Stomatal pattern and evolution
Given the critical role of stomata in plants' successful exploitation of the terrestrial environment, current-day patterns likely are derived from those of fossil land plants, even though the arrays in fossils may represent imperfect solutions to the problem of gas exchange in today's atmosphere. A comprehensive study of fossil stomatal arrays is not available, but would be valuable to compare patterns of plants grown in different atmospheric conditions, in spite of the shortcomings and bias in the fossil record. The range of fossil patterns may yield clues to the regulation of pattern in extant plants. In fact, the patterns could be compared with recent evidence on the interplay between atmospheric gas concentrations and stomatal frequency (Woodward, 1987 ; Beerling and Chaloner, 1993 ; McElwain and Chaloner, 1995 ; Edwards, Kerp, and Hass, 1998 ).
Cooksonia, known from late Silurian deposits (Edwards, Davies, and Axe, 1992 ), is the most ancient vascular land plant with stomata. Stomata appear on the reproductive axis, but because of axis distortion in the compression, their distribution is not known. In other genera from the same deposit, stomata are present in oblique bands at sporangial bases or scattered across sporangial surfaces (Edwards, Kerp, and Hass, 1998 ). Stomata occur in other ancient plants such as Rhynia, Zoosterophyllum, Sawdonia, and Agalophyton, but no information is available regarding order (Willmer and Fricker, 1996 ).
Comprehensive reviews of stomatal evolution (Ziegler, 1987 ) and structure in extant angiosperms (Baranova, 1992 ) are available, for their structure does differ over time, but such changes are peripheral to the central issue of this paper on stomatal pattern. Structural aspects of stomatal development in extant plants are very plastic, up to five different modes of development are reported to occur on a single organ (Baranova, 1992 ). This plasticity will complicate understanding the mechanism of patterning and the evolution of pattern.
Leaves, as opposed to the naked axes and sporangia discussed above, from fossil lycopods (Baragwanathia abitibiensis, Drepanophycus spinaeformis) reportedly have randomly placed stomata (Taylor and Taylor, 1993 ). Photographs of the lyocopod Lepidophylloides show stomata distributed near shallow grooves that flank the single vascular bundle (Taylor and Taylor, 1993 ), similar to some present-day grasses like barley. In the Sphenophyllales stomata reportedly are random (Sphenophyllum multirame) or randomly placed but with their long axes parallel to the long axis of the leaf (Asterophyllites characformis) (Taylor and Taylor, 1993 ). On the other hand, leaves of fossil conifers (Pseudofrenelopsis, Swillingtonia, Podocarpus,and Sciadopityoides microphylla) and a fossil palm (Sabal dortchii) have stomata in linear series (Taylor and Taylor, 1993 ), much like those of Tradescantia. Similarly, surveys of 21 species of fossil Agathis (Stockey and Atkinson, 1993 ) and four species of fossil Dacrydium (Stockey and Ko, 1990 ) report that stomata occur in discontinuous rows, a common feature of monocot leaves. Although there are few reports on fossil stomatal patterns, the published pattern descriptions are not markedly different from those of extant plants.
Stomatal pattern(s) may have evolved as the plant body elaborated into varied forms over geological time. Therefore, patterns on fossil axes or stems (cylinders) may differ from patterns on fossil leaves (sheets). These pattern differences may reflect changes in organ form as well as changes in the ancient atmospheres. Mechanisms of patterning also may have diverged in response to these changes, and given the time frame involved, distinguishing the mechanism or mechanisms may be complicated.
Stomatal pattern and physiology
Although the adaptive responses of stomata have long been of interest to physiologists, stomatal patterns and their relationship to photosynthesis and transpiration have not received adequate attention (Jarvis and Mansfield, 1981 ; Mansfield, Davies, and Leigh, 1993 ). Since photosynthesis operates in one of three fixation pathways (C3, C4, and CAM), stomatal pattern may vary because each pathway has particular advantages or properties that might affect gas exchange. The physiological differences (photorespiration in C3 vs. C4 plants) and the temporal (CAM) and spatial operation (C4) of these pathways may influence or be associated with particular stomatal arrays. Broad surveys of stomatal pattern and fixation pathway are not available. Some plants, maize for example, are C3 as juveniles, but become C4 plants as adults. CAM plants also revert to C3 photosynthesis when well watered. Such plants provide a special opportunity for understanding the relationship between physiology and stomatal pattern.
In addition to differences in photosynthetic pathway, surface features, light penetration (Martin et al., 1989 ; Cui, Vogelmann, and Smith, 1991 ), and the internal architecture of leaves (Terashima and Inoue, 1984, 1985a, b ; Vogelmann, Bornman, and Josserand, 1989 ; Terashima, 1992 ) may impact on stomatal distribution as well as function. These parameters may explain the heterogeneity of stomatal responses in leaves (Omasa and Onoe, 1984 ; Daley et al., 1989 ; Raschke et al., 1990 ; Beyschlag, Pfanz, and Ryel, 1992 ; Malone et al., 1993 ; Weyers and Lawson, 1997 ; Mott and Buckley, 1998 ).
Numerous studies have surveyed the relationship among environmental gasses (CO2, O3, SO3, O2) and stomatal numbers and function (Mansfield, 1976 ; Omasa et al., 1985 ; Fowden, Mansfield, and Stoddart, 1993 ; Ramonell, Crispi, and Musgrave, 1997 ), but not their effect on stomatal pattern. Stomatal frequency (Tichá, 1982 ) and chlorophyll fluorescence (Schriber, Fink, and Vidaver, 1977 ) are known to differ on upper and lower leaf surfaces and across the leaf surface (Croxdale and Omasa, 1990 ). Is this related to environmental conditions, photosynthetic pathway, or internal leaf geometry? As of yet, there are no answers to these fundamental questions for stomatal pattern on leaves or other organs.
How plastic is stomatal pattern on a given species growing in a given habitat? Do plants within an environment share stomatal characteristics? One study of stomata on mangrove species (Stace, 1966 ) concluded that there was no evolutionary convergence of stomatal types on an interfamilial basis. Instead, stomatal characteristics in a species were more similar to species with the same family than species in the same environment. If a community's species were examined throughout the growing season, would each or only some species exhibit changes in stomatal numbers or pattern? Would these changes correlate to season? Can terrestrial species, especially those located near aquatic communities, adapt to inundation? If so, are the changes similar to those exhibited by amphibious species that grow in both environments? Adaptation of stomatal pattern to habitat and changes in stomatal pattern on a seasonal basis are areas ripe for study.
Stomatal pattern and organ form
Stomatal pattern on leaves and stems is likely related to internal and external organ architecture. Given the functional connection between stomata and photosynthesis, the distribution of photosynthetic cells and the topography of the air space system (Turrell, 1936 ; Sifton, 1945 ) may be linked to stomatal pattern. Studies relating internal geometry to different stomatal patterns would be of great value in understanding the relationship—if there is one—among structure, function, and patterning.
Organ geometry and growth (Payne, 1979 ) may be important as well. Cylindrical stems may have arrays of stomata that differ from flattened stems, but mirror those of cylindrical leaves. The stomatal patterns found on flattened stems may be similar to those of bifacial leaves while the unifacial tip of monocot leaves may exhibit a pattern analogous to cylindrical stems. Patterns on species with decurrent leaf bases in which the stem and leaf base extend together would be of special interest. A survey comparing stomatal patterns on different organ forms would provide a measure of pattern plasticity in relation to organ geometry.
Although the molecular mechanisms of patterning in plants and animals are probably similar, the closed growth of animals vs. the open growth of plants has repercussions on understanding and tackling patterning problems. In animals, critical patterning decisions occur during embryogenesis, whereas in plants the timing of these decisions spreads over the lifetime of the organism and the location of these decisions varies by organ. In animals, patterning commonly occurs by three different physical modes: a soluble signal that diffuses to a competent cell, cell matrix contact to induce specification of an adjacent cell, and direct cell-cell contact. Similar modes may exist in plants, but the presence of cell walls, the absence of cell movement, and the status of symplasmic domains may complicate understanding them. In considering molecular mechanisms of patterning, we need to realize the infancy of the field. However, in Drosophila embryos, the presence of Notch and delta, both transmembrane proteins, are required to pattern the epidermis into hypodermal cells (skin) and neuroblasts. When either are missing, all surface cells become neuroblasts. This information is interesting for several reasons—all surface cells can assume the same fate, but Notch and delta are required for some of them to become different. Trichome patterning may work in a similar fashion with two or more genes. Most interesting is the observation that one of the genes (GLABROUS) is expressed throughout the epidermis and later is restricted to trichome precursors (Marks, 1997 ). What turns this gene off in most epidermal cells may be the product of another gene in the pathway. At this writing, equivalent molecular information is not yet available for stomatal patterning.
Stomatal patterning in angiosperms
Angiosperm leaves exhibit two fundamentally different modes of growth and development. Monocot leaves have polarized growth from a single point source of cells at or near the leaf base, creating a leaf blade with the oldest cells at the tip. The epidermis consists of regular longitudinal files of cells, whose cells differentiate basipetally, providing a continuum of stages along the blade length. Stages in the origin of stomatal precursors and their development into mature stomata are readily followed. Dicot leaves on the other hand grow from multiple point sources in a patchwork quilt fashion, with clones of new cells forming throughout growth and development. At maturity, the epidermis consists of irregularly shaped cells interspersed with stomata. Evidence for this type of growth is based on marked mesophyll cells (Poethig and Sussex, 1985 ), and it seems likely that epidermal cell growth is also clonal in order to keep pace with divisions in the underlying cell layers. Superimposed on the pattern of new cell origin in dicot leaves is the basipetal pattern of cell expansion (Wolf, Silk, and Plant, 1986 ). This addition of new cells to the leaf blade and their subsequent expansion complicates understanding stomatal patterning.
Bünning (1956) observed that new stomata developed between mature complexes in dicotyledonous leaves and hypothesized that mature stomata released an inhibitor to prevent new stomata from arising nearby (Fig. 3A). Only when new growth exceeded the inhibitory influence would new stomata arise. Beyond Bünning's original observations, no measurements have established the size of the inhibitory field in any species and no inhibitor has been isolated.
Support for Bünning's inhibition theory comes from studies using a wide variety of species (Pelargonium zonale and Sedum stahlii—Korn, 1972 ; Anagallis arvensis—Marx and Sachs, 1977 ; Aeonium—Sachs and Benouaiche, 1978 ; Vinca major—Sachs, 1979 ; Pilea cadierei—Smith and Watt, 1986 ; Pisum sativum—Kagan, Novoplanksy, and Sachs, 1992 ). However, the presence of aborted stomata in these species indicates that Bünning's theory is imperfect. Inhibition failed to prevent the origin of stomata that ultimately arrest (Anagallis arvensis—Marx and Sachs, 1977 ; Aeonium—Sachs and Benouaiche, 1978 ; Vinca major—Sachs, 1979 ), and their existence is cause to re-examine the foundation of Bünning's theory.
More recently, Sachs (1994) , Larkin et al. (1996) , and M. Geisler and F. Sack (Ohio State University, personal communication) have indicated that stomatal distribution results from an interactive process that occurs during leaf growth and might be attributed to cellular interactions. While it seems likely that some form of cellular communication is involved in stomatal patterning, the evidence to date is limited to cell division patterns that serve to separate stomata from one another. Analysis of tomato chimeras and intergeneric chimeras from members of the nightshade family (Solanaceae) shows that the genotype of the epidermal layer regulates the frequency of specialized cells, trichomes and stomata (Szymkowiak and Sussex, 1996 ; E. Szymkowiak, University of Iowa, personal communication). However, in maize the Knotted genotype in the mesophyll induces divisions in the epidermis that are characteristic of Knotted rather than wild type (Hake and Freeling, 1986 ). These disparate results indicate that cell fate is not dependent solely on cell lineage, but results from the integration of signals within and between cell layers, as is known to occur during floral organ determination (Bouhidel and Irish, 1996 ).
Indeed, many events in dicot leaf development (early trichome initiation, scattered islands of growth, clonal development of photosynthesis, and delayed photosynthetic competence in late arising clones) mean that final stomatal pattern is derived by new stomata being inserted into an existing network. Given this pattern of leaf growth, new stomata may arise according to the development of individual clones and not by exceeding the sphere of influence from existing stomata. Although existing stomata may appear to inhibit new stomata, new stomata may be arising in association with the formation of new epidermal cells (within the clone of new cells), independent of influence from existing cells. Stomatal patterning could be clonal autonomous. Anomalous clusters of Arabidopsis stomata, like those in the mutants tmm and flp (Yang and Sack, 1995 ) or those that occur under certain growth conditions (Serna and Fenoll, 1997 ), are reportedly clonal. However, study of another specialized epidermal cell type, trichomes, concluded that cell lineage or clonal growth was not responsible for trichome distribution (Larkin et al., 1996 ). Considering that dicot leaves exhibit clonal growth scattered across the leaf, the possibility of flawed logic in Bünning's inhibitor theory is obvious.
If we assume that existing stomata regulate the placement of new stomata by cellular interactions, then what is the route of communication? Plasmodesmata between intervening cells might serve as a structural link for easy passage of signals. If a structural bridge between cells is not present, then signals have to traverse cell membranes and cell walls to affect development. In oat, developing stomata are in symplastic continuity, but at maturity are symplastically isolated from their epidermal neighbors (Palevitz and Hepler, 1985 ). The fundamental relationship among leaf growth, symplastic domains, and stomatal commitment in dicotyledonous species must be studied to understand the level at which the regulation of stomatal placement might take place.
Clonal and polar growth in dicot species
The earliest study reflecting the impact of clonal growth and stomatal patterning may be that of Leick (1955) , who reported that in selected dicots (Berberis aquifolium, Sedum confusum, Silene nutans, and Choisya ternata) that stomatal number increases in waves, while in other dicot species (Globularia willkommii, Emex spinosa, Polygonum convolvulus, Penstemon barbatus, Viola, and Jovis), a large number of stomata are present when the leaf is young and the stomatal index or frequency declines as the leaves expand. The latter leaves and at least some members of the Caryophyllaceae (Pappas et al., 1988 ; Fahn, 1990 ) grow in a polarized manner similar to those of monocots. In Arabidopsis leaves stomatal numbers also increase in non-steady-state kinetics, similar to waves, at predictable times (hours after imbibition) or at given numbers of leaf cells (M. Geisler and F. Sack, Ohio State University, personal communication). The non-steady-state kinetics of stomatal initiation likely reflects initiation of new stomata during simultaneous growth of clones.
Polarized growth makes monocot leaves well suited for stomatal patterning studies because the developmental history of the leaf can be traced down its length. New blade cells and stomatal precursors originate in polarized fashion at the leaf base and are present in a continuum of stages from the leaf tip (mature) to the leaf base (immature). Unlike the morning bus example cited previously, stomatal origin in monocotyledons is similar to passenger loading in airplanes, from the rear to the front of the aircraft. Stomatal pattern propagates as a wave down the leaf as the blade grows.
Young monocot leaves have no stomata, although all three axes of symmetry are in place. The proximal-distal axis is established when the blade can be distinguished from the sheath, readily apparent in species with ligules (Sylvester, Cande, and Freeling, 1990 ). The central-lateral axis is apparent internally by the midvein and lateral bundles, and in some species by crystal-containing cell files adjacent to the bundles (Croxdale et al., 1992 ; Croxdale, 1998 ). Externally, the epidermis is organized into longitudinal tracts, each tract bearing similar cells and sometimes papillae (Rasmussen, 1986 ). The dorsal-ventral axis may be designated by a ligule or differential growth of the vascular tissue or mesophyll cells. Divisions responsible for generating stomata occur late in leaf development (Croxdale et al., 1992 ) and, as shown by the corn mutant tangled 1, are under separate regulatory control from generic epidermal divisions (Smith, Hake, and Sylvester, 1996 ).
Stomata in the monocot barley appear in visibly ordered arrays adjacent to cell files near vascular bundles. The predictable placement of stomata led Bünning to suggest that cell divisions in already ordered cells established stomatal distribution. Bünning's theory, which also is called the cell lineage theory, does not address how cells are designated into the stomatal pathway, only with the predictable sequence of divisions that generate stomata from existing cells (Fig. 3B). He is suggesting that stomata are prepatterned, an interpretation disputed by Sachs (1994) , who contends that stomata are not prepatterned in either monocotyledons or dicotyledons because stomatal arrest occurs in each group.
Although results from new studies (Sachs, 1974 ; Rasmussen, 1986 ; Croxdale et al., 1992 ) support the contention that stomata arise by divisions in blocks of ordered cells, Bünning's theory fails to explain several aspects of stomatal distribution. First, not all cells in a block undergo division to yield stomata. Selected cell files have no stomata, while neighboring files are enriched with stomata. Second, sometimes stomata are grouped in linear series, separated from one another by single epidermal cells, and other times stomata are separated by multiple epidermal cells. Lastly, the theory does not explain arrested, immature, or aborted stomata, which are common on these leaves.
Sachs uses the presence of arrested stomata as evidence that stomatal patterning takes place during development, in leaves of Sansevieria and phylloclades of Ruscus hypoglossum for example (Kagan and Sachs, 1991 ; Sachs, Novoplansky, and Kagan, 1993 ). The correctness of his interpretation depends on one's definition of stomatal patterning. If one views stomatal patterning as all events that contribute to the final pattern of stomata, his interpretation is accurate. If one views stomatal patterning as being concerned only with the specification of cells into this specialized pathway, then it might be said that arrest of stomata is a differentiation failure, not an aspect of patterning. Since stomatal arrest yields a more ordered pattern than would have been present had all stomata matured, patterning has been termed Darwinian (survival of the fittest) (Edelman, 1987 ; Held, 1992 ) or epigenetic (Sachs, 1988 ; Kagan and Sachs, 1991 ). However, stomatal development is not complete when they arrest, so fitness is not related to functional ability. Therefore, selection does not occur based on functional fitness, but on position, which is reflected in increasing stomatal order.
Developmental arrest might result from inhibition or from the scarcity of factors essential for continued development. Such factors might be available in differentiating leaf regions and account for epigenetic selection. The stoma that begins to develop first may outcompete those that arise slightly later. The latter is a simpler and, in this author's view, more parsimonious solution to regulating stomatal development because only molecules necessary for stomatal differentiation are required (Boetsch, Chin, and Croxdale, 1995 ). If inhibition regulated stomatal development, both an inhibitor as well as a receptor for the inhibitor would be necessary. The synthesis of these components, and their appropriate temporal expression, would also need to be regulated. Inhibition may be a convenient way of describing stomata that cease to develop, but use of this term tends to restrict one's thinking so that other equally or more likely possibilities are not considered. Stomatal initials that arrest in Tradescantia do not remain undifferentiated; they enter the epidermal cell pathway as evidenced by their expansion and accumulation of secondary metabolites (Boetsch, Chin, and Croxdale, 1995 ).
Korn (1993) uses modeling to understand the growth dynamics of different epidermal cell types, the timing of stomatal origin, and the rate of stomatal induction necessary to generate the spatial pattern present on mature leaves. He established models to account for mature stomatal distribution in Chlorophytum cosmosum, Iris pumila, and Allium cepa. These studies predict spatial aspects of stomatal arrangement, but not how particular cells are targeted to the stomatal pathway. Patterning by inhibition is supported in the dicot Pelargonium by either a mathematical or geometric model (Korn, 1993 ), although the kinetics of pattern was not considered (Sachs, 1994) . Korn's growth models contribute valuable information to our understanding of the epidermal layer. However, my goal is to understand the mechanism of patterning, so my focus must be trained on cells and how they become different from their neighbors.
Charlton (1990) suggested the cell cycle regulated stomatal patterning based on two pieces of evidence: the position of stomatal initials at a common boundary in the leaf base and the tendency of stomata and of epidermal cells to occur in linear groups (Fig. 3C). He postulated that a zone existed at the base of the leaf where cells are targeted for the stomatal pathway and that the fate of a cell entering this area is dependent on the cell's position in the division cycle. Those cells closest to mitosis are more likely to divide forming a stomatal initial than cells distant from mitosis. Since sister cells maintain a high degree of cell cycle synchrony for several cycles (Webster, 1979 ), sister cells and their derivatives would be close to mitosis upon entering this zone and would be patterned to the same fate (Charlton, 1990 ). Thus, a long string of cells patterned to a common fate will be of common ancestry.
In Tradescantia, support for the involvement of the cell cycle was found in precursors of stomatal strings, groups of stomata that occur in linear series (Chin et al., 1995 ). Strings of stomatal precursors can be identified by their position near the leaf base and proximal to the stomatal initial region and by their replicated DNA (Chin et al., 1995 ). Cells in actively dividing regions of Tradescantia rest in G1 (Webster, 1979 ; Kudirka and Van't Hof, 1980 ), while cells with replicated DNA go on to divide. Using DAPI, we identified stomatal strings based on intensity of nuclear staining and location, in support of Charlton's original suggestion that cell cycle position is a critical aspect of stomatal patterning and could result in a string of cells being targeted to the same pathway. Their common position in the cell cycle indicates cells are at a similar point in the division cycle when they are specified to the stomatal pathway and stomatal groups develop in synchrony. Tradescantia leaves, as do maize leaves, bear epidermal cells in linear series (Croxdale et al., 1992 ). The distribution of string lengths is the same in the mature leaf as at the leaf base indicating our correct identification of stomatal precursor strings and the absence of subsequent divisions altering original string length (Chin et al., 1995 ).
A unifying theory of stomatal patterning in angiosperms
Charlton's theory of monocot stomatal patterning explains that cells are selected to become stomata based on their position in the cell division cycle as they are displaced through the patterning region. Now there is evidence of the cycle's importance in Tradescantia stomatal development. The involvement of the cell cycle provides a means of explaining the species-specific variation in stomatal patterning because each array would reflect that species' cell cycle dynamics and the longevity of the patterning signal. The scattered intercalary growth of dicot leaves complicates their use in stomatal patterning studies, but a recent observation is consistent with the possibility that cell cycle position may also be critical in this group (Fig. 4). Arabidopsis cotyledons generate stomata in non-steady-state kinetics (M. Geisler and F. Sack, Ohio State University, personal communication), and it has long been known that dicot meristemoids divide a variable number of times before producing a guard mother cell (GMC). The position in the cell cycle when the patterning signal is present may dictate whether a GMC is produced. This would account for the waves of GMCs throughout the leaf and link their patterning to the monocotyledon type where cell cycle position dictates whether a cell responds when a stomatal-inducing signal is present.
Recent reports highlight unusual features of the cell cycle in stomatal patterning of dicotyledonous leaves. In Arabidopsis,
trichomes and epidermal cells enter the endomitotic cycle and acquire DNA levels up to 64C while stomata remain diploid and remain in the mitotic cycle (Melarango, Mehorotra, and Coleman, 1993
). Even at maturity Arabidopsis
guard cells express genes indicating competency for cell division (Serna and Fenoll, 1997
), an unusual feature of a terminally differentiated cell. Remaining within the cell division cycle may explain why sugar beets were easily transformed from guard cell protoplasts (Hall, Riksen-Bruinsma, and Weyens, 1996
) and why isolated stomata are able to assume several different fates and express different gene products when cultured (Taylor et al., 1998
). While one might argue that the relationship between stomata and the cell cycle is merely circumstantial, the involvement of the cycle in stomatal patterning is a testable hypothesis.
If the cell division cycle is a component of dicotyledon stomatal patterning, then the expression of cycle genes will correlate with GMC formation. Since GMC formation takes place episodically in the Arabidopsis leaf, this expression will occur episodically throughout the leaf epidermis. GMCs also should undergo equal divisions to form guard cells across the leaf in near synchrony. The nature of the molecule that signals or induces particular cells to enter the stomatal pathway is not yet known.
Our understanding of the plant cell division cycle, its regulation, and interaction of gene products is expanding, following advances made primarily in yeast. Although there appear to be differences in cycle regulation by kingdom (Fobert et al., 1996 ; Zhang, Letham, and John, 1996 ), many features of cycle regulation are common among plants, algae, yeast, and animals (John, Sek, and Lee, 1989 ; John et al., 1990 ). Indicators and localization of specific cell-division-cycle products have been detected in plant protoplasts (Ferreira et al., 1994 ), tissue cultures (Savoure et al., 1994 ), and intercalary (Sauter and Kende, 1992 ; Lorbiecke and Sauter, 1997 ) and apical meristems (Fobert et al., 1994 ). Transcriptional regulation of numerous gene products that function in independent regulatory pathways has been associated with the cell cycle (Francis and Halfor, 1995 ; Magyar et al., 1997 ) and cycle products appear in systems developing naturally (Ferreira et al., 1994 ) or when stimulated by specific factors (Savoure et al., 1994 ) or plant hormones (Sauter and Kende, 1992 ; Zhang, Letham, and John, 1996 ; Lorbiecke and Sauter, 1997 ).
Since the basic patterns of cell division and stomatal formation are known in angiosperm leaves, localization of cycle-specific gene products would reveal the relationships between cell division and cell fate. Because stomata are highly specialized cell types that are critical to plant survival and show a unique association to the cell division cycle, they make good candidates for study.