The MAPK substrate MASS proteins regulate stomatal development in Arabidopsis

Stomata are specialized pores in the epidermis of the aerial parts of a plant, where stomatal guard cells close and open to regulate gas exchange with the atmosphere and restrict excessive water vapor from the plant. The production and patterning of the stomatal lineage cells in higher plants are influenced by the activities of the widely-used mitogen-activated protein kinase (MAPK) signaling components. The phenotype caused by the loss-of-function mutations suggested pivotal roles of the canonical MAPK pathway in the suppression of stomatal formation and regulation of stomatal patterning in Arabidopsis, whilst the cell type-specific manipulation of individual MAPK components revealed the existence of a positive impact on stomatal production. Among a large number of putative MAPK substrates in plants, the nuclear transcription factors SPEECHLESS (SPCH) and SCREAM (SCRM) are targets of MAPK 3 and 6 (MPK3/6) in the inhibition of stomatal formation. The polarity protein BREAKING OF ASYMMETRY IN THE STOMATAL LINEAGE (BASL) is phosphorylated by MPK3/6 for localization and function in driving divisional asymmetries. Here, by functionally characterizing three MAPK SUBSTRATES IN THE STOMATAL LINEAGE (MASS) proteins, we establish that they are plasma membrane-associated, positive regulators of stomatal production. MPK6 can phosphorylate the MASS proteins in vitro and mutating the putative substrate sites interferes the subcellular partition and function of MASS in planta. Our fine-scale domain analyses identify critical subdomains of MASS2 required for specific subcellular localization and biological function, respectively. Furthermore, our data indicate that the MASS proteins may directly interact with the MAPKK Kinase YODA (YDA) at the plasma membrane. Thus, the deeply conserved MASS proteins are tightly connected with MAPK signaling in Arabidopsis to fine-tune stomatal production and patterning, providing a functional divergence of the YDA-MPK3/6 cascade in the regulation of plant developmental processes.


Introduction
The mitogen-activated protein kinase (MAPK) cascades are central signaling pathways that regulate a wide range of cellular processes in plant growth, development and stress responses [1][2][3]. They function downstream of the cell-surface receptors to deliver and amplify extracellular stimuli that trigger a myriad of cytoplasmic and nuclear responses [4]. Stomatal development and patterning in the model plant Arabidopsis are tightly regulated by a canonical MAPK signaling cascade composed of the MAPKK kinase YODA (YDA), MAPK Kinase 4 and 5 (MKK4/5) and MAPK 3 and 6 (MPK3/6) [5][6][7]. Mutants and genetic analyses established a pivotal role of this YDA MAPK signaling pathway in suppressing Arabidopsis stomatal production at early developmental stages [5][6][7]. On the other hand, at later developmental stages, a positive regulation on stomatal proliferation was identified that seemed to be achieved by a differently assembled MAPK module of YDA-MKK7/9-MPK3/6 and other unknown MAPKs [7,8].
The MAPK cascades control a diverse variety of biological processes that are achieved by the regulation of a plethora of substrates. In stomatal development, several key factors are modified and regulated by MAPKs. The stomatal lineage initiation is controlled by the bHLH transcription factors SPEECHLESS (SPCH) [9,10] and its partners SCREAM/ICE1 (SCRM/ICE1) and SCRM2 [11]. MPK3/6 phosphorylate SPCH for protein degradation, thus providing a mechanistic link to the suppression of stomatal production [12]. SCRM/ICE1 is also phosphorylated by MPK3/6, so that SCRM/ICE1 protein stability was reduced in cold tolerance [13]. Additionally, recent studies showed that SCRM/ICE1 physically bridges MAPKs and SPCH to initiate the stomata lineage [14].
Besides these strong negative regulation of MPK3/6 signaling in the early stages of stomatal development [6,12], a positive role of the YDA-MKK7/9-MPK3/6 at the late stages was also suggested by the stage-specific manipulation of different tiers of this MAPK cascade in Arabidopsis [7,8]. However, what substrate/s control this cell fate flip remains unknown. Previously, a few collections of putative MAPK substrate proteins were predicted by peptide library screening combined with bioinformatics analysis [15], protein-protein interaction based on yeast two-hybrid screening [16,17], and in vivo phosphoproteomic studies [18]. However, detailed functional characterization of these proteins requires significant endeavor, thus the predicted candidates have been seldom further pursued.
Previously, Sörensson et al. [15] determined consensus phosphorylation sequences for MPK3/6 in Arabidopsis. They found one of the substrates At1g80180 was phosphorylated by MAPKs and overexpression of it induced stomatal overproduction and clustering [15]. Here, we performed in-depth functional genetics to characterize the three MAPK SUBSTRATES IN THE STOMATAL LINEAGE (MASS) proteins that At1g80180 belongs to for their biological functions in Arabidopsis stomatal development. We found that the MASS proteins are associated with the plasma membrane where they promote the stomatal formation and regulate stomatal patterning. We provide experimental evidence supporting that MAPKs can phosphorylate the MASS proteins in vitro and the putative substrates sites may regulate the MASS subcellular localization and protein function, and in turn the MASS proteins interact with YDA at the plasma membrane, possibly suppressing YDA function. Thus, the functional connection between the MASS family and the YDA MAPK cascade provides a new angle to study how external signals through MAPKs fine-tune stomatal development at the plasma membrane.

MASS proteins positively regulate stomata formation
The previous work by Sörensson et al. showed that At1g80180 is a substrate of MPK3 and MPK6 and overexpression of a phosphor-mimicking version of the protein seemed to generate stomatal overproduction and clustering [15]. Inspired by that, we investigated its biological function in stomatal development and possible functional interaction with the core YDA MPK3/6 pathway in Arabidopsis. At1g80180 encodes a short protein (15 kD) with unknown functions and belongs to a small family of three in the Arabidopsis genome (At1g15400 and At5g20100) (Fig 1A). We overexpressed the three genes either in the stomatal lineage cells by using a cell type-specific BASL promoter or ubiquitously by using the CaMV 35S promoter. The results showed that all the transgenic populations produced similar stomatal phenotypes: overproliferated stomatal guard cells in a clustered pattern (Fig 1B-1D and S1A Fig). The elevated transcript levels in these overexpression lines were demonstrated by real-time PCR ( Fig  1E). The phenotypes suggest that all three genes might promote stomatal production and regulate stomatal patterning, therefore they are named as MAPK SUBSTRATES IN THE STOMA-TAL LINEAGE (MASS) 1, 2, and 3 ( Fig 1A).
To functionally characterize the three MASS genes, we analyzed their promoter activity by examination of the transcriptional reporter lines that drive the expression of nuclear YFP (nucYFP) in plants. We found that all three promoters were broadly active in the leaf epidermal cells albeit with preferred cell type-specificity (S1B Fig) and in some other tissues, e.g. the hypocotyl and root, at the seedling stage (S1B Fig). We collected single T-DNA insertional mutants and crossed them to generate double and triple mutants (Fig 2A and S2A Fig). The transcript levels of the three genes in the triple mass1;2;3 mutants were assessed by real-time PCR ( Fig 2E) and the data show that, while mass3 is a knock-down, mass1 and mass2 are null mutations. Quantification of stomatal production in 5-day old cotyledons suggested that the triple mutant produced lowered stomatal index (Fig 2A and 2F), consistent with the overexpression data (Fig 1B-1D and S1A Fig), supporting a positive role of the MASS genes in stomatal production. We also characterized the lower-order mutants (singles and doubles) and found that the double mutant mass 1;3, but not mass1;2 and mass2;3, produced reduced numbers of stomata, though to a less extent when compared with those of the triple mutants ( Fig  2F). None of the three single mutants showed any discernable defects in general growth and To consolidate the loss-of-function phenotypes, we deployed the CRISPR/Cas9-mediated genome-editing strategy [29] to create genetic lesions in all three MASS genes (sgRNA positions shown in S2A Fig). The wild-type plants Columbia-0 (Col) plants were transformed with the CRISPR/Cas9 construct that carried three sgRNAs, each of which should specifically target one of the three MASS genes. Two independent triple mutant lines (T3) were established  which stomatal indices were lowered to around 20% compared to that of the wild-type (23%, n = 14 individual plants) (Fig 2F and 2G).
With respective to genetic complementation, we introduced the expression of N-terminal YFP fused MASS1 and MASS2, both driven by their endogenous promoters, into the T-DNA mass1;2;3 mutants ( Fig 2C, S3A and S3B Fig). We also introduced a CRISPR/Cas9-resistant version of mCherry-rMASS2 (rMASS2 containing nucleotide mutations in the sgRNA targeting site without changing the MASS2 amino acid sequence) into mass1;2;3-crispr#7 plants ( Fig  2D). In both cases, we found that the mutant phenotypes were recovered by YFP-MASS1/2 (Fig 2C and 2D) and the lowered stomatal indices in the crispr mutants were recovered by expression of mCherry-rMASS2 back to the wild-type levels ( Fig 2G). Again, both MASS1 and MASS2 protein expressions were detected in the leaf tissues, with more abundant expression of MASS2 in the epidermis and MASS1 in the mesophyll layer (S3A and S3B Fig). The subcellular distribution patterns of MASS1 and MASS2 (S3A and S3B Fig) in Arabidopsis were comparable with those in tobacco epidermal cells (Fig 3A and 3B); both were localized to the nucleus and the plasma membrane (see below for more details). In parallel, we also generated C-terminal tagged MASS proteins for complementation. Because the orientation of such fusions may disturb the C-terminal motif that is critical for MASS function (see below about MASS2 subdomains), they were therefore abandoned for further analysis. Thus, collectively, our genetic evidence (loss-of-function and overexpression) suggested that the three MASS genes may redundantly contribute to promoting stomatal production in Arabidopsis.

MASS functions at the plasma membrane
By amino acid sequence analysis, no functionally annotated domains can be recognized in the three MASS proteins. To characterize the biological functions of these novel regulators, we analyzed their protein subcellular localization by examining the N-terminal fluorescent protein-tagged MASS proteins in both tobacco epidermal cells and Arabidopsis stomatal lineage cells (Fig 3A-3F). The genomic regions of MASS1 and MASS3 do not contain introns, thus their genomic/coding sequences were amplified for constructing the reporter lines (CFP/ YFP-MASS1/3). As three alternative splicing sites were annotated for MASS2 (S2A Fig), we amplified the genomic region flanking all three variants to generate the reporter line (CFP/ YFP-MASS2g). The localization data in tobacco cells showed that the three MASS proteins were differentially distributed at the subcellular level: MASS1 and MASS2 appeared in the nucleus and at the plasma membrane, whilst MASS3 predominantly localized at the plasma membrane (Fig 3A-3C). When the three genes were expressed in the stomatal lineage cells (driven by the BASL promoter), they showed consistently differential expression patterns as in tobacco epidermal cells, with MASS1 and MASS2 dually localized in the nucleus and at the plasma membrane, while MASS3 mainly at the plasma membrane (Fig 3D-3F).
Because the MASS2 protein showed expression at both subcellular locations and overexpression of it generated elevated number and clustered stomatal lineage cells (Fig 3B, 3E and 3G), we tested where its biological location is, in the nucleus, or at the plasma membrane, or both. As in the MASS protein sequences, no signal peptides for apoplast secretion or transmembrane domains to span the membranes were predicted, we added a myristoylation lipid modification site [30] to artificially tether MASS2 to the plasma membrane. Indeed, myr-GFP-MASS2g was found exclusively at the plasma membrane and, interestingly, overexpression of this membrane-attached version recapitulated, and even slightly enhanced, the stomatal clustering phenotypes caused by the overexpression of the wild-type protein (Fig 3H and 3J). On the other hand, when fused with a nuclear localization signal (NLS), YFP-NLS-MASS2g showed the anticipated nuclear-only pattern, but this version did not induce any obvious stomatal phenotypes (Fig 3I  and 3J). The same strategies were applied to the dually localized MASS1 protein as well, and the data consistently show that plasma membrane-only but not nuclear-localized MASS1 promoted stomatal production (S3C and S3D Fig). Thus, collectively, our data suggested that the positive regulation of the MASS proteins in stomatal development arises from the plasma membrane pool, but not from the nuclear pool. We thus split MASS2.3 into two halves to make GFP-tagged MASS2.3_N76 and MASS2.3_C72, both driven by the BASL promoter ( Fig 4A). The results showed that both truncations failed to localize correctly. The first half GFP-MASS2.3_N76 lost the distinct localization in the nucleus, whereas the second half GFP-MASS2.3_C72 lost the plasma membrane localization (Fig 4B-4D). To further narrow down the critical segments for specific localization at the plasma membrane and in the nucleus, respectively, we deleted the highly conserved regions at the two ends to create GFP-MASS2.3_Δ29N (the N-terminal 29 amino acids deleted), GFP-MASS2_Δ13C and GFP-MASS2_Δ25C (the C-terminal 13 and 25 amino acids deleted, respectively) ( Fig 4A). The subcellular localization data clearly demonstrated that GFP-MASS2.3_Δ29N was only shown in the nucleus, suggesting the N-terminal 29-aa is required for the plasma membrane-association, while GFP-MASS2_Δ13C was only found at the plasma membrane, supporting that the C-terminal 13-aa determines the nuclear accumulation (Fig 4E-4G). Interestingly, the plasma membrane-only GFP-MASS2_Δ13C promoted stomata production and clustering, a phenotype resembling that of the full-length MASS2 overexpression ( Fig 4H and 4I), again supporting the biological function of MASS2 at the plasma membrane. However, the further shortened MASS2.3_Δ25C, albeit successfully localized to the plasma membrane, failed to induce stomatal overproduction (Fig 4J), hinting the critical role of the small region between Δ13C and Δ25C for its biological function in stomatal development (marked in Fig 4A). In addition, none of the other shortened versions, MASS2.3_N76, MASS2.3_C72 and MASS2.3_Δ29N, were sufficient to trigger this phenotype (S4D Fig), probably due to the lack of either plasma membrane association or the critical functional region. Taken together, we established that the MASS proteins promote stomatal formation at the plasma membrane and we defined three specific regions in MASS2.3 that are important for nuclear accumulation, plasma membrane association, and the biological function at the plasma membrane, respectively ( Fig 4A). Sörensson et al. (2012) previously demonstrated that MASS1 is phosphorylated by MPK6 in vitro kinase assays [15]. We further tested MASS2 recombinant proteins. Because one of the splicing variants, MASS2.2, is most similar to MASS1, we purified the MASS2.2 recombinant proteins and found that MASS2.2 was phosphorylated by in vitro constitutively active MKK5 (MKK5 DD )-activated MPK6 (Fig 5A). In parallel, mutating the serine residue (S107) to alanine (A) in the conserved MAPK-substrate P-P-S-P motif abolished the MPK6-mediated phosphorylation of MASS2.2 (Fig 5A), supporting that MASS2 is phosphorylated by MPK6 at the S107 site in vitro, consistent with the previous study that MASS2 was found phosphorylated among isolated plasma membrane-enriched phospho-peptides in Arabidopsis [31].

MPK6-mediated phosphorylation and possible connection with MASS localization and function
Phosphorylation may alter protein subcellular localization. We manipulated the phosphorylation site of MASS1 (S105) by generating a phospho-deficient version MASS1 S105A and a phospho-mimicking version MASS1 S105D , respectively. By examining the YFP-tagged proteins, we found that phosphorylation status is influential to MASS1 subcellular distribution because neither MASS1 S105A nor MASS1 S105D showed robust plasma membrane-association, but both more abundantly accumulated in the nucleus (Fig 5B and 5C and S5A Fig). With respective to the MASS2 localization, because we established that MASS2.2 was phosphorylated by MPK6 in vitro (Fig 5A), we assessed the localization pattern of YFP-MASS2.2 S107A and YFP-MASS2.2 S107D in the stomatal lineage cells. Consistently, the fluorescence intensity profiling results demonstrated that both versions showed reduced abundance at the plasma membrane but elevated accumulation in the nucleus (Fig 5D and S5B Fig). However, we did not detect an obvious change of the PM localization of MASS1 and MASS2 when MPK3/6 signaling was disturbed by overexpression of the dominant-negative MPK6AEF or chemically  inhibiting MPK6 in mpk3;6 null mutants [32] (S6A-S6C Fig). It is possible that some MPK3/6 activity leakage was sufficient to maintain MASS localization or other kinases may mediate MASS phosphorylation. Taken together, the combined data of in vitro phosphorylation and in vivo protein localization suggested that MAPK-mediated phosphorylation may contribute to the MASS proteins to localize robustly at the plasma membrane, though it is still possible that the manipulation of the phosphorylation sites may alter protein conformation thus proteinprotein interaction for localization and function. The failure of D versions (YFP-MASS1 S105D and YFP-MASS2.2 S107D ) to localize robustly at the plasma membrane was not anticipated but suggested that dephosphorylation might be equally important for these proteins to correctly localize.
To further assess the impact of protein phosphorylation on their biological functions, we first overexpressed YFP-MASS2.2, MASS2.2 S107A , and MASS2.2 S107D in the stomatal lineage cells by using the BASL promoter. Our results show that MASS2.2 did not seem to function as effectively as MASS2g in triggering stomatal clusters and neither of the mislocalized phosphovariants MASS2.2 S107A and MASS2.2 S107D produced significant stomatal phenotypes (Fig 5E). Considering the functional location of MASS at the plasma membrane, we modified MASS2.2 and the other two variants with the myristoylation site. When tethered to the plasma membrane, myr-GFP-MASS2.2 induced stronger stomatal clustering. However, neither myr-GFP-MASS2.2 S107A nor myr-GFP-MASS2.2 S107D could function at a comparable level in generating stomatal clusters (Fig 5F). Thus, we suspect that these putative MAPK phosphorylation sites need to stay open for both phosphorylation and dephosphorylation, so that MASS may achieve their function at the plasma membrane in stomatal development. Compared to the previous discoveries by Sörensson et al. that phosphomimick MASS1 trigged more stomatal production, our findings suggested that phosphorylation of MASS2 is critical for protein localization and, more interestingly, the reversible phosphorylation-dephosphorylation might be equally, if not more, important for MASS2 function.

MASS interacts with the MAPKK Kinase YDA
The mass triple mutants occasionally showed a cotyledon-fusion phenotype (S6D Fig), to a certain extent resembling that of a plant expressing the constitutively active MAPKKK YDA (YDA CA ) [5]. In addition, similar to the MASS proteins, YDA is also a peripheral membrane protein in plant cells [28]. To test whether the MASS genes are functionally connected to YDA, we first examined the physical interaction between MASS2 with YDA. Indeed, positive protein-protein interactions were detected between YDA and MASS2 based on yeast two-hybrid and in vitro pull-down assays (Fig 6A and 6B). To test their interaction in plant cells, we assayed the kinase-inactive version of YDA (YDA KI with one point mutation K429R, [7]) because overexpression of the catalytically active enzyme often causes cell death in tobacco [33]. In the bimolecular fluorescence complementation (BiFC) assay in tobacco epidermal cells, the recovered split YFP signals supported that all three MASS proteins may physically interact with YDA KI at the plasma membrane (Fig 6C and S7A Fig). The interaction between YDA and MASS was further confirmed by co-immunoprecipitation (IP) analysis in tobacco leaf cells and by the biolayer interferometry (BLI) assay in vitro (Fig 6D and S7B Fig). In Arabidopsis stomatal lineage cells, we co-expressed mCherry-MASS2 and YDA KI -YFP (driven by the SPCH promoter) and the two proteins co-existed at the plasma membrane. Furthermore, Z-projected confocal images showed that MASS2 forms cortical punctate that overlap with some YDA KI -accumulating dots at the plasma membrane (Fig 7A), though the properties of which have not been characterized yet. Taken together, our data suggested that the MASS proteins might function through their physical interaction with YDA at the plasma membrane.
Based on the phenotypes shown in the loss-of-function and overexpression plants, we hypothesized that the MASS family might promote stomatal production through suppressing the YDA MPK3/6 signal pathway, possibly via directly interacting with YDA. To test this hypothesis, we overexpressed CFP-MASS2g in plants expressing YDA CA -YFP (constitutively active YDA driven by the stomatal lineage-specific SPCH promoter). While YDA CA suppresses stomatal differentiation (Fig 7B and [5]), interestingly, the introgression of CFP-MASS2 suppressed the YDA CA -induced phenotype by restoring the formation of stomatal lineage cells (Fig 7B). Biochemically, by using the p42/44 MAPK antibody that detects activated MPK3 and 6 in Arabidopsis, we found that activated levels of MPK3/6 were elevated in the loss-of-function mutants but lowered in the MASS2 overexpression plants (S7C Fig). However, incubation of the MASS2 protein with YDA CA did not seem to alter YDA kinase activity in vitro (S7D Fig). We suspect that either MASS2 has to be properly modified to function in vivo or other interacting proteins that interact with MASS to participate in the regulation of YDA MAPK signaling. At the plasma membrane, the receptor-like kinases ER could also possibly interact with MASS. When MASS2 overexpression was introduced into er mutants, an additive stomatal phenotype was observed (S8A Fig), suggesting that the MASS function does not seem to rely on the presence of the ER receptor. The plasma membrane-localized polarity protein BASL was also examined in MASS2 overexpression plants but no discernable changes of  Collectively, our results supported that the positive role of MASS in stomatal production might be achieved by physical interaction with YDA thus the suppression of the MAPK signaling pathway in Arabidopsis (Fig 7C). However, since MASS did not seem to suppress YDA activity directly in our in vitro assays (S7D Fig), we propose that MASS may function as a scaffold that recruits other regulators of the MAPK signal pathway.

Discussion
MAPK cascades convert and amplify environmental and developmental cues into adapted intracellular responses. Their functions are particularly important for signal transduction in sessile plants that are incapable of escaping from a stressing environment. With a large number of potential kinase-substrate relationships of plant MAPKs revealed by in vitro and in vivo strategies [15,18], most putative MAPK substrates remained functionally elusive. Plant MAPKs usually are expressed in the cytosol and/or nucleus, therefore their substrates at different subcellular localizations are thought to largely determine signal specificity and the spatiotemporal dynamics of MAPK signaling in a biological process [34]. In Arabidopsis stomatal development, SPCH and SCRM/ICE1 are nuclear transcription factors targeted by MPK3/6 for protein degradation [12][13][14], so that stomatal formation can be modulated by MAPK upstream signals. The polarity protein BASL in the regulation of stomatal asymmetric cell division is phosphorylated by MPK3/6 for its localization at the cell cortex where BASL functions as a MAPK scaffold protein to locally concentrate the YDA MAPK signaling to differentiate daughter cell fates [28,35]. In this study, we established the functions of a newly identified MAPK substrate family, the MASS proteins, in the regulation of stomatal development and patterning in Arabidopsis. Phenotypic analysis of the loss-of-function mutants suggested that MASS functions to promote stomatal production and the overexpression phenotype revealed an additional role of MASS in stomatal patterning (Fig 1 and S1 Fig).

The dual localization of the MASS proteins
The subcellular localization revealed by the fluorescent protein-tagged MASS proteins suggested that MASS1 and MASS2 are dually localized to the nucleus and at the plasma membrane, except for MASS3 that only appeared at the plasma membrane (Fig 3). But overexpression of this plasma membrane-only MASS3 induced comparable stomatal phenotypes as those generated by overexpression of MASS1, MASS2, and their myristoylated versions (Fig 1 and Fig 3), consistently suggesting the biological activities of all three MASS proteins occur at the plasma membrane.
Why doesn't MASS3 localize into the nucleus? This probably can be explained by its protein sequence, in particular at the very C-terminal end (

PLOS GENETICS
MASS3 is excluded from the nucleus, we suspect that this conserved K/R-rich motif is possibly a nuclear localization signal, which usually mediates the interaction with importins for nucleocytoplasmic transport [36], and that the absence of this motif in MASS3 results in the failure of the protein to localize in the nucleus.
With regards to the plasma membrane localization of MASS proteins, the highly conserved N-terminal regions aligned with the first 29 amino acids of MASS2.3 seemed to be required, though no obvious features, e.g. high hydrophobicity, lipid modification, etc., could be recognized to explain the mechanisms for the plasma membrane localization. In parallel, our work also suggested that MAPK-mediated phosphorylation of the highly conserved serine residue in the P-P-S-P motif (S4C Fig) is required for MASS to localize and function at the plasma membrane. Therefore, we propose that the factor/s determining MASS to localize to the plasma membrane may involve protein-protein and/or protein-lipid iterations with both the N-terminal 29-aa and the conserved P-P-S-P phosphorylation site.

Possible functions of MASS at different subcellular localizations
Our findings show that, interestingly, the localization pattern and functional fashion of the MASS proteins, to some extent, mimic those of the polarity protein BASL [37]. For example, both are membrane-associated proteins that appear to be dynamically distributed between the nucleus and the plasma membrane (Fig 3A-3F) [35]. Although MASS is not polarized, both proteins function at the plasma membrane but not in the nucleus [37]. Also, both proteins are MAPK substrates and phosphorylation is important for their subcellular localization and biological function (Fig 5) [28]. In addition, both proteins appear to interact with the MAPKKK YDA at the cell cortical region. Base on the data we collected, we propose that the nuclear pool of MASS proteins, similar to that of BASL, might serve as a storage form that can be quickly targeted by MAPK signaling and redistributed to the target locations, without synthesizing new proteins, to respond to external stimuli. Meanwhile, we should not discount the possibility that MAPKs may phosphorylate MASS proteins that are in the cytoplasm or associated with the PM. Particularly, for MASS3 that is not expressed in the nucleus, the cytoplasm region can be a buffering zone. This possibility can be further tested by the deletion of the N-terminal 29 amino acids and mutating the conserved phosphorylation site.
Our complementation data using the myristoylated versions supported that the MASS proteins promote stomatal production at the plasma membrane. Then, what is the molecular mechanism for MASS to function there? We provided evidence that MASS proteins may interact with the MAPKKK YDA at the cell periphery, where MASS is hypothesized to negatively impact on the MAPK signaling cascade that suppresses the protein abundance of the key stomatal factor SPCH in production and proliferation of the lineage cells [6,12]. However, which form (phosphorylated or dephosphorylated) of MASS may preferentially associate with YDA and how MASS proteins mechanistically suppress YDA functions, directly or indirectly, at the plasma membrane requires more in-depth investigation. There are more questions, for example, what the stability of the MASS proteins is at the plasma membrane and what controls the dynamic nuclear/cytoplasmic partition in the cells, to be addressed for better understanding the potential feedback regulation between MASS and the MAPK signaling pathway (see the model in Fig 7C).
Also, we noted that one major phenotype caused by MASS overexpression was clustered stomata, a phenotype reflecting defective cell-cell communication. In most dicot leaves, stomata are spaced out by at least one nonstomatal epidermal cells to follow the "one-cell-spacing" rule needed for efficient guard cell activity in gas exchange [38,39]. To enforce this patterning rule, it was hypothesized that the developing guard cells release positional signals, e.g. the peptide ligand EPF1 [40], that are perceived by the cell surface receptors, e.g. the receptor-like protein TMM [20] and the receptor-like kinase ERL1 [19,41] with the SERK co-receptors [21], in the neighboring cells, so that the new divisions are reoriented to prevent direct stomatal contact from occurring [42]. In addition, downstream of the ligand-receptor signaling, defective YDA MAPK activities also led to the failures in enforcing the one-cell-spacing rule [5,6]. Therefore, it is possible that the elevated expression levels of MASS at the plasma membrane may directly or indirectly alleviate the ligand-receptor signaling and/or the YDA MAPK cascade activities in the regulation of division reorientation. An expanded survey for the MASS proteins to physically interact with the individual ER and SERK family members is necessary to test this connection.

The MASS genes, plant-specific and deeply conserved in early land plants
Through our sequence comparison and domain analysis, two conserved domains were recognized in the MASS protein family, including one segment at the N-terminus determining localization at the plasma membrane and another one at the C-terminus, containing the K/Rrich motif required for function and the putative MAPK phosphorylation motif P-P-S-P (S4C Interestingly, the typical MASS structure containing two conserved domains that appeared in the earliest flowering plant Amborella trichopoda, AmtMASS, and then the family members expand along with the evolution of angiosperms, e.g. 7 MASS genes in maize (S9 Fig). In addition, based on the phylogenetic assay, we found that MASS3 is more similar to ancient Amt-MASS. We suspect that the full-length MASS might be co-opted from an ancient form (MASS-L) with an additional C-terminal MAPK site and the functional domain to act as a MAPK substrate. Thus, these combined features of the MASS proteins enable them as MAPKresponding regulators for plants to adapt to the developmental and environmental changes.

Plant materials, mutants and transgenic lines
The Arabidopsis thaliana ecotype Columbia (Col-0) was used as the wild-type. In general, Arabidopsis and tobacco Nicotiana benthamiana plants were grown at 22˚C in long days (16 h light/8 h dark). The T-DNA insertional lines mass1 (GABI_902G09), mass2 (SALK_061905), mass3 (SALK_039099) were obtained from Arabidopsis Biological Resource Center (ABRC). The GFP-BASL marker line was described previously [37].

Plasmid construction and plant transformation
In general, the LR Clonase II (Invitrogen)-based gateway cloning technology was used for vector construction. To generate point mutations, the plasmid pENTR/D-TOPO carrying the MASS genomic or coding regions were used as template and specific site mutations were introduced through a QuickChange II XL Site-Directed Mutagenesis Kit (Agilent). The entry clones were then recombined into pMDC43 (the original 35S promoter was replaced by BASL promoter) and pH35CG to make BASLp::GFP/YFP/mCherry-MASS and 35S::CFP-MASS, respectively. Then, the BASL promoter were replaced by the MASS promoter through PmeI and KpnI sites to generate MASS1p::YFP-MASS1 and MASS2p::GFP-MASS2g. The pENTR/ D-TOPO carrying MASS promoters were recombined into pBGYN to make MASSp::nucYFP.
To create CRISPR/Cas9-mediated mutagenesis in Arabidopsis, we adopted the system described in [29]. By following the instructions, the oligos MASS1-CRI-F and MASS1-CRI-R were phosphorylated by T4 PNK (NEB) and annealed in a thermocycler, followed by ligation into the BbsI site of pAtU6-sgRNA-pAtUBQ-Cas9. Then, the chimeric U6-MASS1-Cas9 cassette was cloned into pCambia 2300 through HindIII and EcoRI sites to obtain 2300/crispr_ mass1. U6-MASS3 was amplified by PCR and inserted into 2300/crispr_mass1 through KpnI and EcoRI sites to generate 2300/crispr_mass1;3. Finally, by using the same strategies, U6-MASS2 were inserted at EcoRI site to create the construct 2300/crispr_mass1;3;2 to knockout the three members in the family. The crispr resistant MASS2 version (rMASS2) was generated through two rounds of PCR to introduce synonymous mutations in the Cas9-gRNA targeting site. Primers were listed in S1 Table. Plasmids were transformed into Agrobacterium tumefaciens GV3101, which delivers the desired DNA pieces into Arabidopsis or tobacco cells. Arabidopsis plants were transformed with the standard floral dipping method [43,44] and transgenic seeds were subjected to antibiotic selection. Tobacco cells were infiltrated by the method described in [45]. A. tumefaciens cells harboring 35S::CFP-MASS1/2/3 were infiltrated into N. benthamiana leaves and after 3 days, the leaf epidermal cells were observed under confocal microscope Leica SP5.

Plant cell imaging and image processing
Confocal images of plant cells expressing fluorescence-tagged proteins were taken by a Leica SP5 confocal microscope. 3-dpg (day-post-germination) adaxial cotyledons of Arabidopsis were captured. Cell peripheries were visualized with propidium iodide (PI, Invitrogen). Fluorescent proteins were excited at 488 nm (GFP), 514 nm (YFP) and 594nm (PI). Emissions were collected at 500-528 nm (GFP), 520-540nm (YFP), and 620-640 nm (PI). The confocal images were adjusted using either Adobe Photoshop CS5.1 or ImageJ (Fiji). The fluorescence intensity was measured by ImageJ (Fiji) and the pixel values were export into Excel to generate the histogram graphs.

Quantitative and statistical analysis of stomatal phenotypes in Arabidopsis
The adaxial cotyledons from 5-dpg seedlings were stained with PI and imaged were captured using a Carl Zeiss Axio Scope A1 fluorescence microscope equipped with a ProgRes MF CCD camera (Jenoptik). Stomata index (SI) was calculated as the stomata number versus the total number of epidermal cells. Clustered stomata index was calculated as the percentage of the number of clustered stomata over the total number of stomata. Stomatal clusters in Fig 1C were counted on the adaxial surface of the 10-dpg cotyledons.

Real-time PCR
Total RNAs were extracted from 3-dpg seedlings using an RNeasy Plant Mini Kit (Qiagen). The first-strand cDNAs were synthesized by the SuperScrip First-Strand Synthesis System (Invitrogen) with 2μg of total RNAs as template in a total volume of 20 μl. The fragments of interest were amplified by sequence-specific primers (see S1 Table). Real-time PCR was performed with a SYBR Green Master Mix kit (Applied Biosystems) and amplification was monitored on a StepOnePlus Real-Time PCR System (Applied Biosystems). Gene expression levels were normalized to the reference gene (ACTIN2) expression using the ΔCT method. Data are presented as mean ± SD.

Protein-protein interaction assay in yeast
The yeast two-hybrid assay was performed using the Matchmaker GAL4 Two-Hybrid System according to the manufacturer's manual (Clontech). MASS2 was inserted into pGADT7 and YDA was inserted into pGBKT7, respectively. Plasmids were transferred into the yeast strain AH109 (Clontech) by the LiCl-PEG method. The interactions were tested on SD/-Leu/-Trp/-His plates supplemented with 5 mM 3-amino-1,2,4,-triazole (3-AT). Three independent clones for each transformation were tested.

Pull-down assay
The CDS fragments of MASS2 and YDA were cloned into pET28a or pMAL-c2x for E.coli expression of His-or MBP-tagged proteins, respectively. Constructs were introduced into BL21 (DE3) cells for recombinant proteins expression. The recombinant His-tagged MASS2 and MBP-tagged YDA were purified using Ni-NTA agarose (QIAGEN) or Amylose Resin (New England Biolabs), respectively, according to the manufacturer's protocol. For pull-down assays, 3 mg of MBP-YDA fusion protein was incubated with Amylose Resin at 4˚C for 2 h, the MBP tag was used as a negative control. The beads were cleaned with washing buffer (50 mM Tris-HCl, pH 7.5, 10 mM MgCl2, 150 mM NaCl, and 1 mM DTT) for five times. Then the beads were incubated with 5 mg of His-MASS2 at 4˚C for 2 h. Wash beads five times with washing buffer. Western blot was used to detect the SDS-PAGE separation results of pulleddown mixtures in nitrocellulose membrane with anti-His antibody (Cell Signaling Technology) and anti-MBP antibody (New England Biolabs).

Co-immunoprecipitation of interacting proteins in plants
Agrobacterium tumefaciens strains (GV3101) carrying the 35S::4xMyc-MASS2g and the 35S:: YDA KI -YFP plasmids were co-infiltrated into N. benthamiana leaves. YFP fluorescence was detected 72 h after co-infiltration, and leaves were harvested and ground to powder in liquid nitrogen. Total proteins were extracted with extraction buffer (50 mM Tris pH 7.5, 150 mM NaCl, 10% Glycerol, 1 mM EDTA, 1 mM EGTA, 1 mM NaF, 1 mM Na3VO4, 10 mM DTT, 1 mM β-glycerol phosphate, 1 mM PMSF, 1 tablet/10 ml of Protease Inhibitor cocktail (Roche)). Samples were centrifuged at 14,000 rpm for 30 at 4˚C. Supernatant was subjected to immunoprecipitation by incubating with GFP-Trap Agarose (Chromotek) and rotating for 3 hr at 4˚C. The beads were washed four times with extraction buffer and the immunoprecipitates eluted with 5x Loading buffer by boiling 10 min. Total protein extracts (input) and immunoprecipitated proteins were separated on 10% SDS-PAGE and transferred to polyvinylidene difluoride (PVDF) membranes. Samples were subjected to western blot analysis with anti-Myc (1:1000; Cell Signaling Technology) or anti-GFP (1:1000; Santa Cruz Biotechnology) antibody.

Bio-Layer interferometry (BLI) assay
The binding affinity of MASS2 with YDA was measured using the BLItz system (ForteBio Inc.), as previously reported [46]. The recombinant protein was purified and loaded onto Ni-NTA biosensors (ForteBio Inc.). Ni-NTA biosensors were first equilibrated in 50 mM Tris pH 8.0, 150 mM NaCl buffer for 10 min prior the measurements, then dipped in the buffer with purified YDA protein for the measurement of association and dissociation kinetics. The settings were as follows: initial base line for 30 s, loading for 120 s, base line for 30 s, association for 300 s, and dissociation for 500 s. The kinetic parameters Ka (association rate constant), Kd (dissociation rate constant) and the binding affinity (KD = Kd/Ka) were calculated with the help of data analysis software (BLItZ Pro). All the experiments were performed at room temperature.