A series of novel mutants of Arabidopsis thaliana that are defective in the formation of continuous vascular network: calling the auxin signal flow canalization hypothesis into question

The construction of vascular systems during organ development has provided a model for studying pattern formation as well as tissue differentiation in plants. One can recognize a rich variety of patterns in vascular systems among different species and organs. Despite such diversity in the apparent vascular patterns, it has been thought that a common basic mechanism underlies spatial regulation of vascular tissue formation. With respect to this supposed mechanism, two noteworthy hypotheses have been proposed up to now. One is the auxin signal flow canalization hypothesis (Sachs, 1989, 1991), which is based primarily on experimental observation of the inductive effect of auxin on vascular tissue formation, and the other is the diffusion-reaction prepattern hypothesis (Meinhardt, 1982; Koch and Meinhardt, 1994), which is based on computer modeling of interactions among hypothetical diffusible substances. Either of these hypotheses can account for some but not all aspects of vascular pattern formation and are far from being verified as yet (Nelson and Dengler, 1997).

For the purpose of identifying mechanisms responsible for spatial regulation of vascular tissue formation, genetic analysis, which has played an increasingly important role in recent studies of plant development, would be the most promising approach. In Arabidopsis, several kinds of mutants showing aberrant vascular formation have been reported. In the monopteros mutant, leaf marginal veins are missing or interrupted (Berleth and Jürgens, 1993). In the gnom mutant, tracheary elements are not interconnected to form strands but arranged in clusters or as single cells (Jürgens et al., 1991; Mayer et al., 1993). In the fackel and the fass mutants, lateral veins of the cotyledon are interrupted (Torres Ruitz and Jürgens, 1994; Mayer et al., 1991). These four mutants were identified originally as being impaired in body organization of the seedling. Such morphological abnormalities were traced back to altered embryogenesis. Thus, it is possible that the aberrant venation in these mutants may not be the primary effects of mutations but the secondary effects caused by disorganized body plan. The lop1 mutant, which is defective in basipetal transport of auxin, also exhibits abnormalities in vascular patterning. The leaves of this mutant are narrowed with a bifurcated and twisted midvein (Carland and McHale, 1996). The monopteros mutant was reported to have a reduced capacity of polar transport of auxin (Przemek et al., 1996). Although these findings appear to be consistent with the canalization hypothesis, it is not clear whether the reduced capacity of auxin transport is the cause or the consequence of the vascular pattern abnormalities.

Mutants of Arabidopsis with truly primary defects in vascular formation would be powerful tools for elucidating the molecular basis of the spatial regulation of vascular differentiation. Our strategy for isolating such mutants was to examine vein patterning in the cotyledon for every seedling from each of mutagenized populations of Arabidopsis and to select mutant lines of seedlings exhibiting relatively normal appearance and distinct abnormalities in venation of the cotyledon. The successful application of a similar procedure to isolate Arabidopsis mutants with venation-specific defects, cvp1 and cvp2 was recently reported (Carland et al., 1999). We have screened 3,400 M₃ lines which has led to the identification of 12 mutants that are impaired in vascular pattern formation. We briefly describe the characterization of 8 of them (van1, 2, 3, 4, 5-1, 5-2, 6 and 7) as a first step of genetic analysis on the spatial control of vascular tissue differentiation and discuss the two possible hypotheses for this, based on the phenotypes of the van mutants.

Plants used in the present study were derived from the Landsberg erecta (Ler) strain of Arabidopsis thaliana unless otherwise indicated. Ler is referred to as the wild type in this paper. Wild-type Ler seeds were mutagenized by treatment with 0.3% ethyl methane sulphonate solution for 20 hours at room temperature. After two cycles of self-fertilization, M₃ seeds were collected from individual M₂ plants for construction of M₃ lines. Each M₃ line consists of seeds harvested from only a single M₂ plant. Screening was carried out with some of seeds from each M₃ line. The M₃ seeds were germinated by incubation in water at 22°C for 7 days under continuous white light (30-50 μmol/m² /second). After the resultant seedlings were decolored in 99% ethanol and stained with 1% phloroglucinol in 20% hydrochloric acid, the vessels of the cotyledons were examined under a microscope. M₃ lines in which some seedlings exhibited aberrant patterning of the vascular system were selected as mutant lines.

Ler plants heterozygous for one of van mutations were crossed with wild-type Col plants. Seeds were collected from self-pollinated F₁ plants heterozygous for the van mutation, yielding a polymorphic F₂ population. The genotype of the VAN locus and simple sequence length polymorphism (SSLP) loci or cleaved amplified polymorphic sequence (CAPS) loci were determined for each individual F₂ plant using its F₃ progeny. For SSLP and CAPS analysis, DNA was isolated from each family of F₃ plants with the IsoPlant Kit (Nippon Gene Co., Tokyo, JAPAN). Polymerase chain reaction (PCR) was performed with a thermal cycler (Gene Amp PCR system 9700, PE Applied Biosystems) by repeating 40 times the following reaction: heat-denaturation at 94°C for 30 seconds, annealing at 55°C for 30 seconds, and polymerization at 72°C for 1 minute. PCR products were resolved by electrophoresis on a Nusieve 3:1 agarose gel (FMC Bioproducts, Rockland, Maine, USA) and recombination frequencies between VAN loci and SSLP or CAPS loci were scored.

Detailed histological analysis was performed for van mutants purified genetically through repeating back crosses three times. In the case of analysis on the vascular tissues of cotyledons, hypocotyls and root tips, seedlings that had been incubated in water for 7 days after seed imbibition were used as the source materials. For analysis of rosette leaves, 20-day-old or 26-day-old plants grown aseptically on germination medium were used. The germination medium was prepared as described by Valvekens et al. (1988) but solidified with 0.25% gellan gum instead of 0.8% agar. All cultures were carried out at 22°C under continuous white light (30-50 μmol/m² /second).

For whole-mount observation, samples were fixed overnight in a 9:1 mixture of ethanol and acetic acid or FAA (a 2.5:2.5:45:50 mixture of formalin, acetic acid, ethanol, and water) at room temperature. Fixed samples were cleared in a mixture of chloral hydrate, glycerol, and water solution (8 g:1 ml:2 ml) and observed under a light microscope equipped with a Nomarski optics (U-DICT; Olympus) or a stereoscopic microscope.

Cross sections of 4 μm thickness were made from hypocotyl samples that had been fixed in FAA, dehydrated in a graded ethanol series, and embedded in the Technovit 7100 resin (Heraeus Kulzer, Germany). These sections were stained with toluidine blue and observed under a light microscope.

Transgenic Arabidopsis (Wassilewskija strain), which carried a chimeric gene, pAthb8∷ GUS, consisting of the Athb-8 promoter and the β-glucuronidase (GUS) structural gene (Baima et al., 1995) or a chimeric gene, pTED3∷ GUS, consisting of the TED3 promoter and the GUS structural gene (Igarashi et al., 1998), was used to distinguish vascular tissues. Seeds of transgenic Arabidopsis carrying pAthb8∷ GUS were kindly provided by Dr Baima and Dr Morelli, Unità di Nutrizione Sperimentale.

The van3 mutation was introduced into the pAthb8∷ GUS transgenic plants and pTED3∷ GUS transgenic plants by artificial crossing between heterozygous van3 plants and the transgenic plants. Histochemical GUS staining was performed with seedlings of the F₂ generation or embryos of the F₃ generation. Samples were fixed in 90% (v/v) acetone for longer than 45 minutes at −20°C. After washing in 100 mM sodium phosphate buffer pH 7.2 at least 3 times, they were immersed in a reaction mixture containing 1 mM 5-bromo-4-chloro-3-indolyl glucuronide, 0.5 mM potassium ferricyanide, and 0.5 mM potassium ferrocyanide in 100 mM sodium phosphate buffer pH 7.2 and incubated at 37°C for 2 hours (pAthb8∷ GUS) or 24 hours (pTED3∷ GUS) in the dark. After the reaction, samples were mounted with a mixture of chloral hydrate, glycerol, and water and observed under a light microscope equipped with a Nomarski optics.

The cotyledon of Arabidopsis has a very simple vein pattern; one midvein and three or four lateral veins (Fig. 1A). Taking this pattern as an index, the mutagenized population was examined for mutants impaired in vascular pattern formation. Screening of 3,400 M₃ lines resulted in the isolation of 12 mutant lines, which were characterized by fragmentary lateral veins. Of these, 8 mutant lines (L1315, L1727, L3302, L3409, L3992, L4622, L4668 and L4503) were subjected to further analysis.

In each of mutant M₃ lines, approximately a quarter of seedlings exhibited aberrancy in the vascular pattern of the cotyledon and three quarters appeared normal. Self-pollination of two thirds of those normal plants yielded both normal and abnormal progeny at a segregation ratio of 3 to 1 in the next generation (M₄). These results indicated that alteration of the vascular pattern was caused by a monogenic, recessive mutation in each mutant line.

To determine the number of complementation groups, heterozygous plants belonging to different mutant lines were crossed to each other. Any cross among L1315, L1727, L3302, L3409, L3992, L4668 and L4503 resulted in recovery from mutant phenotype in all F₁ seedlings, indicating that mutations in these lines were located at different loci (data not shown). Only a cross between L3992 and L4622 brought about segregation of wild-type and mutant phenotypes, the ratio of which did not deviate significantly from 3:1 (data not shown). Since L3992 and L4622 were derived from different batches of M₁ plants, they should not be the re-isolated lines of the same mutant but the allelic lines. Thus, these 8 mutant lines were found to represent single mutant alleles at 6 genetic loci and two mutant alleles at one locus. In view of the characteristics of the mutations that affected vascular network, the genes defined by mutations in L1315, L1727, L3302, L3409, L3992/L4622, L4668 and L4503 were designated VAN1, VAN2, VAN3, VAN4, VAN5, VAN6 and VAN7, respectively. Mutant alleles of L3992 and L4622 were designated van5-1 and van5-2 for distinguishing them.

Table 1 summarizes the result of linkage analysis, which localized VAN1, VAN2, VAN3, VAN4, VAN5, VAN6 and VAN7 loci on chromosomes 2, 5, 5, 5, 1, 4 and 1, respectively. Comparison of these VAN loci with genetic loci, mutants of which have been shown to affect vascular patterns, suggested possible relationship between VAN5 and MONOPTEROS and between VAN7 and EMB30/GNOM. Complementation analysis revealed that van7 is allelic to emb30/gnom while van5 is not allelic to monopteros (data not shown). Thus, van7 was renamed emb30-7 as a new mutant allele of EMB30/GNOM. Carland et al. (1999) reported that CVP1, one of newly identified loci participating in vein pattern formation, was located between molecular markers of m235 and m59 in chromosome 1. Reference of van5 to this locus is to be tested.

Vascular patterns in the cotyledons of 7-day-old seedlings of the van mutants are shown in Fig. 1B-I. Although the vascular patterns of the van1, van2, van3, van4 and van6 mutants appeared similar to the wild-type pattern in their overall architecture, the lateral veins of these mutants were incomplete and disconnected (Fig. 1B,C,D,E,H). Fragmentation of lateral vein was most noticeable in the van3 mutant (Fig. 1D). The architecture of the venation of the van5 mutant was rather different from that of the wild type, and highly variable among the van5 individuals (Fig. 1F,G). Lateral veins were not obvious in the cotyledon of the emb30-7 (van7) mutant and many tracheary elements were arranged randomly in its central region (Fig. 1I). An additional feature of the vascular tissue shared by all the van mutants was that tracheary elements were very large as compared with normal tracheary elements in the wild type. In spite of the noticeable abnormalities in the vascular pattern, cotyledons of the van1, van2, van3, van4 and van6 mutants appeared normal in shape. Cotyledons of the van5-1 and van5-2 mutants were distorted and much smaller than those of the wild type. The emb30-7 cotyledon was apparently circular.

Fig. 2 shows the vascular tissues of rosette leaves in adult plants of the wild type and van mutants other than van1, which did not expand rosette leaves. The first rosette leaf of the wild type has a prominent midvein (primary vein), secondary veins which branch from the primary vein, and tertiary veins which branch from the secondary veins (Fig. 2A). In the first rosette leaves of the van2 through van6 mutants, tertiary veins were absent, faint, or fragmentary whereas primary and secondary veins formed a fairly continuous network (Fig. 2B-G). The primary and secondary veins of the van5 mutant consisted of a greater number of tracheary elements (per unit length) than the wild-type veins (Fig. 2E,F). The extra tracheary elements were aligned improperly in the van5 veins. In the case of the emb30-7 mutant, rosette leaves had three-dimensional and complex networks of the vascular system with an increasing number of free ends (Fig. 2H). This phenotype was most obvious in the peripheral zone of the leaf lamia.

All of these van mutations induced a reduction in size of the rosette leaf lamina. In addition to this effect, the van5 and emb30-7 mutations caused remarkable alterations in the overall morphology of the rosette leaves. van5 distorted leaf shape (Fig. 2E,F) and emb30-7 increased the thickness of leaves.

Young seedlings of the van mutants are shown in Fig. 3. The most apparent abnormality in morphology at the whole-plant level was poor growth of the primary root in all of the van mutants. In other aspects such as cotyledon expansion, greening, and hypocotyl elongation, all the van mutants exhibited relatively normal appearances. However, closer observation revealed various effects of the van mutations on the shape and size of cotyledons, as described in the previous paragraph.

At the later stages of development, all the van mutations caused more severe effects on plant morphology and growth. The degree of morphological abnormality of 20-day-old mutant plants increased in the order of van3 < emb30-7 (van7) < van4 < van2, van5-1, van5-2, van6 < van1 (Fig. 4). The van1 plant showing the most severe phenotype formed only two visible rosette leaves, which did not expand (Fig. 4B). The van3 plant showing the slightest phenotype developed more than 7 rosette leaves of almost normal shape though they were significantly smaller than the wild-type leaves (Fig. 4D). The van mutants died after several weeks of vegetative growth and bolting was never observed in any of them.

Vascular systems along the main axes of van mutants

Vascular systems of the hypocotyl (Fig. 5) and root (Fig. 6) were examined in wild-type and van seedlings grown for 7 days. Microscopic observation of cleared, whole-mount specimens of hypocotyls could not detect obvious differences in the longitudinal alignments of vessels between wild type (Fig. 5A) and van1 through van6 mutants (Fig. 5B-H). In the majority of the emb30-7 seedlings, the apical ends of some vessels were found in the middle part of the hypocotyl (Fig. 5I).

Vascular tissues of the hypocotyl were also examined anatomically with transverse sections, which detected some aberrancies of vascular organization in every van mutant except van3 (Fig. 5J-R). In the van1 mutant, radial arrangement of cells in the central cylinder was strikingly disordered, and phloem could hardly be identified (Fig. 5K). Such tendency was observed in the van2, van4, van5-1, van5-2, and van6 mutants (Fig. 5L,N-Q) although their phenotypes were less severe as compared with the van1 (Fig. 5K). In the emb30-7 hypocotyls, lack of one strand of phloem was sometimes observed but otherwise radial organization of vascular cells looked normal (Fig. 5R).

In the primary roots of the van mutants other than the van1, the meristematic zone was reduced in length, and vessel differentiation took place at the position unusually close to the root apex (Fig. 6C-I). With respect to the continuity of vascular strands, however, these mutant roots appeared to be intact. Obvious defects in the root vascular system were observed only in the van1 mutant, where vessels were disconnected and/or improperly aligned (Fig. 6B).

Further experiments were performed with a predominant focus on the van3 mutation, which showed the most restricted effect among van mutations. To decide steps of the vascular tissue differentiation that are affected by the van3 mutation, the expression patterns of pAthb8∷ GUS and pTED3∷ GUS chimeric genes were examined as molecular markers under the influence of the van3 mutation. As the pAthb8∷ GUS and pTED3∷ GUS genes are expressed preferentially in the provascular/vascular tissues (Baima et al., 1995) and differentiating tracheary elements (Igarashi et al., 1998), respectively, their expression patterns can distinguish provascular tissue differentiation and tracheary element differentiation. In the cotyledon of the walking-stick stage embryo of the wild type, pAthb8∷ GUS-expressing cells were aligned in the ‘ϕ’-letter shape, which corresponded to a midvein and a distal loop of lateral veins (Fig. 7A). At the upturned-U stage, pAthb8∷ GUS-expressing cells appeared in the basal part of the lateral veins to complete a circuit of provascular tissues (Fig. 7B). In contrast, pAthb8∷ GUS-expressing cells of the van3 cotyledon did not form continuous lateral loops at any stage of late embryogenesis (Fig. 7E,F) or at the young-seedling stage (Fig. 7G) although the overall tendency of GUS expression, initiating at the distal part and reaching to the basal part, was similar to that of the wild type (Fig. 7A-C). Expression of pTED3∷ GUS was not detectable in either the wild-type or the van3 cotyledon during embryogenesis (data not shown). In the cotyledons of 7-day-old wild-type seedlings, pTED3∷ GUS was specifically expressed in immature tracheary elements and tracheary element precursor cells of the midvein and lateral veins (Fig. 7D). This was also the case for the van3 mutant, and no GUS-expressing cells were observed in the interspace between the fragmented lateral veins (Fig. 7H). These results indicated that the van3 mutation affected very early aspects of vascular patterning prior to provascular network formation without interfering with progression of the vascular tissue differentiation.

In the present study, we screened mutagenized populations of Arabidopsis with the aim of isolating vascular patterning mutants as genetic tools for studying regulatory mechanisms of vascular tissue differentiation. Screening of 3,400 M3 lines with the cotyledonary venation as a guide allowed us to identify 7 genetic loci (VAN1 through VAN7) that are involved in the vascular network formation. Mutations in these VAN loci caused severe defects in minor veins such as lateral veins of the cotyledon and tertiary veins of the rosette leaf while they were less injurious to major veins including midveins of the cotyledon and the rosette leaf. These phenotypes suggest that both the machinery controlling lateral vein formation in the cotyledon and the machinery controlling tertiary vein formation in the rosette leaf involve VAN gene products as common components. In addition, it can be inferred from the van phenotypes that major and minor vasculatures are formed under different genetic controls. This does not necessarily indicate difference in the fundamentals of the regulatory systems but may represent difference in the robustness of the similar systems.

Among the van mutants, emb30-7 (van7) is distinct from the van1 through van6 in several aspects of venation. In contrast to the cotyledons of most of the van mutants, in which lateral veins were incomplete or fragmented but still present in relatively normal architecture (Fig. 1B-E,H), the emb30-7 cotyledon lost lateral veins and had clustered tracheary elements instead (Fig. 1I). This phenotype of the emb30-7 is almost the same as reported for the gnom mutant (Jürgens et al., 1991), which is allelic to emb30. The rosette leaves of the van2 through van6 mutants had simplified venation with fragmentation or (partial) loss of tertiary veins in the rosette leaf (Fig. 2B-G) whereas the emb30-7 rosette leaf had rather complicated venation with an increasing number of freely ending veinlets, particularly near the periphery (Fig. 2H). Such venation of the emb30-7 rosette leaf appears similar to the venation observed for Arabidopsis seedlings that were incubated in the presence of inhibitors of auxin polar transport (Mattsson et al., 1999). The EMB30/GNOM gene encodes a Sec7-like protein, which may be involved in vesicle trafficking (Shevell et al., 1994; Busch et al., 1996). Recently, the gnom mutation was found to interfere with localization of an auxin efflux carrier, PIN1 (Steinmann et al., 1999). Taken together, we can reasonably speculate that the aberrant venation of the emb30-7 mutant may be induced by the reduction in auxin polar transport due to defects in vesicle trafficking and the resultant irregularity in the PIN1 localization.

Noticeable characters other than fragmentary venations common to young seedlings of all the van mutants were poor growth of the primary root (Fig. 3) and reduction in the size of root apical meristem (Fig. 6). This suggests that high meristematic activity of the primary root is linked with integrity of a vein network in the cotyledon in some way. However, this linkage is unlikely to be a simple and direct, considering that root growth and development of cvp1 and cvp2 mutants defective in the cotyledonary venation were indistinguishable from those of the wild type (Carland et al., 1999).

Microscopic examination of vascular tissues along the hypocotyl-root axis with young seedlings of the van mutants discovered additional effects of the van mutations particularly on a radial pattern of vascular tissues. In this respect, only the van3 mutant was normal, and all other van mutants exhibited some aberrancies. As the degree of such aberrancies appeared in almost parallel with the degree of morphological defects of the mutant plants, one can suppose that disorganization of vascular tissues along the main axis has a greater influence on development and growth at the whole-plant level than fragmentation of lateral veins in the cotyledon does.

The van3 mutation seems to affect primarily and specifically vascular patterning of the minor veins since the van3 seedling was almost normal in other morphological aspects including the size and shape of the cotyledon, longitudinal alignment of vessels along the main axis, and radial organization of vascular tissues in the hypocotyl (Figs 1D, 2C, 3D, 4D, 5D, 5M, 6D). Closer characterization of the van3 mutant throughout embryogenesis, using pTED3∷ GUS and pAthb8∷ GUS expressions as indices of vascular tissue differentiation, detected the discontinuity in the provascular tissues of the lateral vein in the cotyledon at the walking-stick stage (Fig. 7E). Accordingly, the VAN3 gene is expected to play an important role in the spatial control of provascular differentiation, which results in a continuous network of vascular strands.

Recently isolated mutants of Arabidopsis, cvp1 and cvp2, are characterized by a defect specific to the formation of lateral loops of the cotyledonary veins (Carland et al., 1999). This phenotype was shown to be traced back to the incomplete network of provascular tissues. In this regard, the cvp mutants are very similar to the van3 mutant. Functional relationships among the CVP1, CVP2 and VAN3 genes would be of great interest in the future analysis of provascular tissue patterning.

With respect to the possible mechanisms determining vein patterns, two hypotheses, i.e., auxin signal flow canalization and diffusion-reaction prepattern, have been proposed and discussed. Mattsson et al. (1999) examined the effects of inhibitors of auxin polar transport on the venation of Arabidopsis and attempted to explain the inhibitor-induced alterations in the venation on the basis of the canalization hypothesis. Przemeck et al. (1996) reported a reduced capacity of auxin polar transport and aberrant vasculatures in monopteros mutant of Arabidopsis, and discussed these features of monopteros in reference to the canalization hypothesis. Mutations in the EMB30/GNOM locus may also influence venation via suppression of auxin polar transport, as discussed above. These studies, together with pioneering works of Sachs (1968, 1969, 1975), seem to support strongly the canalization hypothesis. Strictly speaking, however, any findings reported so far are not really supportive but rather neutral for the following reasons.

(i) The auxin signal flow canalization hypothesis presumes that positive-feedback regulation of auxin polar transport and the resultant formation of canal networks conducting auxin flow direct the positions of vascular tissue differentiation. These core assumptions have not been verified through recent studies.

(ii) Indispensability of auxin polar transport in normal development of vasculatures, which has been demonstrated by recent studies, can be explained in various ways without assuming the canalization hypothesis, and is not contradictory at all to the diffusion-reaction prepattern hypothesis.

(iii) Abnormalities in vascular development undoubtedly influence auxin polar transport since vascular parenchyma takes a major part in auxin polar transport. Thus, poor ability of auxin polar transport in mutants with aberrant vascular patterning may not always be the cause of abnormal development of the vascular system but the result of it in some cases.

Most of the van mutants formed discontinuous networks of the minor veins but maintained relatively normal architecture of venation as a whole. High frequency of these van mutants in the mutagenized population suggests that the continuity of the vein networks is more sensitive to genetic lesions than their overall architecture is. This nature of pattern formation of minor veins is very important when considering the validity of the two hypothetical mechanisms.

A basic model of diffusion-reaction systems hypothesizes interacting substances with different diffusion rates, which can generate stable patterns autonomously in the initially homogenous field through local autocatalysis and long-range inhibition. Meinhardt (1989, 1995) demonstrated that slight changes of this autocatalytic condition readily induced a gradual shift from stripe pattern to spotted pattern via fragmented stripe pattern without destroying the overall architecture. In addition, various alterations of the rate of the basal synthesis, decay, or interaction of the diffusion-reaction substance(s) were also shown to change a pattern from stripes to spots (Asai et al., 1999). In view of such fragility of the continuous patterns developed by diffusion-reaction systems, it is not surprising that the networks of minor veins were fragmented by so many mutations, given that they are controlled on the basis of a version of diffusion-reaction systems as assumed by the diffusion-reaction prepattern hypothesis.

In the vein pattern formation predicted by the canalization hypothesis, in contrast, the overall architecture of the vascular network is supposed to be more fragile than its continuity. The canalization hypothesis assumes that auxin flow builds a canal for conducting auxin flow by itself without any blueprints and that this canal produces a pattern of vasculature. Therefore, all mutations that interfere with auxin flow or canal formation thus preventing continuous formation of vascular networks should necessarily destroy the overall architecture of the vascular pattern. This is apparently not consistent with the discontinuous network of the minor veins with relatively normal architecture found in the van mutants. Thus, the van mutants seem to support the diffusion-reaction hypothesis rather than the canalization hypothesis.

The above argument based on the phenotypes of the van mutants calls the canalization hypothesis into question and indicates that the diffusion-reaction hypothesis still survives as a candidate for a mechanism determining vascular patterns, in particular, minor venations. Isolation and molecular characterization of the VAN genes would provide important information about how vascular tissues are differentiated continuously to form networks and whether the canalization mechanism or the diffusion-reaction mechanism is responsible for vein pattern formation.

We thank Dr Baima and Dr Morelli, the Unità di Nutrizione Sperimentale, for providing transgenic Arabidopsis seeds carrying pAthb8∷ GUS, Dr Berleth, the University of Toronto, for providing monopteros seeds, and the Arabidopsis Biological Resource Center (Columbus, Ohio, USA) for providing gnom seeds. We also thank Miss Matsuda and Miss Furumizu, University of Tokyo, for their help in linkage analysis. This work was supported in part by Grants-in-Aid from the Ministry of Education, Science, Sports and Culture of Japan (No. 10304063, No. 10219201, No. 10182101), and from the Japan Society for the Promotion of Science (JSPS-RFTF96L00605) to H. F.