Isolation of COV1, a gene involved in the regulation of vascular patterning in the stem of Arabidopsis

The vascular system in plants forms the structural architecture required to transport solutes and water. During primary vascular development, the pattern of vascular tissue is first visualised by the appearance of the procambial cells. These procambial cells subsequently differentiate into both the xylem and phloem that constitute each vascular bundle. Proper spatial arrangement of these vascular strands within the plant is essential for normal growth and development. The development of the vascular bundles occurs in a highly ordered and predictable pattern. However, the molecular regulation of this process remains poorly characterised.

In broad flat organs, such as leaves, the vascular tissue consists of an interconnected network that forms during the course of leaf development. By contrast, the pattern of primary vascular development in the stem is laid down in the region immediately below the apical meristem. The differentiation of vascular tissue is closely linked to the activity of the meristem. Early experiments demonstrated that the inductive effects of leaves and organ primordia on vascular development could be replaced by the application of auxin (Jacobs, 1952; Young, 1953). Many subsequent experiments have demonstrated the effect of auxin on both the induction and pattern of vascular tissue (for reviews, see Bennett et al., 1998; Berleth and Mattsson, 2000; Ye, 2002). A large number of experiments investigating vascular tissue induction by auxin were rationalised by Sachs (Sachs, 1981; Sachs, 1991), who suggested that the positive effects of high levels of auxin on the capacity for polar auxin transport (PAT) generate a positive feedback mechanism leading to the canalisation of auxin flow. The high levels of auxin in these regions result in vascular differentiation. Identification of the PIN1 protein has provided evidence to support the canalisation hypothesis. PIN1 is essential for PAT,and is normally localised on the basal membrane of the cell(Galweiler et al., 1998). During embryo development, transition from a uniform localisation of PIN to an orientated localisation occurs along the apical-basal axis, which is consistent with a positive feedback loop for auxin(Steinmann et al., 1999). Furthermore, disrupted localisation of PIN1 in emb30/gnommutants results in a proliferation of vascular tissue that is consistent with auxin delivery to inappropriate cell types(Koizumi et al., 2000; Steinmann et al., 1999).

A large number of experiments by Sachs suggested that auxin may also play a role in the inhibition of vascular bundle development. Pre-existing vascular bundles with high rates of auxin transport inhibited the development of new vascular strands in their immediate vicinity(Sachs, 1981; Sachs, 1991). Interestingly,recent experiments by Reinhardt et al.(Reinhardt et al., 2000)suggest that auxin is involved in both inductive and inhibitory affects in organ formation. Although auxin is able to induce primordia formation, once formed, the region adjacent to these primordia becomes insensitive to further auxin application. The mechanism by which auxin, or some other signalling molecule from the developing primordia, exerts this inhibitory effect remains unclear.

Isolation of mutants that contain islands of vascular tissue that are not connected to the existing vascular tissue are harder to reconcile with the canalisation hypothesis (Carland et al.,1999; Deyholos et al.,2000; Koizumi et al.,2000). Some of these mutants have normal responses to auxin and may be explained more easily by the diffusion-reaction pre-pattern hypothesis. Based on computer modelling, the diffusion-reaction hypothesis proposes the interaction of two diffusible substances with varying diffusion rates that regulate the autonomous formation of veins in an initially uniform field. One of these substances promotes the induction of vascular tissue, whereas the other results in inhibition of vascular development(Koch and Meinhardt, 1994). Although there is a good deal of evidence to support the inductive effects of auxin on vascular development, there is little evidence on what the inhibitory element might be.

In wild-type Arabidopsis stems, the pattern of vascular bundles is very ordered with five to eight bundles developing around the stem, separated by the interfascicular regions. However, the molecular mechanisms that control this patterning remain unclear. There are a limited number of mutants available that show altered vascular patterning in the stem. The amphivasal vascular bundle (avb) mutant(Zhong et al., 1999) has an increased number of vascular bundles in the stem, many of which are abnormally localised in the pith. The vascular bundles also show an amphivasal pattern with the phloem surrounded by xylem. Another mutant, acaulis5(acl5) (Hanzawa et al.,1997), has vascular bundles that are surrounded by an abnormally extended sheath of cells with thickened walls. Although the number of veins in acl5 plants is similar in wild-type plants, the amount of differentiated vascular tissue within the vascular bundles appears to be greater in the mutant.

In this study, we report the cloning of a gene involved in maintaining the ordered patterning of vascular bundles within the stem of Arabidopsis. Mutations in this gene lead to disruptions in the patterning of differentiated vascular tissue in the stem, resulting in a continuous ring-like pattern of xylem and phloem, with very little interfascicular tissue and the loss of defined vascular bundles. Based on the phenotype of the vascular tissue in the stem, this mutant was designated continuous vascular (cov1). COV1 is predicted to be an integral membrane protein that may be involved in the perception or transport of a signalling molecule that negatively regulates the differentiation of vascular tissue in the developing stem of Arabidopsis.

Seeds were sterilised by washing for 10 minutes in sterile water containing 10% bleach and 0.01% Triton X-100, followed by five washes with sterile water. The seeds were then imbibed in 0.1% agar for 48 hours at 4°C. Seeds were germinated on plates containing Murashige and Skoog medium (MS) (4.4 g/l)(Sigma), MES buffer (0.5 g/l) (Sigma) and 1% agar. Plates were placed in a growth cabinet under continuous light at 22°C. Once the seedlings had germinated, they were transferred to pots containing a commercial compost/perlite/vermiculite mixture and grown under continuous light at 22°C in growth chambers. Mutant alleles were identified from a screen of ethyl methane sulphonate (EMS)-mutated M2 seed of wild-type Ler. Scoring was performed by taking hand-cut cross-sections of the base of stems from 4-week-old plants. Sections were stained either with 0.05% Aniline Blue(0.67 M phosphate buffer; pH 8.0) and observed using UV microscopy, or with 0.02% Toluidine Blue and observed using brightfield microscopy.

DNA was isolated from an F₂-mapping population as described by Guidet et al. (Guidet et al.,1991). The testcross population was screened using an alternative rapid DNA extraction procedure. Small sections (6-9 mm²) of the cotyledons or first emerging leaves of young seedlings were removed and ground in 20 μl 0.25 M NaOH. Grinding was performed in 96-well PCR plates using plastic pipette tip ends sealed by passing them through a flame. Once the tissue was homogenised, 7 μl of the sample was removed and added to 150μl 0.3 M Tris-HCl buffer (pH 8.0). Each PCR reaction used 7 μl of this mix, in a total volume of 20 μl. RNA extraction was performed using an RNeasy Plant Mini Kit (Qiagen), and RNA was quantified on a BioSpec-1601 E spectrophotometer (Shimadzu). A full-length cDNA (wild-type Columbia) of COV1 was obtained from a library generated by Asamizu et al.(Asamizu et al., 2000), and also by RT-PCR analysis from wild-type Ler.

PCR was performed in a PTC100 thermal cycler (MJ Research) using Taq polymerase (Life Technologies). Each reaction was performed in a total volume of 25 μl, containing 100 ng DNA with 50 mM KCl, 10 mM Tris-HCl (pH 9.0),0.1% Triton X-100, 200 μM dNTPs (each), 1.5 mM MgCl₂, 0.5 units Taq polymerase and 1 μM each primer. PCR conditions were as follows:94°C for 2 minutes; 35 cycles of 94°C for 30 seconds, 55°C(variable between primer pairs) for 30 seconds and 72°C for 60 seconds;and 2 minutes at 72°C. RT-PCR was performed using a One-Step RT-PCR Kit(ABgene) in a volume of 50 μl with 1 μg of total RNA. Gene-specific primers were used for amplification. PCR conditions were as follows: 47°C for 30 minutes; 94°C for 2 minutes; 35 cycles of 94°C for 20 seconds,55°C for 30 seconds and 72°C for 90 seconds. The positive control(MS2) was provided with the kit. All PCR products were run on 1-2% agarose gels in TBE buffer and visualised using ethidium bromide staining.

Templates generated from restriction fragment cloning, PCR and RT-PCR were gel purified using a High Pure PCR Product Purification Kit (Roche), and cloned into the vector pGEM-T Easy (Promega) for sequencing. Specific primers were designed and synthesised (MWG Biotech) for use along with vector specific T7 and SP6 primers. Plasmid templates were prepared using a QIAprep Spin Miniprep Kit (Qiagen), and sequencing reations were performed using an ABI Prism Sequencing Kit (Applied Biosystems). Sequence data were analysed using the ContigExpress program (InforMax). Sequences used for alignment were identified from BLAST searches of GenBank. Alignments were carried out using CLUSTAL W (Thompson et al.,1994).

Expression analysis was performed using real-time quantitative PCR (qPCR). Primers and probes were designed for the various genes using Primer Express(v1.0) (Applied Biosystems), and probes were labelled with 6-FAM (FAM;5′) and tetramethylrhodamin (TAMRA; 3′). First strand synthesis was performed in a volume of 20 μl, containing 1 μg total RNA with 500 ng poly-(dT) primer and 100 U reverse transcriptase (Promega), at 42°C for 60 minutes. PCR conditions were as follows: 50°C for 2 minutes; 95°C for 10 minutes; 40 cycles of 95°C for 15 seconds and 60°C for 60 seconds. PCR reactions were performed in a volume of 25 μl, containing 12.5μl 2× qPCR Mastermix (Eurogentec), 5 pmol probe and 25 pmol each primer. Data analysis was carried out using the Sequence Detector (v1.7)program (Applied Biosystems).

DNA fragments containing genes of interest were subcloned into the binary vector pc2300 (Cambia) for complementation purposes. Complementation was carried out by transforming plants using Agrobacterium tumefaciens(GV3101) containing various constructs described by Bent and Clough(Bent and Clough, 1998). Transformants were identified by growing seed of the transformed plants on agar plates containing MS (2.2 g/l) and the appropriate selection antibiotic. After 2 weeks, the resistant seedlings were transplanted into pots and grown to maturity to determine their phenotype.

Endogenous auxin levels were quantified by gas chromatography-selected ion monitoring-mass spectrometry with 13C6IAA as an internal standard, as described by Edlund et al. (Edlund et al.,1995). To determine response to auxin, seedlings were grown on MS agar plates containing various levels of 2,4-Dichlorophenoxy-acetic acid(2,4-D; 0.05, 0.1 and 0.25 μM), in a vertical position in a growth cabinet under continuous light at 22°C.

From a screen of ∼5000 EMS-mutagenised M2 Landsberg erecta(Ler) plants, two lines were isolated from independent seed batches that exhibited altered patterns of vascular bundle spacing. These two mutant lines were shown to be allelic by reciprocal crossing and subsequent scoring of the F₁ plants, whereby all plants scored were found to exhibit the mutant phenotype. These mutants were named continuous vascular ring (cov1), and the two lines designated cov1-1 and cov1-2. The heterozygous plants for both mutant alleles show normal vascular patterning in the stem and the mature plants are indistinguishable from wild-type Ler, indicating that the mutation is fully recessive. An F₂ population generated from a cross of cov1-1 back to the Ler parent yielded 57 wild-type plants and 22 cov1plants, which fits the expected 3:1 hypothesis (κ²=0.34, P=0.47) for a single nuclear recessive mutation.

Cross-sections of mature 4-week-old inflorescence stems showed that both alleles had a significant defect in the patterning and development of the vascular bundles compared with wild-type Ler(Fig. 1). In wild-type Ler plants, six to eight vascular bundles develop in a very ordered and predictable pattern around the stem, with bundles of phloem and xylem separated by interfascicular regions made up of differentiated fibres. However, in the cov1 mutants, the ordered patterning of vascular bundles is lost and is replaced by an almost continuous ring of both phloem and xylem tissue around the whole stem. This increase in vascular tissue appears to replace the area normally occupied by the interfascicular cells such that, in plants exhibiting a severe mutant phenotype, the interfascicular region was virtually absent or significantly reduced(Fig. 1). This proliferation of vascular tissue is also accompanied by a reduction in stem thickness.

In order to study further the development of the vascular bundles in the stem of wild-type Ler and cov1-1 plants, cross-sections of 4-week-old plants were taken at various positions along the inflorescence stem to observe the vascular phenotype. Sections were taken just below the apex, a quarter of the way down the stem and half way down the stem(Fig. 2). The results showed that directly below the apical meristem, wild-type plants exhibit a clearly defined pattern of vascular bundles containing both xylem and phloem(Fig. 2). Corresponding bundles are also visualised in cov1-1 plants(Fig. 2). However, in the cov1-1 mutant there was clear evidence of additional differentiation of vascular tissue in the region normally occupied by interfascicular cells. This differentiation is clearly visible as brightly stained yellow phloem cells in between, and unconnected to, the vascular bundles, which exhibit a more normal pattern of development (Fig. 2). It can also be seen that additional differentiation occurs directly adjacent to the developing vascular bundles(Fig. 2). It is unclear at present whether this is actually lateral expansion of the established vascular bundle or the initiation of additional differentiation in the region immediately adjacent to the pre-existing vascular bundle. This patterning,though somewhat variable, continues down the stem such that, at the base, an almost continuous ring of differentiated vascular tissue is visible(Fig. 1).

Under the conditions in which these plants were grown, no evidence of any secondary growth was observed either in the wild type or in the mutant. Although both alleles show similar phenotypes, the second allele(cov1-2) is generally more severe(Fig. 1). Cross-sections of the hypocotyl were also compared between mutant and wild-type plants, and the patterning in this region of the plants appeared to be similar (data not shown). The cov1 mutants were slower growing and shorter compared with wild-type Ler plants. At flowering, cov1-1 was approximately two thirds the height of wild-type Ler and cov1-2 was less than half the height of wild-type Ler when grown under continuous light (Fig. 3). Electron microscopy studies of stems from cov1-1 and wild-type plants showed that the cells in the mutant stems were shorter and smaller compared with wild-type Ler (data not shown), but the number of internodes is similar. The siliques of cov1 were shorter and more wrinkled compared with wild-type Ler, and both alleles showed a variable slightly twisted leaf phenotype, with cov1-2 being more severe in both respects.

Leaves and cotyledons from wild type and both mutant alleles were cleared to observe the vein pattern and to determine the cause of the twisted leaf phenotype. These results showed that there was no difference in vein patterning in cotyledons (data not shown), and that the patterning of primary,secondary and tertiary veins in leaves was similar between wild-type and mutant plants (Fig. 3). However, the shape of the leaves in both mutant alleles appeared to be somewhat asymmetric compared with wild type. This asymmetry, although minor,may give rise to the visible twisted leaf phenotype. There were no apparent defects in phyllotaxis in either of the mutant alleles. When grown on agar plates, the roots were slower growing and shorter in both mutant alleles, but there was little difference in numbers of lateral roots compared with wild-type Ler. There was no difference in the phenotype of etiolated seedlings of cov1-1 and wild-type plants when germinated in the dark(data not shown).

An initial F₂-mapping population of 600 plants was generated from a cross between cov1-1 and wild-type Columbia plants. The population was scored for the mutant phenotype, and 120 mutant plants were selected and screened with polymorphic microsatellite (SSLP) markers spanning all five chromosomes. Linkage was identified with markers on the long arm of chromosome 2. Further SSLP markers were generated in this region for more detailed mapping and the gene was localised to a region of ∼500 kb. Polymorphic co-dominant markers located on BACs F27F23 (SSLP33) and F21P24(SSLP23), which were suitable for rapid PCR screening, were identified that flanked a region of ∼2 Mb containing the COV1 gene. These markers were used to screen a testcross population to identify critical recombinants in this region. The testcross population was generated by crossing F₁ plants generated from the initial cross with Columbia back to cov1-1 to generate a population that was 50% mutant and 50% wild type(heterozygous). Out of 2148 testcross plants screened, 105 showed recombination between the two flanking markers. These critical recombinants were used to finely map the gene further using additional SSLP and cleaved amplified polymorphic sequence (CAPS) markers generated on each of the BACs spanning this region. Based on the mapping data obtained from these plants,the gene was placed within a region of 77 kb, spanned by the ends of two adjacent BACs (T2G17 and F11A3) (Fig. 4).

The two BACs spanning the region to which the gene had been mapped were digested using three restriction enzymes (SacI, SalI and XhoI) and the fragments subcloned into a binary vector for Agrobacterium-mediated transformation. Partial complementation [100%cov1-1 plants (n=30) and 68% cov1-2 plants(n=34)] was achieved when mutant plants were transformed with an 11,644 bp SacI fragment, pGDP1(Fig. 4), that contained four predicted genes. Sequence analysis of the whole fragment from mutant plants did not identify any mutations when compared with wild type. Complementation of cov1-1 was also achieved with a 10,599 bp XhoI/SalI fragment, pGDP6, which also contained four predicted genes, including three of the genes found on pGDP1 and an additional predicted gene (GenBank Accession Number At2g20120). This gene was isolated independently on a 5722 bp BstUI fragment, pGDP7, that contained only a single complete predicted gene (At2g20120). Transformation of cov1-1 plants with pGDP7 restored both the vascular phenotype and the growth phenotypes (Fig. 5). This fragment was also able to complement the cov1-2 allele, with all transformed plants (n=10) showing wild-type vascular patterning in the stem (data not shown).

Comparison of the DNA sequencing of pGDP6 from wild-type Ler and cov1-1 identified a single base-pair mutation in At2g20120. Comparison of the cov1-2 sequence with wild type also revealed the presence of a single base-pair change in this open reading frame. Consequently, both complementation and sequence analysis demonstrated that the cov1 mutation is the result of a defect in At2g20120 (from hereon referred to as COV1).

A full-length EST cDNA clone of At2g20120 (COV1) from a Colombia background was obtained (Asamizu et al.,2000) and the DNA sequence deposited in GenBank (Accession Number AY170845). RT-PCR was performed using mRNA from wild-type Ler, and the coding sequence was identical to the cDNA clone from Columbia. Comparison of the cDNA and genomic sequence revealed that the COV1 gene comprises seven exons and six introns (Fig. 4). The mutation in cov1-1 is located in the third exon(C→T) and results in an amino acid change from proline to serine(P⁷⁶→S). In cov1-2 the mutation occurred in the fourth exon (G→A) and results in an amino acid change from glycine to arginine (G¹³⁰→R). The full-length protein is predicted to be 268 amino acids in length, although there are two possible in frame start codons (Fig. 6) and it is unclear which one is used. The predicted structure of the gene indicates that the N-terminal end of the protein is cytoplasmic, followed by three membrane spanning domains with the C-terminal end predicted to be in the wall(Fig. 4). The mutation in the cov1-1 allele is located in the first membrane spanning domain,whereas the mutation in cov1-2 is located intracellularly, just after the second predicted membrane spanning domain. Based on phenotype, the mutation in cov1-2 is more severe and causes a greater loss of function of the protein, as evidenced by the extreme dwarf growth habit and increased disruption of the vascular patterning in the stem.

Analysis of the genome revealed that there are three other related genes in Arabidopsis. These genes have been designated LIKE COV(LCV). LCV1 (GenBank Accession Number At2g20130; 238 amino acids) lies directly upstream (∼1 kb) of COV1(Fig. 4) and is almost identical at the amino acid level (95%), which indicates that it may be a recent duplication (Fig. 6). LCV2 (GenBank Accession Number At1g43130; 261 amino acids) is located on chromosome 1 and also has a high level of homology at the protein level to COV1 (58% identity); the most variable region is at the 5′ end,which is predicted to be cytoplasmic. LCV3 (GenBank Accession Number At2g18460; 129 amino acids) is located on chromosome 2 and, although being closely related (52% identity), is truncated relative to COV1 and exhibits homology to only the carboxy-half of COV1. EST sequences are available for all four members of the family, which suggests that they are all expressed.

Analysis of protein sequences from other genomes has revealed that there is a homologous gene to COV1 in rice, with the predicted sequence from a cDNA exhibiting 59% identity to COV1 (Fig. 6). COV1 also has clear homologues in a large number of bacterial genomes. Alignment with two of these bacterial sequences, from Mesorhizobium loti and Agrobacterium tumefaciens, is shown in Fig. 6. They both exhibit sequence conservation with COV1 along their entire length. All of the related bacterial genes are annotated as membrane proteins of unknown function. These results indicate that these genes may be involved in cellular mechanisms that are specific to plants and bacteria, and that the COV1 gene may play a common role within these organisms.

RT-PCR experiments demonstrate that the COV1 gene is alternatively spliced in both wild-type Ler plants and the two mutant alleles (data not shown). Sequence analysis revealed that the alternatively spliced product was the result of the retention of intron 2, which leads to a premature stop codon and a truncated protein. Real-time quantitative PCR (qPCR) analysis was carried out to determine the relative levels of the different transcripts in wild-type and mutant plants. In wild-type Ler and cov1-2plants, the amount of alternatively spliced product compared with the correctly spliced product was 0.95%, but in the cov1-1 mutant plants,this was increased to 2.01% (data not shown). Because of the high degree of similarity between COV1 and the three related genes, qPCR was used to determine the expression levels of the specific genes in various plant tissues. The results of the qPCR experiments showed that COV1 was highly expressed in flowers and stems; a reasonable level of expression was also seen in roots, but less in leaves(Fig. 7). The expression levels of LCV1 were also measured and were found to be much lower than those of COV1.

To determine whether mutant plants showed an altered response to auxin,seedlings of wild-type Ler and mutant cov1-1 plants were germinated and grown on MS agar plates containing different concentrations of 2,4-D (0.1 μM, 0.25 μM and 0.5 μM). The results showed that there was no difference in root growth between cov1-1 and wild-type Ler in response to auxin (Fig. 8). In addition, cov1-1 produced wild-type numbers of root hairs and exhibited a normal gravitropic response (data not shown).

Double mutants were generated between cov1-1 and axr1-3:a mutant known to be defective in almost every measurable aspect of auxin signalling (for a review, see Kepinski and Leyser, 2002) and which is resistant to auxin(Lincoln et al., 1990). The double mutants did not show a significantly different plant phenotype compared with either cov1-1 or axr1-3 mutant plants. The cov1-1:axr1-3 plants grew to a similar height compared with cov1-1, and exhibited the disrupted vascular patterning in the stem characteristic of cov1-1 (data not shown). The double mutants were grown on various concentrations of 2,4-D and their root lengths were measured. Both the axr1-3 and cov1-1:axr1-3 double mutants were equally resistant to auxin compared with wild-type Ler and cov1-1 plants (Fig. 8). The double mutant showed no significant alteration in its response to 2,4-D compared with the single axr1-3 mutant. Overall the results indicate that the double mutants were not more resistant to 2,4-D than was the axr1-3 mutant, which suggests that COV1 does not act in the same signalling pathway as AXR1.

To determine whether there was any difference in auxin concentration between wild-type Ler and cov1-1 mutant plants, levels of auxin were measured at various positions along the inflorescence stem. These included the region just below the apical meristem; half way down the stem; at the base of the stem; and in the hypocotyl. The results of the analysis showed that there was no significant difference in auxin concentration in the upper part of the stem (Fig. 8). However, in the base of the stem and in the hypocotyl there was a significant difference: wild-type plants contained almost twice the amount of auxin as did cov1-1 plants.

cov1-1 plants were also transformed with a construct containing the auxin inducible IAA2 promoter coupled with the βglucuronidase (GUS) reporter gene as an indicator of the presence of auxin(Swarup et al., 2001). Transformed cov1-1 plants were then crossed with wild-type Ler plants for comparison, to reduce any variation caused by positional effects. There was little difference in expression patterns of GUS between the wild-type or cov1-1-mutant plants in either the seedlings or developing stems (data not shown), which indicates that there was no difference in the distribution of auxin in the upper part of the stem. There was no detectable IAA2-GUS expression at the base of the stem in either wild-type Ler or cov1-1-transformed plants where the auxin measurements identified a significant difference between wild-type and mutant plants.

cov1 is a novel mutation in which there is a large alteration in the pattern of vascular tissue in the stem. The positioning of the vascular tissue around the outside of the pith is normal, but the vascular tissue proliferates at the expense of the interfascicular tissue that normally separates each vascular bundle (Fig,1). Distinct vascular bundles are clearly visible close to the apex early in development; however, in contrast to wild type, additional vascular tissue is clearly visible between the newly forming bundles in cov1 mutants (Fig. 2). This apparent additional recruitment of new bundles represents the development of vascular tissue unconnected to existing vascular bundles. In addition,vascular tissue also develops adjacent to existing vascular bundles. It is unclear whether this represents the expansion of existing vascular bundles or the formation of new vascular tissue that coincidently adjoins an existing strand. In the mutant, vascular development continues such that, in some mature stems, the interfascicular regions are almost absent (Figs 1, 2). However, cov1-1and cov1-2 mutant plants do not show any significant defects of vascular patterning in cleared cotyledons or leaves compared with wild type,although the shape of the leaves in both mutant alleles appeared to be slightly asymmetric (Fig. 3),which is thought to give rise to the slightly twisted leaf phentotype. In general, leaf shape mutants show pattern defects in the number and placement of the major veins (Dengler and Kang,2001) but these appear to be relatively normal in the cov1 mutants. The fact that any defects in the patterning of veins in leaves were minor compared with the disruption of vascular patterning observed in the stem suggests that vascular differentiation in leaves and stems may, at least in part, be regulated by different pathways. This idea is supported by the recent description of a number of mutants that show altered patterns of vascular development in the cotyledons and leaves, but that show little effect on the pattern of vascular development in other parts of the plant(Carland et al., 1999; Deyholos et al., 2000; Koizumi et al., 2000).

Several of the phenotypic characteristics of the cov1 mutants indicate that the abnormalities might be caused by a defect in auxin signalling. The mutant plants have a stunted growth habit and wrinkled leaves(Fig. 3). Several other mutants defective in auxin signalling and transport also show a dwarf phenotype with wrinkled or curled leaves (Ruegger et al.,1997; Hobbie et al.,2000; Liscum and Reed,2002). Auxin is known to be involved in controlling cell expansion and has been shown to be involved in regulating hypocotyl elongation in light-grown, but not dark-grown, plants(Jensen et al., 1998; Nakazawa et al., 2001). Given the importance of auxin in all aspects of vascular development, cov1-1 was examined for its response to auxin(Sachs, 1981; Sachs, 1991; Carland et al., 1999; Sieburth, 1999; Mattsson et al., 1999). Seedlings of cov1 respond to auxin in a similar manner to wild-type plants. Furthermore, cov1-1:axr1-3 double mutants behave in a manner similar to the single mutants alone (Fig. 8). Both these lines of evidence suggest that auxin signalling is not the primary defect in cov1-1. Direct measurement of auxin levels indicates no difference between mutant and wild type in the upper stem when the vascular defect in cov1 first becomes apparent. There is a small(twofold) difference in auxin levels at the base of the stem, but this may be a consequence of the altered vascular tissue in cov1 plants affecting PAT. It is unlikely to account for the vascular patterning defect that is first apparent close to the apex.

According to the canalisation hypothesis(Sachs, 1981; Sachs, 1991), auxin and auxin transport are responsible for the induction of new vascular tissue, and the subsequent draining of auxin from surrounding cells prevents the formation of new vascular strands close to pre-existing vascular bundles. Although there is a vast body of evidence to support the induction and canalisation of the vascular bundles by auxin, the basis of the inhibitory affect is less clear. The cov1 phenotype is hard to reconcile with this model because a normal pattern of vascular development is observed but apparently fails to inhibit the formation of additional vascular tissue nearby. The cov1phenotype is more consistent with the diffusion-reaction hypothesis in which a second inhibitory component is hypothesised. To date, the nature of this inhibitor has not been identified, but the cov1 phenotype is consistent with a plant defective in the synthesis, transport or perception of such an inhibitor.

Polyamines are also known to affect cell size and play an important role in internode elongation through cell expansion. The acl5 mutant, which encodes a spermine synthase, exhibits a short inflorescence stem similar to cov1-2 (Hanzawa et al.,1997), and, like the acl5 mutant, the cells in the inflorescence of cov1-1 plants are much shorter and smaller compared with wild type (data not shown). The acl5 mutant has much shorter and wrinkled siliques, which is also a characteristic of the cov1mutants. acl5 plants show an increase in vascular tissue within the bundles of the stem, but the spacing and organisation remains relatively similar to the wild type. Polyamines have been hypothesised either to act downstream of auxin signalling or to regulate the interaction of different hormones, such as auxin and gibberellin(Hanzawa et al., 2000). However, the mechanism by which polyamines or their derivatives affect plant development is currently unknown.

The gene responsible for the cov1 mutation has been cloned by a map-based cloning strategy. Two distinct lines of evidence indicate that the defect is caused by a mutation in the predicted gene At2g20120. First, both mutant alleles can be fully complemented using a fragment containing only this predicted gene. Second, comparison of the cov1 alleles with the wild-type sequence revealed single base-pair changes within this gene. The complementation analysis was complicated by the fact that the cov1-1mutation could be complemented using a fragment that lacked COV1. This is explained by fact that COV1 appears to be part of a very recent tandem duplication event. COV1 is adjacent to a very closely related gene (termed LCV1) that exhibits a 95% amino acid sequence identity with COV1. LCV1 has a very short promoter region (∼500 bp) which it shares with a gene coding for one of the subunits of the 26S proteasome (Fig. 4). Presumably, if the gene is inserted into the genome in a position that increases the expression of LCV1, it is sufficiently similar to compensate for COV1. Evidence that LCV1 is not responsible for the cov1 mutation comes from the fact that no mutation was identified in LCV1 in either allele of cov1 and that pGDP1 was only able to partially complement the cov1-2 allele (68% of plants). The presence of LCV1 makes it unlikely that either of the cov1 alleles are functional knockouts, as LCV1 is expressed at low levels in the plant and was shown to be able to partially complement the mutation.

Both the vascular defect and the growth defects were complemented by pGDP7,which indicates that all the phenotypes associated with cov1 were a consequence of a defect in a single gene. Expression of the cov1 gene is high in stems, which is consistent with the observed phenotype. The transcript appears to be alternatively spliced, and a small minority (∼1%)of the mRNA retained the second intron. It is unlikely that this change contributes to the mutant phenotype, because the relative increase is small;however, the exact role of the alternative splicing in wild-type plants remains unclear.

COV1 is predicted to be an integral membrane protein of unknown function in plants and a large number of bacteria. The high degree of homology between the genes in plants and bacteria indicates that COV1 may play an important role in some common pathway, possibly in signalling. Interestingly,the bacterial species that contain the more closely related homologues of COV1 are nearly all species that are associated with plants, including Mesorhizobium loti and Agrobacterium tumefaciens. Whether these genes are used in the perception or transport of a signal generated in plants remains to be investigated. The cov1-mutant phenotype is consistent with that of a plant defective in a component that negatively regulates the differentiation of both xylem and phloem in the developing stem. This component has been predicted in a number of models for vascular development. Further analysis should allow us to determine further the function of COV1 and how it regulates plant vascular development.

We thank the Arabidopsis Biological Resource Centre for the provision of BAC clones and the Nottingham Arabidopsis Stock Centre for the provision of seed stocks. We are also grateful to the Kazusa DNA Research Institute for providing cDNA clones and to Malcolm Bennett for the provision of the IAA2-GUS construct. This work was supported by the 5th EC Framework Programme within the POPWOOD project.