The production of mucilage in the epidermal cells of the Arabidopsis seed coat is part of a coordinated developmental process that begins with cell growth, followed by biosynthesis and polar secretion of large quantities of pectin, formation of a cytoplasmic column through cytoplasmic constriction and vacuolar contraction, and, finally, the synthesis of a secondary cell wall (columella; Beeckman et al., 2000; Western et al., 2000; Windsor et al., 2000). Seed coat mucilage is dispensable under laboratory conditions in Arabidopsis and mum (mucilage-modified) mutants affecting mucilage production can be identified by a simple screening method (Western et al., 2001). Mutations in one of these genes, MUM4, yield a phenotype where mucilage is not released from hydrated mature seeds.

Several genes encoding putative transcription factors have been implicated in seed coat epidermal development because mutations in these genes result in seeds that fail to release mucilage upon hydration. Mutants in AP2 (APETALA2), in addition to their defects in floral morphogenesis, lack differentiation past the growth phase of mucilage secretory cells (Jofuku et al., 1994; Western et al., 2001). TTG1 (TRANSPARENT TESTA GLABRA1), TTG2, and GL2 (GLABRA2), which were originally identified through their role in trichome specification, have defects in both mucilage and columella production in the seed coat (Koornneef, 1981; Penfield et al., 2001; Western et al., 2001; Johnson et al., 2002). It has been shown at the genetic and molecular levels in trichomes and root hairs that TTG1 interacts with a basic helix-loop-helix (bHLH) protein and a tissue-specific MYB protein to activate both GL2 and TTG2 (Payne et al., 2000; Johnson et al., 2002; Schiefelbein, 2003). Recently, mutations in MYB61 have been found to specifically affect both mucilage and columella production in the Arabidopsis seed coat (Penfield et al., 2001).

The objective of this study was to characterize the role of MUM4 in the development of the Arabidopsis seed coat. Using positional cloning of MUM4, we demonstrate that MUM4 encodes a putative NDP-l-Rha synthase. Mutations in this gene lead to reduced mucilage in the seed coat and an altered columella. Expression studies show that MUM4 is developmentally regulated during seed coat differentiation such that its transcript levels are increased at the time of mucilage production. Furthermore, MUM4 appears to be a downstream target of a cascade of transcription factors that includes AP2, TTG1, and GL2. These results demonstrate the importance of mucilage production for the morphology of seed coat epidermal cells and suggest a regulatory framework for the control of seed coat epidermal differentiation.

Figure 1. Structure and development of wild-type (Columbia-2 [Col-2]) and mum4-1 seed coats. A, Whole-mount wild-type seed stained with Ruthenium red. Note red staining capsule of mucilage surrounding seeds. B, Scanning electron micrograph of wild-type seed coat. (more ...)

Viewed with scanning electron microscopy, mature mum4 seed coats are marked not only by the absence of mucilage extrusion upon hydration but also by the absence of volcano-shaped columellae (compare Fig. 1, B with E). Light microscopy of sections of mature seeds revealed that both columellae and mucilage are present in mum4 but reduced in comparison with wild type (compare Fig. 1, C with F). No other phenotypes were observed.

To more fully characterize the mum4 seed coat epidermal cell defects, their development was studied with light microscopy (Fig. 1, G-L). Under our growth conditions, seed maturation to dry seed takes approximately 18 d and has been staged by DPA. In terms of embryo growth, 4 DPA is early to mid globular stage, 7 DPA is mid-torpedo stage, whereas 10 DPA is late upturned-U stage (this study; Western et al., 2000). The early stages of mum4 epidermal cell differentiation are similar to that of wild type: The cells grow, and mucilage starts to accumulate, accompanied by the initiation of intracellular cytoplasmic rearrangement to form a column in the center of the cell (compare wild type, Fig. 1, G and H, with mum4, Fig. 1, J and K). Differences, however, were obvious at 10 DPA. Although wild-type seed coat cells have dark-pink staining mucilage in the extracellular space and a thin secondary cell wall surrounding a central column of cytoplasm (Fig. 1I), the extracellular space containing mucilage in mum4 seed coat epidermal cells is smaller with lighter staining mucilage and the secondary cell wall has formed around a partially formed cytoplasmic column atop a large vacuole (Fig. 1L). Thus, the seed coat appears to have less mucilage, and the flattened columella appears to be the result of reduced cytoplasmic and vacuolar constriction.

The possibility of a change in mucilage quantity or composition in mum4 versus wild-type seed coats was addressed by quantifying monosaccharides using gas chromatography. Because mum4 seed coats release mucilage in the presence of ammonium oxalate, mucilage could be extracted from intact mum4 seeds for direct comparison with wild-type mucilage. Trimethylsilane (TMS) derivatives were used because this method allows detection of GalUA and neutral sugars simultaneously. The results showed that, for equal seed masses, less sugar was extracted from mum4 mucilage, suggesting that less mucilage is synthesized compared with wild-type seeds (Fig. 2). In addition, the decrease in monosaccharides in mum4 mucilage was limited primarily to two sugars, GalUA and a single neutral sugar (identified as either Rha or Fuc; Fig. 2; H₀ μ₁ = μ₂, GalUA, T = 7.897, P < 0.01; Rha, T = 6.960, P = 0.01). Because the identity of the neutral sugar could not be completely resolved using TMS derivatives, alditol acetate derivatization, which gives a single derivative for each monosaccharide, was used to confirm that it is Rha (Fuc found only at trace levels in wild-type mucilage; data not shown). These results are consistent with the suggestion that Arabidopsis mucilage is comprised primarily of relatively unbranched RGI (Penfield et al., 2001).

Figure 2. GalUA and Rha levels of wild-type (Col-2) and mum4-1 mucilage. Comparison of GalUA and Rha levels detected in mucilage isolated from intact mum4-1 seeds (white bars) versus wild-type mucilage (black bars, Col-2) using TMS derivatization and gas chromatography. (more ...)

Figure 3. Immunofluorescence of wild-type (Col-2) and mum4-1 mucilage with antipectin antibodies. A, Immersion immunofluorescence of whole wild-type seeds, labeled with α-PGA/RGI primary and anti-rabbit Alexa 594 secondary antibody. Note autofluorescence (more ...)

To examine the mucilage in situ, samples were cryo-fixed/freeze substituted, resin embedded, sectioned, and then labeled by indirect immunofluorescence. Labeling was seen on thick sections of developing wild-type seed coats stained with α-PGA/RGI (9 DPA; Fig. 3E), with fluorescence in the epidermal cells on either side of a central, unstained space. Control experiments including omitting the primary antibody or substituting nonimmune, normal rabbit serum for the primary antibody, abolished the fluorescent signal (Fig. 3F). The similarity in staining pattern with α-PGA/RGI in sections compared with the pink staining substance in toluidine blue-stained sections (e.g. Fig. 1I) confirms the identity of the pink staining substance as pectinaceous mucilage. When thick sections of mum4 seeds were treated with α-PGA/RGI, only a small amount of labeling was observed on the outer edges of the epidermal cells (Fig. 3H) compared with wild-type seeds (Fig. 3G), once again suggesting a reduced amount of mucilage in mum4 seeds.

Next, we isolated a putative full-length cDNA by reverse transcriptase (RT)-PCR using a series of primers upstream from the predicted translation start site. Sequencing of a putative full-length At1g53500 transcript revealed that the gene consists of three exons and two introns (including one in the 5′-untranslated region [UTR]) and encodes a transcript of 2,477 bp (Fig. 4A).

Figure 4. Structure of the MUM4, RHM1, and RHM3 genes and complementation of mum4-1 with MUM4 coding region. A, Exonintron structures are based on comparison of the genomic and cDNA sequences. The coding region is shown as black boxes, whereas the gray boxes represent (more ...)

To confirm that At1g53500 is MUM4, we sequenced this locus in mum4-1, mum4-2, and mum4-3. All three alleles contain mutations in the second exon. mum4-1 and mum4-2 contain point mutations leading to G to A transitions at nucleotides 595 and 886, respectively (Fig. 4A). These missense mutations should result in a conserved Asp residue being replaced with an Asn in mum4-1 and a Gly being replaced with an Arg in mum4-2 (Fig. 5). mum4-3 contains a T-DNA insertion at nucleotide 1,637 (Fig. 4A). The similar phenotype between the missense alleles and that of the mum4-3 insertional allele that lacks detectable transcript (data not shown) suggests that these changes have a significant effect on protein activity. A database search of the Salk Institute sequence-indexed Arabidopsis T-DNA insertion lines identified two new alleles of mum4, mum4-4, and mum4-5 (Salk_085051 and 038898, respectively), both of which also have insertions in the second exon (Fig. 4A). Based on the Ruthenium red assay, both of these mutants have a phenotype (data not shown) similar to mum4-1 (Fig. 1D). The wild-type At1g53500 sequence plus 2.1 kb upstream and 0.4 kb downstream sequence was cloned to give the plasmid pMUM4g. Transformation of mum4-1 plants with pMUM4g versus an empty vector control showed that the wild-type At1g53500 genomic clone could rescue the mucilage phenotype (Fig. 4, B and C). Together, these results demonstrate that At1g53500 is MUM4.

Figure 5. Alignment of the amino acid sequence of the MUM4 protein with those of RHM1 and RHM3. The putative 4,6-dehydratase and 4-reductase domains within MUM4 are marked with single and double underlines, respectively, whereas the highly conserved NAD(P)+ cofactor-binding (more ...)

MUM4 encodes a protein of 667 amino acids (Fig. 5). Comparison with GenBank sequences suggested that MUM4 contains two domains: an N-terminal domain with similarity to bacterial dTDP-d-Glc-4,6-dehydratases (44% identical, 61% similar to RfbB [COG1088] consensus) and a C-terminal domain with some similarity to bacterial 4-reductases (20% identical, 34% similar to RfbD [COG1091] consensus) (Tatusov et al., 2000; Figs. 4A and 5). Both domains contain the conserved N-terminal NAD⁺ binding (GxxGxxG) and active site catalytic couple (YxxxK) motifs found in members of the reductase/epimerase/dehydrogenase protein superfamily (Graninger et al., 1999; Allard et al., 2001). Both of the missense alleles, mum4-1 and mum4-2, change codons of conserved amino acids (Fig. 5) in the N-terminal, putative NDP-d-Glc 4,6-dehydratase domain.

A comparison with other Arabidopsis genes predicted from genome annotation revealed that MUM4 is a member of a small gene family of putative nucleotide sugar interconversion factors that was initially described in an in silico analysis by Reiter and Vanzin (2001). The gene family consists of three genes, RHM1 (At1g78570), RHM2 (MUM4) and RHM3 (At3g14790), which are 83% to 89% identical at the amino acid level (Fig. 5). This suggests that there are at least three genes with a similar function.

Figure 6. MUM4, RHM1, and RHM3 expression in Arabidopsis plants. A, Saturating, qualitative RT-PCR of RNA isolated from wild-type (Col-2) stem, root, leaf, seedling, inflorescence, and siliques using gene-specific primers for MUM4, RHM1, and RHM3. The loading control (more ...)

Because the only obvious phenotype in mum4 plants is the defect in the seed coat epidermis, we looked at the expression of MUM4 during seed coat differentiation (Fig. 6B). RNA was extracted from siliques at three stages of development: (a) before mucilage synthesis (4 DPA; see Fig. 1G), (b) midmucilage synthesis (7 DPA; Fig. 1H), and (c) after the completion of mucilage synthesis (10 DPA; Fig. 1I), and subjected to northern analysis. This experiment showed that although MUM4 is expressed throughout seed coat differentiation, transcript levels are highest at the time of mucilage synthesis (7 DPA; Fig. 6B). To test if the higher levels of MUM4 detected at 7 DPA may be largely restricted to the seed coat, ap2-7 mutants were used as a control in gel blots (Fig. 6B). ap2 was chosen because the outer two layers of the seed coat of ap2 mutant seeds fail to differentiate, whereas all other cell types of the seed (embryo, endosperm, and seed coat endothelium) appear normal (Western et al., 2001). The low level of MUM4 transcript seen in 7-DPA ap2-7 siliques suggests that a large proportion of the MUM4 transcript at 7 DPA in wild-type siliques may be specific to the outer layers of the seed coat.

Figure 7. Expression of MUM4 and genes encoding putative regulatory transcription factors during silique development. A, RNA gel-blot analysis using MUM4 probe on total RNA from 7-DPA wild-type, ap2-7, myb61-1 (Col-2), wild-type, ap2-1, gl2-1, ttg1-1, and ttg2-1 (more ...)

TTG1 has been shown to be required for maximal expression of both TTG2 and GL2 in leaves (Johnson et al., 2002; Schiefelbein, 2003). To determine if the same is true in seed coats, the expression of both GL2 and TTG2 were tested in ttg1-1 compared with wild-type Ler 7 DPA siliques using semiquantitative RT-PCR. The transcript levels for both GL2 and TTG2 were reduced in ttg1-1 mutants (Fig. 7C), suggesting that TTG1 also acts upstream of GL2 and TTG2 in the seed coat.

The development of the mucilage secretory cells of the Arabidopsis seed coat is a complex process involving the biosynthesis and secretion of a large quantity of pectinaceous mucilage, intracellular cytoplasmic rearrangement, and secondary cell wall production. We have shown that the putative NDP-l-Rha synthase encoded by MUM4 is required for wild-type levels of mucilage biosynthesis and columella formation. Expression of MUM4 is up-regulated during seed coat differentiation and is regulated by several transcription factors required for seed coat epidermal cell differentiation. These data validate the utility of the mucilage secretory cells as a model system for polysaccharide biosynthesis, establish a regulatory framework for seed coat epidermal differentiation, and provide a causal link between pectin biosynthesis and cell morphogenesis.

Gram-negative bacteria such as Escherichia coli encode three separate enzymes (4,6-dehydratase, 3,5-epimerase, and 4-reductase) to convert dTDP-d-Glc to dTDP-l-Rha (Tonetti et al., 1998). The putative MUM4 protein not only contains an N-terminal domain with similarity to the bacterial dTDP-d-Glc 4,6-dehydratases but also a C-terminal domain with some similarity to 4-reductases (this study; Reiter and Vanzin, 2001). There is a precedent in Arabidopsis for a bifunctional 3,5 epimerase, 4-reductase in the synthesis of GDP-l-Fuc from GDP-d-Man (GER1, GER2) (Bonin and Reiter, 2000). Therefore, based on sequence analysis, it is attractive to postulate a single, multifunctional protein acting in the conversion of NDP-d-Glc to NDP-l-Rha in Arabidopsis (Reiter and Vanzin, 2001).

MUM4 is a member of a small gene family consisting of three genes (RHM1, MUM4/RHM2, and RHM3; Reiter and Vanzin, 2001) that exhibit high identity at both nucleotide and amino acid levels. The ubiquitous expression of RHM1 and RHM3 can be postulated to provide a mechanism for the formation of the normal primary cell wall in mum4 mutants and the residual mucilage present in mum4 seed coats. Redundancy in genes coding for putative NDP-l-Rha synthases is unsurprising because of the importance of the Rha-containing pectins RGI and RGII in primary cell walls. In fact, the presence of multiple genes coding enzymes involved in nucleotide sugar interconversions is common in plants (Reiter and Vanzin, 2001). The roles of MUM4 and its paralogs may not be completely redundant, however, because the three genes do not have identical patterns/levels of expression throughout all plant tissues (Fig. 6A; Arabidopsis expressed sequence tag database at The Institute for Genomic Research [http://www.tigr.org]; T.L. Western and G.W. Haughn, unpublished data).

Our data indicate that MUM4 transcription is decreased in ttg1 and gl2 mutants but is essentially wild type in ttg2 and myb61 mutants (Fig. 7A). This result suggests a regulatory framework for mucilage secretory cell differentiation in which there are at least two pathways controlling mucilage biosynthesis in the seed coat. In one of these pathways, it appears that TTG1 and GL2 control the transcription of MUM4 (Fig. 8). In other epidermal systems, such as trichomes and root hairs, TTG1 has been shown to interact with a tissue-specific MYB protein through the bHLH protein GLABRA3 (Payne et al., 2000; Schiefelbein, 2003) to activate GL2. Therefore, we favor a model in which a seed coat-specific complex of TTG1-bHLH-MYB causes activation of GL2 and the subsequent up-regulation of MUM4 during mucilage production (Fig. 8). Recent results suggest that EGL3 (ENHANCER OF GLABRA3) and/or TT8 (TRANSPARENT TESTA8) are the bHLH proteins acting with TTG1 in the seed coat (Zhang et al., 2003). A possible MYB component of this trimeric complex, MYB23, is expressed in developing seeds and groups to the same subfamily of MYBs as the bHLH-TTG1 interacting proteins GLABROUS1 and WEREWOLF (Kirik et al., 2001).

Figure 8. Proposed genetic pathway for the regulation of mucilage production during seed coat secretory cell differentiation. AP2 and a TTG1 complex with a bHLH protein (candidates include EGL3 and/or TT8) and a tissue-specific MYB transcription factor activate (more ...)

Correct regulation of MUM4 transcription at the time of mucilage production does not require TTG2 or MYB61 (Fig. 7A). However, TTG1 is required to activate TTG2 in the seed coat (Fig. 7C) and in other tissues (Johnson et al., 2002). This result places TTG2 downstream from TTG1 in a second pathway that controls mucilage biosynthesis in the developing seed. This alternate pathway, along with another that involves MYB61, may regulate the expression of other proteins involved in RGI synthesis including an NDP-d-GlcUA 4-epimerase (Feingold, 1982; Reiter and Vanzin, 2001), a nucleotide sugar transporter or an RGI-backbone glycosyltransferase. Alternatively, MYB61 and TTG2 may be involved in the transport of newly synthesized RGI to the plasma membrane. Feedback inhibition of mucilage production in the absence of secretion has been demonstrated in the root caps of the maize (Zea mays) mutant Ageotropic (Millar and Moore, 1990).

AP2 encodes a putative transcription factor that is required for the differentiation of the outer two layers of the seed coat (Western et al., 2001). Not surprisingly, ap2 seeds lack mucilage and fail to activate MUM4 transcription beyond its baseline level of expression (Fig. 7A). Although AP2 is required for maximum GL2 and TTG2 transcript levels (Fig. 7B), transcription of TTG1 and MYB61 was independent of AP2 activity (Fig. 7B). Our data are consistent with the hypothesis that AP2 functions in parallel with TTG1 to activate GL2 and TTG2 in the seed coat epidermis (Fig. 8).

mum4-3 was isolated by screening T-DNA mutagenized lines (Feldmann and Marks, 1987; ABRC) by staining with a 0.01% (w/v) aqueous solution of Ruthenium red.

For immunofluorescence, developing seeds were placed in copper hats in a 0.2 m Suc solution and high-pressure frozen using a Balzers HPM 101 (Balzers Instruments, Balzers, Liechtenstein). The samples were freeze substituted over 5 d in 0.25% (v/v) glutaraldehyde/8% (v/v) dimethoxypropane in acetone using a dry ice/acetone bath at -80°C, then gradually warmed to -20°C, 4°C, and then 20°C (3 h at each temperature) before embedding in LR White resin. Sectioning and microscopy were performed as for bright field.

Immersion immunofluorescence was performed according to the procedure of Willats et al. (2001), except dry wild-type seeds were placed directly into the fixative. Alternatively, the seeds were imbibed for 24 h in buffer with gentle shaking before fixation, all subsequent steps also being performed with shaking. Antipectin monoclonal antibodies JIM5, JIM7, and polyclonal α-PGA/RGI at 1:5 and 1:10 (v/v) dilutions were used as primary antibodies. Pre-immune serum controls were rat (JIM5 and JIM7) and rabbit (α-PGA/RGI). Secondary antibodies were 1:100 (v/v) dilutions in 5% (w/v) nonfat dry milk (NFDM) of goat anti-rat IgG conjugated to fluorescein isothiocyanate (JIM5 and JIM7) and anti-rabbit IgG (α-PGA/RGI). Sections were mounted in either anti-fade agent (JIM5 and JIM7; Slowfade Antifade, Molecular Probes, Eugene, OR) or 1:2 (v/v) glycerol:water (α-PGA/RGI) before observation.

For immunofluorescence, 0.5-μm sections were incubated in 50 mm EGTA for 1 h, blocked with 5% (w/v) NFDM for 20 min, then incubated for 1 h with primary antibodies at 1:5 and 1:10 (v/v) dilutions in 5% (w/v) NFDM, and for 1 h in secondary antibodies diluted at 1:100 (v/v). Rinses were performed before and after each incubation with 1× Tris-buffered saline/0.1% (v/v) Tween 20. Primary antibodies, secondary antibodies, pre-immune controls, and mounting were done as described for immersion immunofluorescence.

For alditol acetates, samples were hydrolyzed in trifluoroacetic acid for 2 h at 100°C, dried under nitrogen gas, then reduced for 90 min at 40°C in 245 μL of 2.6 m ammonia and 700 μL of 2% (w/v) sodium borohydride in dimethylsulfoxide. After addition of 175 μL of acetic acid, the samples were acetylated for 10 min at room temperature with 175 μL of 1-methylimidazole and 2.8 mL of acetic anhydride. Water (5.6 mL) was added, and the alditol acetates were extracted with 1.05 mL of dichloromethane (DCM). The recovered phase was dried and resuspended in 200 μL of DCM, then re-extracted with 1 mL of water. This recovered DCM phase was used for analysis. Samples were run on a gas chromatograph (model 5890A, Hewlett-Packard, Mississauga, ON, Canada) on a HP-23 glass capillary column (30-m × 0.25-mm i.d.) with helium as the carrier gas. The temperature program was 2 min at 180°C, increase 10°C min^-1 up to 200°C, 5 min at 200°C, increase 10°C min^-1 to 250°C, and hold at 250°C for 10 min. Monosaccharides were identified through comparison with the retention times obtained with individual sugar standards and a composite standard consisting of Rib, Fuc, Man, Gal, Glc, Ara, Rha, Xyl, GlcUA, and GalUA.

To confirm that the gene was MUM4, mum4-1 and mum4-2 were sequenced. Sequencing of the Salk Institute sequence-indexed Arabidopsis T-DNA insertion lines 085051 and 038898 revealed two further alleles, named mum4-4 and mum4-5, respectively. Molecular complementation of MUM4 was carried out using a 5.2-kb SalI/BstZ17I fragment of T3F20, including At1g53500 plus 2.1 kb upstream and 0.4 kb downstream sequence, into pGREEN0229 (Hellens et al., 2000). mum4-1 plants were transformed either with the genomic clone (pMUM4g) or the empty vector (pGREEN0229) using the Agrobacterium tumefaciens dipping method (Clough and Bent, 1998). Basta-resistant transformants were isolated by spraying young seedlings with 0.1% (v/v) glufosinate (Final EV150, AgrEvo EH, Paris) in 0.1% (v/v) Silwet l-77, and putative transformants were verified by PCR analysis.

The 5′-UTR of MUM4 was located using RT-PCR with primers located up to 900 bp upstream from the predicted ATG in combination with an internal primer. RNA was isolated from developing siliques according to the protocol of Downing et al. (1992), with the addition of a sodium acetate wash to remove excess polysaccharides. One-microgram samples were treated with DNaseI (Invitrogen Life Technologies, Carlsbad, CA) and reverse transcribed with SuperScript II Reverse Transcriptase (Invitrogen) according to the manufacturer's instructions. RT-PCR using two sequential primers identified bands approximately 200 bp smaller than predicted from genomic sequence. The longer fragment was isolated and sequenced.

RNA isolation and RT of DNase-treated RNA was performed as described above. Gene-specific primers surrounding an intron were designed for each gene, with the exception of TTG1, which lacks an intron. Primers were as follows: MUM4, p1-ttcgtgtaaatgtcgcaggt/p2 5′-gagaagctcgtctagtacggtc-3′; RHM1, p1 5′-gagactatccgtgccaatgta-3′/p2 5′-taataactcgtccaacacagtc-3′; RHM3, p1 5′-ccgagtcaacgttgctgga-3′/p2 5′-ggtaagagttcatctagtatggtc-3′; GAPC, p1 5′-tcagactcgagaaagctgctac-3′/p2 5′-gatcaagtcgaccacacgg-3′; MYB61, p3 5′-tggggagacattcttgctg-3′/p4 5′-gatgggcttgtgtgtgtttg-3′; GL2, p1 5′-aacggtcactccaaggtcac-3′/p2 5′-agaaacccgcatgtcttgtc-3′; TTG2, p1 5′-gccatcttgtcctcttccac-3′/p2 5′-ccttgcgatactcctgcttc-3′; AP2, p1 5′-cagggaatcctactactccacaag-3′/p2 5′-atctgattgtgatgatgaggagag-3′; and TTG1, p1 5′-catcctccggtcacagaatc-3′/p2 5′-tttcggctctacatcgttcc-3′. Loading for all RT-PCRs was determined by standardizing against GAPC test PCRs (at 22-23 cycles) for each RT reaction. Saturating, qualitative RT-PCR for MUM4, RHM1, and

RHM3 was done with 30 cycles. Semiquantitative RT-PCR (for AP2, TTG1, TTG2, GL2, and MYB61) was performed a minimum of four times for each set of primers using two to three different cycle numbers (from 25-30 cycles) to confirm results.

We thank Ms. Yeen Ting Hwang (University of British Columbia, Vancouver, BC Canada) and Mr. Adrian Wladichuk (University of British Columbia, Vancouver, BC Canada) for technical assistance, Dr. Michael Bevan (John Innes Centre, Norwich, UK) for supplying myb61-1 seed, Dr. Mark Smith (University of British Columbia, Vancouver, BC Canada) and Ms. Michelle Fawcett (Carnegie Institution, Stanford University, CA) for help with chemical analysis of mucilage, Drs. Andrew Staehelin (University of Colorado, Boulder) and Paul Knox (University of Leeds, Leeds, UK) for gifts of antibodies, and the Salk Institute Genomic Analysis Laboratory (La Jolla, CA, USA) for providing the sequence-indexed Arabidopsis T-DNA insertion mutants. We also thank Drs. Markus Pauly (Max-Planck Institute for Molecular Plant Physiology, Golm, Germany) and Björn Usadel (Max-Planck Institute for Molecular Plant Physiology, Glom, Germany) for communicating unpublished data and delaying publication of their results on RHM2. We thank members of the Ljerka Haughn (University of British Columbia, Vancouver, BC Canada) and George Kunst (University of British Columbia, Vancouver, BC Canada) labs for comments on the manuscript.