Abstract
Trichome development in Arabidopsis thaliana is a well-characterized model for the study of plant cell differentiation. Two genes that play an essential role in the initiation of trichome development are GL1 and TTG. Mutations in either gene prevent the initiation of most trichomes. The GL1 gene encodes a myb-related transcription factor. Mutations in TTG are pleiotropic, affecting anthocyanins, root hairs, and seed coat mucilage in addition to trichomes. Six ttg alleles were examined and shown to form a hypomorphic series. The severity of all aspects of the ttg phenotype varied in parallel in this allelic series. The weakest allele, ttg-10, causes frequent clusters of adjacent trichomes, suggesting a role for TTG in inhibiting neighboring cells from choosing the trichome fate. This allele results from a mutation in the 5′-untranslated region of ttg and creates an out-of-frame upstream AUG codon. The ttg-10 allele shows several unusual genetic interactions with the weak hypomorphic gl1-2 allele, including intergenic noncomplementation and a synthetic glabrous phenotype. These interactions are specific for the gl1-2 allele. The implication of these results for current models of trichome development is discussed.
CELL fate specification is a process of fundamental importance during development. Although much work has been done in plants on differentiation at the tissue level, relatively few systems are available in plants for the study of the differentiation of individual cell types. The development of Arabidopsis thaliana leaf hairs (trichomes) is a well-established model for the study of plant cell fate and differentiation (Larkinet al. 1997; Hülskamp and Schnittger 1998). Arabidopsis trichomes are highly specialized single cells that develop in the plant epidermis. These cells undergo a dramatic program of polarized cell expansion, ultimately elaborating several aerial branches. Trichomes are one of several possible epidermal cell fates; other cell types present in the epidermis include unspecialized epidermal cells, stomatal guard cells, and root hairs. In nature, trichomes may provide protection against insect herbivores (Mauricio and Rausher 1997), but under laboratory conditions trichomes are nonessential. This has allowed the identification of mutations in more than twenty genes involved in trichome development (Marks 1997). The accessibility of trichomes to genetic analysis makes this an excellent system for determining the cascade of developmental events involved in the differentiation of a single cell type.
Current evidence suggests that trichome precursors are selected from a field of developmentally equivalent cells by a mechanism similar to that involved in the development of Drosophila sensory bristles (Artavanis-Tsakonaset al. 1995) and Caenorhabditis vulval development (Sternberg and Horvitz 1984). Developing trichomes are distributed in a minimum-distance spacing pattern more regular than would be expected by chance, indicating the existence of an underlying mechanism controlling the patterning of trichome precursors (Hülskampet al. 1994; Larkinet al. 1996). Cell lineage has been ruled out as a basis for this pattern (Larkinet al. 1996), and current models invoke lateral inhibition of neighboring cells to explain the distribution of these cells (Larkinet al. 1996), although the mechanistic basis remains unknown. Two genes, GLABRA1 (GL1) and TRANSPARENT TESTA GLABRA (TTG), are known to play a central role in the trichome cell fate decision based on the virtually complete absence of trichomes on loss-of-function mutants of either gene. GL1 encodes a protein containing a MYB DNA-binding domain and is expressed most highly very early during trichome development, consistent with its proposed role as a central regulator of the cell fate decision (Oppenheimeret al. 1991; Larkinet al. 1993). In addition, when GL1 is expressed ectopically along with the maize R gene, which encodes a basic-helix-loop-helix (bHLH) transcription factor, trichomes are produced on epidermal surfaces that are normally glabrous (Larkinet al. 1994a). These observations suggest that the GL1 protein, probably in conjunction with an endogenous R-like protein, is a central regulator of the trichome cell fate decision. Mutations in gl1 do not appear to affect any processes other than trichome development.
In contrast to GL1, TTG is required for a striking diversity of processes, as was first noted by Koornneef (1981). In addition to lacking trichomes, ttg mutants are defective in the synthesis of anthocyanin pigments, production of seed coat mucilage, and root hair development (Koornneef 1981; Galwayet al. 1994). The role of TTG in root hair development is particularly noteworthy, because TTG is required for development of the nonhair cells in the root epidermis. Thus, ttg mutants produce extra root hairs.
Two observations suggest that TTG plays a role in the spacing of trichome precursors in the epidermis (Larkinet al. 1994a). First, some hypomorphic alleles of TTG produce frequent clusters of partially developed adjacent trichomes. Clusters of adjacent trichomes are rarely observed on wild-type (WT) leaves. Second, ttg heterozygotes containing a 35SGL1 constitutive overexpression construct also produce a high frequency of trichome clusters. One explanation for these observations is that a process downstream of TTG may be involved in lateral inhibition of cells neighboring a trichome precursor, preventing them from developing as trichomes. A candidate for a gene functioning in this pathway is the TRIPTYCHON (TRY) gene; leaves of try mutants produce numerous trichome clusters (Hülskampet al. 1994).
Several parallels exist between the regulation of Arabidopsis trichome development and the regulation of anthocyanin biosynthetic enzymes in various plant species, in addition to the involvement of TTG in both processes in Arabidopsis. In maize, two gene families have been identified that encode transcription factors required for the expression of anthocyanin biosynthetic enzymes. The R/B family encodes bHLH proteins (Chandleret al. 1989; Ludwiget al. 1989), while the C1/Pl family encodes Myb proteins related to GL1 (Cone et al. 1986, 1993; Paz-Areset al. 1987; Grotewoldet al. 1994). Members of these two gene families have also been identified as regulators of anthocyanin biosynthetic genes in several other plant species (reviewed in Molet al. 1998). In maize, members of these two regulatory protein families cooperate to induce anthocyanin biosynthesis and appear to interact physically (Kleinet al. 1989; Goffet al. 1992). The mode of action of Myb and bHLH proteins in the regulation of anthocyanin biosynthesis thus provides a model for understanding their role in trichome development.
The maize R gene can bypass the need for TTG in all TTG-dependent processes when it is expressed in Arabidopsis ttg mutants (Lloydet al. 1992). This observation initially led to suggestions that TTG might encode an R homologue. Recently, however, TTG has been cloned and shown to encode a protein of 341 amino acid residues containing four WD repeats (A. R. Walker, P. A. Davison, A. C. Bolognesi-Winfield, C. M. James, N. Srinivasan, T. L. Blundell, J. J. Esch, M. D. Marks and J. C. Gray, unpublished results). The petunia anthocyanin regulatory gene an11 also encodes a WD-repeat protein closely related to TTG (de Vettenet al. 1997; A. R. Walker, P. A. Davison, A. C. Bolognesi-Winfield, C. M. James, N. Srinivasan, T. L. Blundell, J. J. Esch, M. D. Marks and J. C. Gray, unpublished results). Apparent homologues of the TTG/an11 genes exist in humans, Caenorhabditis elegans, and Saccharomyces cerevisiae. The WD repeat is a structural motif found in proteins regulating transcription, signal transduction, vesicular trafficking, RNA processing, and cytoskeletal assembly (Neeret al. 1994). Many WD-repeat proteins are known to physically interact with other proteins, and this domain is thought to function as a protein interaction domain.
Previous work has suggested that GL1 and TTG act at essentially the same point in the trichome pathway (Larkinet al. 1994a). Several of the experiments used to draw this conclusion, however, were performed using the maize R gene as a surrogate for TTG function and may reflect the function of an endogenous R-like gene acting downstream of TTG, rather than TTG itself. In this work, we describe allele-specific genetic interactions between ttg and gl1 mutations that suggest that the functions of these two genes are tightly intertwined.
MATERIALS AND METHODS
Growth conditions, plant strains, and genetic methods: Plants were grown under continuous illumination by 40-W Sylvania Cool White fluorescent bulbs (∼100 μE m2 s–1) at 21° in a mixture of a peat-based potting medium and vermiculite and watered from below. Plants were fertilized with a modified Hoagland's solution (Epstein 1972) initially and again when the plants began to bolt.
Trichomes were counted on either of the first two postembryonic leaves, referred to as “first leaves” throughout this report. These leaves are developmentally equivalent and have a similar number of trichomes. To avoid miscounting inconspicuous abortive trichomes on the ttg mutants, trichomes were counted only if they clearly protruded from the surface of the leaf (∼50 μm). In general, trichomes were counted on a single first leaf from 10 independent plants per pot. Root hairs were counted on plants grown on vertically placed agar plates, essentially as described by Galway et al. (1994). Anthocyanin phenotypes were assessed by visual inspection of seedlings grown on both agar plates and soil. Seed-coat mucilage phenotypes were determined by viewing seeds mounted on microscope slides in a dilute suspension of India ink (Rerieet al. 1994).
Trichome clustering phenotypes: Trichome clusters were defined as any instance where two or more trichomes were located immediately adjacent to each other, without intervening adjacent cells. This phenotype has been previously documented (Larkinet al. 1994a) and is readily scored in a dissecting microscope.
Plants homozygous for the ttg-10 allele exhibit a distinctive phenotype, with numerous incompletely developed trichomes that often occur as clusters (Larkin et al. 1994a,b; Figure 1H). Many of these trichomes develop as unbranched spikes, which are very noticeable in the dissecting microscope. In discussion of genetic interactions between ttg-10 and gl1 alleles, the phenotype exhibited by ttg-10 homozygotes in an otherwise WT background is referred to as “spiky.”
Origins of mutant alleles used in this study
Origins of mutant alleles: Unless otherwise noted, WT refers to the Columbia-0 (Col-0) genetic background throughout this paper. The origins and crossing history of the six ttg alleles used in this study are summarized in Table 1. Allelism of all ttg mutations was confirmed by examining both the F1 and the F2 of crosses between the new alleles and the ttg-1 reference allele. Both the defective trichome phenotype and the yellow seed phenotype (caused by the absence of anthocyanin pigmentation) were used to confirm noncomplementation. Linkage of putative ttg alleles to the ttg locus was confirmed by examining the F2 progeny of crosses to the reference allele. The origins of ttg-1, ttg-9, ttg-10, and ttg-11 have been described previously (Larkinet al. 1994a). The ttg-12 allele, isolated in the Col-0 genetic background, was obtained from A. Sessions (University of California, Berkeley, CA). The ttg-13 allele was isolated in the Rschew (RLD) genetic background by D. Oppenheimer in the lab of M. D. Marks (University of Minnesota, St. Paul) in a screen of fast-neutron mutagenized seed purchased from Lehle Seeds (Round Rock, TX). Alleles isolated from a Col background (ttg-9, ttg-11, and ttg-12) were crossed to Col-0 three times before selfing and selecting ttg mutant plants. Alleles isolated from a background other than Col were crossed to Col-0 five times before selfing and selecting ttg mutant plants, with the exception of ttg-13. This allele was added late in the study and was crossed to Col-0 only three times prior to selfing. This reduced level of introgression is unlikely to influence the results, because first leaves of Col-0 and RLD WT plants have similar numbers of trichomes, and no aberrant trichome phenotypes have been observed to segregate from numerous crosses between these ecotypes. To ensure that all potential modifier loci had been eliminated by backcrossing, the WT siblings of each backcrossed ttg mutant were examined to confirm that the number of trichomes per leaf was similar to that of Col-0. Particular care was taken in backcrossing ttg-1 plants to choose families that no longer segregated the er mutation derived from the Landsberg erecta (Ler) background. Ler carries an allele of the RTN locus that reduces the number of trichomes per leaf approximately threefold relative to Col-0, and rtn is tightly linked to er (Larkinet al. 1996).
The gl1-1 allele used in this study was described previously (Oppenheimeret al. 1991). In the line used here, gl1-1 had been introgressed into the Wassileskjia (WS) background by crossing six times before selfing was used in these studies. Crosses to WS were included as controls in some experiments and gave results essentially identical to results obtained with Col-0 (Table 4; data not shown). The gl1-2 allele originated in the Col-1 background (Eschet al. 1994). The 35SGL1 line used here has been described previously as 35SGL1 4-1 (Larkinet al. 1994a). This line was generated by Agrobacterium-mediated transformation of Col-0, and it is homozygous for the T-DNA insert. The try-JC allele was generated in Col-0 and was provided by J. Chien (University of California, Berkeley, CA). Allelism with try was confirmed both by failure to complement the recessive trichome clustering phenotype associated with the try-82 allele (provided by M. Hülskamp, University of Tübingen, Tübingen, Germany) and by the absence of phenotypically WT plants in the subsequent F2, demonstrating linkage of try-JC to the try locus. Of the three try alleles that we have examined, try-JC appears to have the strongest phenotype. The gl2-p1 and gl2-p2 alleles were obtained from M. D. Marks and originated with R. S. Poethig (University of Pennsylvania, Philadelphia). These alleles were generated in the Col-0 background and were not backcrossed before use in the experiments described here. The gl3-1 reference allele in the Ler background was obtained from M. D. Marks and was not back-crossed before use in the experiments described here.
Identification of ttg-10 gl1-2 double mutants: The double mutant was initially identified in the F2 of a cross between ttg-10 and gl1-2 as a completely glabrous plant that produced only completely glabrous plants with yellow seeds in the F3. The gl1-2 genotype of this plant was confirmed by PCR amplification of genomic DNA with the gl1-2-specific primers gl1-2R (5′GCAAAATTCATCATTACGAGTG3′) and gl1-2F (5′TCATCTCAGCAAAAAACTCG3′), using an annealing temperature of 57° with 3 mm MgCl2 for 30 cycles, followed by analysis on a 2.5% agarose gel. These primers flank the gl1-2 deletion (see results) and identify the gl1-2 allele as a length polymorphism relative to WT in the amplified product.
Scanning electron microscopy: Samples fixed in FAA (3.7% formaldehyde, 50% ethanol, 5% acetic acid) were prepared for scanning electron microscopy by standard methods (Irish and Sussex 1990).
Sequence analysis of mutant ttg alleles: Genomic DNA was extracted from 4-wk-old plants using the method of Dellaporta et al. (1983). The DNA was purified further by centrifugation on CsCl gradients (Walkeret al. 1997). Using genomic DNA from mutants ttg-10, ttg-11, ttg-12, and their parental ecotypes (WS and Col) as templates, PCR products from the region of the TTG gene were generated by amplification using a mixture of thermostable polymerases BioTaq (Bioline) and Pfu (Stratagene, La Jolla, CA) in the ratio of 10:1. Four pairs of oligonucleotide primer pairs were designed to give overlapping fragments of about 700 bp. PCR products, which covered the gene from 500 bp upstream from the start of translation to 400 bp into the intron sequence (A. R. Walker, P. A. Davison, A. C. Bolognesi-Winfield, C. M. James, N. Srinivasan, T. L. Blundell, J. J. Esch, M. D. Marks and J. C. Gray, unpublished results), were subjected to electrophoresis on 1% agarose TAE gels (Sambrooket al. 1989) and extracted from the gels using a Qiaex II gel extraction kit (QIAGEN, Valencia, CA). The purified DNA was sequenced directly using both primers by an ABI (Columbia, MD) automated sequencer. Comparisons between mutant and parental WT sequences were carried out using the SeqEd program (ABI) to determine the position of the mutations.
Gel blot analysis of transcripts of TTG from mutants: Total RNA from floral buds of 4-wk-old soil-grown WS and mutant ttg-10, ttg-11, and ttg-12 plants was extracted from 200 mg of tissue using Tripure (Boehringer Mannheim, Mannheim, Germany). RNA (10 μg) was subjected to electrophoresis on a 1.2% agarose denaturing formaldehyde gel and blotted onto Genescreen Plus (Dupont, Wilmington, DE) in accordance with the manufacturer's instructions. Double-stranded probes were prepared as in Walker et al. (1997) by random-primer extension using [α32P]dATP (Amersham, Arlington Heights, IL). The TTG probe contained the entire coding region. Hybridization was carried out at 42° in 50% formamide for 16 hr. Blots were washed in 0.1× SSC, 2% SDS at 65° for 15 min. The filter was exposed to X-Ograph XB-200 film (X-Ograph Imaging Systems, Inc.) with two intensifying screens for 2 days at –80°. The blot was stripped by boiling in 0.1× SSC, 2% SDS for 5 min and checked by exposure to film before hybridization to an Arabidopsis β-tubulin cDNA probe (GenBank accession no. Z25960, 3′-untranslated region of the TUB5 gene).
RESULTS
Genetic characterization of ttg alleles: Koornneef (1981) was the first to describe the pleiotropic phenotype of ttg mutants in detail. He examined eight independent alleles and concluded that the trichome, anthocyanin, and seed-coat aspects of the ttg phenotype were the result of mutations at a single locus. Since this description, however, the trichome clustering and root hair effects of ttg mutations have been described (Galwayet al. 1994; Larkinet al. 1994a). While it is virtually certain that this phenotypic syndrome is caused by mutations in a single gene, it is possible that the TTG gene could have multiple functional domains. If this is the case, some mutations might affect different TTG functions to different extents. For these reasons, the genetic characterization of a series of ttg alleles was undertaken.
Six ttg alleles were characterized in a predominantly Col-0 genetic background (Figure 1; Table 2). All of these alleles were fully recessive to WT. All aspects of the ttg phenotype were found to vary in parallel, with ttg-13 being the strongest allele and ttg-10 the weakest. None of the ttg alleles resulted in visible anthocyanin pigmentation in the seed coat, but seedlings homozygous for the ttg-10 allele produced some anthocyanin pigment on the cotyledons and the upper part of the hypocotyl. Seed coats of ttg-10 mutants also produced mucilage. Seedlings homozygous for the ttg-9 allele did not produce visible anthocyanins but did produce patchy seed-coat mucilage (J. C. Larkin, unpublished observations). None of the other alleles resulted in production of either anthocyanins in the seedling or seedcoat mucilage. As previously reported (Larkin et al. 1994a,b), weak ttg alleles caused clusters of adjacent trichomes at a fairly high frequency (Tables 2 and 3). Recently, ttg-13 has been shown to be a null allele containing a deletion of at least 4 kb that includes the TTG gene (A. R. Walker, P. A. Davison, A. C. Bolognesi-Winfield, C. M. James, N. Srinivasan, T. L. Blundell, J. J. Esch, M. D. Marks and J. C. Gray, unpublished results), consistent with ttg-13 being the strongest allele. Despite the absence of any TTG coding sequences, a few trichomes and clusters of adjacent trichomes were occasionally found on ttg-13 mutant leaves (Figure 1C). These observations demonstrate that TTG is not absolutely necessary for limited trichome development.
All combinations of ttg allele trans-heterozygotes were constructed (Table 3), and all combinations resulted in a trichome phenotype intermediate to that of the parental alleles. In particular, all alleles produced fewer trichomes when in trans to the ttg-13 null than when they were homozygous. Therefore, these six ttg alleles formed a hypomorphic allelic series, with the order of strength of the alleles from strongest to weakest being ttg-13 > ttg-1 > ttg-12 ≈ ttg-11 > ttg-9 > ttg-10. There was no intragenic complementation, and no evidence was found for the existence of multiple functional domains in TTG.
Genetic characterization of the gl1-2 allele: The genetic behavior of the weak gl1-2 allele also was characterized. This allele contains a 14-bp deletion that results in an in-frame stop codon at the point of the deletion (Eschet al. 1994). The mutation would result in a polypeptide missing 27 amino acids from its C terminus. gl1-2 mutant plants produce very few trichomes on their first two leaves (Table 4). On subsequent leaves, the mutants have an increasing number of trichomes, but few trichomes develop near the midvein of the leaf (Figure 2A). The gl1-1 allele is a null allele caused by a 6.5-kb deletion that completely removes the GL1 coding region (Oppenheimeret al. 1991). The gl1-2 allele is fully recessive to WT (Figure 2B; Table 4), and when placed in trans to the null allele, the phenotype is more severe than the phenotype of gl1-2 homozygotes (Table 4). The gl1-2 allele thus behaves as a typical hypomorph.
The ttg-10 allele interacts with alleles of gl1 and try: When mutations in two different genes fail to complement in a double heterozygote, this often indicates that the two genes act in a common pathway (Stearns and Botstein 1988; Simonet al. 1991). Intergenic noncomplementation has been used to identify new interacting mutations, but it also can be used to examine preexisting mutations for genetic interactions (Theisenet al. 1994). To investigate the relationship between ttg and other mutations in the trichome pathway, the weak ttg-10 allele was crossed with alleles of various genes involved in trichome development (gl1-2, gl2-p1, gl2-p2, gl3-1, and try-JC), and the F1 progeny (ttg-10/+ m/+, where m is another trichome mutation) were examined for altered trichome phenotypes. Plants heterozygous for ttg-10 and either gl2-p1, gl2-p2,or gl3-1 were indistinguishable from ttg-10 single heterozygotes (J. Larkin, unpublished observations). In contrast, plants doubly heterozygous for ttg-10 and either try-JC or gl1-2 showed an increase in the number of trichomes present in clusters (Figure 2C; Tables 5 and 6).
—Phenotypes of first leaves of strains carrying the indicated ttg alleles. (A) Col-0 WT. (B) ttg-13. (C) Trichomes on ttg-13 leaf. (D) ttg-1. (E) ttg-12. (F) ttg-11. (G) ttg-9. (H) ttg-10. Bars in A, B, D, E, F, G, and H, 500 μm. Bar in C, 100 μm.
A high frequency of trichome clusters also was observed in the F1 of crosses of several other ttg alleles with gl1-2 (Table 6), indicating that this clustering interaction was not dependent on a particular allele of ttg. In contrast, the high frequency of trichome clusters on ttg/+ gl1-2/+ plants was specific for the gl1-2 allele. When each of the six ttg alleles was crossed with the gl1-1 null allele, less than one-fourth as many trichomes were produced in clusters as were seen on ttg/+ gl1-2/+ plants (Table 6). In fact, fewer trichomes in clusters were produced on ttg/+ gl1-1/+ plants than were produced on any of the ttg allele heterozygote controls (Table 6).
Detailed phenotypes of ttg alleles
The frequency of trichome clusters on leaves of ttg/+ controls (Table 6) was somewhat higher than was reported previously (Larkinet al. 1994a). This difference may be due to the use of a Col-0 genetic background, because in the previously reported experiments the ttg-1 allele was in an Ler background. ttg-1 homozygotes produced trichomes at a higher frequency in a Col-0 background than in an Ler background (J. Larkin, unpublished observations). This suggests that a genetic modifier present in the Col-0 background can increase the frequency of trichome initiation on leaves of ttg-1 mutants. The Ler background contains an rtn allele that results in a reduced number of trichomes relative to Col-0 (Larkinet al. 1996). It is possible that the rtn allele from Col-0 also modifies the frequency of clustering in ttg heterozygotes. However, we cannot rule out the possibility that this difference is due to other modifiers that differ between the two genetic backgrounds.
Trichome counts and percentage clustered trichomes on ttg allele trans-heterozygotes
Allele-specific interactions between gl1 and ttg-10: Further evidence of an interaction between gl1-2 and ttg10 was obtained from the segregation ratio of trichome phenotypes in the self-pollinated F2 of the double heterozygotes. As noted above, the first two leaves of gl1-2 plants produce few, if any, trichomes. The first leaf phenotype of gl1-2 will be referred to hereafter as “glabrous.” Plants carrying the ttg-10 mutation produce numerous clusters of trichomes on the first leaves. These trichomes generally are unbranched spikes or have a reduced number of branches. This distinctive phenotype will be referred to hereafter as “spiky.”
Genetic characterization of the gl1-2 allele
—Genetic interactions between ttg-10 and gl1-2. (A) gl1-2. (B) gl1-2/+. (C) ttg-10/+ gl1-2/+. Arrow indicates a trichome cluster. (D) ttg-10/ttg-10 gl1-2/+. (E) Scanning electron micrograph of ttg-10/ttg-10 gl1-2/+ leaf. Bar, 50 μm. (F) ttg-10/+ gl1-2. Arrow indicates a trichome.
The most obvious expectation for the F2 trichome phenotypes of self-pollinated ttg-10/+ gl1-2/+ plants is that the glabrous gl1-2 phenotype would be epistatic to the spiky ttg-10 phenotype. The predicted segregation ratio in this case would be 9 WT:3 spiky:4 glabrous. Unexpectedly, a significant deficit of spiky plants was noted in this cross (Table 7). The observed phenotypes were consistent, with a 9 WT:1 spiky:6 glabrous ratio (Table 7). This ratio could be explained if plants of the genotype ttg-10/ttg-10 gl1-2/+ were also glabrous, in addition to +/+ gl1-2/gl1-2 single mutants and ttg-10/ttg-10 gl1-2/gl1-2 double mutants. In other words, the gl1-2 allele appeared to be dominant in ttg-10 homozygotes.
Interactions between ttg and try alleles
To test this hypothesis, 15 glabrous F2 plants were allowed to self, and the resulting F3 families were examined for trichome phenotype and anthocyanin production (Table 8). Seven of these F2 plants produced brown seeds, indicating that these plants had been either +/+ or +/ttg-10. As expected in both models, all of the F3 plants derived from these seeds were glabrous, confirming that the F2 parent had been gl1-2/gl1-2. Eight of the F2 plants produced yellow seeds, indicating that they were homozygous for ttg-10. Two of these families produced only glabrous plants in the F3, suggesting that they were ttg-10/ttg-10 gl1-2/gl1-2 double homozygotes. The gl1-2 homozygous genotype of these two families was confirmed by PCR using primers flanking the gl1-2 deletion. The remaining six F2 plants that produced yellow seed segregated 1 spiky:3 glabrous in the F3 (Table 8). This is the expected ratio if the F2 parent is homozygous for ttg-10 and heterozygous for gl1-2, and if gl1-2 is dominant when ttg-10 is homozygous.
Interactions between ttg and gl1 alleles
To confirm that ttg-10/ttg-10 gl1-2/+ plants were glabrous, ttg-10 mutant plants were crossed to the ttg-10 gl1-2 double mutant. The resulting F1 plants were completely glabrous; no trichomes were produced even on subsequent leaves (Figure 2D). No evidence of trichome initiation was detected by scanning electron microscopy (Figure 2E). This phenotype is even more severe than that of gl1-2 homozyotes, all of which produce at least some trichomes on the fourth and subsequent leaves (10 out of 10 plants).
Plants of the genotype ttg-10/+ gl1-2/gl1-2 were also constructed. These plants had a phenotype stronger than that of gl1-2 homozygotes; 4 of 10 plants had no trichomes on any of the first seven leaves, and no plant had more than two trichomes per leaf on any of the first seven leaves (Figure 2F). This result indicates that ttg-10 has a dominant effect on trichome initiation in plants that are homozygous for gl1-2.
In contrast to the gl1-2 allele, the gl1-1 null allele is fully recessive in a ttg-10 homozygous mutant background. The expected 9 WT:3 spiky:4 glabrous ratio was observed in the F2 of a cross between ttg-10 and gl1-1 (Table 7). Analysis of F3 families confirmed that 18 of 22 spiky F2 plants segregated 3 spiky:1 glabrous in the F3. The remainder of these families segregated only spiky plants in the F3. Eleven glabrous F2 plants produced only glabrous progeny. These results are exactly as expected if gl1-1 is acting as a simple recessive mutation that is epistatic to ttg-10 in these crosses.
F2 analysis of crosses between ttg-10 and gl1 alleles
The interaction between 35SGL1 and ttg is not allele-specific: When GL1 is expressed in a WT background from the constitutive cauliflower mosaic virus 35S promoter (35SGL1), the number of trichomes per leaf is reduced (Larkinet al. 1994a; Table 7). In contrast, plants heterozygous for ttg-1 and carrying the 35SGL1 construct also produce a high frequency of trichome clusters, as reported previously (Larkinet al. 1994a). The allele specificity of this interaction was tested in this study. A high frequency of trichome clusters was found on leaves of ttg/+ 35SGL1/+ plants for all three of the ttg alleles tested, including the ttg-13 null allele (Table 7). Thus this interaction is not specific for any single ttg allele.
F3 analysis of the ttg-10 × gl1-2 cross
Molecular analysis of ttg alleles: The recent isolation of the TTG gene (A. R. Walker, P. A. Davison, A. C. Bolognesi-Winfield, C. M. James, N. Srinivasan, T. L. Blundell, J. J. Esch, M. D. Marks and J. C. Gray, unpublished results) allowed the nature of the mutations present in each of the ttg alleles to be determined. The sequences of the ttg-10, ttg-11, and ttg-12 alleles were determined during this study (Figure 3, A and B). All three mutations were single nucleotide substitutions. The ttg-10 mutation wasaGtoA transition 43 nucleotides upstream of the start codon. This mutation was within the 109-bp untranslated leader and results in an out-of-frame upstream AUG codon in the ttg mRNA. Translation from this upstream AUG codon would produce a peptide 77 amino acids long. This new open reading frame was followed by two stop codons. The ttg-11 mutation was a G to A transition, causing a Gly to Arg change in amino acid 149, in the first WD repeat. The ttg-12 mutation was a G to A transition, causing a Gly to Arg change in amino acid 43, upstream of the first WD repeat. As described (A. R. Walker, P. A. Davison, A. C. Bolognesi-Winfield, C. M. James, N. Srinivasan, T. L. Blundell, J. J. Esch, M. D. Marks and J. C. Gray, unpublished results), the ttg-13 mutation is a deletion of greater than 4 kb encompassing the entire locus, the ttg-1 mutation creates a stop codon 25 amino acids from the C terminus of the WT protein, and the ttg-9 mutation causes a Ser to Phe change in the fourth WD repeat (Figure 3A). Transcript levels of the three ttg alleles sequenced in this study were examined by RNA blotting with a TTG probe (Figure 4). Steady state levels of RNA were essentially normal in the ttg-11 and ttg-12 alleles, while the level of mRNA was much lower in the ttg-10 allele.
—DNA sequence changes in ttg alleles. (A) Schematic representation of the locations of five ttg mutations within the TTG mRNA. Noncoding sequences are indicated by lines, and coding sequences are indicated by boxes. WD repeats are indicated as solid or shaded boxes. The positions of ttg-1 and ttg-9 are from the results of A. R. Walker, P. A. Davison, A. C. Bolognesi-Winfield, C. M. James, N. Srinivasan, T. L. Blundell, J. J. Esch, M. D. Marks and J. C. Gray (unpublished results). The ttg-13 allele is a deletion of the entire coding region (A. R. Walker, P. A. Davison, A. C. Bolognesi-Winfield, C. M. James, N. Srinivasan, T. L. Blundell, J. J. Esch, M. D. Marks and J. C. Gray, unpublished results). (B) DNA sequence alterations in ttg-10, ttg-11, and ttg-12. Noncoding sequences are in lowercase, protein coding sequences are in uppercase, and the bases affected by the mutation are in boldface. For coding regions, the singleletter amino acid code is given, and the residue changed by the mutation is numbered.
—Transcript levels of the ttg-10, ttg-11, and ttg-12 homozygotes. RNA gel blots loaded with 10 μg of total RNA per lane were probed with the entire TTG coding region. To control for loading, the blot was stripped and probed with a β-tubulin cDNA.
DISCUSSION
TTG encodes a member of a novel subfamily of WD-repeat proteins that is widely conserved among eukaryotes, including humans (A. R. Walker, P. A. Davison, A. C. Bolognesi-Winfield, C. M. James, N. Srinivasan, T. L. Blundell, J. J. Esch, M. D. Marks and J. C. Gray, unpublished results). The only other member of this subfamily with a genetically identified function is the AN11 gene of petunia (de Vettenet al. 1997). The Arabidopsis TTG gene acts as an important regulator in several diverse pathways, and understanding the role that TTG plays in these pathways should clarify our understanding of the role of related proteins in other organisms. However, the molecular basis of the pleiotropy of ttg mutations remains unclear. One possibility is that TTG encodes a protein with multiple independent functional domains. In this case, mutations affecting each function might be expected to cluster in different regions of the gene, as has been seen in other studies (Brennanet al. 1997). Alternatively, TTG may encode a protein with a single function that is needed by several pathways.
This study reports the genetic characterization of six ttg loss-of-function (LOF) alleles. The alleles include a null allele (ttg-13) and a weak allele (ttg-10) that reduces the ttg mRNA level without altering the protein sequence. Of the four mutations that would produce an altered protein, ttg-12 was a missense mutation upstream of the first WD repeat, ttg-11 was a missense mutation in the first WD repeat, ttg-9 was a missense mutation in the fourth WD repeat, and ttg-1 resulted in a C-terminal truncation of the protein downstream of the last WD repeat (Figure 3). Despite this diversity, all of these mutations affected all aspects of the ttg phenotype that were associated with the ttg-13 null allele. Intercrosses among the alleles demonstrate that these six alleles form a simple allelic series of hypomorphs, with the strongest alleles always severely affecting all aspects of the phenotype and the weakest alleles having moderate effects on all aspects of the phenotype. These results indicate that TTG probably does not encode a protein with independently mutable domains that are specific for subsets of the pathways in which it is involved.
The molecular basis of the ttg-10 allele is particularly significant. Nonsense codons that interfere with translation of a message, including nonsense codons resulting from an introduced upstream reading frame, often result in reduced mRNA levels due to a decrease in mRNA stability (Jacobson and Peltz 1996). Nonsense-mediated mRNA decay also occurs in plants (van Hoof and Green 1996). Although the AUG codon created in the RNA by this mutation is a poor match to the consensus sequence for known plant start codons (Joshiet al. 1997), it seems likely that the reduction in mRNA levels seen in Figure 4 is due at least in part to nonsense-mediated mRNA decay. We cannot at present rule out the possibility that the reduced mRNA level is the result of a defect in transcription caused either by the mutation that we have identified or by another mutation located outside of the region that we have sequenced. Nevertheless, because the sequence of the TTG open reading frame is unaltered by the ttg-10 mutation (Figure 3), it is clear that the phenotype of this allele is ultimately caused by a reduction in the level of TTG protein, not by an altered protein product.
Previous work had suggested that TTG and GL1 act at the same stage in the initiation of trichome development (Larkinet al. 1994a). However, this work involved the heterologous R gene from maize. Additionally, both R and the Arabidopsis GL1 gene were expressed from the strong constitutive 35S promoter from cauliflower mosaic virus. Constitutive overexpression of transcription factors can result in neomorphic phenotypes that may be hard to interpret. Additional evidence based on LOF mutations bearing on the relationship between TTG and GL1 during trichome development would clearly be of value.
Several types of genetic interaction often indicate that two genes function in the same pathway (Huffakeret al. 1987). These include intergenic noncomplementation by mutations in different genes and synthetic phenotypes observed when a double mutant exhibits a phenotype not seen with either individual mutant. Such genetic interactions often occur between genes whose products interact (Hayset al. 1989; Akadaet al. 1996). We have identified both of these types of interaction in combinations of LOF alleles of ttg and gl1. The increased trichome clustering in ttg/+ gl1-2/+ double heterozygotes is an example of intergenic noncomplementation (Figure 2C; Table 7). In addition, ttg-10 gl1-2 double mutant homozygotes have no trichomes, a synthetic phenotype that is more severe than the phenotype of either single mutant. Neither of these interactions is observed in crosses with the gl1-1 null allele, demonstrating that both interactions are allele specific with respect to gl1. The noncomplementation interaction is not allele specific with respect to ttg (Table 7). This lack of allele specificity is not surprising, given that ttg-10 most likely produces a WT protein, albeit at reduced levels. Similar synthetic phenotypes in double mutants and intergenic noncomplementation phenotypes have been seen in clavata1 and clavata3 mutants in Arabidopsis, although no allele specificity was observed in these interactions (Clarket al. 1995). The clavata genes have been proposed to function close together in the same pathway, and it is possible that their products interact.
Another aspect of the interaction between ttg and gl1 shows a similar specificity for the gl1-2 allele. Plants of the genotype ttg-10/ttg-10 gl1-2/+ produce no trichomes (Figure 2, D and E), even though ttg-10 plants have a readily detectable spiky phenotype and gl1-2 by itself is fully recessive. The gl1-1 null allele, which abrogates making of a protein product, is fully recessive in ttg-10 homozygous plants. The dominance of gl1-2 in ttg-10 homozygotes must be due to the product of the gl1-2 allele interfering with the function of the GL1+ allele. This only occurs when the level of TTG product is limiting, however, suggesting that the GL1-2 protein product is competing with WT GL1 protein for access to some TTG-dependent process.
As noted above, the gl1-2 mutation results in a protein with a C-terminal truncation removing 27 amino acids. The Myb DNA-binding domain of GL1 is located in the N-terminal region, and it was hypothesized that the region deleted in gl1-2 was part of a C-terminal transcription activation domain (Eschet al. 1994). Consistent with this possibility, the C-terminal portion of GL1 can activate transcription in yeast (A. Lloyd, personal communication). In some instances, known transcription factors or proteins resembling transcription factors can act as dominant-negative repressors if they lack a transcription activation domain (Keeganet al. 1986; Benezraet al. 1990; Riepinget al. 1994). The recently discovered CAPRICE gene of Arabidopsis is particularly relevant (Wadaet al. 1997). This gene encodes a protein containing a single MYB repeat related to GL1. There is no apparent transcription activation domain. Mutations in caprice reduce the number of root hairs. When CAPRICE is expressed constitutively, excess root hairs are present and trichomes are absent on the leaves. The root hair developmental pathway is also a TTG-dependent pathway in which TTG acts to suppress root hair development. It is possible that CAPRICE and gl1-2, both of which encode MYB-related proteins that lack an activation domain, interfere with trichome development by a similar mechanism.
Recent work on the regulation of anthocyanin pigment biosynthesis in petunia also provides some insight into the mechanisms that may underlie these interactions. In petunia, anthocyanin biosynthesis is regulated by the TTG homologue an11, the GL1 homologue an2, and the R-related bHLH genes an1 and jaf13 (de Vettenet al. 1997; Quattrocchioet al. 1998). The regulation of Arabidopsis trichome development may thus closely resemble the regulation of petunia anthocyanin biosynthesis, with a WD-repeat protein, a MYB protein, and an as yet unidentified R-like bHLH protein. The AN11 protein appears to be located in the cytoplasm and may play a role in activating the AN2 MYB transcription factor (de Vettenet al. 1997).
—Model of TTG-GL1 interactions. (A) Wild type. TTG is present in excess. TTG activates GL1+ in the cytoplasm, and the activated GL1+ enters the nucleus and stimulates transcription of downstream genes. Activation is indicated by *. (B) gl1-2/+. TTG is present in excess. TTG activates both the GL1+ and gl1-2 proteins, but sufficient GL1+ is activated to promote downstream transcription. (C) ttg-10/ttg-10 gl1-2/+. TTG protein is limiting. GL1+ and gl1-2 compete for activation by TTG, and insufficient GL1+ is activated to promote downstream transcription. Additional competition could occur at downstream promoters.
A model consistent with our results is illustrated in Figure 5. TTG is assumed to encode a cytoplasmic protein that activates the GL1 protein outside of the nucleus, which then enters the nucleus and activates transcription of genes downstream in the trichome development pathway (Figure 5A). In plants heterozygous for the gl1-2 mutant allele, there is an excess of TTG, and sufficient GL1+ is activated to induce transcription of downstream genes (Figure 5B). In ttg-10/ttg-10 gl1-2/+ plants, the level of functional TTG protein produced is limiting, and GL1+ and gl1-2 proteins compete for activation by TTG, resulting in reduced transcription of downstream genes. Although a direct interaction between TTG and GL1 is the simplest explanation, we cannot rule out the possibility that GL1 interacts with some factor downstream of TTG in the pathway. The increase in trichome clustering seen in ttg-10/+ gl1-2/+ plants can be explained if the lateral inhibition pathway is more sensitive than the trichome initiation pathway to reductions in TTG levels.
In Figure 5, TTG is shown activating GL1 alone for simplicity. However, it is likely that GL1 functions together with an R-like protein and that these proteins could be activated as a heterodimer. In support of this hypothesis, Szymanski et al. (1998) have shown that GL1 physically interacts with the maize R protein in in vitro translation reactions and that R-like genes have been identified in Arabidopsis (A. R. Walker, unpublished results; A. Lloyd, personal communication). The proposed activation of GL1 could occur through a variety of different mechanisms, including post-translational modifications such as phosphorylation, regulation of transport of GL1 into the nucleus, or assembly of GL1 into a complex with other proteins such as an R homologue. Because WD repeat proteins never have been found to act as enzymes, but are often involved in protein-protein interactions, TTG may act on GL1 in conjunction with other proteins. In previous work, the inability of constitutively-expressed GL1 to bypass a ttg mutation was interpreted as evidence that GL1 acts either at the same step or in parallel to TTG. The current model, placing GL1 downstream of TTG but dependent on TTG for activation, is consistent with the previous data (Larkinet al. 1994a). The ability of R to bypass the requirement for TTG may be due to the use of a heterologous gene or due to the high level of expression from the 35S promoter used in these experiments. These issues cannot be resolved without the analysis of Arabidopsis R homologues that clearly function in trichome development.
The phenotype of 35SGL1 plants remains difficult to explain. As described previously (Larkinet al. 1994a), WT plants carrying a 35SGL1 transgene have a reduced number of trichomes (Table 6). One model proposed to explain this observation is that excess GL1 titrates some essential factor via a “squelching” process (Larkinet al. 1994a). Alternatively, excess GL1 that is not activated by TTG may act as a repressor. It is difficult, however, to reconcile these hypotheses with the phenotype of ttg/+ 35SGL1 plants, which have an increased number of trichomes and a high frequency of trichome clusters (Larkinet al. 1994a; and Table 6). The observations presented here, demonstrating that ttg/+ try/+ double heterozygotes show increased trichome clustering, suggest that TRY acts at a point very close to TTG in the trichome differentiation pathway. One possible explanation is that TRY is under the direct control of TTG. TRY has been proposed to be an antagonist of GL1 function (Schnittgeret al. 1998). Perhaps TTG activates both GL1, a positive regulator of trichome development, and TRY, an inhibitor of trichome development. Such an arrangement could result in the extreme sensitivity to dosage of TTG that has been observed. It is also worth noting that dosage sensitivity can sometimes be a hallmark of direct protein-protein interactions (Meeks-Wagner and Hartwell 1986; Burkeet al. 1989), as we have proposed in this study for TTG and GL1.
The pathway for the differentiation of the trichome cell type has revealed a great deal of complexity. The present work suggests some hypotheses about biochemical interactions between components of the trichome development pathway. The recent isolation of TTG will allow these hypotheses to be tested in the near future. However, a complete understanding of the mechanisms behind this cell fate decision will require a detailed analysis of additional genes, particularly TRY and the hypothesized R-like genes.
Acknowledgments
The authors acknowledge Joy Chien, Martin Hülskamp, M. David Marks, and Alan Sessions for donating seed stocks, Olga Borkhsenious for scanning electron microscopy, and Ron Bouchard and Marcia Duggan for assistance with figures. We also acknowledge Jim Moroney, David Oppenheimer, Mary Pollock, Bryan Rogers, and members of the Larkin Lab for critical reading of the manuscript. J. Larkin and J. Walker were supported in part by National Science Foundation grant IBN-9728047 to J. Larkin and Louisiana Board of Regents grant LEQSF-RCS (1997-00)-RD-A-04 to J. Larkin. A.R.W. was supported by BBSRC (GER00618) and A.C.B.-W. was supported by a BBSRC Research Studentship.
Footnotes
-
Communicating editor: J. Chory
- Received November 10, 1998.
- Accepted January 7, 1999.
- Copyright © 1999 by the Genetics Society of America
