Three Brick genes have distinct functions in a common pathway promoting polarized cell division and cell morphogenesis in the maize leaf epidermis

Plant cells' shapes are defined by their surrounding walls. Consequently,cell shapes are determined by patterns of wall expansion during organ growth. Most plant cells expand by means of anisotropic, diffuse growth, with expansion distributed across the cell surface but oriented preferentially in one or more directions. A few highly elongated cell types such as pollen tubes and root hairs expand by means of tip growth, an extremely polarized mode of cell expansion in which growth is focused at a single site on the cell surface. The mechanisms underlying these different modes of cell expansion remain to be fully elucidated but both depend critically on the cytoskeleton.

Drug studies have demonstrated crucial roles for F-actin in tip growth, one of which is to transport secretory vesicles containing cell wall components along longitudinal F-actin bundles to the vicinity of the growth site(Geitmann and Emons, 2000;Hepler et al., 2001). A fine meshwork of cortical F-actin observed at or near the growth site in tip growing cells is thought to have a function separate from that of the longitudinal F-actin bundles, which is not well understood(Geitmann and Emons, 2000;Hepler et al., 2001). Pharmacological and genetic studies have shown that F-actin also promotes the expansion of diffusely growing cells(Smith, 2003). By analogy to its well-established role in tip growth, it has been proposed that the primary role of F-actin in diffusely growing cells is to guide the deposition of secreted wall components (Baskin and Bivens, 1995; Dong et al.,2001).

As the principal structural component of the primary cell wall, cellulose is thought to constrain cell expansion and thereby function as a key determinant of shape in diffusely growing cells. Drug studies have shown that normal patterns of diffuse growth also depend on microtubules, which, during interphase, are arranged in the cell cortex in a pattern mirroring that of cellulose deposition into the wall (Cyr,1994). Although the precise nature of the relationship linking microtubule organization to cellulose deposition pattern remains unknown, it has been proposed that microtubules help to guide the deposition pattern of cellulose microfibrils into the wall(Giddings and Staehelin, 1991;Baskin, 2001).

Some plant cells acquire complex shapes through multidirectional cell expansion patterns involving both microtubule and actin-dependent mechanisms. Three-branched trichomes that form on the Arabidopsis epidermis provide a good example. Drug studies have established that microtubules, but not actin filaments, play a crucial role in the initial polarized outgrowth of trichomes and trichome branches. However, subsequent outgrowth of trichome branches proceeds by means of an actin-dependent, diffuse growth process(Szymanski et al., 1999;Mathur et al., 1999) (M. Hülskamp, personal communication). Mutations disrupting various aspects of trichome morphogenesis have identified dozens of genes required for the proper shaping of this cell type (Bouyer et al., 2001). Molecular analysis of some of the corresponding genes has begun to shed light on microtubule-dependent mechanisms involved in branch formation (Oppenheimer et al.,1997; Burk et al.,2001; Folkers et al.,2002; Kim et al.,2002). Further molecular analysis of these and other gene products yet to be identified will undoubtedly advance our understanding of both microtubule- and F-actin-dependent mechanisms governing trichome morphogenesis.

The lobed shapes of leaf epidermal and mesophyll cells provide another example of a complex shape resulting from multidirectional cell expansion. Lobes arise as polarized outgrowths at multiple sites along the margins of cells whose overall size is increasing via diffuse growth. In lobe-forming cells from a wide variety of species, emergence of lobes is associated with reorganization of cortical microtubules into bands focused at lobe sinuses(Jung and Wernicke, 1990;Apostolakos et al., 1991;Wernicke et al., 1993;Panteris et al., 1993a;Panteris et al., 1993b;Panteris et al., 1994;Wasteneys et al., 1997;Qiu et al., 2002;Frank and Smith, 2002). Microtubule bands have been proposed to direct the localized deposition of cellulose, creating periodic wall thickenings; intervening thinner regions of the wall are presumed to be more extensible, bulging out under the force of turgor pressure to form lobes.

Far fewer studies have considered the role of F-actin in lobe formation. In expanding wheat mesophyll cells, bands of cortical F-actin were observed to co-localize with microtubule bands; cytochalasin treatments suggested an important role for F-actin in the organization of cortical microtubule bands(Wernicke and Jung, 1992). However, two recent studies have provided evidence of a different role for F-actin in the formation of epidermal cell lobes. These studies showed that,in expanding epidermal cells of maize and Arabidopsis, localized enrichment of cortical F-actin is observed at the tips of emerging lobes,somewhat similar to the F-actin arrangement at the tips of tip-growing cells(Frank and Smith, 2002;Fu et al., 2002). Expression of dominant negative and constitutively active forms of ROP2 (a Rho-related GTPase) disrupted the formation or localization of these cortical F-actin enrichments and also perturbed the formation of lobes(Fu et al., 2002). In maizebrk1 mutants, epidermal lobes completely fail to form, although cells expand to achieve a normal overall size. Loss of lobes on brk1 mutant epidermal cells is associated with a failure of localized cortical F-actin enrichments to form, whereas cortical microtubules bands are still present(Frank and Smith, 2002). Thus,both studies suggest a crucial role for localized F-actin enrichments in epidermal lobe formation separate from simply promoting the formation of microtubule bands.

In addition to its effects on epidermal pavement cell shape, brk1causes 20-40% of stomatal subsidiary cells to form abnormally(Gallagher and Smith, 2000). In wild-type leaves, subsidiary cells arise from the asymmetric divisions of subsidiary mother cells (SMCs), whose premitotic polarization involves the formation of a localized enrichment of F-actin at a specific site in the cell cortex. Interestingly, abnormal brk1 subsidiary cells could be attributed to defects in the polarization of SMCs associated with loss of this localized enrichment of cortical F-actin(Gallagher and Smith, 2000). Thus, our observations have revealed a mechanistic link between the formation of polarized outgrowths on the margins of expanding epidermal cells and the polarization of premitotic SMCs, and suggest a role for Brk1 in an actin-dependent aspect of both processes.

In this study, we present an analysis of two additional mutants,brk2 and brk3, which have essentially the same phenotype asbrk1. Through a combination of phenotypic, double mutant and mosaic analyses, we demonstrate that all three Brk genes have distinct functions in a common pathway promoting lobe formation and polarized cell division in the maize leaf epidermis.

Leaves from EMS-mutagenized lines of maize (a generous gift from L. Harper and M. Freeling, UC Berkeley) were screened by making impressions of the leaf surface in Loctite cyanoacrylate glue and examining them in a stereomicroscope for altered epidermal cell shapes. Nine independent mutations with abrk1-like phenotype were identified, which were found to define two new complementation groups, brk2 (with five alleles) andbrk3 (with four alleles). At least two sequential crosses into the B73 inbred background were carried out prior to use of mutants for phenotypic analysis or for crosses to generate double mutants.

The brk2 and brk3 mutations were mapped to chromosome arms using B-A translocations (Beckett,1994). The brk2 locus is located on chromosome arm 1S,and brk3 on 10S. Linkage analysis with visible genetic markers on 1S showed that brk2 is ∼6 cM from Vp5. More precise mapping relative to restriction fragment length polymorphism markers showed thatbrk2 is ∼7 cM from asg31 at the distal end of 1S, andbrk3 is less than 4 cM from php20075a near the tip of 10S.

Epidermal peels were prepared and stained with Toluidine Blue O as described by Gallagher and Smith(Gallagher and Smith, 1999). For cross sections, mature adult leaf pieces were fixed in 4% formaldehyde/50 mM KPO₄ pH 7 for at least 2 hours at room temperature, rinsed,dehydrated through an ethanol series and then infiltrated and embedded in methacrylate resin as described by Gubler(Gubler, 1989). 3 μm sections were attached to slides coated with Biobond (Electron Microscopy Sciences), incubated in acetone for 10 minutes to remove the resin,rehydrated, stained with Toluidine Blue O and examined with a 20×objective under bright field conditions in a Nikon Eclipse E600 microscope. Visualization of microtubules and F-actin in fixed cells was performed as described by Frank and Smith (Frank and Smith, 2002). For analysis of F-actin in living cells via transient expression of green-fluorescent-protein/talin (GFP-talin), strips of leaf tissue 2-8 cm from the bases of immature adult leaves 20-28 cm long were excised and cultured as described by Gallagher and Smith(Gallagher and Smith, 1999). Within 6 hours of the initiation of tissue cultures, plates were bombarded with 0.3-0.8 μg of pYSC14 (Kost et al.,1998) as described by Ivanchenko et al. (Ivanchenko et al., 2000),except that 1 μm gold particles were used as the microcarrier, and tissue was cultured under continuous light conditions at 22°C for 22-26 hours prior to examining by confocal microscopy as described by Frank and Smith(Frank and Smith, 2002).

To generate double mutants, homozygous brk mutants in a B73 background were crossed in each pairwise combination. F₁ progeny were selfed to generate an F₂ generation. Each mutant in the F₂ generation was test-crossed to both of the appropriate single mutants, and test-cross progeny were analyzed to determine the genotypes of the F₂ mutants. Cyanoacrylate glue impressions of leaf surfaces were examined under DIC conditions in a Nikon E600 microscope using a 10× objective, captured with a DAGE MTI CCD72 camera coupled to a Scion LG-3 framegrabber, and processed with Adobe Photoshop 4.0. To collect data forTable 1, at least four images from each impression were captured from randomly chosen locations.

For mosaic analysis, brk3 Oy1-700/brk3 oyl⁺, brk2 vp5/brk2 Vp5⁺ and brk1 lw2/brk1 Lw2⁺plants were crossed to wild-type plants to generate seeds of the following genotypes: brk1 lw2/Brk1⁺Lw2⁺(n=3500), brk2 vp5/Brk2⁺Vp5⁺(n=1500), and brk3 Oy1-700/Brk3 oyl⁺(n=1750). This crossing scheme ensured that each brk mutant allele would always be linked in cis with the marker mutation in the doubly heterozygous progeny. Seeds were germinated, irradiated and grown as described previously (Walker and Smith,2002). As discussed below, vp5-marked brk2sectors were also generated by means of Ac/Ds-induced chromosome breakage. Glue impressions of the leaf surface spanning the entire width of each sector were inspected for the brk phenotype. Sectors showing thebrk phenotype were then fixed and processed for epidermal peels as described by Gallagher and Smith(Gallagher and Smith, 1999). Epidermal peels and hand-cut cross sections from these sectors were examined in a Nikon E600 microscope under epifluorescence conditions using a rhodamine filter set, and under DIC conditions. Images were acquired and processed as described earlier for double mutant analysis.

The wild-type maize leaf epidermis is composed of linear files of cells with diverse shapes. Pavement cells have marginal lobes, finger-like projections that interlock with those of neighboring cells(Fig. 1A). Interspersed within the pavement cells are stomatal complexes, which are composed of two guard cells flanked by two subsidiary cells; each complex of four cells is contained within a single cell file (Fig. 1A, black arrowheads). A screen for EMS-induced mutations affecting epidermal cell morphogenesis in the maize leaf identified nine recessive mutations with a phenotype similar to that of brick1(Gallagher and Smith, 2000;Frank and Smith, 2002). In all of these mutants, pavement cells lack marginal lobes(Fig. 1B). In addition,stomatal subsidiary cells occasionally invade neighboring cell files and have a variety of abnormal shapes (Fig. 1B, Fig. 3A,D,G,black asterisks). Complementation tests demonstrated that the nine newbrk1-like mutations define two additional genes, Brk2 andBrk3. In spite of the striking alterations observed at the cellular level, the overall morphology of brk mutant leaves and plants appears normal, although brk plants are approximately one-third shorter and reach maturity several days later than their wild-type siblings.

To determine whether brk mutations affect cell morphology in other leaf layers in addition to the epidermis, internal tissue layers were examined. As found previously for brk1(Frank and Smith, 2002),transverse sections revealed no differences in the sizes, shapes or organization of cells in internal tissue layers of brk2 orbrk3 leaves compared with the wild type(Fig. 1C,D). Furthermore,direct examination of isolated mesophyll cells confirmed that their lobed shapes are normal in both brk2 and brk3 mutant leaves (data not shown). However, epidermal cells of brk2 and brk3 mutant leaves appear more rounded in cross section than those of wild-type leaves. This appears to be due mainly to a lack of outer epidermal cell wall material filling the crevices between adjacent epidermal cells in wild type leaves(Fig. 1C,D, arrows).

As discussed in the Introduction, several previous studies have attributed lobe formation in expanding epidermal and mesophyll cells to the organization of cortical microtubules into bands that direct a non-uniform pattern of cellulose deposition. As predicted from these earlier studies, cortical microtubules in expanding wild-type epidermal pavement cells of maize are organized into bands focused at lobe sinuses(Fig. 2A, white arrows). In expanding brk1 pavement cells, cortical microtubules also form bands,although they are less distinct than those in wild-type cells(Frank and Smith, 2002). Similarly, in expanding brk2 and brk3 pavement cells,cortical microtubule bands are observed that are less distinct than those in wild-type cells (Fig. 2B,C compared with 2A) but more pronounced than in those in brk1 cells. Furthermore, cortical microtubules in expanding brk2 andbrk3 pavement cells generally appear more aligned with each other than those in brk1 or wild-type pavement cells(Fig. 2B,C compared with 2A)and are sometimes aligned at an oblique angle to the cell's long axis(Fig. 2C for brk3). Thus, as in brk1 mutants, failure of pavement cell lobes to form inbrk2 and brk3 mutants could not be readily explained in terms of a lack of cortical microtubule bands during cell expansion.

Recent studies of actin organization in expanding epidermal pavement cells of wild-type maize and Arabidopsis revealed local accumulations of F-actin at the tips of emerging and elongating lobes(Fig. 2D, white arrows)(Frank and Smith, 2002;Fu et al., 2002). However, inbrk1 mutants, localized accumulations or `patches' of cortical F-actin were not observed at any stage of pavement cell expansion(Frank and Smith, 2002). F-actin organization is similarly altered in brk2 and brk3expanding pavement cells: no cortical F-actin patches were observed when fixed tissues were stained with FITC-phalloidin(Fig. 2E,F).

To confirm that these observations were not due to an artefact of chemical fixation, F-actin organization was also examined in living cells transiently expressing a GFP-talin fusion protein(Kost et al., 1998) introduced via particle bombardment. As in fixed, FITC-phalloidin-stained cells,expanding wild-type pavement cells expressing GFP-talin showed F-actin patches in elongating lobe tips (Fig. 2G, white arrows). In expanding brk2 and brk3pavement cells expressing GFP-talin, no cortical F-actin patches were observed that were comparable to those seen in the wild type(Fig. 2H,I). Although small puncta of GFP-talin fluorescence were occasionally seen at the cell margins at early stages of mutant pavement cell expansion, they were not localized periodically and were generally restricted to small areas of the cell (data not shown). Thus, all three Brk genes are necessary for the formation of cortical F-actin patches, suggesting that they play an important role in pavement cell lobe formation.

The resemblance of F-actin organization at the tips of elongating pavement cell lobes to that in tip-growing cells led us to consider how brkmutations might affect cells known to expand by means of tip growth. No obvious alterations in the morphogenesis of root hairs were present inbrk2 or brk3 mutants (data not shown). Moreover, when crosses are performed with brk2 or brk3 heterozygous pollen,mutant pollen tubes achieve fertilization at approximately the same frequency as competing wild-type pollen tubes. Thus, as in brk1(Frank and Smith, 2002), tip growth in brk2 and brk3 mutant root hairs and pollen tubes appears to be normal.

Double mutant analysis was used to investigate the relationships between the functions of Brk1, Brk2 and Brk3. In populations segregating both single and double mutants derived from self-pollination of double heterozygotes, no phenotype distinct from the brk single mutant phenotype was observed. The genotype of each plant showing a mutant phenotype was determined by test-crossing to the appropriate single mutants. Pavement cell shapes in brk1;brk2, brk2;brk3 and brk1;brk3double mutants are the same as those in brk single mutants(Fig. 3). The frequency of abnormal stomatal subsidiaries provides a more quantitative measure of phenotypic severity. In this respect, brk1 (20-40% abnormal subsidiaries) is more severe than either brk2 or brk3 (5-10%abnormal subsidiaries; Table 1). Abnormal subsidiary frequencies in brk1;brk3 andbrk1;brk2 double mutants are similar to those in brk1 single mutants, whereas abnormal subsidiary frequencies in brk2;brk3 double mutants are similar to those in both brk2 and brk3 single mutants (Fig. 3;Table 1). Thus, all three classes of brk double mutants have the same phenotypes asbrk single mutants, indicating that all three Brk genes function in a common pathway.

To gain a better understanding of the function of each Brk gene,clonal sectors of brk mutant cells in otherwise wild-type leaves were examined to determine whether each of the Brk genes acts cell-autonomously or non cell-autonomously. For any one of the brkmutations, if brk cells neighboring wild-type cells always show the mutant phenotype then that Brk gene acts cell-autonomously. However,if brk mutant cells next to wild-type cells show a wild-type phenotype then the corresponding gene acts non cell-autonomously.

Mosaic analysis of each brk mutation involved a similar scheme for generation of marked, clonal sectors, which is summarized inFig. 4A. Seedlings heterozygous for a brk mutation linked in cis with a recessive or semi-dominant cell-autonomous marker mutation affecting chlorophyll pigmentation were irradiated to produce random chromosome breaks. Cells in which breaks have occurred between the centromere and the wild-type allele for the marker gene divide to produce white (albino) sectors of brkmutant cells in otherwise Brk⁺, green plants. White sectors of brk1, brk2 and brk3 mutant tissue were analyzed to determine the locations of sector boundaries. Wild-type chloroplasts autofluoresce brightly when examined by fluorescence microscopy using a rhodamine filter set. Sector boundaries in internal tissue layers were located by examining leaf cross sections and in the epidermis by examining epidermal peels. In the epidermis, only stomatal guard cells contain mature chloroplasts and can therefore be scored as mutant or wild type for the albino mutation. Data collected through examination of cross sections and epidermal peels for each sector were compiled to reconstruct the composition of each sector(Fig. 4B).

Brk3 is located near the end of chromosome 10S, distal to theOil Yellow 1 (Oy1) locus(Fig. 4A). Plants heterozygous for the semi-dominant Oy1-700 allele are yellow-green, whereas homozygous or hemizygous Oy1-700 plants completely lack pigmentation and are therefore white. Brk2 is located near the end of chromosome 1S ∼6 cM from Viviparous5 (Vp5). The vp5mutation is recessive; homozygous mutant plants are white. Albino-markedbrk2 and brk3 sectors were generated by irradiating brk3 Oy1-700 and brk2 vp5 heterozygous plants(Fig. 4A) (described in Materials and Methods). A second method was also used to createvp5-marked brk2 sectors. Plants homozygous for brk2and heterozygous for vp5 were crossed to wild-type plants homozygous for a chromosome-breaking Ds transposon on chromosome 1S between the centromere and Vp5 (Neuffer,1995). In the progeny, Ac-induced chromosome breakage at the site of this Ds insertion resulted in the loss of wild-typeVp5 and Brk2 alleles, simultaneously uncovering vp5and brk2. Because Ac/Ds-induced chromosome breaks occur relatively late in development, the resulting sectors are small, sometimes only a few cell files in width, whereas sectors induced by irradiation are much larger (Fig. 4A).

We analyzed 72 brk2 and 64 brk3 lateral epidermal sector boundaries (Table 2;Fig. 4B, boundary types A, C and D). At all boundaries of this type, there was a clear difference between the shapes of brk and wild-type cells. As explained earlier, lateral sector boundaries in the epidermis can be located only to the interval between wild-type and albino stomatal files. Although stomatal files are usually separated by at least one non-stomatal file, we occasionally observed areas where sector boundaries split directly adjacent stomatal files such that one file was wild type and the neighboring file was mutant(Fig. 5A,B for brk2,Fig. 6A,B for brk3). In these areas, we could determine with certainty that wild-type cells were lobed, whereas directly adjacent brk2 or brk3 cells were unlobed. Glue impressions of lateral sector boundaries in the epidermis revealed that, at the junctions between wild-type and brk2 orbrk3 pavement cells, lobes of wild-type cells appear to grow over unlobed, brk cells (Fig. 5C, black arrows, for brk2;Fig. 6C, black arrows, forbrk3). These results demonstrate that both Brk2 andBrk3 act cell-autonomously in the lateral dimension to promote lobe formation in epidermal cells. We also analyzed transverse sector boundaries,where mutant epidermis lies over wild-type mesophyll(Fig. 4B, boundary type B). Forbrk2 and brk3, respectively, 30 and 37 such boundaries were analyzed (Table 2). In each case, brk mutant epidermal cells overlying wild-type mesophyll cells were unlobed (Fig. 5A,B forbrk2; Fig. 6A,B forbrk3), indicating that Brk2 and Brk3 also act cell-autonomously in the transverse dimension.

In all brk mutants, stomatal subsidiaries occasionally form abnormally. Normal stomatal development in maize is illustrated inFig. 7. An asymmetric division gives rise to a small guard mother cell (GMC), which is thought to signal its neighbors to the left and right (SMCs) to become polarized with respect to the GMC and to divide asymmetrically to form subsidiary cells. The GMC later divides longitudinally to give rise to a guard cell pair. Analysis of stomatal development in brk1 mutants showed that abnormal subsidiaries arise from SMCs that failed to polarize properly before dividing(Gallagher and Smith, 2000). The presumed role of GMCs in stimulating the polarization of adjacent SMCs raises the interesting question of whether SMC polarization depends onBrk gene function in the GMC or in the SMC itself. We addressed this question by examining mosaic stomata in which mutant guard cells were flanked by a wild-type subsidiary, or vice versa.

In contrast to its cell-autonomous action in promotion of pavement cell lobe formation, analysis of mosaic stomata showed that Brk3 acts non cell-autonomously to promote polarized SMC divisions. Wild-type subsidiary cells in sectored leaves were never abnormal, whether flanking wild-type orbrk3 guard cells, demonstrating that the presence of a wild-typeBrk3 allele in the SMC alone is sufficient for its normal division. Whereas 15% of brk3 subsidiaries flanking brk3 guard cells were abnormal, only 1.5% of brk3 subsidiaries flanking wild-type guard cells were abnormal (Table 3; Fig. 6A,arrowheads), demonstrating that wild-type GMCs rescued the abnormal divisions of adjacent, mutant SMCs. Notably, however, the abnormal divisions ofbrk3 SMCs were not rescued by adjacent wild-type pavement cells(n=117), so the adjacent GMC appears to be the only cell that can supply Brk3 function to a brk3 mutant SMC.

Similar analysis of brk2 mosaic stomata showed that Brk2acts partially non cell-autonomously to promote the polarized divisions of SMCs. As for Brk3, the presence of a wild-type Brk2 allele in the SMC is sufficient for its normal division, because 3.3% of wild-type subsidiaries in sectored leaves were abnormal whether flanking wild-type orbrk2 guard cells. Whereas 18.4% of brk2 subsidiaries flanking brk2 guard cells were abnormal, 10.5% of brk2subsidiaries flanking wild-type guard cells were abnormal(Table 3;Fig. 5D, arrowheads). Thus,brk2 SMC divisions were partially rescued by an adjacent, wild-type GMC.

Brk1 is located near the end of chromosome 5S. As illustrated inFig. 4A, the recessivelemon white 2 (lw2) mutation, located between the centromere and brk1, was used to mark brk1 mutant sectors. To determine whether Brk1 acts cell-autonomously or non cell-autonomously within the epidermis, we analyzed 63 lateral sector boundaries(Fig. 4B, types A, C and D). In areas where these boundaries split adjacent stomatal cell files, brk1cells in direct contact with wild-type cells appeared to be wild type, with pavement cell lobes and normal stomata(Fig. 8A,B). Moreover, weakly lobed pavement cells were observed in brk1 mutant cells two to three cell files away from the nearest wild-type epidermal cell(Fig. 8A, black arrows). Thus,direct contact with a wild-type cell is not necessary to rescue the phenotype of a brk1 cell. From these observations, we conclude thatBrk1 acts non cell-autonomously in the lateral dimension.

To investigate whether Brk1 also acts non cell-autonomously in the transverse dimension, 53 sectors were identified in which brk1 mutant epidermis was overlying wild-type internal tissue layers. At 28 boundaries with wild-type mesophyll directly underlying brk1 epidermis(Fig. 4B, boundary type B), the mutant epidermis was indistinguishable from wild-type epidermis, with lobed pavement cells and normal stomata (Fig. 8C,D, right-hand side). By contrast, at 25 boundaries wherebrk1 epidermis was separated from a wild-type cell layer by one or more layers of brk1 mesophyll(Fig. 4B, boundary type E), the epidermis appeared to be mutant, lacking marginal lobes and having some abnormal subsidiary cells. These results show that Brk1 is expressed and functions in internal tissue layers even though no mutant phenotype was observed in these layers of brk1 leaves. Furthermore, Brk1expressed in these layers rescues the phenotype of directly overlyingbrk1 mutant epidermal cells. Therefore, Brk1 acts non cell-autonomously in the transverse dimension but its influence does not travel as far in this dimension as it does within the epidermis.

In this study, we have used a variety of genetic approaches to analyze the functions of three Brk genes involved in epidermal cell morphogenesis and division in the maize leaf. All three Brk genes are required for the formation of polarized outgrowths giving rise to lobes on the margins of epidermal pavement cells but not for cell expansion in general, and are also involved in the polarized divisions of SMCs.

Consistent with many previous studies of lobe formation, cortical microtubules are organized into bands focused at lobe sinuses of expanding wild-type pavement cells in the maize leaf epidermis. In spite of the absence of lobe formation, these microtubule bands are also present in all threebrk mutants, although they are somewhat less distinct. Visualization of F-actin in expanding wild-type pavement cells by means of phalloidin staining of fixed cells and GFP-talin expression in living cells showed distinct enrichments of cortical F-actin at lobe tips. These observations are similar to those reported in a recent study in which F-actin organization was visualized in living, expanding Arabidopsis epidermal pavement cells expressing GFP-talin (Fu et al.,2002). Another recent study of lobe formation inArabidopsis pavement cells revealed relatively little enrichment of cortical F-actin in lobe tips (Qiu et al.,2002). In this study, however, antibody staining of fixed cells was used to visualize F-actin, so some cortical F-actin might have been lost during the fixation process. Our observation of cortical F-actin enrichments at lobe tips in both fixed and living maize cells indicates that this feature of cytoskeletal organization is not an artefact of the F-actin visualization method used. In both living and fixed expanding epidermal pavement cells of all three brk mutants, cortical F-actin enrichments were not observed, supporting the conclusion that these F-actin enrichments are important for lobe formation, and indicating that the Brk genes are required for their formation. By analogy to the proposed function of F-actin in both tip-growing and diffusely growing cells, the F-actin enriched in lobe tips might function to guide or promote localized secretion.

Analysis of brk1;brk2, brk1;brk3 and brk2;brk3 double mutants showed that they have the same phenotypes as the corresponding single mutants. These results indicate that all three Brk genes act in a common pathway. Although this might mean that the genes act in a linear sequence, it is also possible that the products of these genes simply act together in a common process, possibly directly interacting with each other. Although the Brk1 gene product has been identified(Frank and Smith, 2002),investigation of its functional relationship to Brk2 andBrk3 gene products awaits the cloning of these other two genes.

Mosaic analyses were carried out to determine whether each Brkgene acts cell-autonomously or non cell-autonomously. Brk1 acts non cell-autonomously in both lateral and transverse leaf dimensions over a short distance. Although the non cell-autonomous influence of Brk1 does not depend on direct cell-cell contact, it travels farther within the epidermal layer (2-3 cell diameters) than it does between leaf layers (1 cell diameter). The non cell-autonomous action of Brk1 suggests that this gene encodes or controls the production of a diffusible or transported molecule. The Brk1 gene was shown to encode a novel protein of ∼8 kDa that is highly conserved in both plants and animals(Frank and Smith, 2002). Neither BRK1 nor BRK1-related proteins in other organisms have the predicted signal peptides expected for secreted proteins. Moreover, recent work has implicated the mammalian BRK1 homolog HSPC300 in the regulation of actin polymerization (Eden et al.,2002), clearly pointing to an intracellular function for BRK1. However, the non cell-autonomous action of Brk1 might reflect movement of BRK1 from cell to cell through plasmodesmata. Particularly at early developmental stages, a variety of proteins considerably larger than BRK1 appear to be transported from cell to cell via plasmodesmata in maize and several other plant species (Jackson,2000). Whether the BRK1 protein moves intercellularly and the functional significance of any such movement remain to be determined.

Unlike Brk1, mosaic analyses showed that Brk2 andBrk3 act cell-autonomously to promote the formation of epidermal pavement cell lobes, indicating that both gene products act in or on each pavement cell to direct its own morphogenesis. Surprisingly, however,Brk3 acts non cell-autonomously, and Brk2 partially non cell-autonomously, to promote the polarized divisions of SMCs. Although polarization of SMCs appears to depend on a signal from the adjacent guard mother cell (GMC), it is unlikely that Brk3 or Brk2 function directly in this signaling process. If they did, we would expect that a subsidiary cell's phenotype would depend solely on the genotype of its adjacent GMC, but we found instead that the presence of a wild-typeBrk3 allele in either the SMC or the adjacent GMC is sufficient for a normal SMC division. The same is true for Brk2, except that a wild-type Brk2 allele in the GMC only partially rescues the polarized divisions of adjacent brk2 SMCs. In summary, the Brk2 andBrk3 function needed for polarized SMC divisions can be provided by the SMC itself, or by the adjacent GMC, but not by surrounding pavement cells. The specificity observed in the non cell-autonomous action of Brk3and Brk2 might be due to diffusion or active transport of BRK3 protein (and to a lesser extent, BRK2 protein) from GMCs to SMCs, but not between other leaf cell types. Alternatively, presence of BRK3 and BRK2 proteins in either the SMC or the GMC might in some way facilitate intercellular communication between these two cell types leading to proper polarization of SMCs.

In summary, genetic and phenotypic analyses of brk mutations show that they define three genes that act in a common pathway promoting specific aspects of maize leaf epidermal morphogenesis and division that involve local actin polymerization. Moreover, mosaic analyses yielding different results for each Brk gene show that the product of each gene has a distinct function in that pathway. The mammalian homolog of BRK1, HSPC300, has recently been directly implicated in regulation of Arp2/3-dependent actin polymerization (Eden et al.,2002). The Arp2/3 complex (putative components of which are encoded by various plant genomes) nucleates polymerization of new actin filaments at specific sites in the cell when activated by proteins that respond to localized intracellular or extracellular cues(Higgs and Pollard, 2001). HSPC300 was identified as a component of a multiprotein complex isolated from bovine brain extracts that also includes WAVE, an Arp2/3 activator whose activity is regulated by Rac. The intact WAVE complex is inactive but, in the presence of GTP-Rac (or Nck, another regulatory protein), WAVE and HSPC300 dissociate from the rest of the complex to form a two-protein subcomplex that stimulates Arp2/3-dependent actin polymerization(Eden et al., 2002). Thus, BRK1 probably functions as part of a complex that also contains an Arp2/3 activator. Brk2 and Brk3 might encode other components of such a BRK1-containing complex, or factors that positively regulate its activity. Indeed, either Brk2 or Brk3 might encode a WAVE ortholog, a particularly exciting possibility given that plant genomes are not predicted by sequence homology to encode WAVE-like proteins or other Arp2/3 activators. Interestingly, the Arabidopsis SPK1 gene is also required for the formation of epidermal pavement cell lobes; this encodes a protein containing a motif that has previously been shown to interact with Rac GTPases(Qiu et al., 2002). Moreover,expression of both dominant negative and constitutively active forms of the Rac/Rho-related Arabidopsis GTPase ROP2 also perturbs the formation of epidermal pavement cell lobes (Fu et al., 2002). In combination with recent findings on the role of HSPC300 in regulation of local actin polymerization by Rac, these findings suggest that BRK proteins might function in the same pathway (probably downstream of) maize SPK1 and ROP2 orthologs. To investigate these possibilities, future work will be aimed at the identification and analysis ofBrk2 and Brk3 gene products.

We thank Challe Woosley, Dan Burnes, Miroslav Dudas and Andy Teng for their contributions to this work. We also thank Lisa Harper and Michael Freeling for F₂ populations segregating EMS-induced mutations, Gerry Neuffer for chromosome-breaking Ds insertion lines, Phil Becraft for theOy1-700 allele and advice for its use as a marker for mosaic analysis, the Maize Genetics Coop. for seeds and RFLP markers, Nam-Hai Chua for the pYSC14 GFP-talin plasmid, the Torrey Mesa Research Institute for use of their particle gun, and Shiping Zhang for his help with use of the gun. Thanks also to Adrienne Roeder, Anne Sylvester, David Jackson and members of the Smith lab for helpful discussions. This work was supported by NSF grant IBN-9817084 to L.G.S.