Initiation of Cell Wall Pattern by a Rho- and Microtubule-Driven Symmetry Breaking

See allHide authors and affiliations

Science  14 Sep 2012:
Vol. 337, Issue 6100, pp. 1333-1336
DOI: 10.1126/science.1222597


A specifically patterned cell wall is a determinant of plant cell shape. Yet, the precise mechanisms that underlie initiation of cell wall patterning remain elusive. By using a reconstitution assay, we revealed that ROPGEF4 (Rho of plant guanine nucleotide exchange factor 4) and ROPGAP3 [ROP guanosine triphosphatase (GTPase)–activating protein 3] mediate local activation of the plant Rho GTPase ROP11 to initiate distinct pattern of secondary cell walls in xylem cells. The activated ROP11 recruits MIDD1 to induce local disassembly of cortical microtubules. Conversely, cortical microtubules eliminate active ROP11 from the plasma membrane through MIDD1. Such a mutual inhibitory interaction between active ROP domains and cortical microtubules establishes the distinct pattern of secondary cell walls. This Rho-based regulatory mechanism shows how plant cells initiate and control cell wall patterns to form various cell shapes.

Cell shape is the basis of behavior and function of specialized cells in multicellular organisms. The plasma membrane plays central roles in the development of the functional and structural polarization of the cell. In plants, plasma membrane polarization leads to the formation of the locally specialized architecture of cell walls, resulting in various shapes of plant cells with specific functions (14). However, how local domains in the plasma membrane are formed and how they control local cell wall architecture in conjunction with the cytoskeleton remain a mystery. The plant-specific microtubule binding protein MIDD1 is anchored to the plasma membrane domains and promotes local microtubule disassembly, forming a specific pattern of secondary walls in xylem vessel cells (5). MIDD1 binds Rho guanosine triphosphatases (GTPases) (68), which regulate cell polarization in diverse organisms (1, 9, 10). We found four ROP (Rho of plants) GTPases that were expressed in xylem cells on the basis of a metaxylem vessel–related gene expression database (11). Because of difficulty in direct observation of subcellular localization of these ROPs in developing xylem cells in situ, we observed green fluorescent protein (GFP)–ROPs expressed under the control of an estrogen-inducible system (12) in in vitro Arabidopsis metaxylem cell culture (5). Only GFP-ROP11 showed colocalization with filamentous structures in secondary wall pits (fig. S1, A to C). Time-lapse observation revealed that GFP-ROP11 accumulated at the end of filaments (fig. S1D), as does MIDD1 (5). Double labeling confirmed the colocalization of GFP-ROP11 and tag red fluorescent protein (tagRFP)–MIDD1 (Fig. 1A). In nonxylem cells, which normally lack MIDD1, GFP-ROP11 was localized along a filamentous structure only when MIDD1 was ectopically coexpressed (fig. S2A). Thus, MIDD1 can recruit ROP11 along cortical microtubules. The N-terminal domain of MIDD1 binds to microtubules, and the C-terminal domain of MIDD1 is anchored to the plasma membrane in the secondary wall pits (5). To determine whether ROP11 can recruit MIDD1 to the plasma membrane, we introduced tagRFP-MIDD1ΔN together with GFP-ROP11, a fusion protein of GFP and a constitutive active form of ROP11 (GFP-ROP11G17V), or a fusion protein of GFP and a constitutive negative form of ROP11 (GFP-ROP11T22N) (13) into nonxylem cultured cells, which normally lack MIDD1. tagRFP-MIDD1ΔN colocalized with GFP-ROP11G17V and GFP-ROP11 on the plasma membrane but not with GFP-ROP11T22N (fig. S2B). Thus, the GTP-bound form of ROP11 can recruit MIDD1 to the plasma membrane. Introduction of a constitutive active ROP11 (GFP-ROP11G17V or GFP-ROP11Q66L), but not GFP-ROP11, into differentiating metaxylem cells disrupted the pit-specific localization of MIDD1ΔN such that distribution of MIDD1ΔN became uniform (fig. S3) and secondary walls were formed throughout the cell surface (Fig. 1, B and C). Transgenic Arabidopsis plants expressing LexA (operator of bacterial repressor)::GFP-ROP11G17V or LexA::GFP-ROP11Q66L also exhibited disorganized secondary walls without obvious pits in metaxylem vessels, although the secondary wall pattern of protoxylem vessels was not visibly affected (fig. S4). A bimolecular fluorescence complementation (BiFC) assay using ROP11 fused with the C-terminal half of yellow fluorescent protein (cYFP-ROP11) and MIDD1ΔN fused with the N-terminal half of YFP (nYFP-MIDD1ΔN) in leaf epidermis revealed that nYFP-MIDD1ΔN interacted with cYFP-ROP11 or cYFP-ROP11G17V but not with cYFP-ROP11T22N (fig. S5), suggesting that MIDD1ΔN binds only active ROP11. The BiFC assay in differentiating metaxylem cells revealed preferential signals in the secondary wall pits (Fig. 1D). Similarly, fluorescence resonance energy transfer (FRET) between MIDD1ΔN-CFP and YFP-ROP11 was detected in the secondary wall pits (fig. S6). These results suggest that ROP11 is present in its GTP-bound form at the plasma membrane beneath the secondary wall pits, where it functions to recruit MIDD1.

Fig. 1

Active ROP11 is localized in secondary wall pits to anchor MIDD1. (A) tagRFP-MIDD1 and GFP-ROP11 in differentiating xylem cells. Both signals are substantially overlapped (white arrowheads). (B) Secondary wall pattern (WGA) of a cell expressing GFP-ROP11G17V or GFP-ROP11Q66L. (C) Percentages of cells that developed flat secondary walls without pits. Data are means ± SD of three experiments (n > 200 cells), *P < 0.01 [analysis of variance (ANOVA) with Scheffe test]. (D) Confocal cross-sections of cell cortex in a differentiating xylem cell expressing nYFP-ROP11 and cYFP-MIDD1ΔN. The intensity plots of BiFC and WGA along the plasma membrane between red arrowheads are shown below the images. Scale bars indicate 5 μm; a.u., arbitrary units.

ROP activation and inactivation are mediated by ROP guanine nucleotide exchange factor, ROPGEF (14, 15), and ROP GTPase activating protein, ROPGAP (16), respectively. From the metaxylem cell–specific gene expression database (11), we found that ROPGAP3, ROPGAP5, ROPGEF4, and ROPGEF7 are up-regulated during xylem differentiation. We examined the localization of these proteins in differentiating xylem cells with GFP-fusion constructs. Although GFP-ROPGAP3 was localized preferentially at the plasma membrane in the secondary wall pits (Fig. 2A), the other ROPGAPs and ROPGEFs were broadly distributed across the plasma membrane or throughout the cytoplasm (fig. S7, A to C). ROPGEF proteins are composed of a catalytic domain named PRONE and variable regions at either or both of their N or C terminal ends (14, 15). The variable region of a ROPGEF can prevent its PRONE domain from accessing ROPs (17). Therefore, we next examined localization of PRONEs of ROPGEF4 and ROPGEF7. GFP-ROPGEF7PRONE and GFP-ROPGEF4PRONE localized weakly and strongly, respectively, at the plasma membrane in the secondary wall pits (Fig. 2B and fig. S7D). GFP-ROPGEF4PRONE accumulated at the plasma membrane in the center region of secondary wall pits as punctuated structures (Fig. 2, D to F). The clustering of GFP-ROPGEF4PRONE occurred before the beginning of secondary wall deposition (Fig. 2C). ROPGEF4 was expressed strongly in metaxylem cell files in Arabidopsis roots (fig. S8). Knockdown and knockout of ROPGEF4 caused reduced density of secondary wall pits in metaxylem vessels (fig. S9). These results suggest that ROPGEF4 plays an essential role in the patterning of pitted secondary wall, probably by local activation of ROPs. In contrast, spiral secondary walls in protoxylem vessels were not affected significantly by knockout of ROPGEF4 or by expression of constitutive active ROP11.

Fig. 2

ROPGAP3 and ROPGEF4 localize preferentially in secondary wall pits in differentiating xylem cells. (A and B) Accumulation of GFP-GAP3 [(A), yellow arrowheads] and GFP-ROPGEF4PRONE [(B), red arrowheads] at the plasma membrane in the secondary wall pits. (C) Accumulation of GFP-ROPGEF4PRONE at the plasma membrane (cyan arrowheads) in early differentiating xylem cells. (D) Accumulation of GFP-ROPGEF4PRONE (green) at the center of pits (white arrowheads) of the secondary wall (magenta). (E) Double labeling of GFP-ROPGAP3 (green) and tagRFP-ROPGEF4PRONE (magenta) in differentiating xylem cells (top). Intensity profile between the orange arrowheads (bottom). (F) Distance between the edge of GFP signals and WGA signals. Values are means ± SD (n = 7 pits), *P < 0.01 (t test). The location of GFP-ROPGEF4PRONE is more central than that of GFP-ROPGAP3. Scale bars, 5 μm.

To understand the relations between ROPGEF4, ROPGAP3, and ROP11, we coexpressed ROP11, GFP-ROPGEF4PRONE, and tagRFP-ROPGAP3 under the estrogen-inducible system in the leaf epidermis of Nicotiana benthamiana, in which no secondary wall–related active ROP domain exists and active ROP domain formation can be analyzed quantitatively. The coexpression induced GFP-ROPGEF4PRONE accumulation as clusters on the plasma membrane, which mimics that in differentiating metaxylem cells (fig. S10A). To examine whether the clusters of ROPGEF4PRONE induce local activation of ROP, we introduced tagRFP-MIDD1ΔN as a marker labeling active ROPs together with ROP11, ROPGAP3, and GFP-ROPGEF4PRONE. tagRFP-MIDD1ΔN localized in and around clusters of GFP-ROPGEF4PRONE (fig. S10, B and C). Thus, clusters of ROPGEF4PRONE induce local activation of ROP11. The clustering of ROPGEF4PRONE needed both ROP11 and ROPGAP3 (fig. S11, A and C). Replacing ROP11 with ROP11G17V or ROP11T22N abolished the clustering of ROPGEF4PRONE (fig. S11, B and C). Replacement of these proteins with other family members reduced the frequency of the GFP-ROPGEF4PRONE clustering except for ROPGAP4 (fig. S12, A to C). The replacement of GFP-ROPGEF4PRONE with GFP-ROPGEF4 did not cause its clustering (fig. S12D), suggesting the absence of a machinery activating GFP-ROPGEF4 in tobacco epidermal cells. Disruption of microtubules or actin microfilaments did not affect GFP-ROPGEF4PRONE clustering (fig. S13). These results indicate that the combination of ROP11, GFP-ROPGEF4PRONE, and ROPGAP3 directs clustering of GFP-ROPGEF4PRONE.

This spontaneously formed complex can induce MIDD1-dependent local microtubule disassembly. Leaf epidermal cells had scattered microtubules that were not suitable for this microtubule-disassembly analysis. Therefore, we used nondifferentiating Arabidopsis cells in vitro (Fig. 3). We established a stable LexA::ROP11 cell culture line and then introduced transiently the LexA::GFP-ROPGEF4PRONE/LexA::ROPGAP3 construct. As observed in the tobacco leaf epidermis, coexpression of these proteins resulted in clustering of GFP-ROPGEF4PRONE (fig. S14). Coexpression of tagRFP-MIDD1 with the three proteins visualized tagRFP-MIDD1–labeled filamentous structures around the GFP-ROPGEF4PRONE clusters (Fig. 3A). Further introduction of a GFP-tubulin gene revealed that the MIDD1-labeled filamentous structures around the clusters of ROPGEF4PRONE were depolymerizing microtubules (Fig. 3B, white arrowheads). Without MIDD1, the cortical microtubules were not depolymerized around the clusters of ROPGEF4PRONE (Fig. 3C). Thus, the clustering of ROPGEF4PRONE produces plasma membrane domains at which MIDD1 is anchored, which in turn cause local microtubule disassembly.

Fig. 3

Reconstitution of the active ROP domains and local microtubule disruption in nonxylem cells. (A to C) Localization of ROPGEF4 PRONE and microtubules in nonxylem Arabidopsis cultured cells. Areas encircled by yellow dotted lines are magnified below. ROP11 and ROPGAP3 were coexpressed with tagRFP-MIDD1 and GFP-ROPGEF4PRONE (A); tagRFP-MIDD1, tagRFP-ROPGEF4PRONE, and GFP-TUB6 (B); or tagRFP-ROPGEF4PRONE and GFP-TUB6 but not tagRFP-MIDD1 (C). Without MIDD1, cortical microtubules are not depolymerized around tagRFP-ROPGEF4 PRONE clusters (C). Scale bars, 5 μm.

Although we showed that the plasma membrane domains can be produced by spontaneous local activation of ROP11, the cortical microtubules and secondary wall patterning are not fully explained, because secondary wall pits and MIDD1ΔN-labeled domains were more elongated in taxol-treated cells than in nontreated cells (fig. S15). Quantification of fluorescence revealed that taxol treatment enhances the fluorescence of GFP-MIDD1ΔN in the secondary wall pits, whereas oryzalin treatment reduces the fluorescence (fig. S16). These results suggest that cortical microtubules are required to retain MIDD1 in the plasma membrane domains, probably by inhibiting the diffusion of the MIDD1-ROP11 complex. To investigate this idea further, we reconstituted local ROP activation domains by introducing ROP11, GFP-ROPGEF4PRONE, ROPGAP3, and tagRFP-MIDD1 into tobacco leaf epidermis, where scattered microtubules allowed clear imaging of the spatial relations between microtubules and ROP activation domains. tagRFP-MIDD1 was located at the plasma membrane in elongated clusters (Fig. 4, A and E). Visualization of cortical microtubules with GFP-TUB6 revealed that the elongated plasma membrane domains were encircled by cortical microtubules that run along the boundary of the plasma membrane domains (Fig. 4F). Disruption of microtubules by oryzalin treatment or transient expression of AtKinesin-13A, whose members are well known to function in microtubule disruption in animals (18), resulted in rounder tagRFP-MIDD1 domains, although inactive AtKinesin-13A did not affect the shape (Fig. 4, B to E). These results indicate that cortical microtubules restrict the localization of the plasma membrane–anchored MIDD1-ROP11 complex. To clarify how cortical microtubules restrict the localization of the MIDD1-ROP11 complex along the plasma membrane, we examined the interaction among YFP-ROP11, cyan fluorescent protein (CFP)–MIDD1ΔN, and tagRFP-TUB6 (microtubules) in tobacco leaf epidermis (fig. S17). ROP11 and MIDD1ΔN were eliminated by the cortical microtubules when the two proteins were coexpressed (fig. S17B). However, microtubules did not eliminate ROP11 or MIDD1ΔN when ROP11 or MIDD1ΔN was solely expressed (fig. S17, A and C). In contrast, cortical microtubules eliminated MIDDΔN-AD, in which a minimal plasma membrane–anchor domain from ROP11 (19) was fused to MIDDΔN (fig. S17D). These results suggest that MIDD1 mediates elimination of ROP11 from the plasma membrane by the cortical microtubules, probably by the mechanism in which plasma membrane–associated cortical microtubules physically interfere the plasma membrane–anchored MIDD1-ROP11 complexes.

Fig. 4

Cortical microtubules restrict the localization of the active ROP11-MIDD1 complex. (A to D) Shape of tagRFP-MIDD1–labeled plasma membrane domains reconstituted in leaf epidermal cells of N. benthamiana coexpressing ROPGEF4PRONE, ROP11, and ROPGAP3. Yellow dotted circles indicate plasma membrane domains with tagRFP-MIDD1. (A) Treated with mock. (B) Treated with 30 μM oryzalin. (C) Coexpressed with AtKinesin-13A. (D) Coexpressed with AtKinesin-13AΔM (inactive form). (E) Circularity (2π × area/perimeter2) of the plasma membrane domains marked with tagRFP-MIDD1 in leaf epidermal cells. Values are means ± SD (n > 300 domains), *P < 0.01 (ANOVA with Scheffe test). (F) Visualization of cortical microtubules (GFP-TUB6) and plasma membrane domains (tagRFP-MIDD1) in leaf epidermal cells coexpressing ROP11, ROPGAP3, and ROPGEF4PRONE. Cortical microtubules run along the borders of the plasma membrane domains (arrowheads). Scale bars, 5 μm.

We show that secondary wall pattern is established by two processes: a ROP-driven symmetry breaking and a mutual inhibitory interaction between cortical microtubules and active ROP domains (fig. S18). The first process may be explained by Turing’s reaction diffusion model (20), where ROPGEF4 and ROPGAP3 act as an activator and an inhibitor, respectively. This model requires a positive feedback of the activator. In yeast, a scaffold protein, Bem1p, mediates a positive feedback of Cdc24p GEF (21). A similar scaffold protein might function with ROPGEF4. In the second process, MIDD1 promotes disassembly at the microtubule tip (5), whereas MIDD1 mediates the elimination of active ROPs at the microtubule sides. MIDD1 may interact with AtKinesin-13A (8) as well as ROP11 and cortical microtubules. MIDD1 may have different functions at the tip and side of cortical microtubules.

We showed that MIDD1-mediated interaction between spontaneously activated ROP domains and cortical microtubules produces pitted pattern of metaxylem cells. Although ropgef4 did not affect significantly protoxylem cell wall patterns, because not only MIDD1 but also members of ROPs are also expressed in protoxylem cells (22), such secondary wall patterns as annular, spiral, and reticulate patterns might be produced by similar MIDD1-mediated interaction between activated ROP domains and cortical microtubules. Indeed, stabilization of microtubules with taxol produced a reticulate-like secondary wall pattern in developing metaxylem cells. Differences among ROPGEF and/or ROPGAP members may also contribute to size differences of activated ROP domains and then of the secondary wall depleted area. Our reconstitution assay may be a powerful tool to test this idea.

MIDD1 is expressed in not only xylem cells but also nonxylem cells (23). Considering the nature of MIDD1, even in nonxylem cells, MIDD1 may function in production of specific patterns of cortical microtubules and of activated ROP domains. As shown in epidermal pavement cells, local ROP domain activation and microtubule organization underlie local polarized growth of the cell (1). Thus, MIDD1-mediated membrane domain establishment may contribute to the formation of various plant cell shapes generally.

Supplementary Materials

Materials and Methods

Figs. S1 to S18

Table S1

References (24, 25)

References and Notes

  1. Acknowledgments: We thank N. Chua of the Rockefeller University for providing the pER8 vector; U. Grossniklaus of the University of Zurich for providing the pMDC7 vector; T. Nakagawa of Shimane University for providing the pGWB vectors; Y. Ohya of the University of Tokyo for critical reading of this manuscript; A. Nakano, T. Ueda, and S. Betsuyaku of the University of Tokyo for technical advice; and Y. Nakashima for technical assistance. This work was supported partly by Grants-in-Aid from the Ministry of Education, Science, Sports and Culture of Japan (19060009) to H.F.; from the Japan Society for the Promotion of Science to H.F. (23227001) and Y.O. (22870005) and the NC-CARP project; and from JST, Precursory Research for Embryonic Science and Technology (PRESTO) to Y.O. (20103).
View Abstract

Navigate This Article