whereas the three other groups have a—TEY—signature. AtMPK18 had been reported to interact with the phosphatase propyzamide hypersensitive 1 (PHS1) and is involved in microtubule functions (Walia et al., 2009).

Here we show that Arabidopsis MKKK20 physically interacts with both MKK3 and MPK18 and phosphorylates these kinases in vitro. Surprisingly, phosphorylated MKK3 did not increase MPK18 phosphorylation status when the three kinases were put together. Previous analysis showed that MPK18 is involved in microtubule-related functions in Arabidopsis (Walia et al., 2009). When tested in the presence of microtubule depolymerizing drugs, mkkk20 and mkk3 mutant plants were found to have shorter roots than wild type, resembling the mpk18 behavior. This, combined with the analysis of mkkk20/mpk18 and mkk3/mpk18 double mutants suggests that MKKK20 modulates cortical microtubule functions through two independent MAPK signaling cascades, including a non-canonical pathway that bypasses the need for an intermediary MKK.

Among the 21 Arabidopsis MEKKs, MKKK19, -20, and -21 form a highly supported clade within the MEKK family. Paralogous MAPKKKs of this clade have been initially studied in Solanaceous species where they were coined Fertilization-Related Kinases (FRKs) from the first three characterized members in Solanum chacoense (ScFRK1, -2, and -3) (Gray-Mitsumune et al., 2006; O’Brien et al., 2007; Daigle and Matton, 2015; Lafleur et al., 2015). This clade has considerably expanded in the Solanaceae family that includes potato and tomato (Daigle and Matton, 2015). In these two species, the FRK clade represents around 40% of all MEKKs (15 and 17 FRKs out of 36 and 39 MEKKs, respectively), compared to 14% (3 out of 21) in A. thaliana. The FRK class is further subdivided in four groups. Arabidopsis MKKK19, -20, and -21 belong to group 4, the most ancient group in dicots, while the Solanaceae family encompass the four groups. While the role of these three Arabidopsis kinases is largely unknown, except for the role of MKKK20 in osmotic stress (Kim et al., 2012) and in abscisic acid responses (Li et al., 2017), the three characterized Solanaceous FRKs show a clear involvement in male and female reproductive development. Here, we employed genetic and biochemical tools to study the biological significance of the A. thaliana MKKK20. Surprisingly, although MKKK20 is strongly expressed in pollen (Figure 3), no obvious reproductive phenotype was observed in the loss-of-function mutant. Instead, MKKK20 single mutants showed a clear defect in MT functions.

The Y2H screen for MKKK20 interacting partners revealed a strong and robust association with MPK18, a result that suggested the operation of an atypical MAPK pathway that bypasses MKKs (Figure 1E, Y2H and Figure 2C, BiFC). Other MKKKs have been shown to interact with proteins other than MKKs. For example, AtMEKK1 not only interact with, and phosphorylate the transcription factor WRKY53, but can also bind to the promoter region of the WRKY53 gene to increase its relative expression (Miao et al., 2007). Although not considered a MEKK, the Raf-like Constitutive Triple Response1 MKKK (CTR1) interacts with and directly phosphorylates Ethylene-insensitive 2 (EIN2), which is not a MKK (Ju et al., 2012). In alfalfa, the MKKK OMTK1 (oxidative stress-activated MAP triple-kinase 1), which is activated by hydrogen peroxide, interacts directly with the MMK3 MAPK, thus bypassing the need for an intermediate MKK (Nakagami et al., 2004). Since MAPK cascades are prototypically activated by three-tiered sequential phosphorylation events, the involvement of a MKK in this putative cascade was addressed. A directed pairwise Y2H assay revealed that, among the 10 AtMKKs, MKK3 showed by far the strongest interaction with MKKK20. However, interaction between MKK3 and MPK18 in two-hybrid assays proved to be very weak compared to the strong MKKK20–MKK3 interaction (Figure 1C). An earlier directed two-hybrid screen between all Arabidopsis MKKs and MPKs, also showed no interaction between MKK3 and MPK18 (Lee et al., 2008), and the weak interaction observed between MKK3 and MPK18 (Figure 1E) was not validated by the BiFC assay (Figures 2E,F) nor by a phosphorylation assay (Figure 5D).

The first results from kinase activity assays demonstrated that MKKK20 autophosphorylates strongly (Figures 5A–D). Since eukaryotic protein kinases (EPKs) catalytic activity depends on phosphorylation of their activation loop, leading to a conversion from inactive to active conformation, the activation loop status of the MKKK20 expressed in bacteria was investigated (Bossemeyer, 1995; Nolen et al., 2004). Interestingly, autophosphorylation of EPKs seemed to be shared by most eukaryotic protein kinases (Beenstock et al., 2016). LC-MS/MS results showed that, inside the activation segment, bordered by the DFG motif in subdomain VII and the APE motif in subdomain VIII, two residues from the MKKK20 activation loop (threonine¹⁶¹ and serine¹⁶⁶) were phosphorylated (Supplementary Figure S2). Such phosphorylation is important for MAPKKK activation as demonstrated for human Mixed Lineage Kinase-3 (Leung and Lassam, 2001). In addition, MKKK20 phosphorylates both MKK3 and MPK18 equally, whereas no phosphorylation was observed when MKK3 and MPK18 are incubated together (Figure 5D). When all three kinases were present, MKKK20 phosphorylated equally MKK3 and MPK18. The presence of a phosphorylated MKK3 did not result in any obvious increase in MPK18 phosphorylation. Thus, the evidence suggests that MPK18 is not a bona fide MKK3 target, and that the pattern of activation of these kinases is an exception to the canonical MAPK signaling pathway. As noted above, this is not without precedent. Indeed, three characterized Nicotiana MAPKKKs (NbMAPKKKα, NbMAPKKKβ, and NbMAPKKKγ) were found to form a linear cascade leading to programmed cell death in tobacco cells (Hashimoto et al., 2012). On the other hand, our two-hybrid analysis and kinase assays demonstrated that MKKK20 interacts with and phosphorylates both MKK3 and MPK18.

Arabidopsis MPK18 had previously been shown to mediate cortical microtubule functions (Walia et al., 2009). Loss-of-function mutation in MPK18 results in more stable microtubule arrays and mpk18 roots display more abundant and better co-aligned microtubule polymers in their elongation zones as a result of MT dynamic instability problems (Walia et al., 2009). When treated with oryzalin, the mkkk20 mutant displayed significantly diminished seedling root length compared to the similarly treated wild type plants (Figures 6A,B). This phenotype is similar to the one observed in mpk18 plants, suggesting that both MKKK20 and MPK18 could be participating in a MAPK signaling pathway that mediates cortical microtubule function. In addition, mkk3 mutant roots, like mkkk20 and mpk18, were significantly shortened when treated with oryzalin, suggesting that these three kinases could be involved in a classical MAPK cascade. Thus, experimental evidence from both genetic and biochemical studies suggested the involvement of MKKK20, MKK3, and MPK18 in cortical microtubule functions in Arabidopsis cells, but in more than one signaling cascade. Since no protein–protein interaction or phosphorylation was observed between MKK3 and MPK18, this suggests that these three kinases act through two independent pathways involved in MT functions. The first would represent a canonical MAPK cascade, consisting of MKKK20, MKK3, and an as yet-unknown MAPK, while the second would be a non-canonical MAPK cascade with MKKK20 and MPK18 interacting directly, thus bypassing the need for a MKK intermediary. MKKK20 would therefore activate MKK3, which could in turn target downstream MAPKs (other than MPK18) involved in microtubule functions. Alternatively, MKKK20 could directly interact with MPK18 that would activate other target(s) (e.g., MAPs), modulating microtubule functions. Such direct MPK18 activation through phosphorylation of its activation loop TDY motif was validated using an anti-p-ERK antibody that specifically targets the [T-X-Y] MPK18 activation loop (Figure 5E). Interestingly, another D-type member, MPK9, was also shown to be activated through an independent MAPK cascade manner (Nagy et al., 2015).

Regarding the first pathway, two MPKs involved in MT functions had been previously found to interact with MKK3: MPK4 and MPK6 (Popescu et al., 2009). MPK4 acts downstream from two MKKKs (ANP1-2; also known as MKKK1 and MKKK2) involved in microtubule organization (Beck et al., 2010, 2011). In addition, MPK4 phosphorylates the microtubule-associated proteins AtMAP65-1 and AtMAP65-2, which are two structural components of microtubule arrays involved in cell division (Sasabe et al., 2011), and possess canonical serine/threonine-proline MAPK phosphorylation sites. MPK6 has been functionally linked to MTs since it localizes to the plasma membrane and is involved in root development and cell division control (Muller et al., 2010). MPK6 has been also shown to interact with γ-tubulin and the MT plus end protein EB1 (Kohoutova et al., 2015). Thus, both MPK4 and MPK6 represent strong candidates for completing the proposed MKKK20/MKK3 module. It is worth mentioning that MKK3 plays an essential role in the activation of MPK6 through the phytohormone jasmonic acid (JA; Takahashi et al., 2007). In a large microarray transcriptional analysis of several MAPK signaling genes, jasmonates were also found to induce transcription of MKKK20 (Menges et al., 2008). Thus, under the influence of JA, MKKK20 could potentially activate the MKK3/MPK6 cascade thereby affecting MT function. Interestingly, JA treatment induces changes in the orientation of cortical MTs during potato tuberization and cell expansion, consistent with the idea that JA might control the direction of cell expansion by changing the arrangement of MTs (Koda, 1997; Cenzano et al., 2003). Recently, a MKKK20-MKK5-MPK6 cascade was also shown to modulate abscisic acid responses, regulating cell division and cell elongation mainly at the root elongation zone (Li et al., 2017). As for MKK5, MKK4 was also shown to interact with MKKK20 in a BiFC assay. Surprisingly, previous results from the Brassica napus MKKK20 ortholog (BnaMKKK20; Sun et al., 2014), showed no interaction between BnaMKKK20 and BnaMKK4 or BnaMKK5, a result also observed in our Y2H assay with the A. thaliana MKKK20 (Figure 1). Furthermore, MKKK20 from A. thaliana (Figure 1) and B. napus (Sun et al., 2014) strongly interacted with MKK3, in both Y2H and BiFC assays. It is noteworthy to mention that in this case, different BiFC assay systems (microparticle bombardment vs Agrobacterium infiltration) were used. The MKKK20–MKK3 Y2H interaction was also recently mentioned in (Colcombet et al., 2016), emphasizing the strong interaction between MKKK20 and MKK3. Taken together, the short root phenotype observed in mkkk20 plants could result in overlapping functions of MKKK20 in microtubule organization and ABA responses. Interestingly, it had been demonstrated that ABA modulates microtubule organization and stability (Sakiyama and Shibaoka, 1990; Sakiyama-Sogo and Shibaoka, 1993; Takatani et al., 2015). Likewise, the same MPK cascade elements can be involved in different biological contexts (Meszaros et al., 2007; Colcombet et al., 2016). Our results indicate that MKKK20 would act in separate cascades in response to different stimuli.

Overall, our results showed that MKKK20 plays a role in plant root microtubule functions, thereby affecting root growth and development. Two distinct processes regulate plant organ growth; production of cells and expansion. Root growth rate is determined primarily by expansion, yet is also influenced by cell division and production (Beemster and Baskin, 1998). Indeed, it has been reported that the growth of Arabidopsis roots was accompanied by increased cell production leading to cell elongation (Baskin et al., 1995). Cortical as well as mitotic microtubules play an essential role in both elongation and division, and a wide range of MAPs (microtubule-associated proteins) have been associated with these developmental stages in plant cells (Hamada, 2014). It has been shown that the phosphorylation state of MAPs influences microtubule stability (Drechsel et al., 1992; Mandelkow et al., 1995), and phosphorylation of MAPs leads to reduced binding to the microtubule cytoskeleton due to a reduction in MAP-microtubule affinity (Drewes et al., 1998). Our results are consistent with the idea that these potentially convergent phosphorylation-based signaling pathways play a role in controlling microtubule dynamics. The binding of specific MAPs to MTs results in structural regulation via interconnected polymerization/catastrophe events collectively referred to as dynamic instability (Hamada, 2014). Some of these MAPs play crucial roles in microtubule orientation or anisotropy. This could be explained by a model in which some structural MAPs (downstream of the canonical and non-canonical MKKK20 cascades) would be lesser phosphorylated in the absence of MKKK20, resulting in their higher affinity for MT polymer chains (Nogales, 2000), and perturbation of the normal array alignment. Therefore, such perturbation could affect directly the normal growth of the plant cell wall, cortical microtubule functions being disturbed, as observed in the elongation zone (Figure 7). The two MKKK20 pathways, one that includes MPK18, and the other MKK3 (possibly with MPK4 or MPK6), are likely to target different MAPs and exert different influences on MT dynamics and organization. Indeed, the phenotype of mpk18 is not as dramatic as that of mpk4 (Komis et al., 2011) and MPK4 was found to co-localize with microtubules, unlike MPK18 (Walia et al., 2009). Moreover, inhibition of phosphorylation via the mkk3 mutation is likely to cause MT hyperstabilization, as seen with mpk18-1 mutant (Walia et al., 2009). The summation of different pathway disturbances in the mkkk20 mutant could therefore lead to more drastic perturbation of microtubule dynamics. This is supported by the observation that, in the absence of oryzalin, the mkk3/mpk18 double mutant root length is significantly shorter than in all other lines (Figures 6A,B). Our understanding of the role played by plant MAPK signaling in regulating microtubule functions still needs to be refined. The involvement of multiple MAPK members in numerous biological processes often compromises a clear understanding of phenotypic specificity. Full characterization of the signaling network, including cross talk between parallel cascades, will ultimately be needed to develop a coherent biological picture.

RB, FB, SD, and AL-H carried out the experiments. BE provided study materials and critical revision of the manuscript. RB and DM designed most experiments and wrote the manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.