The cellulose-deficient Arabidopsis mutant rsw3 is defective in a gene encoding a putative glucosidase II, an enzyme processing N-glycans during ER quality control


* For correspondence (fax +61 2 6125 4331; e-mail


rsw3 is a temperature-sensitive mutant of Arabidopsis thaliana showing radially swollen roots and a deficiency in cellulose. The rsw3 gene was identified by a map-based strategy, and shows high similarity to the catalytic α-subunits of glucosidase II from mouse, yeast and potato. These enzymes process N-linked glycans in the ER, so that they bind and then release chaperones as part of the quality control pathway, ensuring correct protein folding. Putative β-subunits for the glucosidase II holoenzyme identified in the Arabidopsis and rice genomes share characteristic motifs (including an HDEL ER-retention signal) with β-subunits in mammals and yeast. The genes encoding the putative α- and β-subunits are single copy and, like the rsw3 phenotype, widely expressed. rsw3 reduces cell number more strongly than cell size in stamen filaments and probably stems. Most features of the rsw3 phenotype are shared with other cellulose-deficient mutants, but some – notably, production of multiple rosettes and a lack of secreted seed mucilage – are not and may reflect glucosidase II affecting processes other than cellulose synthesis. The rsw3 root phenotype develops more slowly than the rsw1 and rsw2 phenotypes when seedlings are transferred to the restrictive temperature. This is consistent with rsw3 reducing glycoprotein delivery from the ER to the plasma membrane whereas rsw1 and rsw2 act more rapidly by affecting the properties of already delivered enzymes.


Cellulose is the major structural polysaccharide of higher plant cell walls. Chains of β-1,4-linked glucosyl residues assemble soon after synthesis to form rigid, chemically resistant microfibrils. Their mechanical properties together with their orientation in the wall influence the relative expansion of cells in different directions and determine many of the final mechanical properties of mature cells and organs. Mutations in 5 of the 10 genes encoding CesA glycosyltransferases in Arabidopsis thaliana either reduce cellulose production (Arioli et al., 1998; Fagard et al., 2000; Taylor et al., 1999, 2000b) or generate resistance to the cellulose synthesis inhibitor isoxaben (Desprez et al., 2002; Scheible et al., 2001). Mutations in the KOR membrane-bound endo-1,4-β-glucanase (Nicol et al., 1998; Zuo et al., 2000) also cause cellulose deficiency (Lane et al., 2001; Peng et al., 2000; Sato et al., 2001). These glycosyltransferases and the endo-1,4-β-glucanase probably directly participate in glucan synthesis with the endo-1,4-β-glucanase perhaps cleaving a lipid-linked intermediate (Peng et al., 2002) and one or more of the glycosyltransferases assembling that lipoglucan or further elongating the glucan after the endo-1,4-β-glucanase has cleaved off the lipid (Williamson et al., 2001b). Consistent with this, a CesA glycosyltransferase occurs in the rosette terminal complexes which synthesise cellulose at the plasma membrane (Kimura et al., 1999).

Further genetic data point to cellulose synthesis responding to defects in enzymes on the N-glycosylation/quality control pathway. These steps occur in the ER rather than at the plasma membrane, and so, probably act only indirectly on synthesis through the supply of key glycoproteins to the plasma membrane. N-glycosylation begins when the mannose-rich oligosaccharide Glc3Man9GlcNac2 is assembled on dolichol in the ER membrane and transferred to the Asn residue of a newly synthesised protein containing an Asn-X-Ser or Asn-X-Thr motif (where X is any amino acid except Pro). With further processing of the glycoprotein by glucosidases I and II, N-glycosylation intersects with the quality control pathway responsible for ensuring proper folding of newly synthesised proteins (Helenius and Aebi, 2001; Vitale, 2001). Glucosidase I removes the terminal α-1,2-linked glucosyl residue to generate Glc2Man9GlcNac2 and glucosidase II removes the next α-1,3-glucosyl residue. Polypeptides carrying the resultant GlcMan9GlcNac2 specifically bind chaperones (calnexin and calreticulin) and probably other proteins that promote proper folding of newly synthesised proteins. The glycoprotein releases the chaperones when glucosidase II trims off the final Glc residue which is required for chaperone binding. Glycoprotein glucosyltransferase then re-attaches one Glc residue to the Man9GlcNAc2 of improperly folded glycoproteins so that they again bind chaperones and have a further opportunity to fold properly. Properly folded proteins, however, cannot be re-glucosylated by that enzyme and progress through the secretory pathway for further processing and delivery.

Defects at several points in this pathway affect cellulose synthesis. Sequence analysis suggests that the potato MAL1 gene encodes a glucosidase II and antisense suppression reduces glucosidase II activity (Taylor et al., 2000a). MAL1 antisense plants accumulate less cellulose than controls when grown under field conditions, although there is no visible phenotype in glasshouse conditions. The embryo lethal knopf mutant is deficient in glucosidase I and severely deficient in cellulose (Gillmor et al., 2002). Finally, the embryo lethal cyt1 mutant is cellulose-deficient from a defect in mannose-1-phosphate guanylyltransferase, the enzyme generating the UDP-Man required to (amongst other things) assemble the high mannose oligosaccharide that is transferred from dolichol to the nascent protein (Lukowitz et al., 2001). The mutations that affect cellulose synthesis concentrate towards those early steps where the N-glycosylation pathway intersects with the quality control pathway. Quality control, rather than production of mature glycans on critical proteins, seems particularly important since there is no detectable phenotype from a defect in N-acetyl glucosaminyl transferase I that blocks the steps in the Golgi that build mature, N-linked glycans (von Schaewen et al., 1993).

In this paper, we further analyse the rsw3 mutant which shows root radial swelling (Baskin et al., 1992) and a selective reduction in cellulose production (Peng et al., 2000). We show by map-based cloning that the mutation affects a putative catalytic α-subunit of glucosidase II, identify a candidate to be the β-subunit responsible for ER-retention of the holoenzyme, provide further details of the rsw3 phenotype and discuss the linkage between events in the ER and cellulose synthesis at the plasma membrane.


RSW3 encodes the α-subunit of a glucosidase II

The rsw3 allele behaves as a single Mendelian recessive (Baskin et al., 1992) and was identified by a map-based strategy. The F2 progeny from crossing rsw3 with the visual marker line W9 linked RSW3 with yi on the lower arm of chromosome 5. Using the F2 from a cross to the Landsberg erecta ecotype, RSW3 was mapped at 6 cm from the LFY3 locus (4 out of 70 chromosomes showing a cross-over event), so positioning RSW3 between yi and LFY3. Analysis of a further 372 chromosomes identified one recombination event between MBK5/a and rsw3, a notional map distance of 0.27 cm. Several candidate genes in this region were sequenced in rsw3. One (At5g63840) on the P1 clone mgi19 (AB007646) encoded a putative catalytic subunit of glucosidase II and the rsw3 allele showed a T → C substitution predicted to replace Ser599 with Phe in the protein.

We confirmed in two ways that the detected mutation caused the rsw3 phenotype. First, homozygous rsw3 plants were transformed with a binary construct containing a fragment amplified (with one nucleotide change; see Experimental procedures) from the Arabidopsis glucosidase II genomic sequence. Kanamycin-resistant T1 progeny had wild-type roots when grown for 5 days at 21°C followed by 2 days at 30°C (Figure 1a). The inflorescence phenotype (see later) was also complemented. Second, rsw3 was crossed to the tagged mutant, SGT5691 (Parinov et al., 1999) which contains a Ds element in the first exon of the gene encoding the putative glucosidase II enzyme. It presumably represents a null allele and the mutation is homozygous lethal, so hemizygous plants which appear wild type, were used for crossing. The NPTII gene present on the Ds element confers kanamycin resistance to F1 plants receiving the tagged allele from SGT5691. Roots of all kanamycin-resistant F1 seedlings (containing a null allele and a temperature-sensitive allele) appeared wild type at 21°C but swelled at 30°C (Figure 1b). This confirms that the Ds and EMS mutants are allelic and that glucosidase II defects cause radial swelling.

Figure 1.

Mutations in the gene encoding glucosidase II cause radial swelling.

(a) Complementation of root radial swelling in rsw3 transformed with the 5.9-kB fragment amplified from the wild-type genome. Columbia wild type (left), rsw3 (centre) and a kanamycin-resistant T1 seedling of rsw3 transformed with a genomic copy of the glucosidase II gene (right). The wild-type gene suppresses radial swelling. All plants were transferred to 30°C for 2 days prior to photographing.

(b) The rsw3 mutation is allelic to the insertional mutant SGT5691 which contains a Ds element in the first exon of the glucosidase II gene. Columbia wild type (left), rsw3 (centre) and a heterozygous F1 plant from crossing SGT5691 with rsw3. The F1 heterozygote and the rsw3 homozygote show temperature-induced radial swelling. All plants were transferred to 30°C for 2 days prior to photographing.

Conserved features in putative α- and β-subunits

The RSW3 sequence is highly similar from about residue 150 onwards to sequences in the glucoside hydrolase family 31 (Henrissat, 1991; Henrissat and Bairoch, 1993). Monroe et al. identified the RSW3 glucosidase II gene through a search of Arabidopsis ESTs with homology to α-glucosidases and named it Aglu-3 (Monroe et al., 1999). Its protein product formed a clade with several glucosidase II enzymes whose catalytic activities were independently known. They all separated from apoplastic α-glucosidases of Arabidopsis with which Aglu-3/RSW3 shares only 8% sequence identity. Figure 2 shows the two signature motifs for the clade containing Aglu-3/RSW3, which are believed to include catalytic and substrate-binding residues. Aglu-3/RSW3 contains all of the conserved residues within these motifs, as well as the proposed catalytic residues Asp512 and Asp617 (Frandsen and Svensson, 1998). Ser599, which is mutated in rsw3, is likely to be functionally significant, since it is conserved in the homologous gene product from mouse (NP_032086), human (NP_055425), pig (AAB49757), slime mold (AAB18921), potato (T07391) and cotton (Burn et al., unpublished data), and in the more distantly related apoplastic α-glucosidases encoded by the Arabidopsis genes Aglu-1 and Aglu-2 (Monroe et al., 1999). The Arabidopsis Aglu-3/RSW3 gene appears to be a single copy, spans 3.84 kb with five introns and encodes a predicted transcript of 2766 bp, giving a predicted translation product of 104 kDa.

Figure 2.

Alignment of the Aglu-3/RSW3 sequence (Genbank NP_201189) with the sequences of ER-resident glucosidase II enzymes from potato (T07391), mouse (NP_032086) and fission yeast (CAB65603).

The signature motifs for clade 2 of Monroe et al. (1999) are shown to demonstrate the high conservation. They include several residues implicated in catalysis (Asp 512 and Asp 617; *). The site of the rsw3-1 mutation (Ser599; ●) is close to these consensus sequences and is conserved in these and other glucosidase II sequences. Predicted N-terminal signal sequences are boxed. No HDEL ER-retention sequences occur at the C-terminus.

Recent biochemical (Trombetta et al., 1996) and genetic studies (D'Alessio et al., 1999; Pelletier et al., 2000) suggest that native glucosidase II of mammals and yeast consists of a catalytic α-chain (to which Aglu-3/RSW3 is homologous) and a smaller non-catalytic β-chain which retains the heterodimer in the ER. To determine if Arabidopsis contained an orthologue of the β-subunit, a blast search of the NCBI database was carried out with the mouse β-subunit. Unknown protein At5g56360 (protein MCD7.9 on the P1 clone MCD7 (AB009049) from chromosome 5) had 27% amino acid identity and 42% similarity to the mouse β-subunit. A closely related sequence (genbank BAA88186) exists on chromosome 1 in rice but is annotated with a stop codon that truncates it after 496 residues. The conceptual translation of the adjacent 3′ sequence on the PAC clone P0038F12 (AP000836) and re-consideration of proposed splice sites indicate the potential to encode a full length β-subunit that is very similar to the Arabidopsis gene product. The proposed sequence of the gene product is supported by an EST (AU030896), matching the proposed exons. Figure 3 therefore includes our suggestion for the full length rice protein. The Arabidopsis, rice, mouse and Schizosaccharomyces pombe sequences share: HDEL ER-retention signals at their C-termini, predicted leader sequences at their N-termini, a cysteine-rich N-terminal region, a mannose-receptor homology region (MHR) (Munro, 2001) preceding the HDEL sequence at the C-terminus, a central region rich in acidic residues and flanked by regions giving high scores in programs (‘coils’ and ‘paircoil’) predicting the likelihood of sequences forming coiled coils (Berger et al., 1995; Lupas et al., 1991).

Figure 3.

Alignments of the proposed β-subunits of Arabidopsis (At5g56360) and rice (our amendment of BAA88186) with the β-subunits of glucosidase II from mouse (AAC53183) and fission yeast (BAA13906).

Note the predicted N-terminal signal sequences (boxed), C-terminal H/VDEL ER-retention signals and the mannose-receptor homology region (MHR) near the N-terminus. The six cysteines within the MHR (four only in yeast) are numbered and the R and Y residues implicated in substrate-binding (●) and the substrate recognition loop between cysteines 5 and 6 are marked. Elsewhere in the sequence, note the relatively high level of similarity in the N- and C-terminal domains and the much lower similarity and plant-specific inserts in the central region.

Munro (2001) links the MRH domain to carbohydrate recognition. It comprises a region of similarity to the cation-dependent mannose 6-phosphate receptor whose crystal structure is known. Critical conserved features (Figure 3) include the 6 Cys residues forming three disulphide bonds (although the S. pombe protein lacks cysteines 1 and 2), the substrate recognition loop between the cysteines 5 and 6 and the Y and R residues implicated in ligand binding (Roberts et al., 1998). Interaction between mouse α- and β-subunits was mapped to the N-terminal 118 residues of the β-subunit, which are reasonably well conserved in all sequences, and to residues 273–400 (Arendt and Ostergaard, 2000) which are not. Figure 3 shows, however, that all sequences show a high percentage of acidic residues here. RT-PCR-detected expression of the genes encoding the putative α- and β-subunits in all tested tissues of Arabidopsis (Figure 4), but under the conditions used, will not clearly indicate relative expression levels. The low numbers of ESTs in Arabidopsis (13 for the α-subunit, four for the β-subunit), suggest that neither gene is highly expressed. (For comparison, AtCesA1/RSW1, a glycosyltransferase implicated in cellulose synthesis, finds 40 ESTs in a similar search.)

Figure 4.

mRNA for both the α-subunit (a) and the β-subunit (b) occurs in all Arabidopsis tissues tested.

RT-PCR using mRNA from root (lane 1), whole rosette leaves (2), leaf blades (3), mature stem tissue (4), cauline leaves (5), flower buds (6), flowers (7), siliques (8), dark-grown hypocotyls (9). (The presence of the β-subunit in dark-grown hypocotyls was demonstrated in another experiment).

Growth and morphology of rsw3

rsw3 grows like wild type at its permissive temperature of 21°C and the seedling root swells when transferred to 30°C. The bulging cells on the root (Baskin et al., 1992) are often at the base of root hairs suggesting a role for RSW3 in the early stages of root hair development. The swollen primary root only resumes elongation if returned to the permissive temperature within 48 h (not shown), but the root continues to generate laterals (Figure 5a). The laterals – whose primordia were not visible when the transfer to 31°C was made – elongate for several millimetres before they, in turn, swell and stop growing. The root system of mature vegetative plants is consequently short and very highly branched (Figure 5b). The doubly cellulose-defective mutant rsw1–1rsw3 showed only a slightly swollen root tip after 24 h at the restrictive temperature, but since any longer period at the high temperature led to death, swelling was probably already curtailed after 24 h at the restrictive temperature.

Figure 5.

Morphology of rsw3.

(a) Root system of a seedling showing that lateral roots extend some distance before swelling and stopping elongation. Plants grown 5 days at 21°C and 6 days at 30°C. Scale bar = 2 mm.

(b) Continued root growth gives a dense, highly branched root system and a dense mass of very small leaves on plant grown for 21 days at 30°C. Scale bar = 5 mm.

(c) Hypocotyls grown in the dark for 3 days at 21°C and 3 days at 30°C. From the left: wild-type, rsw1-1, rsw2-1, rsw3-1, rsw1-1rsw2-1, rsw1-1rsw3-1. The rsw3 effect on the hypocotyl is weak compared to that of the other single mutants and rsw1-1rsw3 is weaker that rsw1-1rsw2-1. Scale bar = 5 mm.

(d) Light micrograph of rsw3 grown on agar for 35 days at 30°C. Tiny inflorescences with flower buds of near normal size (top right and bottom left) emerge from several of the rosettes. Scale bar = 5 mm.

(e) Scanning electron micrograph of rsw3 plant grown for 21 days at 30°C and showing the presence of multiple rosettes. Scale bar = 1 mm.

(f) Detail of the ringed area in (e) showing the very complex arrangement of the minute leaves, many of which carry trichomes of approximately normal size and morphology. Scale bar = 200 μm.

(g) Scanning electron micrograph of the surface of a wild-type leaf on a plant grown for 10 days at 30°C. Note the clearly defined cell boundaries, stomata and trichomes.

(h) The surface of a rsw3 leaf showing much less clear outlines to the pavement cells, an apparently collapsed trichome (CT) on top of its ring of subsidiary cells and many stomata with their guard cells protruding above the leaf surface. Scale bar for (g) and (h) = 100 μm.

The phenotype in dark-grown hypocotyls is much weaker in rsw3 than in rsw1-1 and rsw2-1 and the phenotype in rsw1-1rsw3 is weaker than rsw1-1rsw2-1 (Figure 5c). Rosette growth of rsw3 in the light is strongly suppressed and many minute leaves are packed in a dense mat in which regular phyllotaxis cannot be recognised (Figure 5d–f). The complex pavement cell shape in wild-type leaves (Figure 5g) is simplified in rsw3, stomata protrude from the leaf surface and some trichomes appear to burst (Figure 5h). Several of the crowded rosettes initiated minute inflorescences (Figure 5d) although these appear much later than wild-type inflorescences (28.6 ± 0.5 days vs. 15.5 ± 0.17 days for agar-grown plants; mean ± SE, n = 98 for rsw3, n = 45 for wild type). The few flowers on the minute rsw3 inflorescences were essentially full-sized although anther filaments, gynoecium and sepals were slightly shortened and buds opened prematurely before the stigma was receptive (similar to the buds from soil-grown rsw3 plants shown in Figure 6(e,f) which are discussed below).

Figure 6.

Growth of the stem and reproductive development in rsw3.

(a, b) Kinetincs of secondary stem elongation in Columbia wild type, rsw3, rsw1 and rsw1rsw3 double mutant at 21°C (a) and 30°C (b). All plants were grown at 21°C until stems began to emerge. These were cut off and re-growth of secondary bolts followed at the indicated temperature. Single mutants show very little difference from wild type at 21°C, although the double mutant elongates more slowly and reaches a significantly shorter final height. The final heights reached at 30°C differ widely as do the trajectories by which they are reached. rsw1 elongates more slowly but elongation continues for at least as long as it does in wild type. rsw3 elongates almost as rapidly as wild type for 4 days but then ceases elongation by about day 6. The rsw1rsw3 double mutant elongates less rapidly and ceases elongation at about day 5.

(c, d). Light micrographs showing well-spaced flowers in wild type (c) and the clustered flowers on rsw3(d) with its early cessation of elongation.

(e, f). Cryo-scanning electron micrographs showing flower buds of wild type (e) and rsw3(f) that are of similar sizes but open prematurely in rsw3. Note the immature state of the stigma (St) and the irregular shapes of the cells on the sepals (Se) in rsw3. Bar for (e) and (f) = 200 μm.

(g, h) Cryo-scanning electron micrographs showing imbided seed of rsw3 that developed on plants held at 21°C (g) and 30°C (h). The 30°C seed is shrunken and lacks the clear cellular pattern of the 21°C seed.

(i–n) Light micrographs of imbided seed stained with ruthenium red to show a surface coat of mucilage. Wild type (i, j), rsw1 (k, l), rsw3 (m, n). Seed in i, k, m developed on plants at 21°C, seed in j, l, n developed on plants at 30°C. Mucilage is secreted normally by rsw1(l) and wild type (j) at 30°C but not by rsw3(n).

To investigate the direct effects of the mutation on stem growth, we grew wild type and rsw3 at 21°C on soil so that subsequent inflorescence development would not be limited by a small rosette supplying little photosynthate. Rosettes of rsw3 were very similar to wild type under these conditions and reproductive growth began at the normal time. Primary bolts were cut off and re-growth of secondary bolts followed at either 21 or 30°C (Figure 6a,b). Re-growth followed a slightly S-shaped curve with rsw3 and rsw1-1 at 21°C showing statistically insignificant reductions in growth rate and final height relative to wild type. rsw1-1rsw3 showed a clear reduction in rate and final height. At 30°C, however, the rsw3 growth rate was similar to wild type for a few days but elongation stopped by about day 5, whereas it continued in wild type until day 16 and even longer in rsw1-1 (Figure 6b). rsw1-1rsw2 (Lane et al., 2001) failed to regenerate secondary bolts at 30°C and rsw1-1rsw3 only grew to about 35 mm (Figure 6b) and produced few flowers and no seed.

We measured daily stem growth increments and the lengths of epidermal cells which had left the elongation zone when the bolts were about half grown (Table 1). This allowed estimation of cell flux (the number of cells leaving the elongation zone per day) at that time since daily growth increment = cell length × cell flux. There was no significant reduction in either cell flux or cell length of rsw3 growing at 21°C. The rsw1-1rsw3 constitutive phenotype at 21°C was entirely due to a reduction in cell length. At 30°C, rsw1-1 showed a 57% reduction in cell length and a 35% reduction in cell flux relative to wild type. Analyses of this type require that the plant is in a near steady state with respect to growth rate, length of the elongation zone, etc. Conditions, however, are far from steady state when elongation is rapidly slowing in rsw3 and rsw1-1rsw3 so that accurate deductions of cell flux for those genotypes are precluded. To get at least an idea of how cell length was behaving when growth was slowing, we measured cell lengths at a height of about 80 mm on the rsw3 stem. (Figure 6b shows that when these cells left the elongation zone, the stem would have been near the end of its growth phase since total plant height at that time would have exceeded 80 mm by the length of the growth zone at that time; 40 mm in wild type according to Fukaki et al., 1996). The cells in rsw3 were, even then, only slightly shorter than wild type (Table 1), suggesting that falling cell production rates were probably more important than reduced cell expansion in slowing stem elongation. In contrast, when we sampled the rsw1-1rsw3 stem at 30 mm for cells maturing when its elongation was slowing (Figure  6b), cell length was reduced by 57% (Table 1). This is consistent with the presence of rsw1-1 in the double mutant, tilting the balance strongly towards reduced cell length.

Table 1.  Analysis of the rate of stem elongation in terms of cell length and, where near steady growth rates occurred, cell flux (number of cells per day leaving the elongation zone)
 Growth rate (mm day−1)Cell flux (per day)Cell length (µm)
  1. Results are given as mean ± SE for n = 5. Statistically significant differences from wild type using the Student's t-test are indicated (*P < 0.05; **P < 0.01; ***P < 0.001).

 Columbia38.7 ± 1.0101 ± 3.5384 ± 4.0
 rsw338.4 ± 1.495.9 ± 4.6402 ± 7.0
 rsw138.9 ± 1.6102 ± 6.9382 ± 9.8
 rsw1rsw330.2 ± 1.9**100 ± 7.6299 ± 8.4**
 Columbia53.8 ± 1.2133 ± 2.7404 ± 3.2
 rsw341.8 ± 3.1**378 ± 22
 rsw115.2 ± 1.4***87.2 ± 7.0**174 ± 5.8***
 rsw1rsw313.6 ± 1.8***173 ± 15***

We checked these conclusions regarding cell division and expansion in a simpler system by using cryo-scanning electron microscopy to examine stamen filaments in flowers showing receptive stigmas (Table 2). The results were similar: rsw3 plants again showed a greater percentage reduction in cell number than in cell length and the double mutant rsw1-1rsw3 showed a further reduction in cell length without an additional reduction in cell number. rsw1-1 showed a much greater reduction in cell length than in cell number (Table 2), a conclusion already reached from a similar experiment reported by Burn et al. (2002).

Table 2.  Cell length and number in mature stamen filaments grown at 30°C
 Total length (µm)Cell numberCell length (µm)
  1. Results are given as mean ± SE for n ≥ 7. Statistically significant differences from wild type using the Student's t-test are indicated (*P < 0.05; **P < 0.01; ***P < 0.001).

Columbia2407 ± 3817.0 ± 1.0152.7 ± 6.2
rsw31458 ± 52***11.4 ± 0.3***127.0 ± 0.1**
rsw1-11050 ± 57***15.0 ± 0.472.7 ± 9.8***
rsw1-1rsw3415 ± 41***12.4 ± 0.5**29.4 ± 2.1***

Flowering and fertility in rsw3

Stems of both wild type and rsw3 regenerating at 30°C reached approximately the same height before initiating their first flower, even though their final heights would be very different (Figure 6b). Wild-type stems generated about 27 well-spaced flowers before elongation ceased, but rsw3 produced only about six closely spaced flowers before elongation ended, leaving a cluster of flowers (Figure 6c,d). rsw3 flower buds opened precociously before the stigma was receptive (Figure 6e,f).

Few flowers and no seed formed on the minute bolts of rsw3 plants grown continuously at their restrictive temperature (Figure 5d). Even flowers on the much larger bolts formed at 31°C on plants which had completed vegetative growth at 21°C (Figure 6d,f) also set very little seed. That seed (Figure 6g,h) was shrunken (probably because of reduced accumulation of seed storage proteins; Boisson et al., 2001), its surface lacked the regular cellular structure of wild type grown at 30°C or of rsw3 grown at 21°C and it showed very little secreted mucilage after imbibition (Figure 6i–n). Reduced mucilage secretion was not typical of cellulose-deficient mutants: rsw1-1 (defective in the CesA1 glycosyltransferase; Figure 6k,l) and rsw2-1 (defective in the KOR endo-1,4-β-glucanase; data not shown) had normal mucilage coats.

To isolate effects on the haploid stages of pollen and ovule development from effects on the diploid stages, we examined seed set in the hemizygous Ds mutant SGT5691 (a presumed null allele in the glucosidase II catalytic subunit). Seed set by self-fertilisation segregates 147 kanamycin-resistant individuals to 153 sensitive individuals. A ratio less than the 2 : 1 expected for a dominant, homozygous, lethal allele shows that the null allele affects post-meiotic development of pollen and/or ovules. We separated the effects on the male and female pathways by reciprocal crosses between the hemizygous-tagged mutant and Landsberg erecta (the appropriate wild type for this mutant). Kanamycin-resistant and sensitive plants will segregate 1 : 1 if pollen or ovule development is unaffected with lower ratios if the null allele reduces pollen or ovule fertility. Pollen from the Ds-tagged mutant gave a segregation ratio of 1 : 16 (six resistant:94 sensitive individuals) indicating a 94% reduction (relative to wild type) in the ability of Ds-tagged pollen to set viable seed. This compared with a 41% reduction when Ds-tagged ovules were crossed to wild-type pollen (ratio of 1 : 1.7, 37 : 63 individuals). The null allele of glucosidase II, therefore affects the haploid stages of pollen development much more severely than it affects post-meiotic ovules.


rsw3 is mutated in the catalytic subunit of glucosidase II

Three lines of evidence indicate that a mutation in the only Arabidopsis gene encoding a putative glucosidase II causes the rsw3 phenotype. First, there is a single amino acid change at a highly conserved site in the glucosidase II of rsw3; second, a wild-type genomic copy of the glucosidase II gene complements the rsw3 phenotype; third, the F1 from crossing rsw3 and SGT5691 (carrying a presumed null allele of glucosidase II; Parinov et al., 1999) has a temperature-sensitive radial swelling phenotype.

There is strong but indirect evidence that RSW3 is the catalytic subunit of glucosidase II. The Arabidopsis gene product falls in a clade with glucosidase II from potato, mammals and yeast which is clearly distinguished from the clade containing plant apoplastic α-glucosidases (Monroe et al., 1999). Other proteins in the clade are experimentally linked to glucosidase II activity by enzyme purification in rats (Trombetta et al., 1996), by gene deletion in yeast (D'Alessio et al., 1999) and by antisense suppression in potato (Taylor et al., 2000a).

Native plant glucosidase II may be a heterodimer

There are conflicting reports regarding the structure of native glucosidase II. The enzyme purified from plant microsomes showed a single band of 110 kDa (mung bean) or 95 kDa (soybean) on SDS-PAGE (Kaushal et al., 1990). These findings paralleled earlier evidence for a single chain structure of the mammalian enzyme (Brada and Dubach, 1984; Hino and Rothman, 1985), but in 1996, Trombetta et al. purified rat liver glucosidase II as a heterodimer comprising a catalytic α-chain of 104 kDa and a β-chain of 58 kDa carrying an HDEL ER retention signal (Trombetta et al., 1996, 2001). Two in vivo studies support the involvement of the β-subunit. First, overexpressing both subunits in mammalian cells increased glucosidase II activity whereas overexpressing only the α-subunit did not (Treml et al., 2000). Second, gene disruption showed that S. pombe required both subunits for in vivo glucosidase II activity (D'Alessio et al., 1999). The reasons why some purifications yield single chains and others heterodimers remain unexplained, although Trombetta et al. (2001) speculated that proteolysis removes the β-subunit unless purification is rapid. No biochemical or experimental genetic evidence currently indicates that plant glucosidase II is a heterodimer, but the highly plausible β-subunits encoded by the Arabidopsis and rice genomes makes this probable.

Selective inhibition of cellulose synthesis in rsw3

Roots of 7-day-old seedlings of rsw3 grown at 31°C contain only 51% of the wild-type cellulose (expressed mg−1 tissue dry weight), a comparable figure to that resulting from single amino acid substitutions in the CesA1 glycosyltransferase (rsw1-1) and the KOR endo-1,4-β-glucanase (rsw2-1) (Peng et al., 2000). The morphological changes indicate that all three genes are needed to make cellulose in primary cell walls but the question of whether they also make some contribution to secondary walls remains open.

Production of Golgi-derived non-cellulosic polysaccharides changes little in rsw3 seedlings (Peng et al., 2000). The selectivity for cellulose production is comparable to that seen with a defect in glucosidase I (Gillmor et al., 2002), the enzyme generating the initial substrate for glucosidase II processing. It exceeds the selectivity seen in the embryo-lethal cyt1 mutants of Arabidopsis (defective in mannose-1-phosphate guanylyltransferase) (Lukowitz et al., 2001) and in potatoes with MAL1 (encoding a glucosidase II α-subunit) downregulated by antisense (Taylor et al., 2000a) where complex changes occur in non-cellulosic polysaccharides and lignin. The differences between the study of Taylor et al. and this study could relate to species (Arabidopsis vs. potato), development (7-day seedlings vs. mature plants), environmental (controlled environment with agar plates vs. field) and genetic (single residue mutation vs. antisense). We therefore conclude that cellulose synthesis is often much more sensitive to N-glycan processing defects than is the synthesis of non-cellulosic polysaccharides in the Golgi, but that this may not hold in all situations.

Secretion of Golgi-derived seed mucilage is strongly reduced in rsw3 but not in rsw1-1 or rsw2-1. Mucilage could be produced but retained intracellularly (perhaps because of structural changes resulting from cellulose deficiency), or mucilage production itself could be reduced. Many developmental blocks reduce mucilage production (Western et al., 2000, 2001), but we cannot yet exclude the possibility that rsw3 has defective processing of Golgi enzymes required to make the particular non-cellulosic polysaccharides making up the mucilage.

rsw3 preferentially inhibits cell division rather than cell expansion

Cell numbers and sizes in stamen filaments indicate that rsw3 affects cell division more strongly than cell expansion. The cell length data for the stem are consistent with this, but rigorous interpretation using our limited measurements is impossible without near-steady state conditions. A strong effect of rsw3 on cell division may explain why its phenotype is rather weak in dark-grown hypocotyls which lack cell division (Gendreau et al., 1997). In more strongly affecting cell division than cell expansion, rsw3 resembles rsw2-1 (Burn et al., 2002) rather than rsw1-1 (Burn et al., 2002 and this report) or plants carrying antisense constructs to RSW1/CesA1 or CesA3 (Burn et al., 2002) which are more severely affected in cell length. (Although CesA1 changes have little impact on division rates, CesA1 is probably expressed in dividing root cells since they show changes in wall ultrastructure (Sugimoto et al., 2001) and swell (Baskin et al., 1992; Beemster and Baskin, 1998) when rsw1-1 is at its restrictive temperature.)

Mechanism by which rsw3 affects cellulose synthesis

Our work does not directly clarify how glucosidase II activity affects cellulose synthesis. However, as noted in relation to a glucosidase I mutation (Boisson et al., 2001), the minimal phenotype shown by a mutant which cannot assemble mature N-linked glycans in the Golgi (von Schaewen et al., 1993) indicates that a lack of mature N-linked glycans on critical proteins will not cause the strong phenotype seen with a glucosidase II defect. Reduced rates of production of Glc1Man9GlcNAc2 and Man9GlcNAc2 (the two products of glucosidase II) would probably slow both the formation and dissociation of the glycoprotein/chaperone complex, creating a bottleneck that may in time reduce the steady state levels of glycoproteins at sites further along the secretory pathway. Because glycoproteins participate in many plant processes, it is not obvious why cellulose synthesis should be much more sensitive to processing defects in the ER than, for example, synthesis of non-cellulosic polysaccharides. One previously suggested basis for glucosidase II effects on the wall can be discounted. Taylor et al. (2000a) suggested an RGD motif in potato glucosidase II (residues 569–571) might mediate wall–membrane interactions. The Arabidopsis protein lacks this motif.

Gillmor et al. (2002) argued that CesA proteins are not glycosylated when they did not detect a mobility shift on SDS-PAGE in knopf (deficient in glucosidase I) or after N-glycosidase F treatment and when they did not see in knopf a change in CesA abundance that was visible by unquantified immunostaining. The KOR endo-1,4-β-glucanase is a better candidate. A soluble fragment of the Brassica napus orthologue of KOR is heavily N-glycosylated when expressed heterologously in Pichia pastoris and the N-glycan is required for in vitro activity (Mølhøj et al., 2001). Further evidence consistent with KOR being a target can be drawn from the rsw3 and rsw2-1 phenotypes affecting cell division more than cell expansion whereas the rsw1-1 phenotype shows the reverse. Attractive as the KOR hypothesis might be, direct evidence is currently lacking that KOR is critical to the rsw3 phenotype.

Phenotype development at the restrictive temperature

The rsw1-1 and rsw2-1 mutations affect genes encoding plasma membrane enzymes that are probably directly involved in cellulose synthesis so that changed enzyme performance at the restrictive temperature will rapidly have an impact on cellulose synthesis. rsw3, in contrast, encodes a processing enzyme in the ER whose changed performance will, we hypothesise, reduce cellulose synthesis only when it restricts the supply of properly folded glycoproteins to the site of cellulose synthesis. The different time courses for the onset of a visible phenotype when the three mutants are transferred to the higher temperature plausibly reflect these different modes of action. Radial swelling starts slowly in rsw3 (latency >24 h compared to <12 h in rsw1-1 and rsw2-1) and the high temperature actually accelerates root elongation during the first 12 h albeit less than in wild type (Baskin et al., 1992). Elongation of rsw1-1 or rsw2-1, in contrast, falls during the first 12 h, roots swell strongly and rsw1-1 shows changed wall ultrastructure within 4 h (Sugimoto et al., 2001).

In conclusion, we have shown that rsw3 is mutated in a gene encoding a putative glucosidase II α-subunit, identified a putative β-subunit encoded by two plant genomes and shown that many aspects of the rsw3 phenotype flow from reduced cellulose synthesis in primary walls. Cell division seems more strongly affected than cell expansion making the KOR endo-1,4-β-glucanase, where mutations also strongly affect cell division, a plausible candidate to be the glycoprotein affected by the processing defect. In addition to its interest for studies of cellulose synthesis, a temperature-sensitive allele of glucosidase II should help experimental studies of N-glycosylation and quality control in the ER and in establishing its links to other developmental and physiological processes.

Experimental procedures

Plant material and growth conditions

The rsw3 mutant of Arabidopsis (Baskin et al., 1992) has been backcrossed to Columbia six times without altering the visible phenotype. Plants were grown either in pots containing a 1 : 1 : 1 (v/v) mix of peat:compost:sand, or aseptically on solid media in Petri dishes. For routine growth, seed was germinated on MS medium (Murashige and Skoog, 1962) containing 0.75% (w/v) agar and, when required, kanamycin (50 µg ml−1). For analysis of root and hypocotyl growth, seed was germinated on Hoagland's medium containing 1.2% agar and plates held vertically (Baskin et al., 1992). Plants were grown in controlled environment cabinets under 16-h photoperiods (100 µE m−2 s−1) at 21°C for permissive conditions and at 30°C for restrictive conditions. Petri dishes containing seeds were wrapped in aluminium foil and placed vertically in another light-proof container inside the growth cabinet to provide dark-grown hypocotyls. To create a double mutant, rsw3 was crossed with rsw1-1 (Arioli et al., 1998) and F1 plants allowed to self. F2 seedlings chosen for a more severe root swelling than either single mutant (grown for 5 days at 21°C and 2 days at 30°C) were backcrossed to the parental mutants to confirm homozygosity. Attempts to create rsw2-1rsw3 are still in progress.

Mapping the rsw3 mutation

An F2 population from crossing rsw3 (Columbia background) with the Landsberg erecta ecotype was screened to give plants showing a root-swelling phenotype. DNA was prepared from two to three rosette leaves per plant using the FastDNA kit (BIO 101, Carlsbad, CA) and mapping carried out using LFY3 (forward primer 5′-GACGGCGTCTAGAAGATTC-3′, reverse 5′-TAACTTATCGGGCTTCTGC-3′, cleavage with RsaI) and MBK5/a (forward 5′-CCCTCGCTTGGTACAAGGTAT-3′ and reverse 5′-TCCTGATCCTCTCACCACGTA-3′). All parts of the allele encoding the catalytic subunit of glucosidase II in the rsw3 mutant were sequenced at least twice using the products of separate amplifications.

Constructs and plant transformation

A genomic copy of the glucosidase II α-subunit including 830 bp of the promoter region was generated by PCR amplification of BAC F20A11 using the forward primer 5′-CCGCTCGAGCGGTTTCACTCACAACTGTGGTCTCT-3′ and the reverse primer 5′-CCGCTCGAGCGGTCTCCTAAGTCCTAACCCCATA-3′. Both primers included a XhoI site (underlined) which allowed the amplified 5.8-kb fragment to be ligated into the SalI site in the binary vector pBin19. The amplified product showed a single base pair change (C to T) from the genomic sequence. This substituted Leu for Ser142, a residue that is conserved in potato but not in other species (Figure 2) and did not impair the ability of the fragment to complement rsw3. The construct was introduced into Agrobacterium tumefaciens strain AGL1 and used to transform the rsw3 mutant by floral dipping (Clough and Bent, 1998). Kanamycin-resistant transformants were selected at 21°C on Hoagland's plates containing kanamycin (50 µg ml−1) and timentin (100 µg ml−1). Healthy seedlings were transferred to vertical Hoagland's plates and placed at 30°C for 2 days to screen for root swelling.


Tissues analysed were roots (including laterals) from 10-day-old seedlings, whole rosette leaves, leaf blades (midrib removed), mature stem tissue (the basal 100 mm of the stems of 5-week-old plants), whole cauline leaves, unopened floral buds (including the stem meristem), open flowers, green siliques and 6-day-old, dark-grown hypocotyls.

RNA (Parcy et al., 1994) was treated with RQ1 RNase-free DNase (Promega, Madison, WI) following the manufacturer's instructions. PCR primers were designed to the 3′ end of the coding region of the α- and β-subunits of Arabidopsis glucosidase II (α-forward 5′-CGTAGTGGTCTACTGGTTCAA-3′, α-reverse 5′-TGAGCTGTGTCCCAAGAGGAT-3′; β-forward 5′-GGTGATGAGGATACCAGCGAT-3′, β-reverse 5′-CCCACTCCCTAACCGGAGTTT-3′). Each spanned an intron, so differentiating RT-PCR products from genomic DNA and mRNA (724 bp vs. 452 bp for the α-subunit, 996 bp vs. 474 bp for the β-subunit). RT-PCR was carried out using the Gibco BRL Superscript one-step RT-PCR kit, following the manufacturer's instructions and an RT-PCR cycle of: 45 min at 48°C; 2 min at 94°C (30 sec at 94°C; 1 min at 54°C; 2 min at 68°C) × 45,7 min at 45–72°C.

Other methods

Measurements of stem growth rate and epidermal cell length (Burn et al., 2002) were used to estimate cell flux (Silk et al., 1989), the number of cells exiting the elongation zone during 1 day when the stem was about half grown. The numbers and lengths of epidermal cells in stamen filaments of mature flowers with receptive stigmas were analysed by cryo-scanning electron microscopy (Williamson et al., 2001a). Plants on agar were photographed using a Wild Photomakroskop 400. Mucilage secreted from seeds imbibed at room temperature for 2 h was stained with 0.01% (w/v) ruthenium red and photographed in colour with the Wild Photomakroskop.


We thank the Cotton Research and Development Corporation for project support, Andreas Betzner for helpful suggestions regarding pollen and ovule analysis and Roger Heady for invaluable assistance with scanning electron microscopy.