The primary root of Arabidopsis thaliana has a remarkably uniform cellular organisation. The fixed radial pattern of cell types in the mature root arises from proliferative divisions within the root meristem. The root meristem, in turn, is laid down during embryogenesis. We have analysed six mutations causing alterations in the radial organisation of the root. Embryonic phenotypes resulting from wooden leg, gollum, pinocchio, scarecrow, shortroot and fass mutations are described. While mutations in the fass gene affect morphogenesis of all cells, the five other mutations cause alterations in specific layers. Wooden leg and gollum mutations interfere with the proper organisation of the vascular tissue. Shortroot, scarecrow and pinocchio affect the endodermis and cortex. The layer-specific phenotypes caused by all five mutations are also apparent in the hypocotyl. All these phenotypes originate from defects in the radial organisation of the embryonic axis. Secondary roots, which are formed post-embryonically, also display layer-specific phenotypes.
The continuous activity of meristems in which cells divide and produce new organs is a remarkable aspect of plant development. Primary meristems of angiosperm plants are laid down during embryogenesis. They contribute to post-embryonic development upon germination by producing organs and secondary meristems. Cells within the primary shoot meristem give rise to various organs such as leaves, the stem, and the different floral organs. In contrast, the primary root meristem elaborates a single organ, the primary root, with a stereotyped radial pattern of tissues. In Arabidopsis thaliana this radial pattern is of elegant simplicity. In the mature region of the root, single layers of epidermis, cortex, endodermis, and pericycle surround a small number of vascular cells (Dolan et al., 1993). Clonal analysis has revealed that the meristematic root can be traced back to ‘initial cells’ originating from a single tier of cells in the heart stage embryo (Scheres et al., 1994). The radial pattern of the mature Arabidopsis root correlates with the spatial arrangement of initials at the early torpedo stage of embryogenesis. This embryonic arrangement of initials results from early cell divisions that create the tissue layers of both root and hypocotyl (Scheres et al., 1994). Histological studies on several other species confirm the idea that cell divisions in the embryonic axis generally determine the radial organisation of the seedling root and hypocotyl (Von Guttenberg et al., 1955; Ramji, 1975; Kuras, 1980).
Experimental studies have pointed to the capacity of a region within the root meristem itself to dictate pattern. Studies in which root apices have been cultured in vitro have been interpreted to demonstrate that pattern information resides in the root apex (Reinhard, 1954; Torrey, 1954). Excision experiments in Zea mays indicate that such information may be present in a subset of slowly dividing cells in the root apex termed the quiescent centre (Feldman and Torrey, 1976). However, these experiments do not determine conclusively where the information to establish the radial pattern of the root originates. The pattern information could be generated in the root apex with the aid of apex-specific gene products. Alternatively, radial pattern information could be generated in the embryo and this information would remain active throughout root development.
To distinguish between these two possibilities, we have identified mutations that cause radial pattern defects in the root. Our aim was to establish whether these defects are specific for roots and therefore provide evidence for the existence of a separate set of genes specifying radial organisation within the root meristem. Six mutant phenotypes are characterised in this study which have defects in pericycle, vascular tissue, endodermis, cortex and overall radial organisation of the root. We show that in all cases the defect in the radial pattern of the root is apparent in the embryonic axis and in both hypocotyl and root of the seedling. Therefore the majority of genetic information determining the cellular organisation of the root appears to be involved in the radial organisation of the embryo axis as a whole.
MATERIALS AND METHODS
Plant growth conditions and mutagenesis
For root and hypocotyl anatomy studies seeds were sterilized in 5% sodium hypochlorite for 10 minutes and imbibed for 2 days at 4 °C in the dark in sterile water. Seedlings were grown on Petri plates containing 1× Murashige and Skoog salt mixture and 0.5 g/l 2(N-morpholino) ethanesulfonic acid, pH 5.8, in 0.8% Duchefa agar. Plates were incubated in a near-vertical position in a plant chamber at 22 °C, 75% humidity with a 16 hours light/8 hours dark cycle.
For embryo anatomy studies, A. thaliana ecotype ‘Columbia’ was grown on soil until seed set became apparent using conditions as described above.
The shortroot (shr), scarecrow (scr) and pinocchio (pic) mutants were recovered as seedlings with slow-growing roots among T-DNA mutagenised ecotype WS seeds (Feldman, 1991; Benfey et al., 1993). The wooden leg (wol) mutant phenotype was recovered within bulk M2 offspring of EMS mutagenised seeds as a seedling with retarded root growth, and outcrossed to determine the segregation ratio. A single-family screening strategy was adapted to allow the recovery of seedling-lethal mutations affecting the radial pattern of root tissues. A. thaliana ecotype ‘Columbia’ dry seeds were mutagenised with 10 mM ethyl methane sulfonate (EMS) in water for 24 hours at 22 °C. Single siliques were harvested from individual M1 lines. The gollum (glm) and fass (fs) mutant phenotypes were identified within M2 offspring. M3 progeny of the non-mutant siblings was re-tested for appearance and segregation ratio of the phenotype.
Lines segregating double mutants were constructed as follows. F1 plants carrying one wol and one shr allele were obtained by crossing parents homozygous for the single allele. Plants homozygous for wol were used as acceptor because shr homozygous plants have low seed set.
Plants homozygous for the fs allele used are sterile. For that reason F1 plants carrying one shr, scr or wol allele together with the fs1213 allele were constructed by using homozygous shr, scr or wol plants as acceptor and donor pollen from plants that were tested to be heterozygous at the fs locus.
For double mutant analysis F1 families segregating both parental phenotypes were selected, numbered and photographed, and subsequently examined microscopically for the occurrence of novel phenotypes. In this way specific features of seedlings with novel microscopical phenotype could be recorded. Families from shr × fs/+ and shr × wol crosses segregated a novel phenotype. Seedlings of putative shr fs double mutants as determined by microscopy had the overall appearance of the fs parent. Six of 36 seedlings with fs appearance segregated the novel phenotype (lacked a casparian strip), consistent with an expected 1:3 ratio of double mutants in the fs phenotype class in the case of unlinked genes (χ2 = 1.3). The seedling appearance of the putative wol shr double mutants resembled the wol single mutant. Five of 22 seedlings with wol single mutant appearance segregated the novel phenotype, consistent with the expected ratio of 1:3 double mutants for unlinked genes (χ2 = 0.06).
In families segregating both parental phenotypes from wol × fs and scr × fs crosses no novel phenotype was detected. Because the phenotype of fs mutants is extremely pleiotropic while the wol and scr mutant phenotype is most prominent in roots, the fs overall seedling phenotype was assumed to be epistatic to the wol and scr seedling phenotype, as observed in the shr × fs cross. For the wol × fs cross 20 seedlings with fs appearance were analysed microscopically and none showed an additional phenotype. Hence the hypothesis that a novel double mutant phenotype is present in the fs phenotype class can be rejected (χ2 at 1:3 expectation = 6.7). 10 seedlings with wol appearance were sectioned as a control and these revealed no novel phenotype. For the scr × fs cross 40 seedlings with fs appearance were analysed microscopically and these revealed no new phenotype (χ2 at 1:3 expectation = 13.3).
Dissected embryos of various stages as well as seedlings were fixed in 1% glutaraldehyde and 4% formaldehyde in 50 mM sodium phosphate buffer, pH 7.2, dehydrated and embedded in Technovit 7100 (Kulzer, Hereaus). For the transverse sections of embryos complete siliques were embedded in an oriented fashion. 3 µm sections from all specimens were made on a Reichert-Jung 1140 rotary microtome carrying a disposable Adamas steel knife. Embryo sections were stained with 2.5% Astra blue (Merck) dissolved in 0.5 M sodium phosphate buffer, pH 4.4, at 50°C for 8 minutes, mounted in DePeX (BDH) and photographed using a Zeiss Axioskop using Kodak Technical Pan film. Root and hypocotyl sections were stained with 0.05% toluidine blue-O (Merck) in water at room temperature for 1 minute, mounted in DePeX and photographed using a Zeiss Axioskop using Kodak Ektar 25 film.
Suberin staining on fresh sections was performed essentially as described by Brundrett et al (1988). Agarose embedded and hand sectioned tissue was incubated overnight in 0.1% (w/v) berberine hemisulphate (Sigma) at root temperature, rinsed with distilled H2O and incubated in 0.5% (w/v) aniline blue WS (Polysciences) for 90 minutes at room temperature. Agarose slices were rinsed as above, transferred to a 0.1% FeCl3 in 50% (v/v) glycerin solution for 5-10 minutes, and mounted on microscope slides in the same solution. The sections were photographed using a Zeiss axioskop fluorescence microscope with excitation filter G 365 nm, chromatic beam splitter FT395 nm and barrier filter LP420, using Ilford XP2 film.
Tissue culture experiments
Explants from wild-type and shr, scr, pic, wol and fs mutant seedlings that were germinated on MS 4.5% agar plates were incubated for 48 hours in callus inducing medium (Valvekens et al., 1988) at room temperature, and subsequently placed on root inducing medium (MS basal salts (Sigma), 0.5 g/l MES (Sigma), 4.5% sucrose, 0.8% agar (BBL, Becton & Dickson), 1mg/l IAA (Sigma), 0.1 mg/l kinetin (Sigma), 0.2% IBA (Sigma), set to pH 5.9 with KOH). The explants were incubated for 7-14 days under the conditions described above. The radial tissue organisation of the regenerated roots was examined using fresh sections.
Ontogeny of the radial organisation of the
Arabidopsis embryonic axis
The radial organisation of the different tissues of wild-type Arabidopsis roots reflects the arrangement of initials in the lower cell tier of the embryonic axis at the torpedo stage of embryogenesis (Scheres et al., 1994). We investigated the cell divisions leading to the radial organisation of the wild-type embryonic axis to provide a framework for mutant analysis. Transverse sections at different levels of the axis in embryos of several developmental stages are shown in Fig. 1. Median longitudinal sections of embryos at corresponding developmental stages are included to facilitate the description.
At the early globular stage the embryo proper is split in an upper and a lower tier (Fig. 1A). In the lower tier, protoderm, ground meristem, and procambium are separated as the progenitors of the major tissue types in the embryonic axis by regular periclinal divisions (Fig. 1C). Anticlinal divisions have determined the number of ground meristem cells at eight and are at this stage elaborating the number of protoderm cells at sixteen (Fig. 1C, arrow shows recently deposited cell wall). At later stages this radial cell number of protoderm and ground meristem is preserved in the cells immediately adjoining the hypophyseal cell. This subtier harbours the progenitors of the root meristem initials (cf. Fig. 1G,L,Q,V). Therefore early anticlinal divisions largely determine the number of epidermal and cortical root meristem initials in the seedling root.
At the late globular/triangular stage the lower tier has split in two subtiers, upper lower tier (ult) and low lower tier (llt) (Fig. 1E). The llt region will produce the embryonic axis, forming root and hypocotyl (Scheres et al., 1994). The four procambium cells have performed a periclinal division to generate the future pericycle and the vascular primordium (Fig. 1F,G; p and v). In the llt region, regular periclinal divisions create four inner and four outer cells (Fig. 1G). In the ult region, cells within the procambium perform less regular tangential divisions (Fig. 1F).
At the early heart stage the three llt subtiers corresponding to the incipient hypocotyl, embryonic root and root meristem initials become separated (Fig. 1I; hy, r, rmi; Scheres et al., 1994). The number of cells in the pericycle increases to 8 –9 by anticlinal divisions, approaching the number of initials counted in the seedling root. Periclinal divisions have doubled the ground meristem in the future hypocotyl region (Fig. 1J, arrowhead) and only in the outer layer additional anticlinal divisions occur. Protoderm cells in the hypocotyl region divide anticlinally to approach the final epidermal cell number in the seedling hypocotyl.
The early torpedo stage embryo contains the root promeristem with all initials and derived cell types (Fig. 1N). Within the vascular primordium additional irregular cell divisions have occurred to establish the number of vascular initials present in the seedling root meristem (Fig. 1O,P). The epidermal initials have given rise to a layer of lateral root cap (Fig. 1N,P,Q). The prospective lateral root cap cells undergo a radial division which doubles their number (Fig. 1P, arrowheads). Characteristic periclinal divisions in the daughter cells of the cortex initials give rise to the endodermis (Fig. 1P, open triangle). The inner ground meristem layer in the hypocotyl also undergoes periclinal divisions, resulting in a continuous endodermis layer in root as well as hypocotyl of the seedling (Fig. 1N, arrowhead). At the mature embryo stage the embryonic axis has elongated considerably by transverse divisions, which preserve the radial organisation of the torpedo stage embryo (Fig. 1S-W).
The radial arrangement of derivatives of the hypophyseal cell (Fig. 1A; h) is established by a sequence of divisions reminiscent of those described above for the lt region in the embryo proper. The prospective quiescent centre cells form a quadrant at the heart stage (Fig. 1M,Q). Within the columella cell tier a quadrant is formed by cell divisions which may be oriented similarly to the quiescent centre quadrant (as in Fig. 1R). However, a shift in orientation of vertical cell walls in the two tiers is frequently observed (data not shown). The elaboration of the final pattern of columella cells ‘4 surrounded by 8’ is completed at the torpedo stage either by periclinal/anticlinal or by tangential divisions (Fig. 1R,W).
A summary of the observed division patterns is shown in Fig. 2. Periclinal cell divisions in centrally located cells of the embryo separate the different tissues. Anticlinal divisions subsequently determine the number of cells of each layer in a regular manner (Fig. 2A). As these divisions proceed in the stele region, the cells become smaller, and tangential divisions can be observed (Fig. 2B). We have no indications that tangential division patterns in the pericycle region lead to abberations in the seedling pericycle layer. Therefore we conclude that, to a certain extent, forces acting on the young cell wall correct the radial organisation (Fig. 2B). Both the typical periclinal/anticlinal and tangential patterns of division occur in the clonally distinct basal subtiers of the hypophyseal cell that form the columella of the root. Hence the radial pattern of all root and hypocotyl cells in the mature embryo of Arabidopsis is the result of remarkably consistent cell divisions exhibited throughout the different stages of embryo development.
Mutations affecting radial organisation of root and hypocotyl
To investigate the nature of genetic controls exerted on radial pattern formation we performed a genetic screen for mutations affecting the radial organisation of root tissues. Slow root growth had been observed in the radial pattern mutant shortroot (shr), perhaps as a consequence of the lack of endodermal tissue (Benfey et al., 1993). The rationale of our screen was that more radial pattern defects would cause retardation of root growth. Seedlings from different batches of mutagenized seeds (see Materials and Methods) were preselected for retarded root growth and microscopically examined for defects in radial organisation. In this way the scarecrow (scr), pinocchio (pic), gollum (glm) and wooden leg (wol) mutants were identified. Fig. 3 displays the phenotype of seedlings homozygous for the different mutations. The major characteristic of all mutant seedlings is retarded root growth (Fig. 3A-F). In addition we have further characterised a mutant previously named dikkiedap, whose increased diameter immediately suggested alterations in radial pattern (Fig. 3G). Complementation analysis revealed this mutation to be an allele of the fass (fs) gene (R. Torres-Ruiz, personal communication; Mayer et al., 1991). All phenotypes segregated as if caused by single nuclear recessive mutations. Complementation and mapping data reveal that the shr, scr, pic, wol and fs mutations define different genes (data not shown).
The radial pattern defects found in mutant seedlings are visualised in the transverse sections depicted in Fig. 4. Wildtype primary roots in the differentiation zone contain single layers of epidermis, cortex, endodermis, and pericycle with constant cell numbers (Fig. 4A; Dolan et al., 1993). Within these layers vascular cells are present containing two protoxylem and two protophloem elements. The wild-type hypocotyl has a similar organisation but contains an additional cortex cell layer (Fig. 4F).
The stereotyped tissue organisation is differently affected in the different mutants. Roots from shr seedlings lack the endodermal layer (Fig. 4B, Benfey et al., 1993). scr and pic roots also lack a ground tissue layer and transverse sections look similar to those of shr roots. Suberin staining indicated that the remaining cell layer in both scr and pic contains the casparian strip characteristic for endodermis cells. We concluded that the cortex layer is absent in these mutants. The hypocotyl in shr, scr and pic contains two irregular layers of cells in the normal location of the endodermis and two cortex layers (Fig. 1G, shr is shown).
Roots from glm seedlings lack the characteristic organisation of the vascular tissue and pericycle (Fig. 4C). The pericycle layer in the hypocotyl of glm seedlings is incomplete and the arrangement of xylem and phloem elements is atypical (Fig. 1H, arrow points to xylem cell in pericycle position). Wol seedlings contain roots with a vascular system in which fewer cells are present compared to wild type, and all these vascular cells differentiate into xylem elements (Fig. 4D). The lower part of the hypocotyl in wol seedlings shows the same deviation in vascular cell number and type observed in the root (Fig. 4I). The number of vascular cells increases in the upper part of the hypocotyl at the site just basal to the separation of the vascular bundles to the cotyledons. It is noteworthy that phloem vessels are present in this region.
Roots of fs seedlings show supernumerary cell numbers. This results in an enlargement of the vascular cylinder and multiple cortical cell layers (Fig. 4E). More epidermis and pericycle cells are present in fs roots but they still appear as a continuous layer. Suberin staining remains restricted to one single layer, indicating that also a single layer of endodermis is present (cf. Fig. 6C). The hypocotyl of fs seedlings contains, like the root, too many cells in all regions, leading to the expansion of the presumptive ground tissue towards five irregular layers in comparison with the two layers in wild type (Fig. 4J). It is interesting to note that the widening of the vascular bundle frequently permits the presence of more then two protoxylem poles. A detailed analysis of fs alleles has shown that the presence of more protoxylem strands is linked to the number of cotyledons in fs seedlings (Torres-Ruiz and Jürgens, 1994).
In summary, the phenotypes caused by six different mutations affecting different aspects of the radial organisation of the root are not specific to the part of the root formed by the root meristem but extend into the hypocotyl region.
All mutants display alterations in radial organisation of the embryonic axis
The occurrence of a radial pattern defect in the root as well as in the hypocotyl of seedlings homozygous for the various mutations strongly suggested a defect occurring during embryogenesis. To detect such a defect we sectioned embryos at different stages that were homozygous for the various mutations. Representative sections are depicted in Fig. 5.
Shr, scr and pic mutants have in common a lack of cell layers normally derived from the ground meristem. Embryos homozygous for the shr, scr, and pic mutation are first recognised as different from wild type at the early heart stage, when the periclinal division that doubles the ground meristem layer in the axis is absent. In sections of mid torpedo stage embryos, only one layer derived from the ground tissue is present (Fig. 5B). The periclinal division giving rise to the endodermis does not occur in the ground meristem of the prospective hypocotyl, nor in the descendants of the root cortex initial (cf. Dolan et al., 1993).
The radial pattern defects observed in the stele of glm and wol seedlings are also apparent in embryos. Stele cells of the mature glm embryo are disorganised compared to wild-type embryos (Fig. 5C). Both the root and the hypocotyl region in mature wol embryos show about 10 cells within the layer of cells resembling the pericycle layer in wild-type embryos. The number of cells within the pericycle of wild-type mature embryos is about 18 (Fig. 5D, cf Fig. 1T). The division pattern elaborating all the other tissues before the heart stage of embryogenesis is normal (data not shown). Therefore the last divisions to complete the number of cells in the stele appear absent in wol embryos.
Embryos homozygous for fs display multiple misoriented cell divisions at essentially all stages of development (Torres-Ruiz and Jürgens, 1994; cf. Fig. 5E). While the orientation of these divisions remains predominantly anticlinal in the presumptive pericycle and epidermis, it is randomised in the ground meristem leading to the formation of additional layers (Fig. 5E).
The formation of a root primordium with the correct radial organisation is not restricted to embryogenesis. Lateral roots are newly formed along the primary root during postembryonic development. Furthermore, the formation of root primordia can be induced in tissue culture. In order to assess whether the mutant phenotypes were specific for the developmental context of the embryo, we examined lateral roots emerging from pericycle cells of the primary root, and of roots induced from callus by treatment with auxins. Fresh sections of lateral roots as well as callus-induced roots from shr, scr, pic, wol and fs mutants revealed in all cases the mutant phenotype (data not shown). Seedlings homozygous for the wol mutation produced no lateral roots from the primary root. Lateral roots emerged only from the upper hypocotyl and these had an intermediate vascular phenotype (L. DiL. and P. B., unpublished data). From the observations on lateral and callusinduced roots we concluded that all mutations described here interfere with radial pattern formation during primary as well as secondary root development.
Are the shr, scr and wol mutations primarily interfering with cell specification or morphogenesis?
Seedling and embryo analysis showed that the absence of specific cell types in shr, scr and wol seedlings is linked to the absence of particular divisions during embryogenesis. The absence of a cell type could be a consequence of a primary defect in cell division. Alternatively, these mutations could have a primary defect in cell specification during pattern formation, leading to the absence of later cell divisions. In fs mutants the frequency and orientation of cell divisions is altered, while pattern formation is unaffected as judged by the presence of all tissues in appropriate positions (Torres-Ruiz and Jürgens, 1994). Therefore we created double mutant combinations with fs. If the lack of particular cell divisions in shr, scr, and wol single mutants was counteracted by additional divisions in the fs background, then the question of whether cell specification defects would still occur could be addressed. Potential double mutants were identified as outlined in materials and methods.
Since shr mutants contain no endodermis the progeny of shr × fs crosses was analysed by suberin staining. Seedlings were found with the appearance of a fs single mutant in which the endodermis-specific casparian strip could not be detected in root and hypocotyl tissue, as shown in Fig. 6B. Nevertheless multiple layers derived from the ground meristem are present in such seedlings. In wild-type seedlings and single fs mutants, suberin staining visualises the endodermis (Fig. 6A,C). The fs mutant seedling appearance of the presumptive double mutant is consistent with the pleiotropic effect of mutant fs alleles on cell morphogenesis. We concluded that the block in periclinal divisions of the ground meristem, characteristic for embryos homozygous for shr, can be released by mutation of the FS gene. However, the occurrence of multiple ground tissue layers in the double mutant does not lead to formation of an endodermis layer. Therefore absence of endodermis appears to be a direct effect of the shr mutation that cannot be suppressed in a fs background.
Scr single mutants contain a single ground tissue layer with a casparian strip and hence appear to lack cortex cells. Double mutants in the scr × fs cross were expected to have the overall appearance of fs seedlings due to the pleiotropic nature of the fs mutation. We analysed seedlings with a fs appearance, from families segregating scr and fs seedlings, by suberin staining. A casparian strip could be visualised in a single cell layer at the same position as in fs single mutants in all seedlings investigated, but not in all other ground tissue layers that were present. We concluded that the presence of additional divisions in the ground tissue of a double mutant leads to the restoration of cell layers that do not contain a casparian strip and display the cell morphology of cortex cells. Hence the fs mutation is epistatic to the scr mutation.
Wol mutants display a narrowed vascular bundle in a specific seedling region while fs mutations affect the morphology of all seedling regions. Therefore we expected wol fs double mutants to have the fs seedling phenotype. In families segregating both fs and wol mutant phenotypes, we could not detect novel features suggestive of a double mutant phenotype, upon microscopic analysis. Seedlings with fs appearance were sectioned and they all contained a wide vascular bundle with all tissue types. We concluded that fs is epistatic to wol and that the lack of phloem in wol single mutants can be restored by the presence of additional vascular cells in the wol fs double mutant. The observation that phloem vessels are present in the widening region of the vascular bundle in the upper part of the hypocotyl is in agreement with this conclusion.
Our data indicate that restoration of cell divisions in the ground tissue of scr mutants as well as in the stele of wol mutants leads to the re-appearance of missing cell types. In both shr and wol mutants specific cell divisions during formation of the embryonic axis are absent. We conclude that the absence of layer-specific cell divisions in these mutants causes the absence of specific cell types. Since both mutations affect cell division we were interested to learn whether shr wol double mutants are strictly additive, which would indicate that the two genes function independently in different regions. If, on the other hand, the mutant alleles influenced a related morphogenetic pathway, enhanced phenotypes in both regions might be expected in the double mutant. Families segregating both wol and shr seedlings contained a distinctive double mutant phenotype in seedlings with wol appearance but with a hypocotyl of smaller diameter. The endodermis layer was lacking as judged by suberin staining (Fig. 6D). Sectioning of these seedlings revealed the absence of any vascular cell type besides xylem, as in wol single mutants, and the presence of only a single ground tissue layer, as in shr single mutants (Fig. 6E). Since the shr and wol phenotypes appear strictly additive in root and hypocotyl, independent mechanisms appear to underly cell division defects in stele and ground tissue.
The novel phenotypes found in the double mutant analyses were in all cases expressed in both the root and the hypocotyl (data not shown). This observation is consistent with the conclusion from the single mutants that cells derived from the root meristem elaborate the radial pattern laid down in the embryonic axis.
Embryonic pattern formation is involved in establishing cell identity of root meristem derivatives
We have shown that six different recessive mutations cause alterations of the radial pattern of root tissues, affecting the numbers and, in several cases, the identity of cells within specific layers. In all cases these radial defects originate during embryogenesis and affect all regions of the seedling derived from the embryonic axis. No mutations were found that affect the root only, so we have not obtained evidence for the existence of genes specifically involved in dictating the radial pattern of the root. Based on these observations we postulate that the radial pattern in the meristematic root is regulated by the same factors as those employed in the embryo. Several other observations lend support to the notion that radial pattern information is not exclusively generated within the meristem. First, dissected carrot somatic embryos show replacement of root tissues by cells of the hypocotyl region before a new root meristem is formed (Schiavone and Racusen, 1991). Second, the first differentiated cells of the Arabidopsis primary root originate from a separate embryonic tier not encompassing the meristem initials (Scheres et al., 1994).
On the basis of our data, two models for the establishment of the radial pattern of tissues in the root can be envisaged. In both models root meristem initials are positioned and specified as a result of radial pattern formation within the embryonic axis, as schematised in Fig. 7A,B. Radial pattern information could be maintained in the root meristem in several ways. One possibility is suggested by recent laser ablation experiments performed in the Fucus embryo which raised the possibility that stable determinants of differentiation are linked to the cell wall (Berger et al., 1994). Similarly, radial pattern information could be laid down in the cell walls of root meristem cells (Fig. 7A). A second possibility is that radial pattern information is passed down from maturing tissues to the meristem in a ‘top-down’ flow of information (Fig. 7B). This information could be provided by continuous expansion of the expression domain of genes required for radial pattern formation in the embryo, but also by basipetal transport of radial determinants. Our collection of mutants does not distinguish between the A and B model since in both cases a mutant pattern is perpetuated in the primary root. We are currently testing by laser ablation experiments the prediction of the B model that meristem initials have a flexible fate. Mosaic analysis, to determine whether genes defined by radial pattern mutations are continuously needed, has to await linked marker genes for detection of mutant tissue sectors. The existence of a separate genetic system for pattern generation at the root apex, as depicted in Fig. 7C, is not in agreement with our observation that the radial pattern defects we observed have an embryonic origin.
By screening for radial pattern alterations in roots with a functional root meristem we did not take into consideration mutations in a theoretical class of genes involved in a ‘separate system for radial pattern formation’ in the apex which, by mutation, would yield inactive meristems. We have screened extensively for mutations leading to an inactive root meristem. The cellular organisation of the remnant meristematic root is abnormal in such mutants but there are no layer-specific defects (V.W. and B.S., unpublished data). We conclude that defects in these mutants are not due to an inability of the meristem to dictate specific elements of the radial pattern.
Information to generate the radial pattern of secondary roots in our view is mediated by the same genes that are first active in the embryonic axis. Radial pattern formation and perpetuation can occur in the new root primordium by mechanisms such as those suggested in Fig. 7A,B. Descriptions of lateral root formation in Zea mays and Arabidopsis are consistent with the notion that the promeristem is not the sole source for pattern information. Such studies suggest direct formation of differentiated cell types from the root primordium without involvement of the new apical meristem (McCully, 1975; H.Nusbaum and I.Sussex, personal communication).
It must be stressed that some pattern information may have to be generated in the meristem. The formation of the root cap is specific for the distal end of the root and is likely to be regulated locally. Distortions in the organisation of the root apex, most notably in the columella root cap, have been observed in ttg mutants (Galway et al., 1994). An as yet unspecified role in cell signalling processes during pattern formation has been assigned to the TTG gene, based on several aspects of the mutant phenotype (lack of trichomes, and ectopic development of root hairs). The observed abnormalities in the root apex of ttg mutants may support the notion that local cell signalling is involved in organising the root cap region.
The models proposed in Fig. 7A,B are not inconsistent with the in vitro culture experiments in which the quiescent centre of maize was explanted and was shown to regenerate a root containing all tissues without passing through the callus phase (Feldman and Torrey, 1976). The cell population called the quiescent centre in maize has been shown to be heterogeneous and polarised (Feldman and Torrey, 1976; Müller et al., 1993). If radial determinants are patterned during embryogenesis then this information can reside in the quiescent centre cell population without having been generated there de novo.
Morphogenesis and pattern formation in the embryonic axis
The shr, scr, and wol mutants described in this study all lack specific cell divisions. In addition they have cell specification defects. In plants, cell division is one of the means of controlling the acquisition of structure and form, defined as morphogenesis. The specification of different cell types at appropriate positions, on the other hand, is referred to as pattern formation. We adressed the issue of whether the mutations were primarily caused by defects in morphogenesis (i.e., cell division) or pattern formation. This was done by bringing the mutations into the fs mutant background, which increases cell numbers in the radial dimension. The availability of only single alleles of the region-specific mutants (the rather indistinct seedling phenotype is an unreliable criterion for efficient screening) warrants extreme caution in deducing precise roles for these genes in embryo axis organisation. Such a deduction has to await the isolation of multiple alleles and molecular cloning of the genes. Nevertheless the suppression of the cell type deletions in scr and wol but not shr mutants in the fs background allows some conclusions to be drawn. Only the SHR gene appears to be involved in the specification of a cell type. The SCR and WOL genes appear to influence pattern formation indirectly by regulating regional aspects of morphogenesis.
How do gene products modify patterns of cell division? The cell divisions that structure the Arabidopsis embryonic axis are remarkably uniform, and not species or family specific. Analysis of embryogenesis of Stellaria media has revealed identical division patterns. Balance between pressure from the inner cell mass and surface tension in the outer cells were postulated to be of major importance for the positioning of cell walls (Ramji, 1975). If this were the case, differences in pressure or surface tension may yield variations of the division pattern. Consequently, products of genes like SCR and WOL may program cell division via direct or indirect modification of the parameters mentioned above.
Our observations imply that distortions in morphogenesis leading to too few cells in scr and wol embryos can cause radial pattern deletions. In contrast, morphogenetic mutations leading to too many cells in the fs mutant have no effect on pattern formation (Torres-Ruiz and Jürgens, 1994). What mechanism underlies the influence of morphogenetic mutations leading to too few cells on pattern formation? Analysis of the wol mutation that affects pattern formation in the vascular cylinder, may provide some answers to this question. A relationship between size of the vascular cylinder and the number of xylem and phloem poles has been observed in pea lateral roots (Torrey, 1954) and in roots cultured from the quiescent centre of maize (Feldman and Torrey, 1976). The absence of phloem in the primary root of seedlings homozygous for the wol mutation may be directly due to a small cell number if it is assumed that xylem poles are specified prior to the specification of the other vascular cell types. If there are too few cells in the vascular cylinder, as in the wol mutant, xylem induction may ‘use up’ all the available vascular cells. The occurrence of phloem in wol seedlings where the widening vascular bundle separates to the cotyledons is consistent with this idea. The observed epistasis of fs in the wol fs double mutant is explained if it is assumed that tissue compartmentalisation in the embryo, separating for example vascular tissues and ground tissue, occurs prior to final specification of cell types. In this scenario the early additional divisions in the fs background lead to specification of a vascular region with more cells prior to the block of late, central cell divisions caused by the wol mutation. Therefore cells are available for other fates after xylem specification. The correlation between xylem pole number and position and the number and position of cotyledons in wild type, and also in fs mutants (Torres-Ruiz and Jürgens, 1994), tempts one to speculate that cotyledon factors are involved in this specification.
We thank F. Kindt, R. Leito, W. van Veenendaal and P. Brouwer for photography; D. Smit for artwork; G. Jürgens, both referees and C. van den Berg for valuable suggestions on the manuscript. We are indebted to R.Torres-Ruiz for performing fs complementation analysis.