The FORKED genes are essential for distal vein meeting in Arabidopsis

The spatial arrangement of the vascular bundles within plant leaves is crucial to plant function as it enables efficient transport of water, minerals and photosynthates and provides mechanical support to the leaf. Two major events are thought to have been fundamental in the evolution of the closed reticulate system present in all higher plant leaves: (1) the evolution of a branching system and (2) the formation of a closed pattern. The most primitive leaf vascular pattern, found in microphylls, consists of only a single vein or a pair of parallel veins running the length of the leaf(Wagner et al., 1982; Gifford and Foster, 1989). Primitive vascular plants bearing megaphylls evolved a more complex, open,dichotomously branching vascular pattern that is proposed to have conferred an advantage in dropping atmospheric CO₂ levels(Beerling et al., 2001). In more advanced vascular plants, including some members of the ferns and gymnosperms and all angiosperms, a closed venation pattern evolved through joining of distal branches (Trivett and Pigg, 1996). Distal meeting is proposed to be advantageous because it provides both redundancy in transportation routes in case of injury or blockage of a vascular bundle and increased mechanical stability and support to the leaf, particularly along the leaf margin (Roth-Nebelsick, 2001).

Both classes of angiosperms, the monocots and dicots, have a complex,branched and closed leaf venation pattern. In monocots the pattern is called striate with a series of longitudinal veins running approximately parallel to each other, meeting at the apex of the leaf and interconnecting through a multitude of smaller transverse veins. In dicots, the pattern is reticulate with a basic pattern of secondary veins branching from the midvein and connecting to one another or to higher order (tertiary and quaternary) veins at their distal ends near the leaf margin.

The primary model to explain vascular patterning, the auxin canalization model, is supported by evidence indicating a central role for auxin in vascular differentiation within the stem(Sachs, 1981; Sachs, 1989; Sachs, 1991), in isolated mesophyll cells (Church, 1993; Berleth and Sachs, 2001) and in leaves (Mattson et al., 1999; Sieburth,1999; Aloni, 2001). According to the auxin canalization model, an initially homogenous field emanates from auxin sources, but random fluctuations expose some cells to increased auxin leading to vascular cell differentiation and increased efficiency in auxin transport. The increased transport results in two classes of cells: (1) those to which auxin is transported form vascular tissue and (2)those from which auxin is drained form nonvascular tissue.

Auxin is thought to be transported in a primarily basipetal manner throughout the plant body from primary sources of synthesis(Wallroth-Marmonr and Harte,1988; Lomax et al.,1995), although recent studies indicate that the shoot apical meristem and very young leaf primordia are auxin sinks to which auxin is acropetally transported (Reinhardt et al.,2000; Avsian-Kretchmer et al.,2002). Basipetal transport from young developing leaves is proposed to direct stem vasculature since removal of leaves eliminates vascular differentiation in the stem below and can be compensated by exogenous auxin (Sachs, 1981). Leaf venation is proposed to be directed by basipetal transport of auxin from the leaf margin (Mattson et al., 1999; Sieburth, 1999; Aloni, 2001; Avsian-Kretchmer et al., 2002),a mechanism consistent with auxin response patterns in regions that will become procambium (Mattsson et al.,2003). Leaves in which polar auxin transport is inhibited either chemically (Goto et al., 1991; Okada et al., 1991; Gälweiler et al., 1998; Sieburth, 1999), or genetically, through eliminating PIN-FORMED 1 (PIN1), a component of the auxin efflux carrier (Gälweiler et al.,1998; Müller et al.,1998; Mattsson et al.,1999; Steinmann et al.,1999), develop increased vascularization adjacent to the leaf margin.

Consistent with a role for auxin in vascular patterning, a number of Arabidopsis mutants known to affect aspects of auxin transport or response have defects in vascular development. Mutations in MONOPTEROS(MP), BODENLOS (BDL) or AUXIN RESISTANT 6 (AXR6) show early defects in embryonic apical basal patterning followed by a reduction in cotyledon venation (Berleth and Jurgens,1993; Hamann et al.,1999; Hobbie et al.,2000). MP encodes ARF5, an auxin response factor that activates auxin response targets, while BDL encodes IAA12, a protein that has been shown to bind to ARF5 and prevent its activity(Hamann et al., 2002). AXR6 encodes the protein cullin, a subunit of the ubiquitin ligase complex SCF (Hobbie et al.,2002). Mutations in SCARFACE (SFC) result in plants that are more sensitive to auxin and show reduction and discontinuity of venation within foliar organs (Deyholos et al., 2000). Plants mutant for PINOID, which encodes a serine-threonine kinase believed to affect either auxin signaling or auxin transport, show altered venation within floral organs, while mutants in LOPPED (LOP) are defective in basipetal auxin transport and alter leaf venation (Carland and McHale,1996; Christensen et al.,2000; Benjamins et al.,2001). Alleles of GNOM (EMB30), such as van7 (vascular network 7)(Koizumi et al., 2000), result in discontinuous venation in cotyledons and increased marginal venation in leaves, a leaf phenotype similar to pin1 and consistent with the proposed role of GNOM in PIN1 localization(Steinmann et al., 1999). While the interactions amongst these genetic factors are not yet well understood, the frequent association of defective auxin signaling or transport with defective vascular patterning clearly points to a primary role for auxin in establishing vascular pattern within shoots.

We report the isolation and characterization of mutants in two novel Arabidopsis thaliana genes, FORKED1 (FKD1) and FORKED2(FKD2), crucial to the formation of the closed leaf vascular pattern characteristic of dicot leaves. Recessive mutations in either FKD1 or FKD2 result in a failure of distal portions of the vascular bundles to form connections with the remaining leaf vascular network, resulting in an open leaf venation pattern reminiscent of primitive vascular plants. Our analysis suggests that FKD1 responds to a particular auxin threshold and allows vascular development, and that this action is redundant to that of FKD2 except at the lowest auxin levels. The function of these genes is of particular interest as the closed leaf vascular pattern is ubiquitous within the angiosperms and appears to have been important in their evolution.

All seed material was from the Arabidopsis Biological Resource Center (Columbus, Ohio) except the ethyl methane sulfonate (EMS) mutagenized seed (Columbia; Col), from Lehle Seeds (Round Rock, TX), pin1-1 and mp^G92 seed provided by T. Berleth (University of Toronto,Toronto, ON) and DR5::GUS seed provided by J. Murfett (University of Missouri, Columbia, Missouri).

Seed were sown on Metromix 200 (W. R. Grace Co., Marysville, OK) in 100 cm² pots or on A. thaliana (AT) growth medium(Wilson et al., 1990) with or without 30 μM naphthylphthalamic acid (NPA; Sigma Chemical Company) in 78.5 cm² Petri plates. Following 4-5 days stratification, seeds were transferred (considered as the day of germination: 0 Day After Germination; DAG) to growth chambers (Percival Scientific, Perry, IA)set for 21°C, 60% relative humidity, and continuous light (130 mol second^-1 m^-2) provided by a combination of Sylvania Cool White, Gro Lux, and incandescent bulbs (Osram Sylvania Inc., Danvers, MA). To assess root growth, 5-day old seedlings vertically grown on plates were transferred to medium containing either 1.0×10^-6,1.0×10^-7 or 1.0×10^-8 M 2,4-dichlorophenoxyacetic acid (2,4-D). Root growth was measured from the position of the root tip at transfer to the position 4 days later.

For isotopic analysis, wild-type and fkd1 plants were greenhouse grown (16-hour days) on soil (as described above) and leaves were harvested when the shoot was 2 cm long. A 1-2 mg subsample of ground leaf material was sealed in a tin capsule and loaded into the elemental analyzer (NC2500, CE Instruments, ThermoQuest Italia, Milan, Italy) and the diatomic nitrogen and carbon dioxide gases generated therein were separated in a gas chromatographic column and passed directly, using a helium stream, to the inlet of the mass spectrometer (Delta Plus, Finnigan Mat, San Jose, CA, USA) for quantification and measurement of stable isotope ratios. Carbon stable isotope ratios(¹³C/¹²C) were expressed in delta notation (δvalues presented in parts per thousand) where the international standard is CO₂ from Pee Dee Belemnite (PDB) limestone(Farquhar et al., 1989).

Approximately 6000 EMS-mutagenized M₂ seed were sown at a density of 50 seed per pot. A cotyledon and first leaf was taken 14 DAG,mounted in low viscosity Cytoseal (Stephen's Scientific, Kalamazoo, MI) and screened for abnormalities in vascular patterning using a dissecting microscope (Stemi 2000, Carl Zeiss Inc., Thornwood, NY). M₃ seed of potential mutants was collected and re-screened. Lines with heritable phenotypes were backcrossed, using wild type as the female, at least twice before characterization.

Wild-type and fkd1 seeds (20 per plate) were sown on AT medium with or without NPA and plants taken every 24 hours from 1 DAG. Histochemical localization of GUS activity and subsequent clearing was performed as described previously (Kang and Dengler,2002). For photography, mature cotyledons (14 DAG) and first leaves (21 DAG) of all genotypes, and fully open fkd1, fkd2 and fkd2 fkd1 flowers were removed and cleared in a solution of 3:1 ethanol:acetic acid for 2-4 hours, 70% ethanol for 1 hour, 95% ethanol overnight and 5% NaOH for 1 hour at 60°C. Dissections were performed in a 50% aqueous glycerol solution and samples viewed with a compound light microscope (Eclipse E600, Nikon, Mississauga, ON).

Cotyledons (14 DAG) and first leaves (21 DAG) of all genotypes were mounted in Cytoseal, images captured with a CCD camera (RS-170, Cohu Inc., Electronics Division, San Diego, CA) attached to a dissecting microscope (Stemi 2000, Carl Zeiss Inc., Thornwood, NY), and assessed using NIH Image(http://rsb.info.nih.gov/nih-image/). Whole cotyledons were scored for numbers of secondaries (veins attached at least at one point to the midvein) and cotyledons and half leaves were scored for number of branch points (two or more veins meeting), vein endings and areoles (any area of the leaf blade completely bounded by veins). Cleared cotyledons and leaves were scored for the percentage showing vascular islands(VIs). Wild-type, fkd1 and fkd2 plants (4 per pot) were scored for germination, total number of leaves, number of secondary stems and time to flowering (days). Statistical differences were determined using Student's t-test.

Cotyledons (14 DAG) and first leaves (21 DAG) of wild type and fkd1 were prepared for sectioning by vacuum infiltration in FAA(18:1:1 70% ethanol:formalin:glacial acetic acid), stored at 4°C overnight, dehydrated through an ethanol series, and embedded in Spurr's resin. Five μm sections were cut using a glass knife and stained with Toluidine Blue before being viewed with a compound light microscope (Eclipse E600, Nikon, Mississauga, ON).

Plants mutant for FKD1 and FKD2 were crossed to Landsberg erecta (Ler) ecotype and DNA was extracted from F₂ plants exhibiting the fkd1 or fkd2 phenotype(Edwards et al., 1991). Mapping using simple sequence length polymorphisms (SSLPs) between the Col and Ler backgrounds was done using standard PCR conditions(Bell and Ecker,1994) and primers (Research Genetics Inc., Huntsville, AL).

Double mutants were generated as described in Table 1. F₃ progeny from axr2 fkd1, aux1-7 fkd1 and fkd2 fkd1 double mutants, F₃ progeny segregating for pin1-1 fkd-1, mp fkd1 or axr6-2 fkd1 double mutants and F₄ progeny from axr1-3 fkd1 were characterized. To generate the fkd1 DR5::GUS line, fkd1 plants were crossed to plants homozygous for DR5::GUS. F₂ progeny were screened on plates with 10μg/ml kanamycin, surviving plants showing the fkd1 phenotype were allowed to self, and three F₃ families entirely resistant to kanamycin were characterized.

All images were captured using a digital camera (Coolpix 990, Nikon,Mississauga, ON) and were prepared for publication using Adobe Photoshop 5.0(Adobe Systems Inc., San Jose, CA).

We screened an M₂ population (gl1-1, Col background) of EMS mutagenized plants for defects in leaf vascular patterning and chose, for further characterization, two mutant lines [forked1 (fkd1)and forked2 (fkd2)] with vascular bundles that fail to meet distally in both the cotyledons (Fig. 1F,G) and leaves (Fig. 2B,C).

Complementation tests between fkd1 and fkd2 yielded all wild-type plants (n=106). Crosses of either fkd1 or fkd2 to wild type yielded an F₁ of wild-type phenotype(n=36 and n=35, respectively). In the F₂ the fkd1 and fkd2 phenotypes segregated from the wild-type phenotype with a ratio of 3:1 (188 wild type: 53 fkd1,χ ²=1.037, P>0.25; 178 wild type: 57 fkd2,χ ²=0.070, P>0.75). We therefore concluded that the fkd1 and fkd2 phenotypes are each the result of a single,nuclear, recessive mutation.

For mapping, fkd1 and fkd2 plants were crossed into the Ler background (Bell and Ecker,1994). The FKD1 gene mapped to 89.48 cM on chromosome III based on recombination with nga112 (1.6%, n=562 chromosomes) and nga6(3.8%, n=210 chromosomes). FKD2 mapped to 30.2 cM on chromosome V based on recombination with ciw8 (11.8%, n=152 chromosomes) and nga106 (14.0% n=50 chromosomes). This places FKD2 near a previously described gene involved in vascular patterning, SFC (Deyholos et al.,2000). Plants mutant for SFC have severely altered morphology and leaves with highly disrupted vascular patterning, including VIs. Unfortunately, sfc seed was not available so we were unable to determine through complementation if fkd2 is a weak allele of sfc or a mutation in a novel gene. For this reason, detailed developmental and double mutant characterization was performed only on fkd1 plants.

In order to assess the differences between wild-type, fkd1 and fkd2 mature cotyledons, we first quantified wild-type numbers of vein branch points, areoles, freely ending veins and secondary veins(Table 2). Most commonly, the vascular pattern of the mature wild-type cotyledon consists of a midvein and 4 secondary veins (two distal and two proximal) that meet with the midvein and one another to generate four closed loops(Fig. 1E).

To determine the sequence of wild-type cotyledon vascular pattern development, we examined seedlings at 24-hour intervals and assessed the timing of developmental landmarks (Table 3, Fig. 1). By 1 DAG, provascular tissue of midvein and distal secondary veins is complete(Fig. 1A), the secondary veins connecting to the midvein at distal (Table 3, Fig. 3A,B) and proximal points (Fig. 3I,J). Midvein maturation (appearance of cell wall thickenings) begins at 1 DAG and is complete by 2 DAG, while distal secondary vein maturation is complete by 3 DAG. Both proximal secondary veins are initiated perpendicular to the distal secondary vein by 2 DAG, and maturation is initiated by 3 DAG(Fig. 1B, Fig. 3F). Completion is variable; at cotyledon maturity (14 DAG), some (23%, n=46) proximal secondary veins never join the midvein. Secondary veins develop either basipetally (66%, n=30 distal veins; 95%, n=758 proximal veins), or bidirectionally.

In mature (14 DAG) fkd1 and fkd2 cotyledons the distal connections between the secondary veins and the midvein, and between the proximal and distal veins often do not form(Fig. 1F,G), resulting in fewer areoles and branch points (Table 2). In fkd1 cotyledons, the complexity of the venation pattern is unaffected relative to wild type, whereas fkd2 cotyledons have a simpler pattern (Table 2).

The lack of distal vein meeting is evident early in fkd1 cotyledon development. At 1 DAG, the provascular tissue of one or both of the distal secondary veins fails to join the midvein(Table 3, Fig. 1C, Fig. 3C,D). Furthermore, fkd1 proximal secondary veins initiate at a point distant from the existing distal secondary vein (Fig. 3G) and initiation is delayed relative to wild type(Table 3). Other aspects of fkd1 vein development are similar to wild type, except that all secondary veins connect proximally with the midvein.

The vascular pattern of wild-type first leaves consists of a midvein from which regularly spaced secondary veins extend to the leaf margin where they join one another (Fig. 2A). Tertiary veins form connections between the secondary veins or occasionally end freely in the lamina, while quaternary veins usually end freely in the lamina. To compare the fkd1 and fkd2 leaf venation pattern to that of wild type, we quantified the number of branch points, freely ending veins and areoles of first leaves 21 DAG(Table 2).

The midvein of the wild-type leaf differentiates acropetally until it reaches the distal tip of the developing leaf 5 DAG where the distal secondary veins are initiated (Table 3, Fig. 4A). These develop basipetally, meeting the midvein between 6 and 7 DAG(Table 3, Fig. 4B). Most remaining secondary and tertiary veins initiate from previously formed, more distal veins and develop basipetally to join the vascular network at their proximal end (Fig. 4C). At least one vein (excluding quaternary) in 53% (n=38) of leaves at 21 DAG fails to rejoin the vascular network proximally.

In fkd1 and fkd2 first leaves the distal ends of most veins fail to join previously formed veins and end freely in the lamina(Fig. 2B,C), resulting in an open venation pattern with fewer branch points and areoles and more vein ends than wild type (Table 2). Proximal non-meeting of veins occurs in fkd1 with a similar frequency(43%, n=43) as in wild type, resulting in VIs in mature leaves(Table 2). VIs occur more frequently in fkd2 first leaves and are distributed throughout the leaf blade (Fig. 2C). Cross sections of fkd1 leaves are indistinguishable from wild type (data not shown).

In fkd1 leaves, development of the midvein, initiation of distal secondary veins and their proximal joining with the midvein is slightly delayed relative to wild type (Table 3). Initiation of proximal secondary veins and subsequent secondary and tertiary veins occurs at a point distant from the previously formed distal secondary veins (Fig. 4G,H), but further development of all veins is normal(Fig. 4G,H).

Given the crucial function of the vascular system, we expected that the loss of the reticulate venation pattern might result in a decreased growth rate or photosynthetic capacity. All characters analyzed were indistinguishable between wild type and fkd1. In contrast, fkd2 plants produced fewer rosette leaves, flowered later, and produced less seed (Table 4). The similar photosynthetic capacities and growth rate in fkd1 and wild-type plants suggests either that the altered vascular pattern has no effect or that the fkd1 plants are compensating for their non-meeting venation by altering another component of the transpiration mechanism. Leaf carbon isotope composition (expressed using δ notation with units of parts per thousand [δ¹³C,%thou]) provides information about the ratio of photosynthetic capacity to stomatal conductance: higherδ ¹³C values (less negative) mean a lower stomatal conductance in relation to photosynthetic capacity(Farquhar et al., 1989). The significantly lower δ¹³C in fkd1 than in wild type suggests that the fkd1 vascular pattern provides less efficient water delivery that is compensated by increased stomatal conductance.

To assess any effect altered vascular pattern might have on leaf shape, we compared shape and size of fkd1 and fkd2 cotyledons and leaves to wild type (Table 5). fkd2 cotyledons are elongated compared to wild type and are somewhat smaller, while fkd1 cotyledons are slightly larger. First leaves of both fkd1 and fkd2 are elongated relative to wild type. The increased length suggests that distal vein connections in wild type may constrain cellular expansion during leaf development, a constraint lacking in fkd1 and fkd2 leaves.

Since leaves are considered the progenitors of floral organs, one might expect a similar loss of the wild-type reticulate venation pattern to occur in fkd1 and fkd2 floral organs. In fkd1 and fkd2 sepals, distal meeting is reduced and VIs are increased compared to wild type (Fig. 5, Table 6), while in petals,distal meeting is reduced.

To determine if FKD1 and FKD2 act in the same or different pathways, we generated plants doubly mutant for fkd1 and fkd2. First leaves of the F₂ progeny of the fkd1× fkd2 cross were scored as wild type if their leaves showed no distal non-meeting, as fkd1 if they showed distal non-meeting and VIs concentrated to the proximal portion of the leaf and as fkd2 if they showed distal non-meeting and VIs throughout the leaf lamina. We identified plants with a vascular phenotype more extreme than either single mutant(Fig. 2D), whose frequency was consistent with the plants being doubly homozygous for both fkd1 and fkd2 (Table 1). Relative to either single mutant, cotyledons of the double mutant were considerably smaller (Table 5)and had a simplified vascular pattern with more VIs. As in both single mutants, distal meeting of secondary veins was rarely seen. These trends result in a significant decrease in both the number of areoles and branch points relative to the single mutants(Table 2). The vascular pattern of double mutant first leaves consisted almost entirely of VIs, resulting in fewer areoles and more free ends than either single mutant(Table 2). Moreover, leaves were smaller and more elongated (Table 5). The increased severity of the venation phenotype was also seen in floral organs (Fig. 5D,E),with reduced distal meeting and increased frequency of VIs(Table 6). The increased severity in double mutant phenotype compared to either single mutant suggests that FKD1 and FKD2 have overlapping and partially redundant functions.

One explanation for the fkd1 phenotype is that the mutation affects a component of auxin canalization and compromises distal vein meeting. If correct, one might expect that fkd1 plants would show altered sensitivity to changes in auxin levels, whether introduced exogenously or through double mutant combinations. Since altered sensitivity of root growth to 2,4-D indicates altered auxin response or transport(Wilson et al., 1990), we compared root growth of fkd1 to that of wild type at four 2,4-D concentrations. No significant difference was observed at any concentration(data not shown), suggesting that fkd1 plants do not show a global change in auxin sensitivity.

While the auxin root assay effectively detects defects in the general auxin pathway, it may not detect defects specific to leaf vascular pattern formation. We therefore grew plants on auxin transport inhibitors and assessed alterations in leaf vascular pattern. As well, we generated double mutants between fkd1 and various auxin mutants known to have alterations in leaf morphology and/or leaf vascular patterning.

The proliferation of veins adjacent to the margin in wild-type leaves treated with auxin transport inhibitors suggests that the leaf margin is a major source of auxin within the developing leaf(Mattsson et al., 1999; Sieburth, 1999). To assess whether, in the presence of an auxin transport inhibitor, fkd1 alters the ability of marginal veins to meet, we grew fkd1 and wild-type seedlings on 30 μM NPA. Mature (14 DAG) fkd1 cotyledons had the same vascular pattern when grown with or without NPA (data not shown). Extensive marginal venation at the distal leaf tip was evident in both wild-type and fkd1 leaves by 5 DAG, and vascular strands began to develop basipetally from the marginal region to the proximal leaf blade at 7 DAG. However, by 8 DAG, the interior of the wild-type leaf blade showed extensive vein branching, whereas in fkd1 no branching was evident in the leaf interior. The marginal regions of fkd1 and wild-type 21 DAG leaves were indistinguishable, however, significantly fewer secondary veins extended from the marginal region to the proximal leaf blade in fkd1(14.74±0.88) than in wild type (20.5±0.78) and tertiary veins were rare (Fig. 2G,H). Furthermore, 20% (n=23) of fkd1 leaves had secondary veins that did not connect proximally, whereas all veins in all wild type leaves were connected (n=25). The ability of veins to meet distally in marginal areas of fkd1 leaves where auxin is increased owing to reduced transport suggests that FKD1 either acts in response to auxin to direct vascular tissue or is necessary for the auxin response that directs vascular tissue.

Like chemical inhibition of auxin transport, the loss-of-function allele, pin1-1, in the auxin efflux carrier results in a proliferation of marginal venation (Mattsson et al.,1999; Galweiler et al.,1998). pin1-1 cotyledons and leaves are similar to wild type in size and shape, although the cotyledons are slightly rounder, the leaves are slightly smaller, and are sometimes fused(Fig. 2E and Table 5). The venation pattern of the cotyledons shows no difference from wild type while the leaf pattern is quite distinct, being simpler (fewer areoles and branch points) and having fewer freely ending veins (Table 2).

pin1-1 fkd1 cotyledons show no significant difference from pin1-1 cotyledons for any of the characters measured(Table 2). Relative to pin1-1, the first leaves of pin1-1 fkd1 have more freely ending veins and fewer areoles, consistent with the increased distal non-meeting of fkd1. Relative to fkd1, they show a decrease in freely ending veins, consistent with the increased vein meeting of pin1-1. Specifically, the increased meeting in double mutants occurs along the margin while the distal non-meeting occurs in the leaf blade(Fig. 2F). Therefore, like treatment of fkd1 leaves with NPA, loss of PIN1 compensates for lack of FKD1 at the leaf margin, but not in the internal regions of the leaf.

Plants mutant for any of AXR1, AXR2 or AUX1 are auxin resistant, with various defects in morphology. AXR1 acts in an ubiquitin-like pathway that responds to the presence of auxin by targeting proteins for degradation (del Pozo and Estelle, 1999; Gray and Estelle, 2000), AXR2 belongs to the IAA family of genes that are inducible by auxin (Nagpal et al., 2000) and AUX1 encodes a membrane protein that is believed to be a component of the auxin influx carrier(Bennett et al., 1996; Marchant et al., 1999; Swarup et al., 2000). We found that both axr1-3 and axr2 cotyledons and leaves are smaller than wild type (Table 5) with a simpler vascular pattern (Table 2, Fig. 2M) that in axr1-3 is combined with frequent proximal non-meeting(Table 2). In contrast, aux1-7 cotyledons and leaves are significantly larger but maintain the wild-type vascular pattern, and aux1-7 first leaves are more elongate (Table 5).

In double mutant combinations between these auxin resistant mutants and fkd1, both leaf morphology and leaf vascular patterning show essentially additive phenotypes (Tables 2, 5). Relative to the respective single mutants, cotyledons and first leaves of axr1-3 fkd1, axr2 fkd1and aux1-7 fkd1 have significantly fewer branch points and areoles,consistent with the double mutant phenotypes combining the distal non-meeting of fkd1 with the simplified vascular pattern of the auxin mutants(Table 2 and Fig. 2N).

Plants with loss-of-function MP alleles or gain-of function AXR6 alleles do not produce the basal embryonic structures, hypocotyl and root, but produce normal apical embryonic structures, including shoot apical meristem and cotyledons (Berleth and Jurgens, 1993; Przemeck et al., 1996; Hobbie et al.,2000). AXR6 mutants rarely form a small number of rosette leaves; mp seedlings do not normally form leaves, but seedlings carrying weak MP alleles, such as mp^G92, can be induced by wounding to form roots, and the rooted seedlings go on to form leaves and inflorescences (Przemeck et al., 1996). Within these structures, defects occur in the leaf venation, including simplification of pattern and discontinuities in vascular strands. We found that mp^G92 and axr6-2cotyledons have few distal secondaries arising from the midvein, and that these do not connect proximally with the midvein(Fig. 1I).

Many of the shoot defects seen in mp^G92 and axr6-2 plants are suppressed in the mp^G92fkd1 or axr6-2 fkd1 double mutants, while the morphology and development of double mutant basal structures is indistinguishable from mp^G92 or axr6-2 respectively. The vascular pattern of mp^G92fkd1 cotyledons is simpler than that of mp^G92, while the vascular pattern of axr6-2 fkd1 cotyledons is slightly more extensive than axr6-2(Fig. 1J,L, Table 2) with increased frequency of VIs.

The remainder of the double mutant shoot structure diverges markedly from that of mp^G92 or axr6-2: while mp^G92 or axr6-2 plants grown under the same conditions only rarely (mp^G92 -16.2%, n=37; axr6-2 -10%, n=19) develop a very small first leaf with highly reduced venation (Fig. 2I,K), 64.8% (n=54) of mp^G92fkd1 double mutants and 39% (n=33) of axr6-2 fkd1double mutants show one to four small but well developed leaves, with extensive venation (Fig. 2J,L). Because mp^G92fkd1 and axr6-2 fkd1double mutants rarely survived to 21 DAG, at which point we consider first leaves to be fully expanded, we could not make a quantitative comparison of their vascular pattern to either wild type or fkd1. However,examination of 20 mp^G92fkd1 and 12 axr6-2 fkd1 first leaves indicate that they show a simplified vascular pattern with reduced numbers of tertiary and quaternary veins(Fig. 2J,L), and that, as in fkd1 leaves, the distal ends of veins often do not meet. Moreover, if mp^G92fkd1 plants survive on medium for 21 days,70% (n=10) form inflorescences, usually consisting of a single terminal flower, often reduced to a pistil.

To test the hypothesis that the fkd1 phenotype is the result of reduced ability to respond to auxin, we introduced the synthetic AuxRE reporter gene construct (DR5::GUS) into fkd1 plants. The DR5::GUS line contains a composite promoter with seven tandem repeats of the AuxRE TGTCTC fused to a minimal cauliflower mosaic virus 35S promoter-GUS reporter gene construct(Ulmasov et al., 1997). Initial reporter gene expression in the subapical cells of the 3 DAG fkd1 leaves was indistinguishable from wild type (not shown). However, reporter gene expression in fkd1 cells destined to be secondary veins was reduced in both intensity and duration relative to wild type (Fig. 6). Moreover,whereas in wild type, reporter gene expression at the distal junctions of secondary veins was somewhat reduced or delayed relative to the rest of the vein (see arrows, Fig. 6A,C)(Mattsson et al., 2003), in fkd1, such cells never expressed the reporter gene. In support of the hypothesis that auxin transport inhibition alleviates the fkd1phenotype by increasing auxin levels and allowing increased auxin response,both wild-type and fkd1 leaves treated with NPA showed a broad band of intense reporter gene expression (Fig. 6I,J) that preceded the formation of vascular tissue(Fig. 6K,L).

We have identified two mutants of Arabidopsis, forked1(fkd1) and forked2, that form an open vascular pattern in foliar organs because newly initiating veins do not join with previously formed veins. Owing to their recessive nature, we will assume both alleles represent loss of function of the respective genes FORKED1(FKD1) and FORKED2 (FKD2). While both genes are necessary at distal vein junctions, they act redundantly throughout the leaf veins. Auxin responsive reporter gene expression (DR5::GUS) is reduced in fkd1 plants. DR5::GUS expression and vein junctions are restored in areas of fkd1 leaves where auxin concentrations are increased. We therefore propose that FKD1 and FKD2 are necessary for the auxin response that directs vascular development, and that distal vein junctions, being sites of low auxin concentration, are particularly sensitive to their reduced function.

Several models for the role of FKD1 in vascular pattern formation are consistent with the fkd1 phenotype: (1) FKD1 is a component of the auxin synthesis or transport pathway that controls distribution of auxin; (2) FKD1 is necessary for differentiation of vascular tissue in response to auxin; (3) FKD1 is necessary for the auxin response that results in vascular differentiation; or (4) FKD1acts independently of, but in conjunction with, auxin canalization to induce vascular tissue in areas where auxin alone is insufficient. To distinguish amongst the possibilities, we assessed auxin responsive reporter gene expression (DR5::GUS) in fkd1, the response of fkd1to synthetic auxin and auxin transport inhibitors and the response of fkd1 to mutant backgrounds that affect auxin response or transport.

Our data provide the most support for Model 3. The expression of DR5::GUS is reduced in duration and extent throughout fkd1leaves, suggesting that FKD1 is required for the auxin response. The reduced auxin response is sufficient to direct vascular development throughout most of the leaf, although it results in a slight developmental delay. At distal junctions, the level of auxin is sufficiently low that no auxin response occurs in the absence of FKD1 and distal junctions do not form. When fkd1 seedlings are grown on medium containing an auxin transport inhibitor or when double mutants are made between fkd1 and the auxin transport mutant pin1-1, distal meeting of marginal veins in either cotyledons or leaves is more similar to wild type. Moreover, when auxin transport is inhibited, DR5::GUS expression in fkd1leaves is indistinguishable to that in wild type. Thus, the increased auxin at the margin following auxin transport inhibition reestablishes auxin response even in the absence of FKD1 function. Our data suggest that FKD1 activity is restricted to the leaves and is not present globally; however, the additive phenotypes of double mutants between fkd1 and the auxin response mutants axr1-3, axr2 and aux1-7, as well as the root response of fkd1 plants grown in 2,4-D, may be the result of redundant activities in other tissues.

The hypothesis that FKD1 is necessary for the auxin response is further supported by the effect of fkd1 on both mp^G92 and axr6-2 phenotypes. Plants homozygous for mp^G92 or axr6-2 show a similar,seedling lethal phenotype, forming no root, an abnormal, stubby hypocotyl,cotyledons with reduced venation and rarely a first leaf. Double mutants with fkd1 show no change in the root or hypocotyl phenotype, but like mp sfc double mutants, mp^G92fkd1 show enhanced vascular defects in the cotyledons. More surprisingly, axr6-2 fkd1 or mp^G92fkd1 double mutants exhibit suppression of the mp^G92 or axr6-2 leaf phenotypes, producing several well-expanded leaves and in the case of mp^G92fkd1, an inflorescence. MP encodes the auxin response protein ARF5 (Hardtke and Berleth, 1998), while AXR6 encodes the protein cullin, a subunit of the ubiquitin ligase complex SCF(Hobbie et al., 2002), The similarities in mp, axr6 and bdl phenotypes suggest that AXR6 may degrade IAA12 (product of BDL) and allow activity of ARF5. The enhancement of the mp^G92 cotyledon phenotype by fkd1 suggests that in cotyledons, FKD1 acts redundantly with MP. While it is difficult to find a simple explanation for the suppression of mp^G92 and axr6-2 leaf phenotypes in a fkd1 background, the most likely interpretation is that, like MP and AXR6, FKD1 is involved in the auxin response. While it is tempting to speculate that FKD1, MP and AXR6 act as a complex required for auxin response, the function of which is restored if pairs of genes are mutated, the proposed activities of MP and AXR6 products and the fact that mp^G92 introduces a nonsense codon indicate that this is unlikely. A second possibility is that the suppression is due to indirect changes in auxin distribution caused by altered source/sink relationships resulting from mp^G92 and axr6-2morphological defects. In wild-type seedlings, the initial source of auxin is probably the cotyledons (Ljung et al.,2001; Bhalerao et al.,2002; Marchant et al.,2002), while the region below the quiescent centre (QC) acts as a sink for auxin (Sabatini et al.,1999) that controls a gradient of auxin distribution(Casimiro et al., 2001; Marchant et al., 2002) and auxin-dependent patterning (Friml et al.,2002). Disrupting either auxin source or sink alters auxin distribution. In stm1 seedlings, disrupting the auxin source alters auxin gradients within the root (Casimiro et al., 2001). One might predict a corresponding alteration in seedlings lacking an auxin sink, such as the rootless mp^G92 and axr6-2 seedlings. The lack of a sink could cause higher than normal auxin gradients within the shoot that would disrupt formation of vascular tissue and leaf development. Such a disruption would provide a possible explanation for the suppression of the mp^G92 and axr6-2 phenotypes in a fkd1background: while auxin levels are too high to allow the function of wild-type FKD1, they allow the product of fkd1 to function and enable differentiation of vascular tissue in developing leaves.

Our analysis suggests that FKD1 is necessary to induce an auxin response at low auxin levels and enable differentiation of vascular tissue. The product of fkd1 cannot initiate a response to low auxin concentrations but can to higher levels as demonstrated by increased marginal vein meeting and DR5::GUS expression when auxin transport from the leaf margin is prevented. In contrast, one explanation consistent with the defective venation in mp^G92 and axr6-2 plants and suppression of this defect in a fkd1 background is that the product of FKD1 cannot function if auxin concentrations are too high.

If FKD1 is necessary for responses to auxin concentrations below a particular threshold and direct vascular cell differentiation, then the fkd1 phenotype may represent a leaf in which only higher auxin concentrations are driving vascular formation and provide insight into how auxin is distributed within the developing leaf. Based on mutant phenotypes,we suggest that three different sources of auxin are sequentially required for vascular pattern development within the leaf: (1) acropetally directed auxin directs midvein development; (2) a dynamic marginal source directs secondary vein development; (3) internal auxin sources direct tertiary and quaternary vein development.

Auxin from the first source is transported acropetally into the meristem and young leaf primordia (Reinhardt et al., 2000; Avsian-Kretchmer et al., 2002) where it directs differentiation of the midvein. The control of midvein differentiation by a different mechanism to that of subsequent veins is supported by its unique, acropetal developmental pattern and by the observation that all vascular pattern mutants described thus far affect subsequent veins but do not affect midvein development(Carland et al., 1999; Deyholos et al., 2000; Koizumi et al., 2000). Moreover, exposure to auxin transport inhibitors eliminates acropetal midvein development (Mattsson et al.,1999; Seiburth, 1999; Avsian-Kretchmer et al.,2002).

The second source arises at the margin, beginning at the distal tip and proceeding proximally with marginal cell differentiation, and directs differentiation of secondary veins. This idea is supported by this study and a number of previous studies. If auxin transport from the marginal source is inhibited, auxin accumulates at the distal leaf tip(Avsian-Kretchmer et al., 2002)and vascular development begins with a band of vascular tissue parallel to the margin of the leaf near the distal tip which lengthens along the margin as the leaf develops (Mattsson et al.,1999; Sieburth,1999). Moreover, high levels of auxin have been found to move basipetally during leaf development(Avsian-Kretchmer et al.,2002), coincident with highest rate of cell division and onset of vascular differentiation (Ljung et al.,2001). The sequential development of distally unconnected secondary vascular bundles originating at the margin of fkd1 leaves also supports a dynamic auxin source. Moreover, the fkd1 phenotype combined with reduced DR5:GUS expression at distal junctions in wild-type leaves suggest that auxin depletion of neighbouring cells by secondary veins results in low auxin concentrations at the point of distal vein connection. Interestingly, despite changes to marginal venation in pin1, fkd1 and fkd2 mutants, marginal vascular tissue always forms at a point distant from the margin, suggesting (1) that a maximum is reached at a point distant from the margin and/or (2) that cells at this point have prior competence for the vascular fate.

The third source of auxin is within the leaf blade, and directs formation of tertiary and quaternary veins (Aloni,2001; Aloni et al.,2003). In fkd1 and fkd2 leaves the marginal auxin source alone is sufficient to explain the generation of these veins,since the open venation pattern would allow auxin from the margin to reach the interior leaf blade. However, the wild-type leaf blade interior is separated from the marginal auxin source by a continuous loop of vascular tissue,prompting the proposal that an internal source of auxin directs formation of tertiary and quaternary veins (Aloni,2001; Aloni et al.,2003). The reduction in these veins in both axr1-3 and axr2 suggests that the internal source of auxin is lower than the marginal source so that the partial loss-of-function gene products can respond to the marginal source but not the internal source.

We have identified a second gene, FKD2, whose mutant phenotype is very similar to fkd1, being distinguished only by a higher frequency of VIs in the distal portion of the leaf blade. FKD2 maps very close to a previously characterized gene, SFC(Deyholos et al., 2000). Plants carrying mutations in SFC show severely disrupted vascular pattern with a very high frequency of VIs, and correspondingly severe defects in morphology, a phenotype consistent with fkd2 being a weak allele of SFC. Furthermore, double mutants between fkd1 and fkd2 exhibit a phenotype that is very similar to that reported for SFC mutants. The double mutant phenotype suggests that both FKD1 and FKD2 are essential for distal vein meeting,presumably because auxin concentrations are extremely low in these areas. Within other regions of developing veins where auxin levels are higher, the two genes act redundantly such that partial loss of function of either gene is insufficient to alter vein morphology, but partial loss of function of both genes results in severe disruption to the vascular integrity. We suggest that both FKD1 and FKD2 are required to respond to auxin and allow vascular differentiation. In the partial absence of one gene product,only the lowest thresholds are not recognized, disrupting vascular differentiation at vein junctions. In the absence of both gene products,response to a range of thresholds is faulty, and vascular discontinuities occur throughout the leaf veins.

If the fkd1 and fkd2 leaf pattern, which is similar to that of lower vascular plants, represents a leaf in which only high levels of auxin are able to initiate vascular differentiation, it is plausible that the vascular pattern seen in lower plants results from auxin canalization but that these plants have decreased sensitivity to auxin thresholds. Sztein et al.(Sztein et al., 1995) have shown that primitive vascular plants have only simple IAA conjugates and the increasing conjugate complexity occurring in higher vascular plants is associated with increasingly complex vascular tissue. It has been well established that leaves of conifers, Gingko and ferns are sources of auxin and that this auxin is at least partly responsible for the vascular pattern found within the stem of these plants(Gunckel and Wetmore, 1946; Steeves and Sussex, 1989; Wang et al., 1997). Additionally, lycopod leaves have been shown to have an effect upon the stem vasculature but it has not been conclusively demonstrated that auxin is responsible for this phenomenon (Freeberg and Wetmore, 1967). In conifers, it has also been shown that a concentration gradient of auxin is required for proper differentiation and patterning of xylem and phloem from the cambium(Nix and Wodzicki, 1974; Uggla et al., 1996). Given that auxin in primitive plants has a similar, if not identical, role to that in more advanced plants, it is likely that an auxin canalization mechanism is responsible for leaf vascular patterning in these groups, but that higher plants have evolved mechanisms to respond to a wider range of auxin thresholds. We suggest that FKD1 and FKD2 enable distal meeting of the vascular bundles in areas of low auxin concentration to form a complete, closed vascular pattern.

We thank members of the Flanagan lab for performing isotopic analysis,Beverly Burton, Larry Flanagan, Pat Kubik and James Meservy for technical assistance and John Bain, Beverly Burton, Jasmine Garrett, George Haughn and James Meservy for critical reading of the manuscript. This work was funded by an Operating Grant (E.S.) and a Postgraduate Scholarship (Q.S.) both from the Natural Science and Engineering Research Council.