One-celled spacing pattern.

A consistent feature of patterning is that stomata are separated from each other by at least one cell (Sachs, 1991). In Arabidopsis, as well as in other species, the frequency of stomata in contact is much lower than would be found in a random distribution (Figures 1 and 3; Korn, 1972; Sachs, 1978, 1991; Geisler et al., 2000). The stoma-free zone reduces the overlap between gaseous diffusion shells from nearby stomata. This increases the efficiency of each stoma, avoids unnecessary evaporation, and establishes an optimal ratio between CO₂ uptake and photosynthetic capacity. The minimal one-celled spacing pattern also ensures the presence of neighbor cells for ion exchange.

Higher order spacing patterns.

In some cases, different stomatal densities result from higher order patterns. For example, more cells separate adaxial (upper epidermis) than abaxial (lower) stomata in Arabidopsis leaves and cotyledons (Serna and Fenoll, 2000a; Geisler et al., 2000; Geisler and Sack, 2002). Variations in spacing may optimize the ratios between CO₂ uptake, evaporation, and internal leaf temperature. The presence of some adaxial stomata would allow evaporative cooling (as well as gas uptake), while higher densities might permit too much water loss.

Modified stomata.

Arabidopsis also has modified stomata such as in the teeth of leaves (hydathodes), in anthers, and in nectaries at the base of stamens (Figure 4; Bowman and Smyth, 1999; Baum et al., 2001). Modified stomata may not function in gas exchange; some may release fluids such as excess water (guttation) or nectar (Davis and Gunning, 1993). Some modified stomata in wild-type Arabidopsis occur in contact (Figure 4) and may be patterned differently than normal stomata.

MMCs.

The first precursor cell is the first cell in the stomatal pathway to divide asymmetrically. This division produces a smaller, usually triangular meristemoid and a larger sister cell (Figure 5). As far as is known, all Arabidopsis stomata are produced by at least one asymmetric division, that of the MMC. MMCs that are committed to this division can be recognized by their polarized cytoplasm with the nucleus and preprophase band of microtubules located at one pole, and a vacuole at the other (Zhao and Sack, 1999). MMCs typically originate from relatively small epidermal cells that do not have very sinuous cell walls.

Meristemoids.

Some newly formed meristemoids convert directly into guard mother cells, the third and final precursor. Other meristemoids divide one to three times asymmetrically before conversion. As in the asymmetric divisions of MMCs, meristemoids that are about to divide can be recognized by their polarized cytoplasm (Figure 9; Galatis and Mitrakos, 1979). Each time a meristemoid divides, it regenerates a meristemoid and a larger sister cell. In this way, a meristemoid can be considered a “stem” cell. We were not able to predict the number of times a meristemoid will divide by its size, position, or any other parameter that can be detected by the dental resin method. Thus, the simplest assumption is that all meristemoids have the same potential for divisions, and that the factors that regulate the number of divisions need to be determined. It is, however, useful to separate out one class of meristemoids, satellite meristemoids, based on where they form. Satellite meristemoids originate from MMCs that are located next to a pre-existing stoma or precursor and, as will be discussed, are central to the generation of the spacing pattern.

GMCs.

Guard mother cells divide symmetrically to produce two guard cells. The GMC division site, the location of cell plate fusion with the parental cell wall, is marked by a preprophase band of microtubules, a marker found in almost all types of cells (Zhao and Sack, 1999 Smith, 2001). In Arabidopsis GMCs, the division site is also marked by wall thickenings at opposite ends of the cell (Figure 9; Zhao and Sack, 1999). While similar thickenings have been found in GMCs in the Fabaceae (Galatis and Mitrakos, 1979; Galatis et al., 1982), they have not been found in any other cell type. It is not clear whether these thickenings function in guiding cytokinesis and/or whether they help the GMC attain an oval shape.

GMC and stoma.

The final stage of stomatal development involves the differentiation of the stoma itself. Differentiation in Arabidopsis is comparable to that described for other dicots (Galatis and Mitrakos, 1980; Sack, 1987; Zhao and Sack, 1999). Stomatal morphogenesis actually begins in the GMC. The oval outline of the stoma is first established in the GMC, perhaps by the end wall thickenings that may restrict local elongation of the cell wall (Figure 9). Other features of mature stomata, such as starch accumulation and vacuolar enlargement also initiate in the GMC, and then continue until the stoma matures.

Stomatal pore.

Pore formation starts with a lens-shaped thickening located at the middle of the ventral wall (the new wall formed in the GMC; Figure 9). This thickening may be organized by radial arrays of microtubules that have gamma-tubulin at their focus (Marc, 1997). It is striking that each guard cell develops with a mirror-like symmetry with respect to the other. For example, the radial microtubule arrays and the wall thickenings in each guard cell are precisely opposite each other (Wasteneys et al., 1997). After the pore thickening reaches a critical stage, the anticlinal walls separate in the region of the thickening. This creates the stomatal pore. Obviously, the location of wall separation is regulated spatially, a trait under strong selection pressure because a separation elsewhere could result in desiccation.

Wall specialization.

Stomata have an elaborate architecture. In Arabidopsis, there are distinct differences in the thickness of the different cell walls in the mature guard cell (Zhao and Sack, 1999). In addition, cellulose is deposited in arrays that radiate out from the pore, arrays that presumably result from microtubule-directed deposition. The collective distributions of cellulose, wall thickenings, and hinges presumably translate changes in hydrostatic pressure into pore opening and closing. The elimination of plasmodesmata during stomatal development helps maintain turgor by making each guard cell an island of symplasm in the epidermis.

Anisocytic complexes.

Arabidopsis and other members of the Brassicaceae exhibit anisocytic complexes where the stoma is surrounded by three cells, one of which is smaller (Figures 5 and 9B; Metcalfe and Chalk, 1950; Pant and Kidwai, 1967). This arrangement can result from three consecutive asymmetric divisions, one of the MMC and two of the meristemoid (Berger and Altmann, 2000; Geisler et al., 2000). Because the divisions are arranged in an inward spiral, the larger daughter cells become located to the outside; as a result they completely surround the stoma and their relative size correlates with their age. Here the complex is monoclonal because all five cells (three neighbor cells plus two guard cells) originated from the same MMC. Anisocytic complexes make up 40–66% of all complexes in leaves and cotyledons (Geisler et al., 2000; Serna and Fenoll, 2000a).

The inward spiral of asymmetric divisions can occur in a clockwise (Figure 5) or in a counterclockwise direction (Serna et al., 2002). Moreover, the angle of the symmetric division of the GMC is often roughly parallel to that of the last asymmetric division of the meristemoid (Galatis and Mitrakos, 1979; Serna et al., 2002). Nothing is known in this system about how division planes are positioned with respect to each other over four sequential divisions.

Complexes other than anisocytic.

Other complexes in Arabidopsis are not anisocytic either because the stomata are surrounded by 2, 4 or 5 neighbor cells, or because the three cells are of equal size, or because two complexes share one neighbor cell (Figure 3; Geisler et al., 2000; Serna and Fenoll, 2000a). Using the dental resin impression method, we found that this variability results in part from how many times meristemoids divide (Figure 5; Geisler et al., 2000). We estimated that about a third of all meristemoids in a lineage divide once or not at all before converting to a GMC. The resulting complexes would contain at least one non-clonal neighbor cell and would thus be polyclonal. Other estimates of the number of polyclonal complexes are lower (13 to 23%) and were derived from the analysis of transposon-induced sectors marked by GUS-staining (Larkin et al., 1996; Serna and Fenoll, 2000a; Serna et al., 2002). Thus, about 13–34% of Arabidopsis stomatal complexes are polyclonal, and 33–60% are not anisocytic.

These findings show that unlike the one-celled stomatal spacing pattern, the anisocytic pattern is not obligate. Because not all cells adjacent to stomata are distinct morphologically, a requirement for the term “subsidiary cell” (Esau, 1977), we refer to these cells generally as neighbor cells. More importantly, the existence of polyclonal complexes has implications for how the stomatal spacing pattern is generated (see below).

Hypotheses for spacing.

As mentioned the central feature of patterning is that stomata are separated by at least one intervening cell. Several hypotheses have been proposed to explain how this pattern is generated in dicotyledons (Figure 11; Sachs, 1979, 1991; Sylvester et al., 1996; Larkin et al., 1997; Croxdale, 2000).

The classical lateral inhibition hypothesis invokes the presence of an inhibitory field around developing stomata that prevents new stomata from forming in neighbor cells (Bünning, 1953; Korn, 1993, 1994). This implies that a neighbor cell should be prohibited from functioning as an MMC.

In the cell lineage hypothesis, a series of stereotyped asymmetric divisions creates a monoclonal boundary of neighbor cells (Sachs, 1991; Larkin et al., 1997; Serna and Fenoll, 2000a; Serna et al., 2002). According to this hypothesis, the placement of neighbor cells is generated within the cell lineage and does not require communication with surrounding cells. This is thought to prevent the formation of stomata in contact by generating intervening cells.

A third view, based on study of the same cells through time, is that stomata are mostly randomly placed, but oriented divisions correct occasional patterning mistakes (Sachs, 1991; Sachs and Kagan, 1993; Sachs and Novoplansky, 1993). For example, Sachs and co-workers found that when meristemoids formed in contact, one meristemoid divided away, behavior thought to result from cell signaling.

Stomata are spaced by the placement of satellite meristemoids.

We studied the development of many Arabidopsis stomata using the dental resin impression method (Geisler et al., 2000). This made it possible to analyze how Arabidopsis stomata are patterned. We found that many neighbor cells become MMCs, including cells adjacent to a stoma, a GMC or a meristemoid. The asymmetric division of the neighbor cell MMC is invariably oriented so that the new meristemoid, the satellite meristemoid, is placed away from the pre-existing stoma/precursor (Figure 12).

This is a central event in stomatal patterning in Arabidopsis in the abaxial leaf epidermis. The majority of stomata in these tissues originate from satellite meristemoids. Satellite meristemoid placement creates the one-cell spacing around this precursor and maintains the space around the pre-existing stoma or precursor so that both become patterned.

It is likely that cell signaling influences the orientation of the asymmetric divisions that produce satellite meristemoids. Satellite meristemoids were correctly placed regardless of whether the new MMC was or was not clonally related to the adjacent stoma/precursor. This argues against the idea that division orientation requires the inheritance of mitosis allocated factors which organize the site of subsequent divisions as is found in yeast budding (Chen et al., 2000). The simplest explanation is that the neighbor cell receives signals from the pre-existing stoma or precursor, positional information that is used to select the location of the division site.

Relationship to other patterning hypotheses.

Our results support Sachs' (1991) concept that the positions of the first generation of stomata are generally not regulated with respect to each other. We found that MMCs formed in contact with each other with a frequency predicted by chance (Figures 10 and 13; Geisler et al., 2000). This suggests that the first precursor cells are randomly placed and that MMCs do not laterally inhibit each other.

As noted, Sachs (1991) suggested that oriented divisions could correct patterning mistakes. We also found that when meristemoids were in contact, that either or both divided away (Geisler et al., 2000). While this contributes to stomatal patterning, we found it occurs less frequently in Arabidopsis than satellite meristemoid formation, and thus it is a secondary mechanism of patterning.

Bünning's (1953) formulation of the lateral inhibition hypothesis (Figure 11) implies that the presence of a stoma prevents adjacent cells from becoming stomata. Lateral inhibition usually means that one cell prevents an adjacent cell from assuming a particular fate (Appel et al., 2001). But the only way that a neighbor cell can form a stoma is by functioning as an MMC, a frequent event in Arabidopsis. Thus, in this sense, patterning does not involve lateral inhibition. However, the presence of a stoma or precursor does seem to inhibit the formation of a meristemoid nearby. But the mechanism is the control of the placement of the MMC division site, not the fate of the neighbor cell itself.

Our data suggest that cell lineage does not generate the one-celled spacing pattern in the leaf epidermis. First, the lineage hypothesis doesn't provide a mechanism for how stomata in polyclonal complexes are spaced. Up to a third of all complexes are polyclonal in the abaxial epidermis of the first leaf and cotyledon (Geisler et al., 2000). The cell lineage hypothesis as formulated fails to explain why the non-clonal cells in polyclonal complexes are not stomata.

Second, even when a monoclonal complex is produced, neighbor cells do not provide a boundary or buffer, because many divide and produce stomata. It is not the existence of the neighbor cell itself, but the correct placement of the satellite meristemoid that establishes the minimal one-celled spacing pattern.

Third, even when adjacent complexes are each monoclonal (Serna and Fenoll, 2000a; Serna et al., 2002), it remains to be demonstrated that a series of stereotyped divisions within a cell lineage actually establishes the one-cell spacing pattern. The key question is when in development the separation between meristemoids is first generated. A single asymmetric division can be sufficient to create a one-celled spacing such as in the placement of the satellite meristemoid or in the division of meristemoids in contact. In addition, many meristemoids are correctly spaced by chance (Geisler et al., 2000). In all three cases, subsequent asymmetric divisions in a lineage would increase the number of cells between stomata but would not affect the already established minimal one-celled spacing. Although cell lineage may not generate the one-celled spacing pattern, divisions within a cell lineage are crucial for establishing the position, arrangement and higher order spacing of stomata and other epidermal cells.

Cell position and events in the stomatal pathway.

Some events in the stomatal pathway seem independent of cell position. For example, MMCs can form in contact, and the divisions of these MMCs appear to be randomly oriented (Figure 13A–E). It is unclear whether other events require positional cues communicated via intercellular signaling. For example, the divisions of meristemoids in an inward spiral could result from an endogenous patterning program such as through the allocation of spatial cues through mitosis.

Other events seem to require the communication of information about adjacent cell position. Examples mentioned include the orientation of asymmetric divisions in neighbor cells and in adjacent meristemoids (Figure 13). Cell signaling also appears to regulate the fate of cells in two specific positions in the stomatal pathway in Arabidopsis (Figure 13 H and I; Geisler et al., 2000). First, when meristemoids accidentally form in contact, sometimes one meristemoid arrests and does not divide or become a GMC. Second, cells next to any combination of two stomata, meristemoids, and GMCs seem to be prohibited from dividing asymmetrically. Neither event seems to play a major role in stomatal patterning (Geisler et al., 2000).

In domains that have fewer stomata and satellite meristemoids, such as the adaxial epidermis, spacing may rely more upon chance or minor spacing mechanisms. But even in the adaxial epidermis, we estimate that the majority of all stomata are patterned by satellite meristemoid placement (Geisler and Sack, 2002).

In summary, some stomata are placed by chance. The remainder is spaced largely by oriented asymmetric divisions of neighbor cells. Other position-dependent events play either a secondary role or are not directly involved in generating the minimal one-celled spacing pattern.

Cell competence and selection as MMCs.

Stomatal formation appears prohibited in several positions. In some cases, such as around trichomes, this prohibition occurs before asymmetric divisions start in the developing leaf (Larkin et al., 1997; Martin and Glover, 1998; Glover and Martin, 2000). In other cases, specific cells become prohibited even in regions actively forming stomata. For example, we found that cells next to two stomata and/or precursors in any combination usually do not divide asymmetrically (such as the cell to the right of red asterisk in Figure 12 bottom; Geisler et al. 2000).

The differentiation of pavement cells also removes cells from the pool of those competent to form MMCs. Large, sinuous cells that are known to undergo endoreduplication have not been observed to divide asymmetrically (Melaragno et al., 1993; Geisler et al., 2000; Glover, 2000). Conversely, smaller less wavy cells with 2C levels of DNA are most likely to be division competent and form MMCs (Donnelly et al., 1999; Geisler et al., 2000). Such cells are abundant in the young, developing epidermis. Later in leaf development, smaller cells are mostly located next to stomata (Figure 14). Thus the generalized capacity of the protoderm to initiate a stomatal lineage later becomes progressively restricted to small, less sinuous cells. Multiple generations of stomata can develop over the course of leaf development when sister cells to satellite meristemoids themselves divide asymmetrically and so on (Landré, 1972; Berger and Altmann, 2000).

Other small cells arise from symmetric divisions. Symmetric divisions produce cells of comparable size and developmental potential (Geisler et al., 2000). Symmetric divisions occur in neighbor cells (Figures 7 and 10) or in cells removed from stomata or their precursors. The pool of cells competent to form stomata is filled by asymmetric and symmetric divisions, and drained by cell enlargement and differentiation. Stomatal number is therefore partly a function of the balance between cell proliferation and cell differentiation.

In summary, stomatal number is regulated by the availability of division competent cells and by how many of those cells are induced to divide asymmetrically. Many of the cells capable of forming MMCs originate from prior asymmetric divisions in the stomatal pathway.

Stomatal index.

This ratio depends upon how many stomata and other cells arise during leaf development. MMC divisions directly increase the number of both stomata and epidermal cells. Symmetric divisions of epidermal cells and asymmetric divisions of meristemoids directly increase only the number of epidermal cells. Thus, the stomatal index may be modulated by the differential regulation of these three types of divisions, a possibility that has not been studied. Such an analysis could identify the mechanisms by which environmental parameters influence plant productivity.

Longitudinal gradients in organs.

In developing Arabidopsis leaves most stomata form and mature in a roughly tip to base gradient (Pyke et al., 1991; Donnelly et al., 1999). Successive waves of stomatal initiation, satellite meristemoid formation, and generalized cell enlargement start at the tip and proceed basipetally.

Grasses also display a tip-to-base gradient of stomatal formation. But the gradient in stomatal development is predictable in time and space within cell files. Meristemoids appear to be absent in grasses (Larkin et al., 1997). Stomatal number in monocots is controlled partly by the production of new regions of the leaf blade. In Arabidopsis leaves the initial longitudinal gradient in stomatal formation is much less strictly defined and is disrupted by mosaic growth. Stomatal formation is asynchronous and variable in the number of asymmetric divisions. Stomata are produced iteratively and progressively including in regions that are otherwise mature (Figures 12 and 14).

Relatively little is known about stomatal development in Arabidopsis stems and other cylindrical organs such as petioles, carpels and pedicels. In hypocotyls, a series of longitudinal and transverse divisions seems to create a pool of cells competent to form MMCs (Berger et al., 1998b). In stems and hypocotyls stomata appear to mature acropetally (Nadeau and Sack, unpublished). These observations imply that longitudinal waves of competence to form MMCs pass through developing leaves and stems.

Leaves.

There seems to be no obligate relationship between the placement of stomata and substomatal cavities in dicot leaves. In one study, 13% of mature stomata were located directly over a mesophyll cell rather than a substomatal cavity (Sachs, 1979). Stomatal precursor cells also can be positioned over mesophyll cells instead of cell junctions (Figure 15; reviewed in Sachs, 1978).

Serna and Fenoll (2000a) analyzed whether stomatal placement is coordinated with internal tissues in Arabidopsis. Because substomatal cavities are small or absent in the adaxial side of the leaf, they used junctions rather than cavities between mesophyll cells for scoring. Over 90% of stomata and meristemoids were located over cell junctions. This percentage differed significantly from their estimate of a randomly generated placement. However, meristemoids only form at cell edges, a feature that would differ from a randomly generated control. Also, epidermal cell outlines may be positionally out of register with the cells below (Figure 15). Using these parameters, it would be valuable to confirm that the distribution over cell junctions is non-random. If so, then radial signaling may take place between these two cell layers. One scenario could be that a signal from mesophyll cells could help orient asymmetric divisions so that meristemoids tend to form over cell junctions (Serna and Fenoll, 2000a).

Hypocotyls.

Hypocotyl stomata are present in files of short, flattened cells and excluded from files of cells that are elongated and bulging (Gendreau et al., 1997). The protruding epidermal cells are located over the periclinal wall of a single underlying cortical cell. The files that produce stomata are above anticlinal walls (Schiefelbein 2002; Berger et al., 1998b; Hung et al., 1998). Thus, the competence to form MMCs is restricted in a position-dependent manner to hypocotyl cells that are radially aligned with subjacent landmarks.

A similar mechanism controls root hair formation and considerably more is known about root hair patterning. Root hairs form in cells (H cells) located over anticlinal cortical cell walls rather than in cells (N cells) over periclinal walls (Schiefelbein, 2000, 2002). The signal that coordinately establishes root hairs over anticlinal walls is unknown. The cellular pattern is set up in the embryo and is maintained in derivatives of the root apical meristem (Berger et al., 1998a). These authors killed subjacent root cortical cells using laser ablation before the fate of recent epidermal derivatives was fixed so that the derivatives no longer contacted living cortical cells. The isolated derivatives continued to express markers appropriate to their position relative to underlying cells. These data argue against the need for continuous communication between cell layers, and are consistent with the possibility that the outer cell walls of dead cortical cells retain some factor that maintains the fate of the overlying epidermal cells (Berger et al., 1998a; Benfey, 1999).

The formation of stomata and root hairs over anticlinal cortical walls is a striking example of the conservation of a radial patterning mechanism. It places stomata over future intercellular spaces and root hairs over the shortest route to the xylem. This conservation involves common molecular pathways, the subject of the next section.

Clusters result from disruption of the number, plane and polarity of asymmetric divisions.

TMM is essential for the establishment of stomatal patterning (Yang and Sack, 1995; Larkin et al., 1997). A major tmm phenotype is the presence of clusters of adjacent stomata in leaves and cotyledons (Figure 16). Stomata in clusters are arranged at different angles with respect to each other, and the stomata vary in size and stage of development. These traits suggest that the clusters develop progressively and iteratively.

Analysis of cluster development using both dental resin impressions and static images shows that tmm disrupts several spacing mechanisms (Figure 17; Geisler et al., 2000).

First, instead of satellite meristemoids being placed away from pre-existing stomata, some tmm meristemoids form in contact with the stoma (Figure 17; Yang and Sack, 1995). These ectopic satellite meristemoids result from the misorientation of the plane of asymmetric divisions in neighbor cells. Analysis of division angles shows that tmm randomizes the division plane. This appears to be the major defect resulting in cluster formation.

Second, tmm disrupts a secondary stomatal spacing mechanism, the oriented divisions of meristemoids in contact (Figures 11 and 13; Geisler et al., 2000). Unlike the wild type, tmm meristemoids fail to divide away from each other.

Third, stomatal clusters also develop from alterations in the frequencies of cell divisions. Satellite meristemoids, like all meristemoids, can divide asymmetrically, divisions that often add to the separation between the pre-existing stoma and the new one. However, tmm meristemoids divide fewer times compared to the wild type resulting in an acceleration of when meristemoids convert into GMCs. This reduces the likelihood that meristemoid divisions will correct patterning mistakes.

Fourth, clusters form partly from an overproduction of satellite meristemoids. While the number of asymmetric divisions of meristemoids is reduced in tmm, the number of neighbor cells that divide asymmetrically increases. Thus, TMM acts as a negative regulator of asymmetric divisions in neighbor cells, and a positive regulator of divisions in meristemoids. The consequence of disrupting this regulation is an overproduction of satellite meristemoids, many of which are misplaced and which may not divide.

Fifth, cells that do not divide in the wild type do so in tmm. Cells located between two stomata or precursors appear to be prohibited from dividing asymmetrically in the wild type. The asymmetric divisions of these cells in tmm combined with randomized division planes produces more stomata in contact (Figure 17). Thus, cluster formation also results from the disruption of the position-dependent control of cell fate.

Sixth, tmm also occasionally alters the fates of daughter cells of divisions that normally would be asymmetric. Usually tmm asymmetric divisions are clearly polar meaning that they produce daughter cells of different sizes and fates. However, sometimes both products of a division in a neighbor cell develop into stomata. This shows that TMM is also necessary for the polarity of the asymmetric division, and that tmm may disrupt the segregation of different fate factors.

Thus, stomatal clusters result from alterations in the orientation and polarity of asymmetric divisions, from disruptions in cell fate, and from defects in the regulation of division frequency. Many of these events probably require cell-cell signaling to communicate spatial cues. Both tmm alleles are recessive and are presumed to be loss-of-function mutations. We hypothesize that TMM functions in an intercellular communication pathway that receives and conveys information about the identity and location of nearby cells. Positional cues may play several roles. These may orient and fix the plane of asymmetric divisions in neighbor cells. They could be required for the correct distribution of cell fate determinants that result in the asymmetric fates of daughter cells. Also spatial signals may guide the divisions of meristemoids in contact, and ensure that cells adjacent to two stomata or precursor cells do not divide. Thus, TMM may help interpret spatial signals so as to regulate the number, plane and polarity of asymmetric divisions. Further study of TMM may provide access to genes and events that regulate the placement of asymmetric divisions in plants.

TMM is a negative or positive regulator of entry into the stomatal pathway.

In addition to a role in stomatal patterning, TMM is also a key regulator of stomatal production. tmm stems have virtually no stomata (Yang and Sack, 1995; Geisler et al., 1998). Thus tmm has two completely opposite stomatal phenotypes, clusters and the elimination of stomata.

Sometimes these phenotypes are found in separate organs. As mentioned, leaves have clusters and stems lack stomata (Figure 18). Other organs display both phenotypes. The abaxial sepal epidermis has clusters, but no stomata are present in the adaxial epidermis (Figure 19; Geisler et al., 1998). The most striking case is in the flower stalk (pedicel) where clusters are present at the floral end, stomata are absent at the stem end, and there is a gradient in between (Figure 20).

TMM is thus essential for stomatal formation in some contexts, and a negative regulator in others (Figure 21). TMM appears to control how many cells are induced to divide asymmetrically (to function as MMCs). The direction of that regulation depends upon organ and domain. Thus, TMM is a major determinant of stomatal formation as well as patterning. It remains to be seen whether these are separate functions of the same protein, or whether they result from the same key function such as regulation of the frequency and execution of asymmetric divisions.

The diversity of tmm phenotypes suggests that TMM interacts with partners in a molecular complex or pathway. In addition, cues conveying tissue identity and position may signal the direction of the interaction thereby increasing or decreasing the number of MMCs as appropriate to location.