The HVE/CAND1 gene is required for the early patterning of leaf venation in Arabidopsis

The venation patterns of plant leaves provide two-dimensional models for dissecting the generative mechanisms of branched structures. Although detailed descriptions and consistent theoretical models account for venation pattern formation (Meinhardt, 1976; Meinhardt, 1984; Mitchison, 1980; Mitchison, 1981), the mechanism by which such structures are built is still not fully understood at the genetic level. Considerable experimental evidence suggests that indole acetic acid (IAA, the main auxin) transport is involved in the patterning of plant vasculature. The spatial distribution of IAA correlates well with the sites of vascular differentiation in shoot apices and young leaf primordia(Avsian-Kretchmer et al.,2002), and the inhibition of auxin polar transport causes a variety of developmental defects in leaves, including an abnormal vascular network (Mattsson et al.,1999; Sieburth,1999). These observations support the so-called `canalization hypothesis', which states that the patterning of plant vascular systems depends on a positive feedback loop acting on the transport of auxin(Sachs, 1991a; Sachs, 1991b).

Mutants with aberrant venation patterns in leaves or cotyledons have been isolated using Arabidopsis thaliana as a model(Carland and McHale, 1996; Carland et al., 1999; Deyholos et al., 2000; Koizumi et al., 2000; Steynen and Schultz, 2003). Some of the corresponding genes have already been cloned, including COTYLEDON VASCULAR PATTERN1 (CVP1), which encodes a sterol methyltransferase involved in the biosynthesis of sterol and brassinosteroid compounds (Carland et al.,2002), and CVP2, which encodes an inositol polyphosphate 5′ phosphatase and confirms a role for inositol 1,4,5-trisphosphate(IP3)-mediated signal transduction in vein ontogeny(Carland and Nelson,2004).

In a search for naturally occurring variations of the leaf venation pattern in Arabidopsis, we found that the spontaneous hemivenata-1(hve-1) allele reduces venation complexity in leaves and cotyledons(Candela et al., 1999). Here,we analyze the developmental effects of hve-1 and two other recessive alleles of HVE, which we positionally cloned and found to encode a CAND1 protein. Our results demonstrate a role for ubiquitin-mediated auxin signaling in Arabidopsis vein patterning.

Seeds of most Arabidopsis thaliana (L.) Heynh. lines were supplied by the Nottingham Arabidopsis Stock Centre (NASC) and the Arabidopsis Biological Resource Center (ABRC). These included the Wassilewskija-2 (Ws-2; NASC Accession Number N1601), Col-0 (N1092) and Eifel-5(Ei-5; N1128) wild-type accessions, the N599479 and N610969 T-DNA insertion lines from the Salk Institute Genome Analysis Laboratory(Alonso et al., 2003), and the pin1^cay (N324) and axr1-12(Leyser et al., 1993) mutants. The pin1^cay mutant was isolated by G. Röbbelen and later named cay by Serrano-Cartagena et al.(Serrano-Cartagena et al.,1999). Allelism of cay and pin1-1 was suggested by their similar phenotypes and confirmed in complementation crosses. We found that pin1^cay is a null allele carrying a G-to-A substitution that produces a truncated protein lacking 489 amino acids. lop1-65 and cvp2-1(Carland and McHale, 1996; Carland et al., 1999) were kindly provided by Francine Carland; mp^T370(Przemeck et al., 1996) by Thomas Berleth; and the ATHB-8-GUS(Baima et al., 1995), CAND1-GUS (Chuang et al.,2004) and DR5-GUS(Ulmasov et al., 1997)transgenic lines by Simona Baima, William Gray and Thomas Guilfoyle,respectively. We obtained mp^T370/mp^T370;hve-3/hve-3 and mp^T370/mp^T370;HVE/- seedlings, all of which were lethal and failed to develop roots and produce leaves. In our hands, the induction of adventitious roots by hypocotyl basal bisection(Przemeck et al., 1996) led to an extremely low survival rate, which made not possible the isolation of a significant number of mp^T370/mp^T370 rosettes for morphometry of leaf vascular patterns.

Plants were grown on agar medium, at 20±1°C and 60-70% relative humidity and with 7000 lux of continuous fluorescent light(Ponce et al., 1998). Crosses and allelism tests were performed as described previously(Berná et al., 1999). ATHB-8-GUS and CAND1-GUS transgenic seeds were sterilized and kept in water at 4°C in the dark for 4 days to ensure synchronized germination.

To study the effects of the synthetic auxin 2,4-dichlorophenoxyacetic acid(2,4-D), seeds were sown on non-supplemented agar medium and the seedlings were transferred to supplemented media 4 days later. Root lengths were measured 4 days after the transfer(Lincoln et al., 1990). The gravitropic response of roots was studied on Petri dishes containing 1.5% agar medium, which were kept vertically in a Conviron TC30 growth chamber for 8 days before being rotated 135° and observed 4 days later.

Histological sections were obtained as described previously(Serrano-Cartagena et al.,2000). For venation pattern visualization and micrography, lateral organs were cleared as described previously(Candela et al., 1999). Histochemical visualization of β-glucuronidase (GUS) activity was performed as described previously(Donnelly et al., 1999). We combined the null hve-3 allele with null alleles of other genes, in most cases in the same genetic background (Col-0), to obtain double mutants,the venation of which was compared with those of their siblings. Data corresponding to leaf lamina area, venation density and number of branching points/mm² were collected as described previously(Candela et al., 1999) and analyzed using the SPSS for Windows version 10.0.6 statistical software package (SPSS). Kolmogorov-Smirnov tests with Lilliefors correction were applied to assess the normality of our data. When the data fitted a normal distribution, t tests were used to compare mean values. When normality was not accepted, the non-parametric Mann-Whitney U test was used instead.

Linkage analysis (Ponce et al.,1999) was used for the positional cloning of HVE. We have developed several novel molecular markers that are used for high-resolution mapping of the HVE gene (Table 1). Two RFLP markers that flank nga1145(Lister and Dean, 1993) were converted to PCR-based CAPS markers(Konieczny and Ausubel, 1993): MnlI restriction of the ve012 PCR product rendered a single band in Ei-5 (339 bp) and two in Ws-2 (300 and 39 bp), whereas EcoRI restriction of the m246 PCR product rendered a 1.6 kb band in Ws-2 and two 0.8 kb bands in Ei-5. In addition, we found a polymorphic (AT)_nmicrosatellite, which we named T17M13b, based on the available sequence of the T17M13 BAC clone. Sequencing the candidate region allowed us to identify five novel single nucleotide polymorphisms (SNP), named T8K22-4, P450, T8K22-7,T8K22-1 and E2, which were used for mapping the HVE gene. The synthetic oligonucleotides were purchased from Sigma-Genosys (UK). Genomic DNA was extracted, PCR-amplified and sequenced as described previously(Pérez-Pérez et al.,2004).

For gene expression and splicing analyses, RNA was extracted using the RNeasy Plant Mini Kit (Qiagen). RNA samples (2.5-5.0 μg) were reverse transcribed using random hexanucleotide primers and SuperScript II (Gibco-BRL,Cleveland). The resulting cDNA was used as a template in amplification reactions of the HVE gene with primers from Table 1, and in control amplification reactions of the housekeeping OTC gene(Quesada et al., 1999).

We have previously identified the hemivenata-1 (hve-1)allele of the HVE gene as being responsible for a monogenic,recessive trait displayed by the Eifel-5 (Ei-5) wild-type accession(Candela et al., 1999). This spontaneous allele strongly reduced venation pattern complexity in vegetative leaves and cotyledons.

We positionally cloned HVE(Fig. 1A) using an F2 mapping population derived from a Ws-2 × hve-1/hve-1 (in an Ei-5 background) cross. After finding linkage between hve-1 and the nga1145 marker, near the upper telomere of chromosome 2, we developed eight molecular markers (Fig. 1A)that were used to screen for recombinants in 1545 F2 plants. Informative recombinants narrowed down the candidate interval to a 61 kb genomic region including 14 annotated genes (Fig. 1A).

To identify HVE among the candidate genes, we studied publicly available T-DNA and transposon alleles. Forty-nine lines carrying insertions within or close to 11 candidate genes (At2g02550-At2g02590,At2g02610-At2g02640, At2g02660 and At2g02670) were searched for phenotypic traits reminiscent of those of hve-1/hve-1 plants. Two of them,N599479 and N610969, carried T-DNA insertions in the At2g02560 gene and displayed a simple leaf vasculature and a bushy inflorescence. Complementation tests confirmed that both lines carried alleles of hve-1. The presence and position of their T-DNA insertions was confirmed by PCR and by sequencing (Table 1). These insertions lie at nucleotides 1195 (N599479) and 5764 (N610969) of the At2g02560 transcription unit (numbering from the ATG; Fig. 1B), as described at http://signal.salk.edu. We named the alleles carried by N599479 and N610969 as hve-2 and hve-3, respectively.

To determine the molecular lesion of the hve-1 allele, we sequenced the gene in several wild-type accessions, including Ei-5, Col-0 and Ws-2. To distinguish silent polymorphisms from the mutation causing the phenotype, we only considered the differences between Ei-5 (hve-1)and Col-0 or Ws-2 that did not correspond to differences between Col-0 and Ws-2. Four single nucleotide changes in the hve-1 allele met this condition, affecting the 5th, 20th and 24th introns, and the 16th exon (which substitutes a methionine for an isoleucine). We also found a 24-nucleotide deletion in the 16th intron and a 16-nucleotide deletion spanning the 3′end of the 14th exon and the 5′ end of the 14th intron(Fig. 1B). The latter deletion removes a splicing donor site and causes mis-splicing, as shown by PCR amplification of Col-0, Ws-2 and Ei-5 cDNAs using primers flanking the deletion (Fig. 1C). Three splice forms were detected in Ei-5, but only one in Col-0 and Ws-2. Two of the three corresponding gel bands (numbered as 1 and 3 in Fig. 1C) were sequenced and confirmed to originate from mis-spliced HVE transcripts(Fig. 1C,D). The corresponding cDNAs were out of frame and were predicted to encode truncated proteins. We were not able to isolate molecules from the third band obtained from hve-1 cDNA (band number 2 in Fig. 1C), which had a molecular weight greater than that of the single band found in the wild type.

HVE encodes a predicted protein of 1217 amino acids (see Fig. S1 in the supplementary material) and a molecular weight of 134.6 kDa(http://mips.gsf.de/cgi-bin/proj/thal/search_gene?code=At2g02560),closely related to the mammalian TATA-binding protein-interacting protein 120(TIP120A) (Yogosawa et al.,1996). TIP120, also known as CAND1 (Cullin-Associated and Neddylation-Dissociated) (Liu et al.,2002; Zheng et al.,2002), regulates the formation of the SCF complexes of ubiquitin E3 ligases. Although two paralogs have been described in the rat and human genomes (Aoki et al., 1999), HVE/CAND1 is a single-copy gene in Arabidopsis.

Since mammalian CUL1 is known to interact with CAND1(Liu et al., 2002; Zheng et al., 2002; Hwang et al., 2003; Oshikawa et al., 2003), we generated a recombinant Glutathione-S-transferase-HVE (GST-HVE)protein that was used in pull-down experiments. CUL1 was translated in vitro in rabbit reticulocyte lysate in the presence of ³⁵S-Met and then incubated with GST or GST-HVE. As shown by SDS-PAGE gel autoradiography (see Fig. S2A in the supplementary material), GST-HVE, but not GST, was able to interact with CUL1, suggesting that, as their mammalian counterparts, the HVE/CAND1 protein of Arabidopsis regulates the activity of the SCF complexes by binding CUL1. While this work was being carried out, other groups demonstrated that CAND1 and CUL1 physically interact(Chuang et al., 2004; Feng et al., 2004).

The hve mutants displayed vegetative leaves of slightly reduced size (Fig. 2A-D), and a bushy,extremely branched inflorescence (Fig. 2E-G). Fertility was reduced in hve-1/hve-1 plants and was very low in hve-2/hve-2 and hve-3/hve-3, which produced less than 10 seeds per plant. The inflorescence of the hve-1 mutant was of normal size, whereas those of hve-2 and hve-3 were dwarf (Fig. 2F,G). Flowering time and senescence were delayed in all hve mutants. Together with the molecular characterization of the alleles, these results suggest that hve-1 is a hypomorphic allele, whereas hve-2 and hve-3 are null.

We also found that hve mutations determine an abnormal pattern of root growth, as shown in hve-3 homozygotes and in 19 F3 families derived from hve-1/hve-1 individuals of the F2 of a Ws-2 × Ei-5 cross. The `wavy' pattern of growth characteristic of Col-0 and Ws-2 wild-type roots was absent from the mutant plants(Fig. 2H-J).

To determine the extent to which HVE is required for the development of the vascular pattern, we cleared cotyledons, vegetative and cauline leaves, sepals and petals of the hve mutants and the Col-0 wild-type accession, which is the genetic background of hve-2 and hve-3 (Fig. 3). The venation patterns of the three hve mutants were similar to one another and clearly distinct from that of Col-0. The cotyledons contained three or four areoles (regions of the lamina completely bordered by veins) in the wild type (Fig. 3A), but only two in the hve mutants, often incompletely closed(Fig. 3H,O,V). The reduction in vein numbers was also apparent in the first two rosette leaves, which had fewer secondary veins than the wild type. No quaternary veins and only a few tertiary veins, which ended blindly within the areoles, were present in mutant leaves, the intramarginal vein of which was occasionally interrupted(Fig. 3B,I,P,W). Similar observations were made in vegetative leaves of the third(Fig. 3C,J,Q,X) and seventh(Fig. 3D,K,R,Y) nodes. The lower structural complexity of the mutant venation patterns was in all cases a consequence of a reduced number of secondary and tertiary veins and the absence or shortening of higher-order veins (quaternary and higher). Cauline leaves, by contrast, were smaller in the hve-2 and hve-3mutants than in Col-0, but did not show an obvious reduction in vascular density (Fig. 3E,L,S,Z). In wild-type sepals, two secondary veins diverged from the apical end of the primary vein to form two arches (Fig. 3F), whereas a single primary vein and two incomplete secondary veins were observed in the mutants (Fig. 3M,T,AA). Wild-type petals also showed a more elaborate venation pattern than the corresponding organs of the hve mutants(Fig. 3G,N,U,AB).

As seen in paradermal and transverse sections (see Fig. S3 in the supplementary material), the primary vein was thinner in hve-3/hve-3leaves than in the wild type, but the secondary veins looked normal. The structure of bundle sheath cells, sieve tubes and tracheary elements was also normal in the mutant. No differences with the wild type were found in the cell shape and size of the mutant epidermis, palisade mesophyll and spongy mesophyll. However, the spongy mesophyll had larger air spaces and the palisade mesophyll was partially disorganized in the mutant. A midrib was clearly distinguishable abaxial to the midvein in the wild type, but seemed to be absent in hve-3.

The various organs affected in hve mutants suggested that the HVE/CAND1 gene is required throughout the life cycle, from embryogenesis to floral development. To verify this, Col-0 RNA was extracted from assorted organs and RT-PCR amplified, and HVE was found to be expressed in every organ studied (see Fig. S2B in the supplementary material). To further define the postembryonic spatial expression pattern of HVE/CAND1, we characterized the wild-type expression pattern(Fig. 4) of a CAND1promoter-GUS fusion (P_ETA2-GUS) that includes 2.7 kb of upstream sequence (Chuang et al.,2004). Consistent with the pleiotropic phenotype of loss-of-function hve mutants and with our semi-quantitative RT-PCR results, the gene was found almost ubiquitously expressed in aerial and underground organs of 10-day-old plants(Fig. 4A,B). At this stage, the gene was widely expressed in rosette leaves, at the highest level in the vasculature (Fig. 4A). By contrast, GUS staining was mostly confined to the leaf vasculature in 21-day-old plants, indicating that HVE expression changes dynamically throughout leaf development (Fig. 4C). For cotyledons and vegetative leaves, the expression in mesophyll cells was apparently more intense at the beginning of leaf expansion, disappearing progressively as the leaves grew(Fig. 4D,E-H,P). Closer inspection revealed that the gene was expressed before xylem differentiation in developing veins at the basal actively dividing region of rosette leaves(Fig. 4R) and, after the differentiation of tracheary elements, in the living cells of the vascular bundles (Fig. 4Q). The strong association of HVE expression and the vascular tissues is consistent with its role in vascular development (Fig. 4E-K,O-P) and with the defective venation of hve mutants. By contrast, the reporter was expressed throughout young cauline leaves before being relegated to their margins, at which time no expression could be detected in the veins (Fig. 4I). This observation correlates with the absence of vascular mutant phenotype in the cauline leaves of hve mutants. Similarly, the expression of the gene changed dynamically in developing flowers, being general in immature flowers (Fig. 4M) and restricted to the distal tip of sepals, the veins of petals, and the vasculature, filament and anthers of stamens in mature flowers(Fig. 4J,K). Expression was also detected in other organs, including siliques(Fig. 4L,N), stems(Fig. 4L), the root meristem and vascular cylinder (Fig. 4B).

ATHB-8 encodes a class III homeodomain/leucine zipper (HD-Zip III)transcription factor specifically expressed in procambial cells(Baima et al., 1995). Consistent with a positive role in vascular differentiation, it is induced during revascularization of injured stems and after the exogenous application of auxin (Baima et al., 1995),and its overexpression causes the premature differentiation of procambial cells into primary xylem (Baima et al.,2001). We have used an ATHB-8-GUS reporter(Baima et al., 1995) to trace back the perturbations caused by hve mutations in venation pattern formation. The expression of ATHB-8-GUS is restricted to the veins of wild-type Ws-2 (Baima et al.,1995) and Col-0 leaves (Fig. 5I-P). Previous results indicated that, at the stage shown in Fig. 5 (21 days after sowing),the first leaf is almost fully expanded, whereas the tenth leaf is just beginning to expand (Candela et al.,1999; Pérez-Pérez et al.,2002). In all these wild-type leaves, the expression of the reporter uncovered a complex, reticulate pattern consisting of mature and differentiating vascular strands. By contrast, the expression of the reporter revealed a much simpler GUS pattern in hve-3/hve-3 leaves than in wild-type leaves of the same age (compare Fig. 5A with 5I, 5B with 5J, 5C with 5K,and 5D with 5L). The GUS pattern was restricted to the fully differentiated venation pattern of the mutant, which lacked most tertiary and higher order veins. The absence of GUS staining at the sites where the veins normally differentiate indicates that ATHB-8-GUS expression is dependent on the earlier acquisition of vascular fate at specific locations within the leaf. Identical results were obtained in comparisons of Ws-2 with hve-1/hve-1 (see Fig. S4 in the supplementary material).

The reduction of ATHB-8-GUS expression due to loss of HVE/CAND1 function suggested that HVE/CAND1 is required earlier to determine the sites of ATHB-8 expression, and prompted us to compare the timing of CAND1-GUS (P_ETA2-GUS)and ATHB-8-GUS expression throughout the development of wild-type first leaf primordia. Their expression was not detected in leaf primordia 2.5-3.0 days after germination (DAG; Fig. 6A,L). The onset of CAND1-GUS was observed in 3.5-DAG primordia (Fig. 6B), when their tips and presumptive midvein regions showed a mild GUS staining, and became stronger 4 or 4.5 DAG in the cells that will differentiate as the midvein(Fig. 6C,D). Later on (5 DAG),when the characteristic cell wall thickenings were apparent in the tracheids of the developing midvein, the promoter turned on in the secondary veins,which were arranged in the form of two loops(Fig. 6E). The veins in these loops displayed differentiated tracheary elements at 5.5 DAG, when two additional loops of secondary veins were developing in the basal region of the lamina. Expression of the transgene at this stage also revealed tertiary veins differentiating within the first two loops(Fig. 6F). GUS staining persisted in mature veins, beyond the procambial stage(Fig. 6F-K), and in young trichomes, and was diffuse in the leaf lamina. The veins of the first leaves of CAND1-GUS plants remained GUS-positive during the studied period,between 3.5 and 21 DAG.

GUS staining was stronger and less diffuse in ATHB-8-GUS than in CAND1-GUS plants. ATHB-8-GUS was strongly expressed in procambial strands and veins with differentiating tracheary elements in 3.5-to 7.5-DAG primordia (Fig. 6M-U). The region of the developing midvein appeared stained in 3.5- and 4-DAG primordia (Fig. 6M,N). The expression was then detected in the secondary veins of the most distal loops 4.5 DAG (Fig. 6O), and subsequently (5 DAG) in the basal loops(Fig. 6P). Only later (5.5 DAG), did the ATHB-8-GUS transgene reveal the differentiation of tertiary veins (Fig. 6Q).

To ascertain whether the reduced vein number of the hve mutants results from reduced responsiveness to auxin, we studied the expression of the DR5-GUS construct (Ulmasov et al., 1997), which drives GUS expression under the control of auxin response elements of the synthetic DR5 promoter. Previous studies have shown that DR5-GUS expression precedes and coincides with the appearance of procambial strands, and then disappears as the veins mature(Mattsson et al., 2003). The DR5-GUS reporter was not expressed in the veins of 21-day-old Col-0 and hve-3/hve-3 leaves from the first, third and seventh nodes (see Fig. S5 in the supplementary material), although it was in the leaf margin,the hydathodes and some mesophyll regions. For actively developing leaves of the 9th and upper nodes, by contrast, the DR5-GUS reporter was expressed at the tip, hydathodes, and tertiary and higher order veins of the proximal regions of the lamina in the wild type, but only at the tip in hve-3/hve-3 leaves.

We found no differences between the hve mutants and their wild types with regard to the root gravitropic response and skotomorphogenesis. Moderate 2,4-D resistance was displayed by these mutants (see Fig. S6 in the supplementary material), providing further evidence for the involvement of HVE in auxin responsiveness.

A number of genes required for vascular development and auxin transport and perception have been identified, including the abovementioned CVP1and CVP2; the recently cloned LOPPED1 (LOP1; also known as TORNADO1), which encodes a putative leucine-rich repeat protein of unknown function (Carland and McHale, 1996; Cnops et al.,2006); AUXIN RESISTANT1 (AXR1), which encodes a subunit of the RUB1-activating enzyme that promotes the modification of CUL1 with RUB1 (Lincoln et al.,1990; Leyser et al.,1993; del Pozo and Estelle,1999; del Pozo et al.,2002); and PINFORMED1 (PIN1), which encodes an auxin efflux carrier that contributes to the basipetal transport of auxin(Goto et al., 1987; Goto et al., 1991; Gälweiler et al., 1998). To investigate the functional relationships between HVE and these genes, we obtained double mutants and quantitatively analyzed their venation patterns. Leaf area, venation length and number of branching points in the vascular network were measured and used to calculate vascular density and number of branching points per surface unit(Table 2), as described previously (Candela et al.,1999).

We performed an HVE/hve-3 × PIN1/pin1^caycross, and identified hve-3/hve-3;pin1^cay/pin1^cay double mutants with an additive phenotype (Fig. 7, Table 2), suggesting that the minimal venation pattern conditioned by the null hve-3 allele is still dependent on auxin, as it is disrupted by the defective polar transport of this hormone. Phenotypic additivity suggests that auxin perception and transport operate independently but coordinately to pattern leaf veins. Identical conclusions were drawn from the study of hve-1/hve-1;pin1-1/pin1-1 double mutants (data not shown).

Loss of AXR1 function causes auxin insensitivity and a bushy inflorescence, the latter due to a decrease in the inhibition of apical dominance mediated by auxin (Estelle and Somerville, 1987; Lincoln et al., 1990; Stirnberg et al.,1999). The phenotypic similarities between hve/hve and axr1-12/axr1-12 plants, including their bushy inflorescence and low fertility, point to a functional relationship between HVE and AXR1. We did not find significant differences in vascular density and number of branching points per surface unit between axr1-12/axr1-12and hve-3/hve-3 first leaves[Table 2; this aspect of the axr1 phenotype has been reported by Deyholos et al.(Deyholos et al., 2003)]. However, the leaf epinasty of axr1-12 mutants (see Fig. S7 in the supplementary material) is not shared by hve mutants. We identified hve-3/hve-3;axr1-12/axr1-12 double mutants among the F2 progeny of an HVE/hve-3 × axr1-12/axr1-12 cross(Fig. 7, see Fig. S7 in the supplementary material). The hve-3/hve-3;axr1-12/axr1-12 leaves were epinastic, like those of HVE/-;axr1-12/axr1-12 plants (Fig. S7), and significantly smaller than those of hve-3/hve-3;AXR1/- individuals(Table 2). The vascular length and the number of branching points were similar in hve-3/hve-3;axr1-12/axr1-12 and hve-3/hve-3;AXR1/- plants,indicating similar levels of vascular development and the functional relationship of HVE and AXR1. However, the vascular density and number of branching points per surface unit were higher in the double mutant, the leaves of which were smaller than those of the parentals.

We identified hve-3/hve-3;lop1-65/lop1-65 double mutants in F2 families from an HVE/hve-3 × LOP1/lop1-65 cross in a 9:3:3:1 segregation (χ²=0.995; P=0.802; the presence of hve-3 was tested in phenotypically Lop1 and Hve Lop1 plants). The double mutants showed significantly smaller leaves (see Fig. S7 in the supplementary material, Table 2), with fewer branching points, fewer branching points per surface unit, and a shorter vascular network than those of their hve-3/hve-3;LOP1/- and HVE/-;lop1-65/lop1-65 siblings(Figs 7, S7 and Table 2). Their vascular densities, however, were found to be similar to those of hve-3/hve-3;LOP1/- plants. Vascular islands were also observed in some double mutant individuals (Fig. 7).

An additive phenotype was found in the hve-3/hve-3;cvp2-1/cvp2-1double mutants, which were identified in F2 families as plants with a simple venation pattern and disconnected secondary veins(Fig. 7, see Fig. S7 in the supplementary material). These disconnections occurred more frequently in the double mutants than in HVE/-;cvp2-1/cvp2-1 single mutants, an observation that suggests that HVE contributes to normal vascular connectivity, possibly by ensuring a certain level of vein thickness. Phenotypic characterization of HVE/-;cvp2-1/cvp2-1 plants and their hve-3/hve-3;cvp2-1/cvp2-1 F2 siblings demonstrated that leaf size,vascular network length and the number of branching points per surface unit were reduced in the double mutants and very similar to those of hve-3/hve-3;CVP2/- plants.

We identified hve-3/hve-3;cvp1-3/cvp1-3 double mutants in the F2 of an hve-3/hve-3 × cvp1-3/cvp1-3 cross. The cotyledons of these double mutants displayed an additive phenotype consisting of the simple vascular pattern characteristic of hve-3, combined with the vein disconnections characteristic of cvp1-3. We sequenced CVP1 in these plants to confirm that they were hve-3/hve-3;cvp1-3/cvp1-3. The venation patterns of hve-3/hve-3;cvp1-3/cvp1-3 and hve-3/hve-3;CVP1/- first rosette leaves were not significantly different(Table 2, Fig. 7). By contrast, the venation length, vascular density, number of branching points and the number of branching points per surface area were significantly smaller for HVE/-;cvp1-3/cvp1-3 plants than for their HVE/-;CVP1/-wild-type siblings (Table 2, Fig. 7). This role of CVP1 in the venation patterning of vegetative leaves had remained unnoticed before.

The pleiotropic phenotype of hve mutants includes a simple vascular pattern, an extremely branched inflorescence, altered root waving and reduced fertility. Impaired auxin homeostasis may well explain these phenotypic traits, which are reminiscent of mutants deficient in auxin perception or biosynthesis (Lincoln et al., 1990; Mullen et al.,1998; Rutherford et al.,1998). According to the canalization hypothesis(Sachs, 1991b), the spacing of veins is the outcome of a mechanism that requires that auxin exerts a positive feedback on its own transport. Cells that transport the hormone more efficiently become prospective vascular bundles after reaching a threshold of auxin concentration, while the resulting depletion of auxin in neighboring cells has the same effect as a lateral inhibition mechanism. This model predicts that decreased auxin sensitivity should result in a sparse vascular network, such as the one we observed in hve leaves.

Like axr1 mutants, although to a lesser extent, the hvemutants are insensitive to the synthetic auxin 2,4-D, in agreement with previous observations (Cheng et al.,2004; Chuang et al.,2004; Feng et al.,2004). By contrast, root gravitropism, skotomorphogenesis and sensitivity to the auxin transport inhibitor TIBA were normal in hve/hve plants. Unlike some agravitropic mutants that are insensitive to exogenous auxin, such as auxin resistant4 (axr4)(Hobbie and Estelle, 1995) and auxin-herbicide resistant1 (aux1)(Yamamoto and Yamamoto, 1998), hve/hve roots did not display altered gravitropism.

Following a map-based approach, we cloned HVE and found it to encode the Arabidopsis ortholog of mammalian CAND1. The important role of the HVE/CAND1 gene in vascular development has remained unnoticed so far, even though its cloning and the characterization of several mutant alleles have been reported by three research groups(Cheng et al., 2004; Chuang et al., 2004; Feng et al., 2004). Human CAND1 regulates the ubiquitination of target proteins by binding to unneddylated CULLIN1 (CUL1). CUL1, SKP1 and F-box proteins form SCF complexes that ligate ubiquitin moieties to proteins that will be degraded by the 26S proteasome. It has been proposed that the binding of CAND1 to unneddylated CUL1 hinders the interaction of SKP1 with CUL1 preventing the assembly of the SCF complex. After neddylation of CUL1, CAND1 is released and the SFC complex is assembled (Liu et al.,2002; Zheng et al.,2002; Hwang et al.,2003; Oshikawa et al.,2003).

Ubiquitin-mediated protein degradation is essential for auxin signaling(Dharmasiri and Estelle,2004). Auxin triggers the rapid degradation of AUX/IAA transcriptional repressors through the action of the SCF^TIR1complex (Gray et al., 2001). AUX/IAA proteins act by sequestering members of the ARF family of transcription factors (Ulmasov et al.,1997; Guilfoyle et al.,1998) that bind auxin response elements (AREs), a conserved sequence motif found in the promoters of auxin primary responsive genes(Guilfoyle et al., 1998). The identification of the F-box protein TIR1 as the auxin receptor thus provides a simple explanation for the rapid responses to auxin(Dharmasiri et al., 2005; Kepinski and Leyser, 2005). In Arabidopsis, CAND1 binds to unneddylated CUL1, as demonstrated in vivo (Feng et al., 2004) and in vitro (Chuang et al., 2004)(this work), and hence regulates the formation of SCF complexes involving the F-box proteins UFO, TIR1, COI1 and SLY1.

The fact that not only hve but also axr1 mutants have impaired leaf venation confirms the role of the ubiquitin pathway in the patterning of plant vascular tissues. AXR1 encodes a protein similar to the E1 ubiquitin-activating enzymes and is required for the conjugation of RUB1 to CUL1 (Leyser et al.,1993; del Pozo and Estelle,1999; del Pozo et al.,2002). As inferred from their similar number of branching points,the leaves of the hve and axr1 single mutants and hve axr1 double mutants show similar structural complexity, suggesting that both genes act in the same developmental pathway. As noticed previously(Cheng et al., 2004) for other phenotypic traits, the similar venation defects of axr1 and hve loss-of-function mutants are paradoxical, as one would expect them to have opposite phenotypes if the only role of CAND1 is to negatively regulate the assembly of SCF complexes. The recent finding that the deneddylation of CUL1 by the COP9 signalosome is enhanced by CAND1(Min et al., 2005) is equally paradoxical, as extra neddylation of CUL1 in loss-of-function hvemutants should also lead to a phenotype opposite to that of axr1mutants. Alternatively, by sequestering unneddylated CUL1, HVE/CAND1 may cause that only neddylated CUL1 is incorporated into functional SCF complexes. The incorporation of unneddylated CUL1 into the SCF complexes may impair their functionality through a dominant-negative effect, explaining the similar phenotypes of axr1 and hve mutants, the compromised functionality of the SCF complexes in cand1-1 mutants(Feng et al., 2004), and why,unlike null CUL1 alleles (Shen et al., 2002), null hve alleles are not lethal.

It has been proposed that the concentration of auxin that leads to the patterning of the primary and secondary veins is higher than the concentration required for other veins (Aloni et al.,2003). If a certain threshold of an auxin signal must be surpassed for the ground cells to adopt vascular fates, the co-occurrence of decreased auxin sensitivity and the lower auxin concentration required for the patterning of higher-order veins may explain why higher-order veins are more severely affected in hve mutants. We found no expression of the DR5-GUS reporter in hve-3/hve-3 leaves at the sites where higher-order veins are normally differentiating in the wild type. This suggests that loss of HVE function leads to a reduction in GUS expression driven by the auxin response elements of the synthetic DR5promoter, and further supports a correlation between an impaired auxin perception and the failure to form tertiary and higher-order veins.

Based on the correlation between venation density and venation branching found for different rosette leaves, we previously proposed that a common patterning mechanism operates in all of them(Candela et al., 1999). This hypothesis is reinforced by the simple vascular network of all the rosette leaves and laminar floral organs of hve mutants. The extent of the venation abnormalities caused by hve alleles indicates that HVE is a component of such a common mechanism. Nevertheless, the apparently normal venation pattern of hve cauline leaves also indicates the existence of organ-specific elements. The expression of CAND1-GUS in the hydathodes and margins of fully expanded cauline leaves suggests additional roles for HVE.

The dwarfism and extremely low fertility caused by hve-2 and hve-3 suggest that they are null alleles, as also proposed by Feng et al. (Feng et al., 2004). By contrast, hve-1 is probably hypomorphic, as its effects on fertility and overall plant and inflorescence size are not so severe. Mis-splicing of hve-1 transcripts generates at least two truncated predicted proteins that may keep some remnant function. Similarly, the phenotype caused by mis-splicing of another allele, Atcand1-1, is weaker than that of null alleles (Cheng et al.,2004). The similar vein pattern phenotypes conferred by hypomorphic and null alleles of HVE suggest that venation patterning is more sensitive to HVE loss of function than are plant size and fertility.

We studied the expression of ATHB-8-GUS, an early marker of vascular cell identity (Baima et al.,1995), in hve-1/hve-1 and hve-3/hve-3 plants. The reporter highlighted a simple vascular pattern with reduced vascular density and lacking higher order veins. Consistent with their early role, the expression of both HVE and ATHB-8 was detected before xylem differentiation in the midvein of the first leaves, and they evolved in a very similar way. Our results suggest that, at least in higher order veins, HVE acts to determine the location of new strands. Lack of HVE expression, however, does not preclude the formation of the primary vein, and only partially inhibits the formation of the secondary and tertiary veins. ATHB-8 expression seems to be required for vein-specific events after vein initiation sites are established, and depends on prior HVE activity in higher order veins, only partially so in the case of secondary and tertiary veins, and not at all in the primary vein. As this conclusion is solely based on the behavior of an ATHB-8-GUStransgene, further research will be required to shed light on the order of action of HVE and ATHB-8. Cleared leaves and paradermal and transverse leaf sections showed no traces of aborted or incompletely developed vascular strands at the sites where higher-order veins are lacking in the hve mutants, an observation indicating a perturbation of vascular initiation rather than final differentiation. The thin primary vein of the hve-3 mutant is a likely consequence of impaired auxin perception on the recruitment of provascular cells.

We investigated the functional relationship of HVE with several genes involved in venation pattern formation or auxin perception or transport,such as CVP1, CVP2, LOP1 and PIN1. The hve-3 pin1^cay double mutant was particularly informative, given that the simple vascular pattern of hve-3 was affected by the null pin1^cay allele. As PIN1 is an auxin transporter, this result shows that the minimal vascular pattern of hve leaves still depends on auxin for its development and that alternative mechanisms of auxin perception are at work in the mutant. The additive hve-3 pin1^cay phenotype was as to be expected if auxin perception and transport operate independently to pattern the vascular network. As the architecture of the venation pattern seems to be very sensitive to alterations in either of these two factors, it follows that they must act coordinately. Further support for these conclusions came from the application of the auxin transport inhibitor TIBA to the hve-1 mutant, which caused the same range of abnormalities as in the wild type.

The phenotypes of hve-3 lop1-65, hve-3 cvp1-3 and hve-3 cvp2-1 double mutants were also considered to be additive, suggesting that additional independent processes are required for vein development. The free-ending higher-order veins of cvp2-1 single mutants indicate a role for CVP2 in the connectivity of vascular strands. The secondary veins of hve-3 cvp2-1 double mutants were often disconnected,suggesting that the secondary veins of hve-3 are equivalent to the higher-order veins of the wild type, as inferred from their similar sensitivity to the loss of CVP2 function. This equivalence suggests a gradual establishment of the venation pattern and may be explained by an impaired auxin sensitivity leading to a premature arrest of the patterning process in the mutant, which is consistent with the mechanism proposed by the canalization hypothesis.

We thank F. Carland, S. Baima, W. Gray, T. Guilfoyle, T. Berleth, the NASC and the ABRC for providing seeds; P. Robles and V. Quesada for comments on the manuscript; and J. M. Serrano and V. García-Sempere for technical assistance. This work was supported by grants BMC2002-02840 and BFU2005-01031 to J.L.M., and by a fellowship to M.M.A.-P. from the Ministerio de Educación y Ciencia of Spain.