Mucilages are largely composed of pectins, a heterogeneous group of complex, acidic polysaccharides that also comprise the majority of the plant cell wall matrix. Dicotyledonous pectins largely consist of poly-GalUA (PGA) and rhamnogalacturonan I (RG I; Brett and Waldron, 1990; Carpita and Gibeaut, 1993; Cosgrove, 1997). PGA is composed of an unbranched chain of α1,4-linked GalUA residues, whereas RG I is a highly branched polysaccharide with a backbone of alternating α1,4-linked GalUA and α1,2-linked rhamnose (Rha), with sugar side chains attached to the Rha residues (Brett and Waldron, 1990). The degree of gelling of pectins is largely dependent on ionic bonding between PGA molecules and free divalent calcium. Thus, cell wall fluidity is affected by the degree of methyl esterification of PGA carboxyl groups and the frequency of interruptions of homogalacturonan chains with RG I (Brett and Waldron, 1990; Carpita and Gibeaut, 1993). Studies of PGA and RG I production have shown that they are manufactured in the Golgi apparatus, then transported to the extracellular matrix via secretory vesicles (Brett and Waldron, 1990; Zhang and Staehelin, 1992; Driouich et al., 1993; Staehelin and Moore, 1995; Dupree and Sherrier, 1998). Although some pectin biosynthetic enzymes have been identified biochemically (e.g. Rodgers and Bolwell, 1992; Piro et al., 1993; Doong and Mohnen, 1998; Edwards et al., 1999; Perrin et al., 1999; Seitz et al., 2000), little is known about the regulation of complex polysaccharide biosynthesis and secretion.

The genetic model species Arabidopsis, a member of the Brassicaceae, is myxospermous. In addition, its epidermal cells are marked by a central volcano-shaped secondary cell wall known as the columella (Vaughn and Whitehouse, 1971; Koornneef, 1981). Differentiation of the outer integument epidermal cells to form the seed mucilage cells involves a highly regulated series of events, including growth, morphogenesis, mucilage biosynthesis and secretion, and secondary cell wall production (Beeckman et al., 2000; Western et al., 2000; Windsor et al., 2000). Although the presence of mucilage in Arabidopsis seeds is dispensable under laboratory conditions, only a few genes affecting seed coat morphology have been identified: TRANSPARENT TESTA GLABRA1 (TTG1), GLABRA2 (GL2), APETALA2 (AP2), ABERRANT TESTA SHAPE, and ABSCISIC ACID DEFICIENT1. In each case, the seed coat defect has been noted as a pleiotropic effect of a mutation in the gene and the seed specific cellular defects have not been investigated in detail. ttg1 and gl2 mutants were both originally identified as trichome mutants, which lack leaf hairs (Koornneef, 1981; Rerie et al., 1994). Further study of each revealed epidermal cell defects throughout the plant in the form of extra root hairs and lack of seed coat mucilage and columellae (Koornneef, 1981; Bowman and Koornneef, 1994; Galway et al., 1994; Rerie et al., 1994; Di Cristina et al., 1996; Masucci et al., 1996). AP2 is involved in the regulation of both flower and ovule development (Bowman et al., 1989, 1991; Kunst et al., 1989; Modrusan et al., 1994; Western and Haughn, 1999). Observation of ap2 seeds showed altered seed shape and the absence of mucilage and columellae (Bowman and Koornneef, 1994; Jofuku et al., 1994). Mutants defective in ABERRANT TESTA SHAPE, a gene involved in ovule integument development, have heart-shaped seeds that have a reduced amount of mucilage (Léon-Kloosterziel et al., 1994). Finally, abscisic acid deficient1 mutants, as part of their syndrome resulting from lack of abscisic acid, produce a reduced amount of mucilage (Karssen et al., 1983).

As a first step in analyzing the complex process of Arabidopsis seed epidermal cell differentiation, we and others have made a detailed study of wild-type development and mucilage composition (Beeckman et al., 2000; Western et al., 2000; Windsor et al., 2000). In this paper, we report the use of a novel screen for the isolation of mutants specifically defective in the mucilage-containing seed coat epidermal cells. Our characterization of the seed coat defects in five such mutants (mucilage-modified [mum] 1–5), as well as the previously identified mutants ap2, ttg1, and gl2, indicate that the products of these mutated genes act at multiple steps in seed coat epidermal cell differentiation, ranging from the regulation of outer integument differentiation to mucilage biosynthesis and post-deposition cell wall modification.

Figure 1 Scanning electron micrographs showing whole seed and seed coat details of wild-type and mutant seeds. A, Wild-type seed. Note hexagonal epidermal cells with thickened radial cell walls and volcano-shaped columellae in the center of each cell. B, Detail (more ...)

Figure 2 Ruthenium red staining of wild-type and mutant seeds. A, Wild-type seed placed directly into stain without agitation. Two layers of mucilage are present, an outer, cloudy layer, and an inner, intensely staining layer. B, Wild-type seed first shaken in (more ...)

The late differentiation of the integuments of the wild-type ovule into the seed coat of Arabidopsis has been divided into five dynamic stages (Western et al., 2000). The first stage, passing from Figure 3, A to B, is a period of cell growth driven by vacuolar expansion to give cells with a large central vacuole surrounded by cytoplasm. Stage 2 is marked by the accumulation of large starch granules (amyloplasts) and later by the drawing of the cytoplasm away from the edges of the cell. Stage 3 has further rearrangements such that the cytoplasm becomes a discrete column the center of the cell. In addition, the cells manufacture and secrete large quantities of pectic mucilage between the plasma membrane and primary cell wall on the outer tangential face of the cell as shown in Figure 3, C and D, and Figure 4A. In stage 4, mucilage production has ceased and a new, secondary cell wall is produced that replaces the column of cytoplasm, resulting in the columella (Figs. 3E and 4C). Desiccation of the seed in Stage 5 leaves a central columella surrounded by shrunken mucilage, with the entire cell overlaid by the primary cell wall (Fig. 3K). Upon hydration of the seed, the hydrophilic mucilage swells rapidly, leading to the rupture of the primary cell wall at the thin radial cell walls and release of the gel-like mucilage to surround the seed (Fig. 3F).

Figure 3 Structure and development of wild-type and mutant seed coats. Plastic sections of tissue fixed in 3% (v/v) aqueous glutaraldehyde (unless otherwise stated) and stained with Toluidine blue. A, Mature wild-type ovule (0 d after pollination [DAP]). (more ...)

Figure 4 Transmission electron micrographs of developing epidermal cells of wild-type and ttg1-1seeds. A, Wild-type seed at 7 DAP. The cytoplasm is drawn into the center of the cell in a sharply defined column above the vacuole. Note large amyloplasts. B, ttg1-1 (more ...)

Table I Results of reciprocal backcrosses to wild-type plants and segregation analysis

Light and scanning electron microscopy (SEM) were used to categorize ttg1, gl2, ap2, and the mum mutants into distinct groups based on mature seed phenotypes. Ruthenium red staining after shaking in water divided the mutants into two categories depending on their mucilage capsule phenotype: no mucilage capsule (mum1-1, accession no. CS3903; mum2-1, accession no. CS3904; mum2-2, accession no. CS3905; mum4-1, accession no. CS3907; ttg1-1; gl2-1; ap2-1; and ap2-6; Fig. 2C), and a very reduced mucilage capsule just seen around the seed periphery (mum3-1, accession no. CS3906; mum4-2, accession no. CS3908; and mum5-1, accession no. CS3909; Fig. 2D). SEM of dry seed determined that the seed coat epidermal cells of mum2-1, mum2-2, mum3-1, and mum5-1 have wild-type cell surface features, mum1-1 cells have slightly irregular columellae (data not shown), and mum4-1, mum4-2, gl2-1, and ttg1-1 have absent or reduced columellae (Fig. 1, C and D). ap2-1 seeds are heart shaped and have epidermal cells that are often rectangular rather than hexagonal and lack or have reduced columellae (Fig. 1, E and F); ap2-6 seeds were much more severe than ap2-1 (Fig. 1, G and H; compare with Fig. 1, E and F, respectively). Thus, ttg1, gl2, ap2, and mum1 through 5 can be divided into four groups based on their seed coat defects: (a) no mucilage capsule (mum1 and mum2), (b) no mucilage capsule and reduced columellae (mum4, gl2, and ttg1), (c) reduced mucilage capsule (mum3 and mum5), and (d) no mucilage and aberrant seed coat (ap2). In the following sections, we describe further characterization of these four categories to determine the origin of their seed coat defects.

To determine whether these changes could be due to an alteration in mucilage composition and/or amount compared with wild type, the carbohydrate content of the mutant seeds was investigated using coupled gas chromatography (GC)-mass spectrometry (MS). To use a single assay to reveal both the basic neutral sugar profile and give a gross estimate of the relative uronic acid levels, trimethylsilane derivatives were used. We have shown previously that the mucilage extruded from wild-type (Columbia [Col]-2) Arabidopsis seeds gives a consistent monosaccharide profile (Western et al., 2000; see Fig. 5C). However, mucilage is retained within the epidermal cells of mum1-1 and mum2-1 seeds. Thus, as a control for these mutants, we repeated the sugar analysis, this time by grinding wild-type (Col-2) seeds and going through extraction and hydrolysis in the presence of crushed seeds, rather than just hydrolyzing and derivatizing the ammonium oxalate-soluble sugars found in the supernatant after shaking intact seeds. Once again, a consistent sugar profile was obtained (Fig. 5A), though considerably more complex than that from mucilage alone (compare Fig. 5, A with C) because the whole seed extraction also includes sugars from cell walls and other compartments of cells of the seed coat, endosperm, and embryo. Similar to wild type, derivitization of ground mum1-1 and mum2-1 seeds yielded reproducible sugar profiles (Fig. 5A). A comparison of the monosaccharide profile obtained from ground mum2-1 seeds with that of wild-type seeds revealed that there is no significant difference in either sugar composition or amount (Fig. 5A). In a similar manner, mum1-1 seeds had a similar composition, but an increased amount of most sugars compared with wild-type seeds. The increased level of sugars in mum1-1 seeds is not correlated with a gross increase in seed size as viewed externally (data not shown).

Figure 5 Monosaccharide composition of wild-type and mutant seeds and mucilage. The bars represent the average quantity of monosaccharide found by GC over multiple experiments. The data were standardized and converted to micrograms per 100 seed through comparison (more ...)

Developmental differences from wild type were also investigated for mum2-1 using Toluidine blue stained sections, SEM, and transmission electron microscopy (TEM). The timing and production of mucilage and columellae appears to be identical to wild type, with the exception that mucilage is retained within the cells upon maturity (Fig. 3, G–J; compare with Fig. 3, C–F). In developing wild-type seeds exposed to aqueous conditions (i.e. fixation), mucilage release can occur starting at the beginning of Stage 4, once most of the mucilage has been deposited and the new secondary cell wall of the columella is initiated (data not shown; Fig. 3E). Some mucilage release was also observed in mum2-1 seeds at the beginning of Stage 4 (Fig. 3H), but by the end of this stage, the ability to release mucilage had been lost (Fig. 3, I and J). These results suggest that cell wall or mucilage modification is occurring in the mum2-1 and possibly mum1-1 seeds during Stage 4 that prevents subsequent mucilage release. Because a common modification of pectins is demethylation of the PGA carboxyl groups after secretion (Brett and Waldron, 1990; Carpita and Gibeaut, 1993), we performed methylation analysis on ammonium oxalate extracts of intact, wild-type mum1-1 and mum2-1 seeds. Our results revealed a significant, 6% to 8% increase in methylation for the mutants compared with wild-type seeds (average values of 25.6% and 23.9% for mum1-1 and mum2-1, respectively, versus 18.3% for Col-2 wild type; measurements were done in triplicate; mum1-1 versus Col-2, H₀: μ₁ = μ₂, T = −33.792, P 0.01; mum2-1 versus Col-2, H₀: μ₁ = μ₂, T = −17.431, P 0.01), which is consistent with an altered methylation state of the mucilage and/or primary cell wall pectins. Because the degree of pectin methylation could affect both cross-linking and hydrophilicity, it is conceivable that an increase in the proportion of methyl esterified PGA molecules could be responsible for the retention of mucilage within the primary cell wall of the seed coat epidermis of these mutants. However, we cannot exclude the possibility that the observed difference in methylation could be simply a consequence of differential extraction of wild-type seeds bearing extruded mucilage versus mum1-1 and mum2-1 seeds where the mucilage is contained within intact epidermal cell walls.

The possibility of a reduced amount or altered chemical composition of mucilage in these mutants was explored using GC-MS. Once again, due to mucilage retention in the epidermal cells, ground seeds were used for analysis and reproducible profiles were obtained for Ler wild type, ttg1-1, gl2-1, mum4-1, and ap2-1 (see Fig. 5B for Ler, ttg1-1, and ap2-1; gl2 and mum4-1 data not shown). When ttg1-1, gl2-1 and mum4-1 seeds were compared with wild type, it was found that all the monosaccharides were present, but the peak for GalUA was significantly reduced (ttg1-1 versus Ler, H₀: μ₁ = μ₂, T = 11.758, P < 0.01; ap2-1 versus Ler, H₀: μ₁ = μ₂, T = 12.616, P < 0.01). The peak for Rha and Fuc (ttg1-1 versus Ler, H₀: μ₁ = μ₂, T = 8.769, P < 0.01; ap2-1 versus Ler, H₀: μ₁ = μ₂, T = 11.143, P < 0.01), and the Gal peak were also smaller (ttg1-1 versus Ler, H₀: μ₁ = μ₂, T = 8.319, P < 0.01; ap2-1 versus Ler, H₀: μ₁ = μ₂, T = 8.642, P 0.01). Thus, all three mutants have a decrease in the monosaccharides (GalUA and Rha) that are major components of pectin. Because similar decreases were seen in ap2-1 seeds where little or no mucilage appears to be synthesized (see below), these results suggest that less mucilage may be made in ttg1-1, gl2-1, and mum4-1 seed coats.

The origin of the flattened columellae was studied in gl2-1, ttg1-1, and mum4-1 by following the differentiation of the seed coat epidermis using both Toluidine blue-stained sections (Fig. 3, L–O) and TEM (Fig. 4). Similar results were seen for all three mutants; thus, only ttg1-1 will be described in detail. During the first 4 d after pollination, ttg1-1 development was similar to wild type, with the enlargement and vacuolization of the epidermal cells together with the accumulation of amyloplasts (data not shown). This was followed by rearrangement of the cytoplasm and accumulation of pink-staining mucilage (Fig. 3L). Differences between wild type and ttg1-1 development became evident just prior to Stage 4, where new cell wall is laid down around the cytoplasm. In wild-type seeds, the cytoplasm is first pulled into the center of the cell over a large basal vacuole, followed by narrowing of the column and reduction of the vacuole; the secondary cell wall is then laid down, forming a tall, volcano-shaped structure (Figs. 3, C–E, and 4, A and C). In ttg1-1, gl2-1, and mum4-1 seeds, however, only the first stage of the cytoplasmic rearrangement occurs and the secondary cell wall is laid down in a peaked dome over a large vacuole (Figs. 3, L–N, and 4, B and D). Desiccation leads to the crushing of the vacuole and a flattened columella (Fig. 3O). These data suggest that in all three mutants, there is a defect in the later stages of cytoplasmic rearrangement and/or constriction.

It is interesting that a phenocopy of mum3-1 and mum5-1 mutant seeds resulted from wild-type seeds shaken in the presence of EDTA or EGTA prior to staining with Ruthenium red (compare Fig. 2, F with D). Ruthenium red dye stains molecules with two negative charges 0.42 nm apart (Sterling, 1970). Both EDTA and EGTA are heavy metal chelators, with EGTA being relatively specific for Ca²⁺. The removal of Ca²⁺ ions could lead to disruption of the ionic cross-linking of pectin (PGA) carboxyl groups, allowing for separation of PGA molecules under agitation and thus loss of Ruthenium red staining. By a similar argument, the defect in mum3-1 and mum5-1 could be due to increasing methyl esterification and consequent neutralization of the PGA carboxyl groups. Methylation analysis of ammonium oxalate soluble sugars extracted from intact seeds, however, showed a slight decrease in methylation in these mutants (average values of 15.8% and 14.1% for mum3-1 and mum5-1, respectively, versus 18.3% for Columbia; measurements were done in triplicate with no measurement more than 2.2% from the mean). In an alternate manner, the ability to phenocopy with chelators may reflect a more basic change in mucilage composition. To investigate this possibility, the sugar composition of mum3-1 and mum5-1 mucilage was analyzed by GC-MS of ammonium oxalate-soluble sugars obtained from intact seeds. The results (Fig. 5C) showed that for both there is a consistent, significant decrease in the peak containing both Rha and Fuc (mum3-1 versus Col-2, H₀: μ₁ = μ₂, T = 4.188, P < 0.01; mum5-1 versus Col-2, H₀: μ₁ = μ₂, T = 5.823, P < 0.01), suggesting that mum3-1 and mum5-1 mutants have defects in mucilage sugar composition, which may affect its basic structure and thus its branching and cross-linking properties.

In some seed batches, the seed coat phenotype of ap2-6 varied in penetrance such that seeds with either mutant or wild-type appearance were produced. Aberrant and normal seeds were generally found in different siliques. Although normal or mutant-bearing siliques both arise from flowers with an ap2 phenotype, there was a correlation between the severity of morphological defects of the ap2-6 flower and defects in the seed coat of the seeds derived from it. The “wild-type” ap2-6 seeds appeared normal in all aspects of seed and seed coat shape and development (data not shown).

The seed coat phenotypes of plant lines homozygous for several other ap2 mutant alleles were examined by SEM and light microscopy. Seed and seed coat epidermal cell shape defects similar to ap2-6 have been observed in two other strong alleles of ap2 (ap2-7 and ap2-2; data not shown). Seeds from plants homozygous for the weak allele ap2-1 had phenotypes similar to but less severe than ap2-6 but did not display variation in penetrance. ap2-1 mutant seeds, like those of ap2-6, are heart shaped (Fig. 1E). The epidermal cells, however, were variable in shape ranging from hexagonal cells with almost normal columellae to rectangular, thin-walled cells lacking columellae (Fig. 1F). Sectioning of a developmental series of ap2-1 seeds showed results consistent with similar but less severe effects on epidermal cell development compared with ap2-6 seeds (data not shown).

The development of the epidermal layer of the seed coat from the outer integument of the ovule in Arabidopsis is a complex process, involving cell growth, cytoplasmic rearrangement, biosynthesis and secretion of pectinaceous mucilage, and production of a secondary cell wall (Beeckman et al., 2000; Western et al., 2000; Windsor et al., 2000). We have shown here that mutants specifically defective in the seed coat epidermis can be isolated and should be useful in dissecting many of these processes. The loss of the seed coat epidermal cells and those of the palisade layer in the ap2 mutants had no obvious effect on viability and germination, demonstrating that these cell types are completely dispensable under laboratory conditions. Therefore, the seed coat epidermal cells represent an excellent model system for the use of genetics to study carbohydrate synthesis and secretion, secondary cell wall biosynthesis, and cell morphogenesis.

A hint as to the type of defect preventing extrusion in the mutants comes from methylation analyses suggesting an increase in overall methylation of mum1 and mum2 seed pectin. The degree of methylation affects the number of free carboxyl groups available in the pectin, which in turn affects both pectin hydrophilicity and its ability to form Ca²⁺ bridges between PGA molecules (Bolwell, 1988; Brett and Waldron, 1990; Carpita and Gibeaut, 1993; Reiter, 1998). An increase in methyl-esterified PGA may reduce the affinity of the mutant's mucilage for water, thereby lessening either the speed or the extent of hydration and decreasing the ability of the hydrated mucilage to break the primary cell wall. This idea is supported by the observation that some mucilage release can be obtained by treating mum1 and mum2 seeds with the Ca²⁺-specific chelator EGTA (T.L. Western, W.L. Tan, and G.W. Haughn, unpublished data). Treatment with EGTA might remove Ca²⁺ as a competitor with water for the limited number of free carboxyl groups within the mutant's mucilage, thereby partially suppressing the extrusion defect. However, our data do not allow us to determine if the increased methylation is limited to the mucilage or affects pectin of the primary cell wall as well. Thus, we cannot eliminate the possibility that an increase in pectin methylation in the mutant in some way strengthens the primary cell wall, thereby preventing extrusion.

Studies of pectin biosynthesis in root caps have shown that the degree of PGA methyl-esterification differs between cell types and the degree of methyl esterification is likely controlled by the secretion of pectin methyl esterases into the cell wall (Moore et al., 1991; Zhang and Staehelin, 1992; Sherrier and VandenBosch, 1994; Staehelin and Moore, 1995). Assuming an analogous situation in the epidermal cells of Arabidopsis seeds, it is possible that MUM1 and MUM2 encode either pectin methyl esterases or positive regulators of such an enzyme in the seed coat.

Because the gl2, ttg1, and mum4 phenotypes all include changes in both the amount of seed coat mucilage deposited in the extracellular space and the cytoplasmic rearrangements during seed coat epidermal cell differentiation, these two processes are likely to be related. This functional relationship between the late events of cellular morphogenesis and mucilage deposition, events that occur simultaneously during differentiation, could be explained in several ways. First, both processes probably rely on the cytoskeleton. Recent studies have demonstrated the important role of the cytoskeleton in trichome morphogenesis (Oppenheimer et al., 1997; Mathur et al., 1999; Szymanski et al., 1999; Mathur and Chua, 2000). In a similar manner, movement of the cytoplasm in the seed coat epidermal cells from the edges of the cell to a precise column in the center could be an active process that involves the cytoskeleton. Mucilage deposition requires polar secretion, which is likely to be dependent on directional trafficking of vesicles via the cytoskeleton (Hyde, 1970; Van Caeseele et al., 1981; Staehelin et al., 1990; Lynch and Staehelin, 1992, 1995; Western et al., 2000). Thus, the gl2, ttg1, and mum4 mutant phenotypes could be due to defects in the cytoskeletal rearrangements required for cell differentiation. Second, the complete compression of the cytoplasm and expulsion of water from the vacuole of differentiating seed coat epidermal cells may be dependent on pressure created by the accumulation of mucilage between the plasma membrane and the outer tangential cell wall. According to this hypothesis, the improper restriction of the cytoplasm and vacuole in the mutant seed coats would be a secondary effect of the reduced amount of mucilage produced in these mutants. Third, mucilage secretion may be dependent on cell morphogenesis to provide an optimal amount of membrane surface relative to the amount of extracellular space. If so, the lower mucilage deposition in the mutants could be a secondary effect of a defect in cell morphogenesis. Identification of a seed coat mutant defective in either cell morphogenesis or mucilage deposition but not both would help distinguish between these possibilities.

Both GL2 and TTG1 have been cloned. GL2 encodes a putative homeodomain transcription factor (Rerie et al., 1994; Di Cristina et al., 1996), whereas TTG1 encodes a protein of unknown function with a putative WD40 protein-protein interaction domain (Walker et al., 1999). In addition to their roles in the seed coat epidermis, GL2 and TTG1 are required for normal development of both trichomes and root hairs (Koornneef, 1981; Galway et al., 1994; Rerie et al., 1994; Di Cristina et al., 1996; Masucci et al., 1996). In a converse manner, MUM4 appears to be seed specific, making it a candidate downstream target or specificity factor for these genes.

The molecular structure of Arabidopsis mucilage is unknown, but similar to other mucilages, it is likely to consist of a network containing not only PGA and RG I, but other complex polysaccharides, including hemicelluloses (Lynch and Staehelin, 1992, 1995), and possibly even structural cell wall proteins. This network is held together through Ca²⁺ bridges between PGA molecules and bonding among complex polysaccharides and between proteins (Bolwell, 1988; Brett and Waldron, 1990; Carpita and Gibeaut, 1993; Reiter, 1998). Therefore, weakening of the network in mum3 and mum5 seeds could be due to anything from neutralization of PGA by increased methyl esterification to reduced bonding between complex polysaccharides or proteins. The former possibility is unlikely due to the slight decrease in methylation state in these mutants. In a converse manner, the alteration of relative monosaccharide levels in mum3 and mum5 mucilage would seem to reflect a change in the polysaccharide composition itself. This could be due to a change in the number or types of polysaccharides present, which could in turn lead to decreased cross-linkages between molecules within the gel. Under this assumption, MUM3 and MUM5 could encode biosynthetic enzymes or regulators of complex polysaccharide production.

Very few biosynthetic and regulatory genes have been identified for complex polysaccharide biosynthesis in plants. These include the recently identified MUR genes of Arabidopsis, which were identified through altered leaf sugar composition (Reiter et al., 1993, 1997; Zablackis et al., 1996; Bonin et al., 1997; Burget and Reiter, 1999). Because several of the mur mutants have lower Fuc and/or Rha levels (mur1, mur2, mur8, and mur11), we tested them with our Ruthenium red staining assay to see if any had a phenotype similar to mum3 and mum5. Under these conditions, all the mur mutants tested appeared wild type (T.L. Western, W.L. Tan, and G.W. Haughn, unpublished data), suggesting that MUM3 and MUM5 represent novel loci affecting complex polysaccharide biosynthesis.

The AP2 gene is required for ovule morphogenesis (Modrusan et al., 1994; Western and Haughn, 1999). A small percentage of ap2 mutant ovules develop as carpel-like structures with the remaining ovules developing normally. Because the seed coat is derived from the ovule integuments, it is possible that the failure in seed coat differentiation is a consequence of earlier defects in ovule development. In an alternate manner, AP2 may regulate ovule morphogenesis and seed coat differentiation independently of one another. Given the variability of the ovule defects and the specificity of the seed coat defects, we tend to favor the latter hypothesis.

Figure 6 A time course of mucilage secretory cell development showing the timing of TTG1, GL2, AP2, and MUM1 through 5 action as determined from their phenotypes. Mucilage secretory cell development can be divided into five stages: (1) growth, (2) amyloplast accumulation (more ...)

Due to the maternal origin of the seed coat, mutant screens were performed on M3 lines. The M3 lines were derived from plants randomly chosen from several independent ethylmethane sulfonate-mutagenized M2 populations of Arabidopsis ecotype Col-2 (M1 population size > 40,000). Seeds were screened by first shaking seeds in water, then placing them in a 0.01% (w/v) aqueous solution of Ruthenium red before inspection under a dissection microscope. Suspected mutants were then rescued through germination on Arabidopsis minimal medium (Haughn and Somerville, 1986), then transferred to soil at the two-true-leaf stage.

Compounds were identified initially through comparison with the retention times obtained with individual sugar standards, and then confirmed through GC-MS. GC-electron impact MS was performed by the University of British Columbia Mass Spectrometry Centre (Vancouver). Individual sugar standards and a composite standard were made from the following monosaccharides: myo-inositol (used as internal standard), Fuc, Man, Gal, Glc, Ara, Rha, Xyl, GlcUA, and GalUA (Chaplin, 1986).

We thank Drs. Elaine Humphrey, Lacey Samuels, Mary Berbee, and Mr. Réza Shahidi (University of British Columbia, Vancouver, Canada) for assistance with microscopy; Ms. Yeen Ting Hwang (University of British Columbia) for valuable technical assistance; and Dr. Gunter Eigendorf (University of British Columbia Chemistry Mass Spectrometry Facility) and Dr. Anthony Millar (University of British Columbia) for help with chemical analysis of mucilage. We also thank Drs. Ljerka Kunst, Linda Matsuuchi, Jennifer Klenz, Mr. Mark Pidkowich, Ms. Yeen Ting Hwang, and Mr. Theodore Popma (University of British Columbia) for helpful discussions and comments on the manuscript.