- © 1999 American Society of Plant Physiologists
Abstract
Trichomes in Arabidopsis are single-celled hairs that exhibit a regular spacing pattern. Here, the role of TRIPTYCHON (TRY) in the generation of this spacing pattern is studied. By using genetic mosaics, we demonstrate that the formation of trichome clusters in try mutants is not correlated with cell lineage, indicating that TRY is required to single out trichome cells in a process involving cellular interactions. The genetic interactions of TRY, GLABRA1 (GL1), and TRANSPARENT TESTA GLABRA (TTG) in trichome patterning are assessed by determining the cluster frequency in various genetic combinations. It is shown that TRY acts as a negative regulator of GL1- and TTG-dependent pathways. Furthermore, it is demonstrated that trichome initiation in ttg-1, a strong ttg allele, is rescued almost to wild-type levels in a try background in which GL1 is expressed under the control of the cauliflower mosaic virus 35S promoter, indicating that TTG acts upstream of GL1 and TRY. These findings are incorporated into a model to explain the generation of a trichome spacing pattern from a homogeneous population of epidermal cells.
INTRODUCTION
During the development of higher organisms, the spatial arrangement of differentiated cells is generated by a process called pattern formation. In most of the cases described to date, pattern formation integrates spatial information that is already present. In some cases, a pattern is generated de novo, meaning that initially all cells have the same competence and that during development, individual cells are singled out to acquire a different cell fate. De novo pattern formation could involve a stereotyped cell division program that has the intrinsic property of segregating different cell fates (cell lineage scenario). Alternatively, a pattern could arise from equivalent cells competing with each other by intercellular interactions to produce asymmetries that eventually result in the selection of individual cells (cell–cell interaction scenario).
Pattern formation in the leaf epidermis of the flowering plant Arabidopsis is a model system suitable for studying de novo patterning (Sylvester et al., 1996; Larkin et al., 1997; Marks, 1997; Hülskamp and Schnittger, 1998; Hülskamp et al., 1999). Two cell types, stomata and trichomes, show a characteristic spacing pattern: they are regularly distributed on the epidermal surface with a specific distance between the respective cell types. In both cases, pattern formation takes place in a two-dimensional field of cells. Stomatal patterning is mainly established by a series of highly oriented asymmetric cell divisions placing the two guard cells in the center of the future stomatal complex (Yang and Sack, 1995; Serna and Fenoll, 1997). As a result, guard cells are always surrounded by epidermal pavement cells and, thus, are separated from other stomatal cells. Although several observations suggest that lateral inhibition of stomatal cell fate can occur, the observed stereotyped division pattern is in principle sufficient to account for the observed spacing pattern (reviewed in Larkin et al., 1997).
Trichomes are large single cells that are surrounded by approximately eight accessory cells of characteristic morphology but of unknown function (Larkin et al., 1997; Marks, 1997; Hülskamp and Schnittger, 1998; Hülskamp et al., 1999). In contrast to stomatal patterning, the trichome–accessory cell complex does not appear to be derived from a stereotyped division pattern. This was shown by Larkin and co-workers, who demonstrated that early-induced β-glucuronidase sectors may divide the trichome–accessory cell complex, indicating that trichomes and accessory cells are not closely related (Larkin et al., 1996). This finding suggests that cells in immediate contact with a trichome cell become recruited as accessory cells after the trichome pattern has been established. This rules out a cell lineage scenario whereby a stereotypic series of cell divisions forms a complex in which the trichome cell becomes completely surrounded by its precursor cells. However, other cell lineage scenarios or trichome patterning based on cell–cell interactions are still consistent with these findings.
To date, four genes have been identified as playing a role in trichome initiation and/or trichome patterning. These include REDUCED TRICHOME NUMBER (RTN), GLABRA1 (GL1), TRANSPARENT TESTA GLABRA (TTG), and TRIPTYCHON (TRY) (Koornneef et al., 1982; Hülskamp et al., 1994; Larkin et al., 1994, 1996). Whereas RTN appears to control the “developmental time window” during which trichome initiation occurs, GL1, TTG, and TRY are thought to be involved in the generation of the spacing pattern. All three share, either alone or in certain genetic combinations, a characteristic patterning defect: the formation of trichome clusters rather than individual trichomes. Mutations in the TRY gene result in the formation of trichome clusters containing up to four trichomes, suggesting that TRY functions to inhibit trichome cell fate in neighboring cells (Hülskamp et al., 1994). In addition, trichomes in try mutants show an increased number of endoreduplication cycles (Hülskamp et al., 1994).
By contrast, strong gl1 and ttg alleles lack virtually all trichomes, suggesting that GL1 and TTG act as positive regulators of trichome development. Genetic evidence suggests that both genes are involved early in trichome patterning. Several weak ttg alleles are known to produce trichomes that are often found in clusters (Larkin et al., 1994). GL1 appears to function differently because none of the weak gl1 alleles described to date exhibits trichome clusters (Esch et al., 1994; this study). However, a role of GL1 in trichome patterning was suggested by Larkin et al. (1994) when they analyzed GL1-overexpressing plants that were heterozygous for a strong ttg allele. Whereas plants overexpressing GL1 did not exhibit trichome clusters, the additional reduction of TTG dosage resulted in a large number of clustered trichomes, suggesting that the relative dosage of both genes is important during pattern formation (Larkin et al., 1994).
GL1 encodes a protein with sequence similarity to myb-related transcription factors (Oppenheimer et al., 1991). GL1 expression is initially found uniformly in all epidermal cells at the base of young leaves and accumulates in developing trichome cells to high levels while transcription ceases in surrounding epidermal cells (Larkin et al., 1993). The analysis of genetic mosaic patterns suggests that GL1 acts locally because gl1 mutant sectors are not rescued by neighboring wild-type tissues (Hülskamp et al., 1994). Most of the molecular analyses directed toward the function of the TTG gene have focused on the R gene, because the overexpression of the heterologous R gene from maize can rescue all aspects of the phenotype of ttg mutants (Lloyd et al., 1992). The product of the maize R gene shows sequence similarity to myc-related transcription factors and, like TTG in Arabidopsis, is required for the regulation of anthocyanin biosynthesis (Ludwig et al., 1989). The recent cloning of the TTG gene has shown that TTG does not encode an R gene homolog (Walker et al., 1999), suggesting that the R gene acts downstream of TTG if it has a functional homolog in Arabidopsis. The TRY gene has not been cloned.
In this study, we used genetic mosaics to show that trichome clusters in try mutants are not derived from clonally related cells, suggesting that TRY functions to single out trichome cells in a process that does not involve cell lineage. To determine the role of TRY, GL1, and TTG in trichome patterning, we studied the genetic interactions among the three genes in various double and triple mutant combinations. Our results suggest a genetic model for de novo trichome patterning in Arabidopsis.
RESULTS
Formation of Trichome Clusters in try Mutants Does Not Involve Cell Lineage
The phenotype of try mutants is characterized by the development of up to four trichomes in the place of a single trichome. This phenotype is in keeping with TRY being involved in either of two patterning mechanisms: cell lineage or cell–cell interactions. In a cell lineage scenario, TRY would be involved in the control of cell fate decisions during cell divisions. Hence, clustered trichomes would have the same clonal origin, because they were derived from one common ancestor cell. In a cell–cell interaction scenario, TRY would function to prevent neighboring cells from adopting the fate of trichome cells. In this scenario, clustered trichomes would develop from cells that may or may not be clonally related.
To determine the clonal relationship of trichomes in try clusters, we induced somatic stichel (sti) clones by mutagenizing the F1 seed of sti/+ try/try plants with ethyl methanesulfonate. If a new mutation is induced in the wild-type allele, the corresponding cell is a homozygous mutant for sti; if this cell contributes to mature vegetative epidermal tissues, it will give rise to a mutant sti sector. Epidermal sti sectors can easily be recognized as fields of unbranched trichomes on mature leaves (Figure 1A) (Schnittger et al., 1996). Borders of mutant sti sectors were inspected for try clusters through which a sector border passed (Figures 1B and 1C). Of 173 sti sectors, we identified three try clusters, each consisting of one unbranched trichome and two branched trichomes (Figure 1B). This finding argues against the cell lineage scenario and suggests that TRY is involved in the regulation of cell fate decisions mediated by intercellular interactions.
Criteria Used to Study the Role of TRY, TTG, and GL1 in Trichome Patterning
Patterning mechanisms not involving cell lineage either use spatial information already provided by other cells or generate positional information de novo. Because trichome patterning does not appear to be correlated with the spatial distribution of other epidermal (e.g., stomata) or subepidermal cell types (e.g., vascular cells), it is likely that trichomes are initiated solely with reference to each other.
Sector Analysis of Cluster Formation in try Mutants.
(A) Overview of a mature rosette leaf with a sti sector in a try mutant background. Red lines indicate the boundary of the sti sector.
(B) try cluster consisting of two branched and one unbranched sti mutant trichomes.
(C) Schematic representation of a sti sector in a try mutant background. The sti sector encompasses part of the leaf with the sector border passing through a try cluster (circled).
To study the role of TRY, TTG, and GL1 in trichome patterning, two different criteria can be used to evaluate mutant phenotypes: (1) the frequency of clustered trichomes and (2) the distance between trichomes. The latter criterion, however, is ambiguous because changes in the distance between trichomes in mature leaves do not necessarily reflect alterations in the initial trichome patterning in different mutants. For example, mutants affecting the division rate of leaf epidermal cells would alter trichome density after trichome initiation has already taken place. In this study, we examined the trichome cluster frequency under the assumption that it reflects quantitative changes in the interactions of the respective genes. The cluster frequency is the percentage of trichome clusters per total number of trichome initiation sites. Trichome initiation sites are leaf positions with one trichome or a cluster of trichomes. To exclude secondary patterning events (for example, the transformation of accessory cells into trichomes, see below), we restricted our analysis to trichomes visible in the “trichome initiation zone” at the base of developing leaves. This region contains young trichomes that have not initiated secondary branching (Figure 2).
TRY Acts as a Negative Regulator of Overexpressed GL1 and R Gene Activity
The GL1 and R genes can be considered positive regulators, whereas TRY appears to act as a negative regulator of trichome development. To test whether TRY acts as a negative regulator of GL1 and/or R gene function, we studied two try mutant lines overexpressing GL1 or the R gene under the control of a strong constitutive cauliflower mosaic virus 35S promoter.
In 35S::GL1 plants, virtually no trichome clusters were observed to be similar, as found in wild-type plants (Larkin et al., 1994) (Figure 3A and Table 1). By contrast, try 35S::GL1 plants showed an enhanced phenotype (Figure 3B): the cluster frequency on try 35S::GL1 leaves (13.9%) was found to be increased approximately twofold compared with the frequency on try mutants (8.3%) not overexpressing GL1 (Table 1). This observation suggests that TRY limits the activity of overexpressed GL1.
For the analysis of TRY function in plants overexpressing the R gene, we used 35S::R-GR, an inducible form of the R gene with its N terminus fused to the vertebrate glucocorticoid receptor (Lloyd et al., 1994). Whereas try 35S::R mutants were difficult to grow, the inducible try 35S::R-GR line could easily be maintained under uninduced conditions. In our study, 35S::R-GR plants showed residual R gene activity under uninduced conditions, because ttg mutants are partially rescued by uninduced 35S::R-GR (data not shown). 35S::R-GR plants have a cluster frequency of 0.4% (uninduced) and 7.3% (induced), respectively (Table 1). Compared with try or 35S::R-GR, uninduced try 35S::R-GR plants showed a significant increase in cluster frequency (31.5%), probably due to the residual activity of R-GR (cf. Figures 3C and 3D; Table 1). Similar to try 35S::R plants containing the uninducible R gene, induced try 35S::R-GR plants exhibited a strong increase in trichome density (Figures 3E and 3F). The cluster frequency and cluster size were determined in cleared whole-mount preparations of try 35S::R-GR leaves (Figure 3E). In this line, 83% of all trichome initiation sites contained two or more trichomes (Table 1). In addition, in induced try 35S::R-GR plants, the average number of trichomes per trichome initiation site was greatly increased. Induced try 35S::R-GR plants produced a significant number of larger clusters, with some having up to 40 trichomes (Table 2).
Wild-Type Trichome Initiation on a Young Leaf.
Scanning electron microscopy of a young leaf shows trichomes at different developmental stages. Trichome initiation is restricted to the trichome initiation zone, the border of which is marked by a white line. Trichomes become further separated due to cell divisions of intervening epidermal cells in more distal parts of the leaf (Hülskamp et al., 1999).
Cluster Formation in Different Genetic Combinations of 35S::GL1, 35S::R-GR, and try Mutants.
Scanning electron microscopy ([A] to [D] and [F] to [J]) and light microscopy (E) show the different mutant phenotypes.
(A) Overview of a 35S::GL1 leaf.
(B) Overview of a try 35S::GL1 leaf.
(C) Overview of an uninduced 35S::R-GR leaf.
(D) Overview of an uninduced try 35S::R-GR leaf.
(E) Young induced try 35S::R-GR leaf.
(F) Young try 35S::R leaf.
(G) Uninduced 35S::R leaf.
(H) Uninduced 35S::GL1 35S::R-GR leaf.
(I) 35S::GL1 35S::R leaf.
(J) High-magnification view of a 35S::GL1 35S::R leaf. Note that trichomes are still separated by intervening epidermal cells.
Overexpression of the GL1 and R Genes in the Absence of TRY: Saturation of the System?
If GL1, TTG, and TRY are sufficient for trichome patterning, then overexpression of the GL1 and R genes in the absence of TRY should result in the transformation of all epidermal cells into trichome cells.
Rosette leaves of Arabidopsis plants overexpressing GL1 and R genes appear densely covered with branched trichomes; however, the leaves do not expand, making it difficult to assess the actual trichome pattern (Figure 3I) (Larkin et al., 1994). Mature 35S::GL1 35S::R leaves exhibit an increased number of trichomes when compared with the single lines (cf. Figures 3A, 3G, and 3I). Trichomes at the leaf margin were usually clustered, whereas trichomes on the leaf blade still showed a spacing pattern, with individual trichomes or trichome clusters separated by several epidermal cells (Figure 3J). Thus, overexpression of the GL1 and R genes in the presence of TRY is not sufficient to trigger trichome fate in all epidermal cells.
Cluster Frequency in try, gl1, ttg, 35S::GL1, 35S::R-GR, and the Respective Double Mutant Combinations
To test whether TRY is required for trichome patterning in 35S::GL1 35S::R plants, we constructed a line overexpressing the GL1 and R genes in a try mutant background. Because 35S::GL1 35S::R plants stop growing after the first two rosette leaves are formed, we used the inducible 35S::R-GR line for these experiments. Compared with uninduced try 35S::R-GR (Figure 3D) or 35S::GL1 35S::R-GR (Figure 3H) lines, uninduced try 35S::GL1 35S::R-GR plants exhibited a marked reduction in growth and produced fewer rosette leaves that never fully expanded and a reduced number of flowers (Figure 4A). In addition, uninduced try 35S::GL1 35S::R-GR plants exhibited a striking trichome phenotype. Similar to those in 35S::GL1 35S::R plants, ectopic trichomes formed on various plant organs, including the abaxial surfaces of the first rosette leaves, the stamen, the pistil, and even between papillar cells (Figures 4A and 4B). In addition, trichome density was greatly increased on all plant organs. Thus, uninduced try 35S::GL1 35S::R-GR plants show a phenotype similar to plants overexpressing the GL1 and R genes at high levels. These findings indicate that at least part of the trichome patterning found in 35S::GL1 35S::R plants is mediated by TRY. However, trichome patterning is not completely abolished in these plants. In high magnifications of rosette leaves, most trichomes were grouped in large clusters, and a significant number of trichomes were still separated by intervening epidermal cells (Figure 4C).
To test whether a further increase of R gene activity would trigger the complete transformation of all epidermal cells into trichome cells, try 35S::GL1 35S::R-GR seeds were germinated on agar plates containing 10 μM dexamethasone. All seeds germinated but then stopped growing after the initiation of the first rosette leaves. The incipient rosette leaves of these seedlings showed trichomes at various stages. These trichomes were separated by intervening epidermal cells (Figure 4D). Thus, even induced try 35S::GL1 35S::R-GR triple mutants do not exhibit a complete transformation of all epidermal cells into trichome cells. This result suggests the existence of at least one additional gene with a partially redundant role in the inhibitory pathway of trichome formation.
Genetic Interactions between TRY and TTG
TTG function is required for two aspects of trichome patterning. TTG acts positively on trichome initiation, as indicated by the lack of trichomes in plants carrying strong ttg alleles. In addition, TTG plays a role in the lateral inhibition of neighboring cells, similar to the role postulated for the TRY gene. The latter role is suggested by trichome cluster formation in plants carrying weak ttg alleles. To test whether TRY and TTG act in the same pathway during lateral inhibition, we studied try/+ and ttg/+ double heterozygous plants. If TRY and TTG act in the same pathway, one might expect that a reduction in the activity of both genes would be sufficient to cause a phenotypic effect. A similar strategy has been used to study the genetic interactions between CLAVATA1 and CLAVATA3 (Clark et al., 1995). Several double heterozygous combinations of try with strong or weak ttg alleles were analyzed (Table 1). All double heterozygous plants revealed a significant increase in cluster formation, suggesting that both genes act in the same pathway. Additional support for this idea comes from the finding that both the try ttg-w and the try ttg-10 double mutants did not show an additional increase in cluster frequency or maximal cluster size compared with ttg-w and ttg-10, respectively.
Trichome Cluster Size in try 35S::R-GR
We performed a similar analysis using genetic combinations of try and different gl1 alleles. For this analysis, we used one new strong gl1 allele, gl1-EM1, and two newly identified weak gl1 alleles, gl1-EM2 and gl1-EM3. The two weak gl1 alleles are characterized by a reduced number of trichomes in the center of the leaf and do not show any trichome clusters. All double heterozygous combinations of try and all gl1 alleles tested showed no increase in cluster formation (Table 1).
Genetic Interactions between GL1 and TTG
The very similar trichome phenotype of gl1 and ttg mutants suggests that both genes have a very similar function during trichome formation. To test the genetic interaction between GL1 and TTG, we analyzed different double heterozygous combinations of weak and strong gl1 and ttg alleles. No increase in cluster frequency was found in double heterozygous combinations of strong alleles (Table 1). However, double heterozygotes of weak or strong ttg alleles with one weak gl1 allele, gl1-EM2, produced significantly more trichome clusters than did plants heterozygous for either single allele (Table 1). The formation of clustered trichomes was highly specific for the respective allelic combinations (Table 1). Double heterozygous combinations of the weak gl1-EM2 allele showed clustered trichomes together with the strong ttg-1 allele and the weak ttg-10 allele but not with the weak ttg-w allele, whereas all combinations with the weak gl1-EM3 allele showed no increase in cluster formation. These observations could reflect that trichome patterning is sensitive to changes in the relative activity of GL1 and TTG. Alternatively, the results could be interpreted as allele-specific effects, which may point to a direct molecular interaction of GL1 with TTG.
Overexpression of GL1 in the Absence of TRY Rescues the Strong Trichome Phenotype of the ttg-1 Mutant
It has been reported previously that the constitutive expression of GL1 cannot bypass strong ttg mutations, indicating that GL1 and TTG act at the same point in trichome development (Larkin et al., 1994). However, it is also conceivable that in this rescue experiment the function of GL1 is limited by TRY. To test this assumption, we constructed the try 35S::GL1 ttg-1 triple mutant, with ttg-1 being one of the strongest ttg alleles available. The triple mutant showed a rescue of the trichome phenotype of the ttg-1 mutant with differentiated branched trichomes near the leaf margin and unexpanded gl2-like trichomes in the leaf blade (Figures 5C to 5F). The total number of trichomes was restored to that observed in try 35S::GL1 mutants, although it was significantly reduced when compared with the wild type (Table 3). Also, the cluster frequency in the triple mutant was similar to that observed in try 35S::GL1 plants (Table 1). A partial rescue of the trichome phenotype of the ttg-1 mutant was also found in 35S::GL1 ttg-1 double mutant plants (Table 3). Whereas ttg-1 mutants produced no trichomes on rosette leaves, the ttg-1 35S::GL1 double mutant exhibited on average five trichomes per leaf. Most of these were not expanded (Figures 5A and 5B).
Phenotype of the try 35S::GL1 35S::R-GR Mutant.
Light microscopy (A) and scanning electron microscopy ([B] to [D]) were used to show the mutant phenotype.
(A) Overview of an uninduced try 35S::GL1 35S::R-GR plant.
(B) Flower of an uninduced try 35S::GL1 35S::R-GR plant.
(C) High-magnification view of an uninduced try 35S::GL1 35S::R-GR leaf. Note the intervening epidermal cells.
(D) Rosette leaves of an induced try 35S::GL1 35S::R-GR plant grown on dexamethasone-containing plates.
Rescue of the ttg-1 Mutation.
Scanning electron microscopy was used to show the mutant phenotype.
(A) Overview of a 35S::GL1 ttg-1 leaf.
(B) High-magnification view of a reduced 35S::GL1 ttg-1 trichome.
(C) Overview of a young try 35S::GL1 ttg-1 leaf.
(D) Overview of mature try 35S::GL1 ttg-1 leaf.
(E) Differentiated try 35S::GL1 ttg-1 trichome.
(F) High-magnification view of clustered gl2-like try 35S::GL1 ttg-1 trichomes.
Trichome Initiation Sites (TIS) in try, 35S::GL1, ttg-1, and the Respective Single and Double Mutant Combinations
Maintenance of the Trichome Pattern
Initial trichome patterning results in the selection of single cells in a regular spacing pattern. How is this pattern fixed? Several double mutant combinations resulted in secondary trichome initiation immediately next to developing trichomes. Plants overexpressing the R-GR gene in the absence of TRY displayed a transformation of accessory cells into trichome cells under uninduced conditions (Figures 6C and 6D). Developing trichomes in the trichome initiation zone of the leaf were initially surrounded by a ring of apparently normal accessory cells. Later in leaf development, a variable number of accessory cells started to grow out, and most of them differentiated into mature trichomes (Figures 6C and 6D). This was observed for virtually all trichome initiation sites. A similar phenotype was observed for try 35S::GL1 plants (cf. Figures 6A and 6B). Also, 35S:: GL1 35S::R-GR plants under uninduced conditions exhibit this transformation of accessory cells into trichome cells. Thus, overexpression of both activators or overexpression of either of the activators in the absence of the inhibitor results in this phenotype. In summary, these data suggest that TRY is also involved in the lateral inhibition of trichome formation at later developmental stages.
Transformation of Accessory Cells into Trichomes.
(A) Scanning electron microscopy of a single mature trichome in 35S::GL1 with surrounding accessory cells.
(B) Whole-mount 4′,6-diamidino-2-phenylindole staining of a trichome cell in try 35S::GL1. Some of the adjacent accessory cells had initiated trichome differentiation.
(C) Scanning electron microscopy of a trichome cell with accessory trichomes in uninduced try 35S::R-GR plants.
(D) Overview of an uninduced try 35S::R-GR leaf.
DISCUSSION
Two lines of evidence suggest that trichome patterning in Arabidopsis does not involve cell lineage. The clonal analysis of wild-type patterning ruled out a cell lineage scenario in which the trichome cell is positioned in the center of its precursor cells (Larkin et al., 1996). The clonal analysis of trichome patterning in try mutants in this study shows that one of the key genes is not involved in cell lineage. Both experiments render it likely that the trichome spacing pattern is established by a competition mechanism that selects single trichome cells from otherwise equivalent epidermal cells.
Generation of a Spacing Pattern by Competition
Trichome initiation is confined to a relatively small region at the base of young leaves (Hülskamp et al., 1994; Larkin et al., 1996). At the time when developing trichomes first can be recognized, they are separated by approximately three or four epidermal cells (Hülskamp et al., 1994; Larkin et al., 1996). How is this spacing pattern generated? A simple scenario is that all cells within the trichome initiation zone are competent to become trichomes and compete with each other. The minimal requirements for this kind of cell-type selection have been formulated in a mathematical model (Meinhardt and Gierer, 1974). In this model, each cell produces an activator that stimulates its own production and that of an inhibitor. The activator acts locally and must have a self-enhancing property to allow the amplification of small stochastic local fluctuations. The inhibitor acts over long distances. A local increase of the activator would result in a locally increased production of the inhibitor, which in turn suppresses neighboring cells. According to this model, cluster formation would be expected either when the inhibitor levels are reduced or when the level of the activator is limited. Discussed below is a possible role for GL1, TTG/R, and TRY in the light of the predictions made by this model.
GL1 and TTG/R are considered to function as the activator and TRY to act as the inhibitor (Figures 7A and 7B). If TRY functions as the inhibitor, one would predict that TRY (1) would be able to negatively regulate GL1 and/or TTG/R and (2) would be activated by GL1 and/or TTG/R. In this study, we provide evidence for both predictions. A negative regulation of GL1 and TTG/R by TRY is suggested by the finding that plants overexpressing GL1 or the R gene show a strongly enhanced phenotype in the absence of TRY. This is most obvious in uninduced try 35S::GL1 35S::R-GR plants. Uninduced 35S::GL1 35S::R-GR plants show only little extra trichome formation, whereas the additional removal of TRY activity results in a strongly enhanced patterning phenotype.
The second prediction is that the function of TRY is coupled to the activity of the positive regulators. This prediction is supported by the increased cluster frequency in plants double heterozygous for try and different ttg alleles. Thus, a reduction of the inhibitory activity together with a reduction of the activating activity cause an enhanced phenotype. This finding establishes a link between the activating and the inhibitory pathways. Because this phenotypic enhancement was not observed in any of the double heterozygous combinations of try and gl1 alleles, it is unlikely that cluster frequency in try/+ ttg/+ double heterozygous plants is simply due to a change in the relative dosage of the activator and inhibitor concentrations. Thus, it is conceivable that the function of TRY is regulated by the TTG/R pathway. It should be emphasized that the proposed inhibitory role of TRY is based solely on genetic arguments, and therefore, TRY is not necessarily required to act non–cell autonomously.
The proposed role of GL1 and TTG/R is to promote trichome formation. But, how do these genes interact? We favor a model in which the R gene and GL1 act at the same regulation level downstream of TTG to initiate trichome formation. This model accommodates several lines of genetic and molecular evidence. It has been shown previously that overexpression of the R gene cannot rescue gl1 mutants (Larkin et al., 1994). This rules out a linear regulatory pathway and suggests that the GL1 and R genes act at the same regulatory level. Recently, Szymanski and co-workers demonstrated a direct molecular interaction of GL1 and the maize R gene (Szymanski et al., 1998a). The recent cloning of the TTG gene showed that TTG is not the R gene homolog (Walker et al., 1999). Therefore, it is likely that the R gene acts downstream of TTG, although the functional R gene homolog from Arabidopsis has not been identified.
Tipping the Balance: A Genetic Model for Trichome Patterning in Arabidopsis.
(A) Two epidermal cells in a balanced situation. The R and GL1 genes act together to promote trichome initiation, with TTG regulating the R gene. The double heterozygous combinations of gl1 and ttg alleles suggest an interaction between GL1 and TTG, probably with GL1 functioning as a positive regulator of TTG. TTG and TRY act together in the inhibition of neighboring cells, with TRY acting as a repressor of trichome initiation.
(B) Two epidermal cells in a biased situation. The black cell has gained more competence, and as a result, this cell expresses more GL1 and R and produces more TRY activity to inhibit trichome initiation in surrounding cells. The dashed line represents reduced interactions.
In (A) and (B), arrows indicating activation of TTG by GL1 are dotted to indicate that this interaction is postulated based on theoretical considerations and that evidence for this interaction is only circumstantial. Solid-line arrows indicate positive interactions.
That TTG acts upstream of GL1 is suggested by our finding that the strong ttg-1 phenotype is partially rescued by overexpression of GL1 and even more restored in a try 35S::GL1 background. However, according to the theoretical predictions, it is not possible that TTG acts simply upstream of GL1/R because no feedback loop would be established. A likely scenario is that TTG activity is dependent on GL1 activity (Figure 7). This view is consistent with two observations. Plants overexpressing GL1 produce a high number of trichome clusters when heterozygous for TTG (Larkin et al., 1994). A regulatory interaction between GL1 and TTG is also suggested by the allele specificity of cluster formation in double heterozygous gl1 and ttg mutants. Thus, in our model, TTG is not considered to be a general activator but to be involved in mediating the inhibitory signal and in the proposed positive feedback loop. In this scenario, the lack of trichome formation in ttg mutants would be due to the inability to locally enhance the activator concentration to levels high enough for trichome initiation. The phenotypic rescue of the ttg mutation in a try 35S::GL1 background would be due to elevated levels of GL1 that are sufficient for trichome initiation. Trichome spacing in the triple mutant would be mediated by the activity of other unidentified factors. The existence of additional factors is suggested by the finding that we were not able to transform all epidermal cells into trichome cells in try 35S::GL1 35S::R-GR plants. A candidate for such a redundant gene is COTYLEDON TRICHOME1 (COT1), which has been identified as an enhancer of the GL1 overexpression phenotype (Szymanski et al., 1998b).
Maintenance of the Trichome Pattern
When the leaf matures, developing trichomes become separated by cell divisions of intervening epidermal cells (Hülskamp et al., 1994). The initiation of new trichomes ceases in mature sections of the leaf (Hülskamp et al., 1994). It is likely that epidermal cells gradually lose the competence to become a trichome cell. This view is supported by the finding that R gene–induced trichome initiation is found only in young developing tissues but not in mature tissues (Lloyd et al., 1994). In contrast, overexpression of the GL1 and R genes or overexpression of either one in the absence of TRY results in trichome initiation in more mature leaf areas. New trichomes, however, are initiated in immediate contact with already existing trichomes rather than between trichomes. This suggests that in plants overexpressing the GL1 or R gene, cells immediately neighboring a trichome are competent to develop into trichomes and are subject to lateral inhibition by the central trichome cell. Because this late transformation of accessory cells into trichome cells was observed only in lines overexpressing positive regulators of trichome development, the extent to which lateral inhibition of neighboring cells is relevant in later stages of wild-type trichome development remains to be determined.
Multiple Roles of GL1, TTG/R, and TRY in Different Aspects of Trichome Development
Although the proposed role of GL1, TTG/R, and TRY in trichome patterning formally resembles well-characterized patterning processes in animals, the underlying molecular mechanisms are likely to be different. In animal development, pattern formation and cell fate determination are usually separate processes. A well-characterized example is the patterning of sensory bristles in the epidermis of Drosophila. A class of transcription factors, products of the achaetescute gene complex, promote bristle development. Cellular interactions are mediated by a receptor–ligand—based mechanism that involves the Delta and Notch gene products (Campos-Ortega, 1993; Ghysen et al., 1993; Hinz et al., 1994; Artavanis et al., 1995). However, these genes appear to act exclusively in producing different cell fates within a given tissue, because the same set of genes acts not only in ectodermal patterning but also in endodermal patterning. The actual cell fate is determined by tissue-specific cues of the respective tissue layers (Yan and Jan, 1993; Tepass and Hartenstein, 1995).
By contrast, GL1, TTG/R, and TRY are involved not only in trichome patterning but also in the regulation of organ and tissue layer specificity of trichome formation (Hülskamp and Schnittger, 1998). Overexpression of the R gene triggers trichome initiation on organs that normally do not exhibit trichomes (Lloyd et al., 1992). Similarly, try 35S::GL1 plants produce ectopic trichomes on different organs and in subepidermal tissue layers (Schnittger et al., 1998; Szymanski and Marks, 1998). This suggests that in trichome development, cell-type specification and patterning are mediated by the same set of gene functions. A possible entry point in cell differentiation is the regulation of endoreplication or DNA synthesis, which marks the transition from epidermal to trichome cell fate (Hülskamp et al., 1994). Interestingly, GL1 and TRY are also involved in the regulation of endoreplication. try mutants show an increased DNA content compared with that of the wild type (Hülskamp et al., 1994). A role of GL1 in the regulation of endoreplication is suggested by the finding that try 35S::GL1 plants show an additional increase in nuclear DNA content (Schnittger et al., 1998; Szymanski and Marks, 1998). The dual role of GL1 and TRY in trichome patterning and the regulation of endoreplication could reflect a functional link between the two processes.
TTG also appears to have a dual role in trichome patterning and differentiation. This is suggested by our finding that try 35S::GL1 ttg-1 plants show a significant rescue of trichome initiation; however, trichome differentiation is still impaired in >50% of the developing trichomes. Strikingly, the phenotype is reminiscent of that of unexpanded trichomes in gl2 mutants, suggesting that TTG is also required for the regulation of genes acting later in trichome morphogenesis.
In summary, GL1, TTG/R, and TRY are involved in multiple aspects of trichome development. It is likely that these different aspects reflect a complex molecular network underlying trichome patterning. Determining the roles of these genes in such a network is necessary for us to test our model of trichome patterning.
METHODS
Plants and Plant Culture
The Arabidopsis thaliana Landsberg erecta ecotype was used as a wild-type reference. The strong ttg-1 allele was obtained from M. Koornneef (Agricultural University, Wageningen, The Netherlands). The weak ttg-w allele was obtained from G. Haughn (University of British Columbia, Vancouver, Canada), and the weak ttg-10 allele was from D. Marks (University of Minnesota, St. Paul, MN). The try-EM1 allele was isolated as a recessive mutation in a Landsberg erecta background (Hülskamp et al., 1994). The strong gl1-EM1 allele and the two weak gl1 alleles gl1-EM2 and gl1-EM3 were identified by screening M1 progeny of ethyl methanesulfonate–mutagenized Landsberg erecta seeds. The 35S::GL1 line (Larkin et al., 1994) and the pGGE4 line (Larkin et al., 1993) were obtained from D. Marks. The transgenic line containing the 35S::R-GR was obtained from A. Lloyd (Lloyd et al., 1994). Plants were grown under constant illumination, as described previously (Mayer et al., 1993).
The try 35S::GL1 and try 35S::R-GR lines were constructed as described previously (Schnittger et al., 1998). The try 35S::R mutant was identified based on the new phenotype among the progeny of selfed F1 plants from a cross between the homozygous 35S::R line and try. The try/try 35S::GL1/+ 35S::R-GR/+ plants were obtained from a cross between try 35S::GL1 and try 35S::R-GR. The try 35S::GL1 ttg-1 triple mutant was constructed by selecting plants from the F2 progeny of a cross between try 35S::GL1 and ttg-1 displaying the phenotype of the try 35S::GL1 mutant. The triple mutant was identified among the F3 progeny of single plants exhibiting reduced trichomes and the transparent testa phenotype of the seeds. Homozygosity for try was confirmed by backcrosses with try. The presence of 35S::GL1 was confirmed by kanamycin selection. The ttg-1 35S::GL1 double mutant was identified among the F2 progeny of a cross between 35S::GL1 and ttg-1 as glabrous, kanamycin-resistant plants. All double heterozygous combinations were F1 plants of crosses between the respective parental lines.
For dexamethasone induction, 35S::R-GR plants were grown on soil, and whole pots with plants were immersed in 10 μM aqueous dexamethasone solution for 15 min. For in vitro germination under inductive conditions, seeds were grown on agar plates (0.5% agar, 0.5 × Murashige and Skoog medium [Duchefa, The Netherlands], 3% sucrose, and 100 μM ampicillin) containing 10 μM dexamethasone.
Calculation of the Confidence Interval
The 95% confidence interval is defined as the interval that contains the mean value with a probability of 95%. The calculation of the confidence interval is based on the following formula (Kreyszig, 1979):
where n is the number of trichome initiation sites. Trichome initiation sites are defined as leaf positions containing one trichome or a trichome cluster. The cluster frequency is p. The number of trichome initiation sites containing more than one trichome is k. The chosen probability is c with c = 1960 at 95% probability (for details, see Kreyszig, 1979). Because n − k was always large in this study, a simplified formula was used for the calculation of the confidence interval (Kreyszig, 1979): KONF {k/n − a = <p = <k/n + a} with a = c/n((k(n − k))/n)0.5.
Microscopy and Graphics Work
4′6-Diamidino-2-phenylindole staining was done as previously described (Hülskamp et al., 1994). For scanning electron microscopy, plant material was fixed in 100% methanol for 5 min, transferred to 100% ethanol, and further processed as previously described (Laux et al., 1996). In some cases, epoxy replicas of epidermal cells or immature trichome cells were used for scanning electron microscopic analysis (Williams and Green, 1988). Dental impression material (Colténe PRESIDENT light body; Coltène AG, Altstätten, Switzerland) was applied to the leaf surface for 5 min and peeled off. The negative molds were filled with Spurr's resin, polymerized at 70°C, and processed as described for fixed material.
Images were processed using Adobe Photoshop 3.0 (Adobe Systems Inc., San Jose, CA) and Aldus Freehand 7.0 (Aldus Corp., Seattle, WA) software.
ACKNOWLEDGMENTS
We thank Niko Geldner, Thorsten Hamann, Maren Heese, Hilmar Ilgenfritz, Michael Lenhard, and Kathrin Schrick for helpful suggestions and critical reading of the manuscript. We thank Heinz Schwarz and Jürgen Berger for generous hospitality and help with the scanning electron microscopy. U.F. is a recipient of a Boehringer Ingelheim Fonds fellowship. B.S. is a Friedrich Naumann Stiftung fellowship recipient. This work was supported by a Sonderforschungs bereich 446 grant to M.H. and a Leibniz award to G.J. from the Deutsche Forschungsgemeinschaft.
- Received December 14, 1998.
- Accepted March 31, 1999.
- Published June 1, 1999.
REFERENCES
