The TCP domain: a motif found in proteins regulating plant growth and development


*For correspondence (fax +44 1603 456844; e-mail


The cycloidea ( cyc) and teosinte branched 1 ( tb1) genes code for structurally related proteins implicated in the evolution of key morphological traits. However, the biochemical function of CYC and TB1 proteins remains to be demonstrated. To address this problem, we have analysed the predicted secondary structure of regions conserved between CYC and TB1, and looked for related proteins of known function. One of the conserved regions is predicted to form a non-canonical basic-Helix-Loop-Helix (bHLH) structure. This domain is also found in two rice DNA-binding proteins, PCF1 and PCF2, where it has been shown to be involved in DNA-binding and dimerization. This indicates that the conserved domain most probably defines a new family of transcription factors, which we have termed the TCP family after its first characterised members (TB1, CYC and PCFs). Other plant proteins of unknown function also belong to this family. We have studied two of these in Arabidopsis and have shown that they are expressed in rapidly growing floral primordia. This, together with the proposed involvement of cyc and tb1 in influencing meristem growth, suggests that many members of the TCP family may affect cell division. Some of these genes may have been recruited during plant evolution to generate new morphological traits.


The cycloidea (cyc) and teosinte branched 1 (tb1) genes have been implicated in the evolution of key morphological traits. The cyc gene is involved in the control of floral symmetry, a character that has changed many times during plant evolution ( Carpenter & Coen 1990; Luo et al. 1996 ; Stebbins 1974). The tb1 gene controls developmental switches that contributed to the evolution of maize from its wild ancestor teosinte ( Doebley et al. 1995 ; Doebley et al. 1997 ). Although both cyc and tb1 have been isolated, the biochemical function of their encoded proteins is unclear ( Doebley et al. 1997 ; Luo et al. 1996 ). To address this problem, we have analysed the predicted secondary structure of these and related proteins, and compared some of the gene expression patterns.

The cyc gene is required, together with a related gene, dichotoma (dich), to establish dorsoventral asymmetry of the Antirrhinum flower. In flowers mutant for both cyc and dich, differences between dorsal, lateral and ventral organs are eliminated, rendering the flower radially symmetrical. The cyc gene is expressed in the dorsal region of wild-type floral meristems, from very early through to later stages of development. The initial activity reduces the growth rate in the dorsal region of the wild-type meristem and controls primordium initiation. Late expression prevents the full development of the dorsal stamen and affects the asymmetry, size and cell types of the dorsal and lateral petals ( Luo et al. 1996 ).

The tb1 gene affects the fate of maize axillary meristems; at lower nodes it prevents the outgrowth of buds and at upper nodes it promotes the development of female inflorescences (ears). In tb1 mutants of maize, axillary buds of lower nodes grow out to give basal branches (tillers), and the buds of upper nodes give branches tipped with male inflorecences (tassels), a phenotype reminiscent of the ancestor of maize, teosinte ( Doebley et al. 1997 ).

Although the processes controlled by cyc and tb1 appear to be unrelated, there are some common themes. First, both genes are involved in the development of axillary structures, either flowers or branches. Second, both genes affect petals and stamens, organs whose development is regulated by the B class of floral organ identity genes ( Coen & Meyerowitz 1991). Third, both genes have been proposed to function, at least in part, as modifiers of organ growth.

To understand the biological role of cyc and tb1, we have investigated the possible functions of their proteins. In particular, we have studied regions conserved between cyc and tb1 that may act as functional domains. The predicted secondary structure of one of the regions is a basic-Helix-Loop-Helix (bHLH). This region is unrelated to the bHLH structure found in canonical bHLH transcription factors ( Murre et al. 1989 ), but is closely related to a bHLH domain found in two rice DNA binding proteins, PCF1 and PCF2 ( Kosugi & Ohashi 1997). Based on this homology we define a new class of proteins, the TCP family, that most likely act as transcription factors. We have also characterised two further members of the TCP family from Arabidopsis and have shown that they are expressed in floral organs undergoing rapid growth. Taken together with the phenotypic effects of cyc and tb1, this suggests that members of the TCP family may influence cell division and growth. Recruitment of some of these genes to play new roles during plant development may underlie some key morphological changes during angiosperm evolution.


CYC and TB1 contain a basic-Helix-Loop-Helix domain

To investigate the biochemical action of the CYC and TB1 proteins, we analysed their predicted secondary structures. In particular, we studied the regions showing sequence conservation, as these might constitute functional domains.

The first conserved region is predicted to form a basic-Helix-Loop-Helix (bHLH, Figs 1a and 2a). This bHLH domain is defined by structural criteria and is unrelated in sequence to the bHLH domain found in MyoD, E12 and related proteins ( Murre et al. 1989 ). The basic region of the bHLH domain of both CYC and TB1 is 21 residues long and includes a putative bipartite nuclear localisation signal (NLS) ( Dingwall & Laskey 1991; Doebley et al. 1997 ; Luo et al. 1996 ). The helical regions are amphipathic and comprise alternating conserved hydrophobic residues and partially conserved hydrophilic residues ( Fig. 1b). The second helix contains a LXXLL-motif (indicated in Fig. 2a), which has been shown to mediate binding of transcriptional co-activators to liganded nuclear receptors in animals ( Heery et al. 1997 ). Helix II also contains three potential sites of phosphorylation (serine or threonine), two of them in conserved positions. The region linking the two helices has conserved glycine, aspartate and serine residues found with high frequency in loops ( Lesczynski & Rose 1986), as well as proline in the case of CYC. In addition to the bHLH domain, CYC and TB1 have a second conserved region termed the R-domain ( Fig. 2b), which is rich in polar residues (arginine, lysine and glutamic acid). The R-domain is predicted to form a hydrophilic α-helix (not shown). Both the bHLH and R domains are also predicted to form coiled coils, similar to those formed by leucine zippers ( Lupas et al. 1991 ; Lupas 1996).

Figure 1.

Predicted secondary structure of the CYC/TB1 bHLH domain.

(a) Plot representing secondary structure probabilities of the CYC bHLH domain based on DSM (Discrete State-Space Models) predictions. The consensus for CYC and TB1 is shown below the plot. The probabilities of each residue being in each of the states (turn, helix or loop) are depicted using contour lines of constant probability, in increments of 0.1. The strand state is improbable for the CYC bHLH domain (not shown).

(b) Helical wheels representing the residues from proposed helix I and helix II. Divergence angle between adjacent residues being 100°. Conserved hydrophobic residues are represented by white letters inside black circles. Conserved hydrophilic residues are represented by black letters inside white circles. Empty circles correspond to positions with non-conserved residues. Note that the conserved hydrophobic residues are at one side of the helix and conserved hydrophilic residues mainly at the opposite side. The grey circle corresponds to a tryptophan residue on the hydrophilic side. Ψ stands for any hydrophobic residue.

Figure 2.

The TCP gene family.

(a) Alignment of the predicted amino acid sequence from members of the TCP family. Only the TCP domain is shown and founder members of the TCP family are indicated in bold. Identical amino acids are in black boxes, while amino acids with similar charge or hydrophobicity are in grey. Circles along the top indicate conserved basic residues; underlined circles indicate residues forming part of the putative bipartite NLS; asterisks indicate conserved hydrophobic residues in the helices; underlined asterisks indicate the LXXLL motif; black arrowheads point to residues (glycine or proline) that disrupt α-helix formation. The variable predicted length of helix II is indicated by a dotted line.

(b) Alignment of the R-domain in the CYC/TB1 subfamily. Note that TCP1 has several glycines in these region that are likely to prevent the formation of a helix.

(c) Consensus tree showing relationship between various TCP proteins. The percentage of bootstrap samples in which particular clades were monophyletic is indicated for those clades with support of 70% or more.

Amino acids are expressed in the standard single letter code; Ψ stands for any hydrophobic residue. In the PCF subfamily the N-terminus of the bipartite domain lies outside the conserved region ( Kosugi & Ohashi 1997). The predicted length of the helices varies slightly for each member of the TCP family. Note that the LXXLL motif is not fully conserved, however, the VXXLL sequence found in TB1 has also been shown to tolerate interactions with liganded nuclear receptors ( Heery et al. 1997 ). TCP1 (AC002130) corresponds to gene FIN21.14, BAC FIN21, chromosome 1. TCP2 (AF072691) corresponds to cDNA clone 161 l4T7 (EST R30409). TCP3 (AF072134) corresponds to cDNA clone 75H5T7 (EST T45419). TCP4 corresponds to the maize clone php20581. TCP5 corresponds to an ORF found in the genomic clone MSL3 (AB008269), chromosome 5. TCP6 corresponds to an ORF found in the genomic clone MEE6 (AB010072). TCP7 (N97289) is an EST from cDNA clone 241F10T7. TCP8 (H36511) is an EST from cDNA clone 174B16T7. TCP9 (T88453) is an EST from cDNA clone 156 J1T7. Only non-redundant ESTs or BACs have been used for the alignment.

The TCP domain

When CYC and TB1 were first described, no sequence similarity to proteins of known biochemical function was found. However, more recent database searches revealed two additional proteins with similarity to CYC and TB1 in the bHLH region: PCF1 and PCF2. These proteins were isolated on the basis of their ability to bind specifically to promoter elements of the rice gene for the proliferating cell nuclear antigen, PCNA, a protein involved in DNA replication and cell cycle control (for a review see Zophonías & Hübscher 1997 ). The PCF proteins may bind to DNA as homo- or heterodimers. A region containing the bHLH domain has been shown to be sufficient for DNA binding and necessary for dimerization ( Kosugi & Ohashi 1997). This suggests that CYC and TB1 may also function as DNA-binding proteins and may interact with other proteins through their bHLH domain.

In addition to CYC, TB1 and the PCFs, several sequences of unknown function from Arabidopsis and maize show homology to the bHLH domain ( Fig. 2a; Doebley et al. 1997 ; Kosugi & Ohashi 1997; Luo et al. 1996 ). Alignment of these protein sequences shows a high conservation of key residues in the bHLH domain: two short stretches of residues in the basic region, hydrophobic residues along the apolar face of both α-helices, a tryptophan in helix II and a helix-breaking glycine in the loop between the helices ( Fig. 2a). However, the residues in the loop and the hydrophilic residues of the helices are not as well conserved.

The alignment shows that the proteins form two subfamilies: one closely related to CYC and TB1, and another more related to the PCFs. The basic region of the CYC/TB1 subfamily contains a bipartite NLS ( Dingwall & Laskey 1991; Doebley et al. 1997 ; Luo et al. 1996 ) while the basic region of the PCF subfamily contains only a portion of a bipartite NLS. Although largely conserved within each subfamily, the residue composition of the loop is different between the two subfamilies. Helix II of the CYC/TB1 subfamily is longer than that of the PCF subfamily, which is interrupted after nine residues by a proline ( Fig. 2a). Each subfamily shares specific regions outside the bHLH domain: members of the PCF subfamily share regions adjacent to the bHLH domain ( Kosugi & Ohashi 1997), while some members of the CYC/TB1 subfamily share an R-domain located at various distances from the TCP domain ( Fig. 2b,D. Luo, M. Chadwick and P. Cubas, unpublished results). The tree obtained from this alignment confirms that the proteins fall into two clusters ( Fig. 2c).

Thus, CYC, TB1 and the PCFs belong to a large family of proteins sharing a common motif that we propose to call the ′TCP domain’, based on the initial letters of the founding members (TB1, CYC, PCF). According to the evidence presented for the PCFs ( Kosugi & Ohashi 1997), the TCP domain is probably involved in DNA binding and protein–protein interactions.

Role of TCP proteins in meristem proliferation

A common feature of the TCP proteins characterised thus far is a role in meristem growth: CYC controls the growth of the floral meristems and primordia; TB1 affects axillary meristem growth; PCF1 and PCF2 bind to the promoter of a gene (PCNA) involved in meristematic cell division. To investigate whether this correlation with growth held for other members of the TCP family, we analysed two Arabidopsis ESTs, here termed TCP2 and TCP3, that are closely related to CYC and TB1. If these were involved in cell growth and proliferation, they might be expected to be expressed in dividing tissue. The TCP2 and TCP3 cDNAs were sequenced ( Fig. 3a,b; GenBank accession numbers AF072691 and AF072134, respectively), genetically mapped ( Fig. 3c) and their expression patterns characterised by in situ hybridisation.

Figure 3.

Complete sequence and genetic map positions of TCP genes in Arabidopsis.

Sequence of cDNAs for TCP2 (a) and TCP3 (b) with predicted amino acid sequence in standard one-letter code and conserved residues in the bHLH and R domains highlighted in grey. (c) Map positions for four TCP genes. In addition to the positions of TCP2 (chromosome 4) and TCP3 (chromosome 1), two more putative TCP genes (T45419b and T45419c) that show weak hybridisation to TCP3 were mapped in chromosome 2 and chromosome 3.

Sections of plant tissue were collected and embedded 4, 6, 8, 10, 12, 16 and 22 days after sowing, and hybridised with digoxigenin-labelled probes from TCP2 and TCP3 templates. The transcription patterns of the two genes were qualitatively very similar. At 4–8 days, when the apical meristem was producing vegetative leaf primordia, there was no detectable mRNA from TCP2 and TCP3. After 8 days, the apical meristem started producing flower primordia. At this time, weak expression of both genes was detected in vegetative primordia (not shown).

Transcripts were also detectable from the earliest stages of flower development ( Smyth et al. 1990 ). At stages 1–2, the floral meristem was dome-shaped and the signal was diffuse throughout ( Fig. 4a,e). From stage 3, when the floral organ primordia started to form, expression patterns were monitored separately for each of the four whorls. (i) In sepals, mRNA accumulation was highest at stages 3–4, during which sepal primordia grew to enclose the bud ( Fig. 4a,e). After this, the signal progressively decayed. (ii) In petals, weak signal was detected when the petal primordia first became visible (stages 5–6, not shown), and peaked during stages 8–12 when they were growing rapidly ( Fig. 4c,d,h). During this time, most signal was observed at the tips and the margins of the petal primordia ( Fig. 4h). Dorsal (adaxial) and ventral (abaxial) petals showed similar levels of expression. (3) In the stamens, signal increased during stages 7–8 when the stamen primordia were growing rapidly and anthers were starting to become distinguishable ( Fig. 4b). By stages 8–9 signal was restricted to the developing anthers, being mainly detectable in the pollen sacs where microspore mother cells were undergoing meiosis ( Fig. 4c,g). Transcript levels were similar in dorsal (adaxial), lateral and ventral (abaxial) stamens. By stage 10, when pollen grains were mature, transcripts were no longer detectable in stamens ( Fig. 4c). (4) In carpels, signal was detected in the placental tissue during stage 9 when ovule primordia were forming ( Fig. 4g). The expression pattern in roots was not analysed.

Figure 4.

Expression patterns of TCP2 and TCP3 during Arabidopsis floral development.

RNA in situ hybridisation of wild-type Arabidopsis sections probed with TCP2 (a–d) or TCP3 (e–h). (a,b,e,f,h) show longitudinal sections; (c,d,g) are transverse sections. (a) Stage 2–3 floral meristem, when sepal primordia are yet not visible. TCP2 mRNA accumulates at the positions where adaxial (right) and abaxial (left) sepal primordia will form (arrowheads), although some signal is detected throughout. (b) Stage 7–8 floral meristem when petals are small but the stamens are growing rapidly. Signal accumulates in the anther region. (c) Stage 8–9 floral meristem. Signal accumulated in the pollen sacs where pollen grains are forming and in the growing petals. (d) Stage 10 flower. In the anthers, the pollen grains are mature and TCP2 mRNA is no longer detectable. The mRNA accumulates in the growing petals. (e) Young floral meristems of stages 2 and 3 when TCP3 signal is diffuse throughout. (f) Stage 4 the signal accumulated in the growing sepal primordia. (g) Stage 9 flower meristem. Notice that the expression pattern of TCP3 is similar to that of TCP2 (c), although TCP3 is expressed at higher levels in the carpel primordia. In the petal primordia the TCP3 signal is stronger at the margins. (h) Stage 10 petal primordia. TCP3 mRNA mainly accumulates at the margins of the growing petal primordia. The section is parallel to the plane of the petal primordium, i.e. diagonal to the flower bud. Controls using TCP2 or TCP3 sense probes gave no signal (not shown). s, sepal; p, petal; st, stamen; c, carpel. The numbers indicate developmental stages according to Smyth et al. (1990) .

In summary, the Arabidopsis TCP2 and TCP3 genes are most strongly expressed during flower development; expression being highest in petal and stamen primordia, but also being detectable in sepals and carpels. Expression coincides with the stages when primordia are growing rapidly, consistent with a role for these genes in primordial growth, but by no means conclusive as many other functions may also be compatible with such expression patterns. Unlike cyc, there is no distinction in expression levels between dorsal and ventral primordia.


We have shown that CYC and TB1 belong to a family of proteins sharing a common region, the TCP domain, that is predicted to form a basic-Helix-Loop-Helix (bHLH) structure. This region is unrelated in sequence to the canonical bHLH domain found in transcription factors such as MyoD ( Murre et al. 1989 ). However, it is similar to the bHLH domain found in PCF1 and PCF2, plant DNA-binding proteins that most probably act as transcription factors ( Kosugi & Ohashi 1997). The main conserved features of the TCP domain are: two short stretches of residues in the basic region, hydrophobic residues along the apolar face of both α-helices, a tryptophan in helix II, and a helix-breaking glycine in the loop between the helices. So far, members of the TCP family have only been found in plants.

What is the biochemical function of the TCP domain? Important clues have been obtained from the analysis of PCF1 and PCF2. In these proteins, the basic region of the TCP domain is necessary for specific binding to promoter elements of the PCNA gene ( Kosugi & Ohashi 1997). Basic regions have also been shown to be involved in DNA binding in the case of bHLH, bHLHZ and bZIP proteins ( Hurst 1994; Littlewood & Evans 1994). In these transcription factors, the basic domain adopts an α-helical structure that interacts with the major groove of the DNA. In contrast, the basic region of the TCP domain contains residues that prevent helix formation, suggesting that in this case the basic region binds to DNA through a different mechanism ( Kosugi & Ohashi 1997). It is possible that other TCP proteins bind DNA through their basic domain in a similar way. In addition, the basic region of the TCP domain may target these proteins into the nucleus as it contains a complete (CYC/TB1 subfamily) or partial (PCF subfamily) bipartite NLS ( Dingwall & Laskey 1991; Doebley et al. 1997 ; Luo et al. 1996 ). This would fit the observation that NLSs often overlap or flank nucleic acid-binding domains ( LaCasse & Lefebvre 1995; Littlewood & Evans 1994)

The role of the HLH region of the PCFs and other TCP proteins is less clear. One possibility is that the amphipathic helices mediate protein–protein interactions through their hydrophobic surfaces. This would be similar to the proposed role of the amphipathic helix (K domain) of MADS box genes ( Davies & Schwarz-Sommer 1994; Shore & Sharrocks 1995). Amphipathic helices also mediate homo- and heterodimerization in bZIP and bHLH proteins of the MyoD type ( Landschulz et al. 1988 ; Murre et al. 1989 ). In the PCFs, a region containing the TCP domain is essential for homo- and heterodimerization, although the role of the helices has not been tested ( Kosugi & Ohashi 1997). Amphipathic helices might also be involved in interactions with non-TCP proteins. For instance, helix II of the TCP domain contains a conserved sequence that resembles the LXXLL motif shown to be involved in protein interactions with liganded nuclear receptors ( Heery et al. 1997 ). It is possible that TCP proteins interact with as yet unidentified plant nuclear receptors or other proteins through this sequence.

Most members of the TCP family contain up to three potential phosphorylation sites in serine and threonine in the basic domain and helix II. Phosphorylation has been shown to affect nuclear localisation, DNA binding and transcriptional activation of regulatory proteins ( Hunter & Karin 1992), raising the possibility that the activity of the TCP proteins might be regulated by a similar mechanism.

The TCP proteins fall into two subfamilies (one including CYC and TB1 and the other including the PCFs) based on features both within and outside the TCP domain. Within the TCP domain, each subfamily has a different linker for the bipartite NLS, a distinct residue composition in the loop and hydrophilic faces of the helices, and a different length for helix II. Outside the TCP domain, most members of the CYC/TB1 subfamily have an R-domain, predicted to form a coiled coil that may mediate protein–protein interactions ( Lupas et al. 1991 ), and all members of the PCF subfamily share short regions flanking the TCP domain. These differences between subfamilies may reflect differences in the DNA-binding specificities and/or protein–protein interactions.

Unlike the homeo-domain containing genes of metazoans, many of which occur in tandem clusters in the genome ( Holland et al. 1994 ), the Arabidopsis TCP genes are dispersed throughout the genome. In this respect they resemble the MADS box family of higher plants, which are also dispersed ( Hauge et al. 1993 ; Rounsley et al. 1995 ).

All the members of the TCP family investigated thus far function in processes related to cell proliferation. CYC retards growth of the dorsal region of young floral meristems, affecting the number of primordia initiated. Later, it arrests dorsal stamen development and promotes dorsal petal growth ( Luo et al. 1996 ). TB1 is involved in arresting growth of some axillary buds, repressing internode growth in branches, and arresting petal (lodicule) and stamen development in the female flowers ( Doebley et al. 1995 ; Doebley et al. 1997 ). PCF1 and PCF2 most probably control the transcription of PCNA, a gene expressed only in meristematic tissue where it is involved in DNA replication and cell cycle control. Here we show that the expression patterns of two Arabidopsis members of the CYC/TB1 subfamily, TCP2 and TCP3, also correlate with actively dividing regions of the floral meristem, suggesting a possible involvement of these genes in regulating cell growth and division. However, such expression patterns may also be compatible with other biological roles. Further experiments, such as inactivation of the TCP2 and TCP3 genes, will be needed to determine whether these genes are indeed involved in regulating growth.

The TCP2 and TCP3 genes are most probably not orthologues of CYC or TB1, as CYC and TB1 are more closely related to each other than to either of these genes ( Fig. 2c). However, they do share with CYC and TB1 some features in their expression patterns: the TCP2 and TCP3 genes are upregulated in petal and stamen primordia, similar to cyc; the tb1 gene is also likely to be expressed in these organ primordia as tb1 mutants affect the development of petals and stamens. One distinctive feature, however, is that cyc is expressed only in dorsal primordia whereas TCP2 and TCP3 are expressed uniformly along the dorsoventral axis.

It is possible that many members of the TCP family function in proliferating tissues where they may act in combination with other proteins to influence cell division and growth, and perhaps recruitment of some of these regulatory genes for new developmental functions has been involved in generating key changes in plant morphology during angiosperm evolution.

Experimental procedures

Secondary structure analysis

The secondary structure prediction for CYC was obtained by probabilistic Discrete State-Space Models analysis, DSMs ( Stultz et al. 1993 ; White et al. 1994 ), submitting the protein sequences to the Protein Sequence Analysis (PSA) server ( Helical wheels were plotted with PROTEAN (DNASTAR for Windows 3.10a) and refined by hand. The prediction of coiled-coil regions ( Lupas et al. 1991 ) was obtained by submitting the protein sequences to the Network Protein Sequence analysis server (

Multiple sequence comparison

The TCP and R-domain alignments were constructed with MEGALIGN (DNASTAR for Windows 3.10a), using CLUSTALW ( Thompson et al. 1994 ) and PAM250 weighted distances and were manually refined. Phylogenetic analysis was performed with the phylip package ( Felsenstein 1985). Two hundred bootstrap resamplings of the original data were generated with the SEQBOOT program. Distance matrices were made for each bootstrap dataset using the PRODIST program with the Dayhoff PAM distance method, and 200 trees constructed from these by the Neighbor-joining method.

In situ hybridisations

Plant material was grown under long days and collected for in situ hybridisation as described by Ratcliffe et al. (1998) . Digoxigenin labelling of RNA probes, tissue preparation and in situ hybridisation were done as described by Coen et al. (1990) .


Plasmid clones of the R30409 and T45419 ESTs (corresponding to TCP2 and TCP3, respectively) were obtained from the ABRC at Ohio University. These clones were completely sequenced at the Advanced Genetic Analysis Center (University of Minnesota) using gene specific internal oligonucleotide primers.

Genetic mapping

Four TCP genes were placed on the Arabidopsis genetic map using the R30409 (TCP2) and T45419 (TCP3) EST clones as probes on DNA blots of a set of 99 recombinant inbred lines (ABRC Stock Number CS1899). For R30409, there was one strongly hybridising locus that corresponds to TCP2 and maps to chromosome 4 between markers g3845 and m600. For T45419, there was one strongly hybridising locus that corresponds to TCP3 and maps to chromosome 1 between markers m213 and nga128. T45419 also hybridised to two other loci more faintly: one between markers m283 and nga361 on chromosome 2 and the other between markers EW18E10 l and nga162 on chromosome 3. In addition, TCP1 (AC002130) maps in chromosome 1, TCP5 (AB008269), TCP6 (AB010072) and TCP7 (AB007648) in chromosome 5, and TCP8 (H36511) and TCP9 (AC003680) in chromosome 2. DNA isolation and DNA blot analysis were performed as described by Doebley & Stec (1993). Linkage maps were assembled using Mapmaker 2.0 ( Lander et al. 1987 ).


We thank Oliver Ratcliffe for providing the Arabidopsis material for in situ hybridisation. Thanks also to Rosemary Carpenter, Desmond Bradley, Oliver Ratcliffe, Utpal Nath and Sandra Doyle for comments on the manuscript and Coral Vincent and Theresa Warr for help with the final version of the manuscript.

GenBank accession numbers AF072691 and AF072134.