Cellulose biosynthesis and deposition in higher plants


Author for correspondence: Neil G. Taylor
Tel: +44 1904 328756
Fax: +44 1904 328762
Email: ngt2@york.ac.uk



II.Structure of cellulose2
III.Cellulose synthase2
IV.Mutations affecting cellulose synthesis4
V.Cellulose synthase complex assembly6
VI.Transcriptomic approaches to identify genes involved in cellulose synthesis7
VII.Purification of the cellulose synthase complex8
VIII.Regulation of cellulose synthesis8
IX.Feedback responses to cell wall defects9
X.Regulation of orientation of cellulose deposition10
XI.Cellulose as a source of renewable energy10


The plant cell wall is central to plant development. Cellulose is a major component of plant cell walls, and is the world's most abundant biopolymer. Cellulose contains apparently simple linear chains of glucose residues, but these chains aggregate to form immensely strong microfibrils. It is the physical properties of these microfibrils that, when laid down in an organized manner, are responsible for both oriented cell elongation during plant growth and the strength required to maintain an upright growth habit. Despite the importance of cellulose, only recently have we started to unravel details of its synthesis. Mutational analysis has allowed us to identify some of the proteins involved in its synthesis at the plasma membrane, and to define a set of cellulose synthase enzymes essential for cellulose synthesis. These proteins are organized into a very large plasma membrane-localized protein complex. The way in which this protein complex is regulated and directed is central in depositing cellulose microfibrils in the wall in the correct orientation, which is essential for directional cell growth. Recent developments have given us clues as to how cellulose synthesis and deposition is regulated, an understanding of which is essential if we are to manipulate cell wall composition.

I. Introduction

Cellulose is a central component in plant cell walls. In the primary cell wall (deposited in cells that are still expanding), it is a vital component of the load-bearing network and, because of its physical properties, is important in determining the orientation of cell expansion. After a period of expansion, some cells lay down a thick secondary cell wall inside the primary wall. The secondary cell wall provides plants with the mechanical properties that allow them to stand upright, and is a major component in properly functioning xylem vessels. Cellulose is highly abundant in the secondary cell wall. The importance of cellulose in plant cell walls is reflected in its being the world's most abundant biopolymer.

Given the importance of cellulose, it may seem surprising that our understanding of its synthesis is not more complete. Cellulose synthesis and deposition has been under investigation for several decades, but there have been many technical problems that have hindered progress. In the past decade, however, significant progress has been made, in particular using the model plant Arabidopsis thaliana, as described in several recent reviews (Doblin et al., 2002; Somerville, 2006; Joshi & Mansfield, 2007). It has been found that components necessary for cellulose synthesis are conserved across plant species, potentially allowing information from Arabidopsis to be transferred to more commercially important species. Because of the progress made using Arabidopsis and the apparent conservation of components between species, this review concentrates on cellulose synthesis and deposition in this species. Given current interest in developing the use of plant cell walls as a bioenergy source to combat global climate change, a significant research effort has been directed towards understanding cellulose synthesis. While the synthesis of cellulose is clearly very important, it is only with correct and organized deposition that the physical properties necessary for its role in plant cell walls can be utilized. Thus this review covers not only recent developments in our understanding of the synthesis of cellulose, but also our current knowledge of how the synthesis and orientation of deposition of cellulose is controlled.

II. Structure of cellulose

Cellulose is a simple polymer of unbranched β-1,4-linked glucan chains (Fig. 1). Successive glucose residues are inverted 180°, forming a flat ribbon in which the repeating unit is cellobiose. These parallel chains are then able to form extensive hydrogen bonds between individual cellulose chains. This results in crystallization of multiple cellulose chains into microfibrils-insoluble, cable-like structures conferring the physical properties required for their role in the cell wall. Microfibrils are thought to be composed of c. 36 glucan chains, and this is reflected in the structure of the machinery responsible for their synthesis. The molecular weight of individual glucan chains has been difficult to measure, possibly because of degradation of the chains during extraction. It has been suggested that in the secondary cell wall in cotton, the degree of polymerization (DP) is c. 14 000–15 000, as reviewed by Brett (2000). A number of fractions of cellulose with a lower molecular weight were also found in this study. This may be a result of surface chains having a lower DP than those within the core of the microfibril. Primary cell wall cellulose may have a lower molecular weight, with a measured DP of c. 8000 as reviewed by Brown (2004). The reason for this apparent difference in DP between primary and secondary cell wall cellulose chains is unknown, but may be the result of differences in the higher-order aggregation of microfibrils.

Figure 1.

A fragment of β-1,4-glucan showing inversion of adjacent sugar residues. The repeating unit cellobiose is marked.

III. Cellulose synthase

Freeze fracture of plant plasma membranes reveals the presence of hexameric ‘rosette’ structures approx. 25–30 nm in diameter. These protein complexes have been shown to contain cellulose synthase by use of a cellulose synthase-specific antibody (Kimura et al., 1999). It has been suggested, based on measurements of the dimensions of microfibrils, that each of the six ‘lobes’ of a rosette may synthesize six β-1,4-glucan chains which then cocrystallize to form a 36-chain microfibril (Herth, 1983) (Fig. 2).

Figure 2.

Model of cellulose synthase rosette. Cross-section through a rosette, each ‘lobe’ containing a number of catalytic subunits. The active sites of these subunits are cytosolic, and the extending cellulose microfibril crystallizes as it enters the cell wall.

To date, the only components identified as part of the cellulose synthase complex in higher plants are the cellulose synthase (CesA) proteins. These were originally identified in plants by comparing sequences from a cotton fibre cDNA library with bacterial cellulose synthase (Pear et al., 1996). The real breakthrough in the field, however, was the discovery that the gene affected in the temperature-sensitive radial swelling 1 (rsw1) mutant was a member of the CesA gene family (Arioli et al., 1998). At the restrictive temperature, it was found that rosettes disassociated into individual lobes, demonstrating for the first time a definitive link between CesA proteins and the structure of the rosette, and proof that a complete rosette was required for proper cellulose synthesis. Genome sequencing has revealed that all higher plants have multiple CesA genes, which share a conserved structure. The completed Arabidopsis genome reveals a complement of 10 CesA genes (Holland et al., 2000; Richmond, 2000), maize has at least 12 (Appenzeller et al., 2004), barley at least eight (Burton et al., 2004), rice at least nine (http://waltonlab.prl.msu.edu/CSL_updates.htm) and aspen (Populus tremuloides) at least seven (Joshi et al., 2004). Naming of CesAs follows the convention of species initials followed by gene number (e.g. AtCesA1 is A. thaliana Cellulose Synthase 1). Unfortunately, because of the order in which these genes were cloned, there is little or no conservation in the numbering of the genes between species.

Arabidopsis CesA proteins range from 985 to 1088 amino acids in length, and share a conserved structure. They contain eight predicted trans-membrane domains, and their topology within the membrane, predicted by computational methods, has been confirmed by proteomic analysis of ‘inside-out’ membrane vesicles (Nuhse et al., 2004). The structure of a typical CesA protein is shown in Fig. 3. The amino terminus contains a cytosolic domain of between c. 60 and 100 amino acids followed by two predicted trans-membrane domains. The central portion of the protein is cytosolic, with a further six predicted trans-membrane domains at the carboxy-terminus of the protein. The amino-terminal region of the protein contains a cysteine-rich domain that contains four CxxC motifs, which fits well to the consensus for a RING-type zinc finger. These domains have been shown to be involved in a variety of protein–protein interactions in protein complexes (Saurin et al., 1996). It has been shown that this purified domain from a cotton CesA protein (GhCesA1), when overexpressed in Escherichia coli, can bind zinc, as predicted from its structure (Kurek et al., 2002). In addition, it was shown that the zinc domain from GhCesA1 interacted both with itself and with the zinc domain of a second cotton fibre cellulose synthase, GhCesA2, both in a yeast two-hybrid system and in vitro using isolated recombinant proteins (Kurek et al., 2002). This suggests that the zinc RING finger domains may be involved in CesA protein dimerization, and this may lead to higher-order structures. The assembly of the cellulose synthase complex, and its importance in synthesizing cellulose of the correct microfibrillar structure, are discussed below.

Figure 3.

Membrane topology of a CesA protein. Black bars indicate trans-membrane domains. Zn, zinc-binding region; VR1, variable region 1; CSR, class-specific region. Approximate positions of D, D, D, QXXRW processive glycosyl transferase motifs are shown.

The central cytosolic ‘catalytic’ domain contains the signature motifs of processive glycosyltransferases, which are collectively known as the D, D, D, QXXRW motif (Saxena et al., 1995). These motifs have been mutated in other processive glycosyltransferases and shown to be essential for activity (Nagahashi et al., 1995; Saxena et al., 2001). The central cytoplasmic domain is highly conserved between all the CesA proteins, apart from one region. This region was initially referred to as a hypervariable region, as it was highly divergent between the sequences that had been identified. The availability of CesA sequences from many more species led to the recognition of sequence conservation in this region between CesAs from different species, but not within CesAs from the same species, leading to this region being renamed the Class Specific Region (Vergara & Carpita, 2001). This, along with the increasing number of sequenced genomes, has led to the identification of orthologs between species that are more closely related to one another than are the paralogs within a species. (Orthologs – genes originating from a single ancestral gene in the last common ancestor of the compared genomes; paralogs – genes related via duplication; Koonin, 2005) An example of this can be seen in Arabidopsis, wheat, barley and rice CesAs (Burton et al., 2004). One current exception to this rule of conservation being higher between CesAs of the same class than within the same species has been shown by the sequencing of the Populus trichocarpa (black cottonwood) genome. Analysis of this genome reveals the presence of 18 CesA genes, which fall into seven pairs, one set of three genes and a single gene. The proteins within the pairs show similar levels of identity at the amino acid level to each other as they do to their orthologs from hybrid poplar (Populus tremula × tremuloides) (Djerbi et al., 2005). Both members of some of these gene pairs are actively transcribed, but further functional studies are required to establish the role of these CesAs in plant development. One possibility is that the intensive periods of cellulose synthesis during wood formation may require additional copies of cellulose synthases (Djerbi et al., 2005).

IV. Mutations affecting cellulose synthesis

Much of our knowledge about cellulose synthesis has been derived from the identification of mutants in A. thaliana. These mutants have been isolated from a variety of genetic screens that reflect the importance of cellulose synthesis in plant growth and development. It has become clear that cellulose synthesis in the primary and secondary cell walls, while sharing some components, use a different set of cellulose synthases. Hence cellulose-deficient mutants can be conveniently divided into mutations in CesA genes that affect either primary or secondary cell walls, and mutations in other genes affecting cellulose synthesis. Because of the diversity of mutant screens that has yielded mutations in CesAs, there is a wide range of mutant allele names. To minimize confusion, the nomenclature for specific mutants used here will be the CesA number with the particular allele following (e.g. AtCesA7irx3-1).

1. Mutations in primary cell wall CesAs

A number of mutations that affect cellulose synthesis in the primary cell wall have been identified from a variety of mutant screens. The broad range of screens that have yielded these mutants reflects the importance of primary cell wall cellulose synthesis in plant development. Severe mutations, such as strong alleles of AtCesA1 (Beeckman et al., 2002; Gillmor et al., 2002) are embryo-lethal, with the homozygous mutant embryos severely deficient in cellulose. It has been shown recently that null alleles of AtCesA1 and AtCesA3 are gametophytically lethal due to defective pollen development (Persson et al., 2007). The severe phenotype of mutations in these two genes demonstrates that they are not redundant with one another (Desprez et al., 2007; Persson et al., 2007). The isolation of less severe mutations has, however, allowed the study of primary cell wall CesAs. A temperature-sensitive mutation in AtCesA1 (AtCesA1rsw1-1 (radial swelling 1-1)) exhibits swelling of cells when grown at the nonpermissive temperature. This is a result of a reduction in cellulose synthesis caused by disassembly of the cellulose synthesizing rosettes (Arioli et al., 1998). This mutant does, however, produce a ‘noncrystalline’β-1,4-glucan. It is not known whether this is an activity of the disassembled rosettes, or a response of the plant to the lack of cellulose in the primary cell wall. Signalling of a defective cell wall and compensatory polymer synthesis is becoming increasingly clear as a phenomenon, and is discussed below. Less severe mutations in AtCesA3 are also viable, and have been identified from a number of mutant screens, such as resistance to the cellulose synthesis inhibitor isoxaben (AtCesA3ixr1 (isoxaben resistance 1, Scheible et al., 2001)), enhanced disease resistance (AtCesA3cev1 (constitutive expression of VSP1, Ellis et al., 2002)) and lignin deposition in normally unlignified cell walls (AtCesA3eli1 (ectopic lignin, Cano-Delgado et al., 2003)). The phenotypes of these latter two mutants that led to their identification are also likely to be caused by a response of the plant to a defective cell wall. Null mutations in AtCesA6 show a reduced elongation of roots and dark-grown hypocotyls as a result of cellulose deficiency (AtCesA6prc1-1 (procuste, Fagard et al., 2000)), but mutant plants are viable and the effects of the mutation are clearly not as severe as null mutations in AtCesA1 and AtCesA3. Another mutation (AtCesA6ixr2, Desprez et al., 2002) was identified in the screen for resistance to the cellulose synthesis inhibitor isoxaben, which also yielded the AtCesA3ixr1 mutation. Two recent studies demonstrate that AtCesA2, AtCesA5 and AtCesA9 are partially redundant with AtCesA6 at different stages of growth (Desprez et al., 2007; Persson et al., 2007). AtCesA2 mutants show a slight decrease in the length of dark-grown hypocotyls (depending on the individual mutant used), but otherwise AtCesA2, AtCesA5 and the corresponding double null mutants display no clear cellulose-deficient phenotypes. AtCesA5/6 double mutants are seedling-lethal, however, indicating that for seedling establishment AtCesA5 and AtCesA6 are redundant with one another (Desprez et al., 2007). Double AtCesA2/6 knockout mutants show an enhanced phenotype compared with AtCesA6 single mutants, as seen by the reduced dark-grown hypocotyl and light-grown seedling length and dwarfed adult plants (Desprez et al., 2007), suggesting that these two genes are only partially redundant with one another. As described earlier, null AtCesA1 and AtCesA3 mutants are defective in pollen development and are hence gametophytically lethal. AtCesA2, AtCesA6 or AtCesA9 single mutants and AtCesA2/6, AtCesA6/9 and AtCesA2/9 double mutants are all normal in pollen development. Pollen from plants homozygous for two of the three mutations and heterozygous for the third (AtCesA6/9/2+/−) has half normal and half deformed pollen, as is the case for pollen from heterozygous AtCesA1 and AtCesA3 null mutants, demonstrating that at least one of AtCesA2, 6 or 9 is required for correct pollen development (Persson et al., 2007). Taken together, these data suggest that AtCesA6 is partially redundant with other members of the CesA gene family, AtCesA2, 5 and 9 (which, by virtue of their close relationship to CesA6 in a phylogeny of CesA proteins (Burton et al., 2004), can be described as AtCesA6-related). The differential redundancy between the isoforms appears to be tissue-dependent. This has led to the idea that AtCesA1, AtCesA3 and an AtCesA6-related protein are all essential for cellulose synthesis in primary cell walls.

2. Mutations in secondary cell wall CesAs

Mutations affecting secondary cell wall synthesis in Arabidopsis are characterized by their collapsed xylem cells owing to the inability of xylem vessels to withstand the negative pressure generated by the transpiration stream. These mutations have been named irregular xylem (irx) mutants (Turner & Somerville, 1997). (These should not be mixed up with the isoxaben-resistant (ixr) mutants: confusingly, they all encode different CesAs.) Mutants irx1, irx3 and irx5 are caused by lesions in the AtCesA8, AtCesA7 and AtCesA4 genes, respectively (Taylor et al., 1999; Taylor et al., 2000; Taylor et al., 2003). Stems of these mutants contain approx. 30% of the cellulose compared with the wild type, because of the absence of cellulose in the secondary cell wall (Ha et al., 2002). Cellulose in the primary cell walls appear unaffected in these mutants (Turner & Somerville, 1997; Ha et al., 2002). That mutations in three different members of the same gene family led to almost identical phenotypes and a similar severe reduction in stem cellulose content indicates that these three genes are nonredundant and are active in the same cells. Both promoter–GUS constructs and specific antibodies to the three proteins revealed that they were indeed expressed in the same cells at the same time (Taylor et al., 2003). Mutations in AtCesA7 and AtCesA8 have also been isolated from a screen to identify mutants with interfascicular fibres of reduced strength. These mutants were named fragile fibre (AtCesA7fra5 and AtCesA8fra6, Zhong et al., 2003). Interestingly, three transposon mutations in rice (Oryza sativa) resulting in a brittle culm phenotype with reduced tissue strength have been shown to be caused by lesions in three different members of the CesA gene family (OsCesA4, OsCesA7 and OsCesA9). These genes are orthologs of the CesAs responsible for secondary cell wall synthesis in Arabidopsis, although because of differences in the numbering of genes between the species, OsCesA4, -7 and -9 are orthologs of AtCesA8, -4, and -7, respectively (Tanaka et al., 2003).

3. Mutations in other genes affecting cellulose synthesis

Because of the importance of cellulose synthesis in plant development, many mutants have been isolated that have a cellulose deficiency. Despite this, many of these mutants are affected in general ‘housekeeping’ roles and are not specific to cellulose synthesis. These include mutations in steps in the pathway for N-linked glycosylation (Boisson et al., 2001; Lukowitz et al., 2001; Burn et al., 2002; Gillmor et al., 2002). Only mutations directly affecting cellulose synthesis will be described here.

KORRIGAN   Mutations in a predicted β-1,4-glucanase, KORRIGAN, have been identified from a number of screens including dwarfism (Nicol et al., 1998), radial swelling of root tips (Lane et al., 2001; Sato et al., 2001) and collapse of xylem vessels (Szyjanowicz et al., 2004). These mutations result in reduced cellulose in both primary and secondary cell walls. The most severe mutations in KOR are defective in cytokinesis (Zuo et al., 2000), whereas mild alleles are unaffected in primary cell wall cellulose synthesis, but show reduced cellulose in the secondary cell wall (Szyjanowicz et al., 2004). This shows that KOR is required for proper cellulose synthesis in both primary and secondary cell walls; there appears to be sufficient KOR activity in weak alleles to support the relatively low rates of primary cell wall synthesis, but not the high rates of cellulose synthesis in secondary cell walls (Szyjanowicz et al., 2004). Strong alleles, however, will not support any level of cellulose synthesis and are hence cytokinesis-deficient (Zuo et al., 2000). KOR is predicted to be a membrane-bound β-1,4-glucanase, and the soluble domain of a KOR-like protein from Brassica napus heterologously expressed in Pichia pastoris showed cellulase activity (Molhoj et al., 2002), as did an ortholog purified from poplar (Master et al., 2004). The precise role of KORRIGAN is unknown, and the subcellular localization of the protein could provide a clue to its role in cellulose synthesis. There are, however, many inconsistencies in the reports as to its localization. A C-terminal GFP fusion to KOR expressed in tobacco BY-2 cells accumulated in intracellular organelles in interphase cells and to the phragmoplast in dividing cells (Zuo et al., 2000). It was not shown, however, whether this protein fusion was functional. A more recent study with an N-terminal–KOR fusion protein that complemented the cell elongation defect of the kor1-1 mutation was localized to endosomes and Golgi membranes, but could not be detected at the plasma membrane, although some GFP–KOR1 was seen in motile compartments near the plasma membrane (Robert et al., 2005). Colocalization of KOR with the secondary cell wall CesAs in developing xylem vessels was not observed (Szyjanowicz et al., 2004). It is apparent that KOR is not an integral part of the cellulose synthase complex, as it was not copurified with cellulose synthases responsible for either secondary cell wall or primary cell wall synthesis (Szyjanowicz et al., 2004; Desprez et al., 2007). Despite the fact that a β-1,4-glucanase is also required for cellulose synthesis in bacteria (Molhoj et al., 2002), its exact function is unclear. It has been suggested that KOR may have a role in cleaving glucan residues from a sitosterol-linked primer that is then incorporated into the growing cellulose chain (Peng et al., 2002), or in the recycling of these sterol glucoside primers (Robert et al., 2004). An alternative function for KOR is at the plasma membrane, either ‘editing’ growing microfibrils to ensure proper packing of individual chains, or severing cellulose chains to terminate microfibrils. These roles later in cellulose synthesis, in either editing or severing microfibrils, are consistent with the presence of secondary cell wall deposition at the corner of cells (which are the earliest sites of secondary cell wall deposition, Altamura et al., 2001) in the relatively weak irx2 alleles of korrigan (Turner & Somerville, 1997).

COBRA   Mutations at the COBRA locus were identified from a screen for plants with abnormally expanded root tips. COBRA encodes a glycosyl-phosphatidylinositol (GPI)-anchored protein that is polarly targeted to the plasma membrane and longitudinal cell walls of elongating root cells (Schindelman et al., 2001; Roudier et al., 2005). COBRA is distributed in a banded pattern perpendicular to the longitudinal axis via a microtubule-dependent mechanism (Roudier et al., 2005). Null alleles of COBRA are defective in polarized (elongation) cell growth that occurs after the initial cell expansion responsible for morphogenesis (Roudier et al., 2005). cobra mutants have reduced cellulose, but the use of conditional cobra alleles has shown that swelling of cells occurs as a result of disorganized cellulose microfibril deposition. This disorganization occurs before there is a significant decrease in the amount of cellulose in the cell wall. Thus COBRA is involved in the orientation of deposition of cellulose microfibrils, and the reduction in cellulose in cobra mutants is probably caused by a feedback mechanism reflecting the already disordered deposition of cellulose microfibrils (Roudier et al., 2005). Arabidopsis contains a family of 12 COB genes (consisting of COB and 11 COB-LIKE (COBL) genes, and while mutations in COB itself appear to affect only the second phase of maximal and developmentally regulated anisotropic cell expansion (during cell elongation), mutations in other members of the gene family affect other cell walls. AtCOBL4 is required for cellulose synthesis in the secondary cell wall, and was identified from a transcriptomics approach that will be discussed in a later section (Brown et al., 2005; Persson et al., 2005). The rice ortholog of AtCOBL4 was previously identified as being required for secondary cell wall cellulose synthesis in rice (Li et al., 2003). AtCOBL9 is required for tip-directed growth in root hair development and was identified by its enrichment in the root hair transcriptome (Jones et al., 2006).

KOBITO   Mutations in the KOBITO gene result in cellulose-deficient, dwarfed mutants. These mutants exhibited randomized cellulose microfibril orientation on recently deposited cell wall layers compared with the wild type, which had microfibrils well ordered transverse to the direction of elongation (Pagant et al., 2002). Another allele of KOB was identified from a cellulose-deficient elongation-deficient (eld) mutant (Lertpiriyapong & Sung, 2003). A further allele was identified in a screen for abscissic acid-insensitive mutants (abi8, Brocard-Gifford et al., 2004). It is not clear, however, what the link is between the insensitivity to abscissic acid and cellulose synthesis. KOBITO is a protein of unknown function, and is predicted to be a type II membrane protein with the N-terminus exposed to the cytosol. Its subcellular localization has not been definitively proven, however, with different studies showing plasma membrane localization in the cells in the elongation zone of the root, but a punctuate intracellular distribution in the cell division zone at the root tip (Pagant et al., 2002), a punctuate cytoplasmic pattern in the root elongation zone (Brocard-Gifford et al., 2004) or cell wall localization (Lertpiriyapong & Sung, 2003). As with KORRIGAN and COBRA, the precise role of KOBITO in cellulose synthesis is currently unclear.

V. Cellulose synthase complex assembly

The role of plasma membrane-bound rosettes in the synthesis of cellulose has long been clear (Herth, 1983). The first proof that CesA proteins were central to the rosette structure was gained from the temperature-sensitive allele AtCesA1rsw1-1. At the restrictive temperature, freeze-fracture studies revealed that the intact hexameric rosette structures were lost (Arioli et al., 1998). Antibodies specific to the conserved domain of cotton cellulose synthase have also been shown to label rosette structures in freeze-fracture studies (Kimura et al., 1999).

It is clear from studies of mutants of Arabidopsis that three different CesAs are essential for normal cellulose synthesis. In the secondary cell wall, these are AtCesA4, 7 and 8 (Taylor et al., 2003), and in the primary cell wall they are AtCesA1, 3 and a CesA6-related protein (Desprez et al., 2007; Persson et al., 2007). All three subunits must be present for the assembly of the cellulose synthase protein complex. The most detailed study of this has been made using specific antibodies to the secondary cell wall CesAs. Using a functional epitope-tagged version of CesA7 (IRX3), which complements the AtCesA7irx3-1 mutation, it was shown that CesA7, CesA4 and CesA8 could be copurified as part of the same protein complex. This was confirmed in wild-type plants by coimmunoprecipitation experiments (Taylor et al., 2003), and it was further demonstrated that in developing xylem elements all three proteins were colocalized at the surface of the cell, presumably in the plasma membrane (Gardiner et al., 2003). In the absence of one of the subunits, however, the remaining two subunits no longer interact, indicating that there are specific and defined interactions between the three proteins necessary for correct assembly of the protein complex (Taylor et al., 2003). This lack of interaction was accompanied by an intracellular localization of the remaining two subunits in developing xylem elements (Gardiner et al., 2003). This is entirely consistent with the protein complex being assembled in an intracellular compartment (presumably the endoplasmic reticulum) and complexes being transported to the plasma membrane intact. Indeed, apparently intact rosettes have been visualized in electron micrographs in what were thought to be Golgi vesicles (Haigler & Brown, 1986). The assembly of the cellulose synthase rosette is dependent on the presence, but apparently not the activity, of these multiple subunits. The AtCesA8irx1-1 mutation alters a highly conserved aspartate residue and is presumed to have no activity (Taylor et al., 2000). Levels of this protein are comparable with the wild type, and interactions between the three CesAs are indistinguishable from wild type (Taylor et al., 2003). The protein complex containing this inactive protein is also transported to the plasma membrane normally (Gardiner et al., 2003). All these data support the idea that three different CesA proteins are essential for cellulose synthesis in the secondary cell wall.

In the primary cell wall, it has also been assumed that multiple CesAs make up the cellulose synthase complex, based on the similarity of the mutant phenotypes, and that resistance to the cellulose synthesis inhibitor isoxaben can be achieved through mutation of either AtCesA3 or AtCesA6 (Scheible et al., 2001; Desprez et al., 2002). Recent data have confirmed that the primary cell wall CesAs AtCesA3 and AtCesA6 also interact with one another in the same protein complex, probably with CesA1 (Desprez et al., 2007). It has also been observed both in the primary and secondary cell walls that, in the absence of one of the subunits, the remaining subunits accumulate to reduced levels compared with wild-type plants (Taylor et al., 2003; Desprez et al., 2007). This is consistent with all three proteins being required for assembly of the cellulose synthase complex. There are several examples of oligomeric protein complexes in which the correct folding and assembly of protein subunits in the endoplasmic reticulum depends on the presence of other subunits of the complex (Geering, 1997; Beggah et al., 1999).

While all the available data support the assembly of three different CesAs into a protein complex, there are no data describing how the rosette is arranged. A commonly depicted model of a cellulose synthase rosette shows a hexameric structure of six lobes, which is well supported by electron micrographs of rosettes (Arioli et al., 1998), each containing six catalytic subunits (Fig. 4). While there is no evidence against this model, it is based mainly on the observation that a cellulose microfibril contains c. 36 chains. There are a number of questions that need to be answered before the finer detail of the structure of a rosette can be resolved. For example, what is the basic ‘catalytic unit’ within a complex? Analysis of predicted protein sequence suggests that there is only one active site in each subunit. How does this fit with the fact that adjacent glucose residues are oriented 180°, this rotation being essential for correct packing of the glucan chains into a microfibril? While it has been suggested that there is enough flexibility in the bond to allow this rotation (Delmer, 1999), one possibility is that CesA proteins work as dimers, with the active sites oriented such that two glucose residues are added in their correct orientation simultaneously. The fact that the zinc-binding domain of cotton CesA can bind to another, similar domain suggests that the formation of dimers is a possibility. There is also the question as to what is the role of the different CesA proteins in rosette architecture – what makes a type 1 CesA different from a type 2 CesA? One way to answer this question would be to undertake a comprehensive domain-swapping experiment between the three CesAs required for secondary wall synthesis and determining which domain defines the CesA type. The availability of antibodies and epitope-tagged lines for these CesAs now makes this a possibility (Taylor & Turner, 2007). Rosette structure is clearly central to the correct synthesis of cellulose microfibrils that have the properties required for their role in cell walls, and a more detailed analysis of the structure of the rosette and the interactions between the different CesA proteins is required.

Figure 4.

Model of a cellulose synthase complex. Each of the six ‘lobes’ are predicted to contain multiple copies of three different CesA proteins. In the Arabidopsis primary cell wall, these are CesA1, 3 and a CesA6-like protein, and in the Arabidopsis secondary cell wall these are CesA4, 7 and 8. The exact number of CesA proteins contained within a lobe, their stoichiometry and their specific interactions are currently unknown.

VI. Transcriptomic approaches to identify genes involved in cellulose synthesis

As described previously, mutant screens for plants defective in cellulose synthesis have led to the identification of a number of genes involved in cellulose synthesis. It is likely, however, that mutant screens will not have identified all the components necessary for cellulose synthesis because of mutations being lethal or exhibiting severely affected development due to their essential role in primary wall cellulose synthesis. In addition, secondary cell wall mutant screens may have missed some milder phenotypes because of this screen preferentially identifying severe phenotypes (Turner & Somerville, 1997).

It has been described previously that primary and secondary cell wall cellulose synthesis requires a separate complement of CesA genes, and these genes are expressed in a very tightly coordinated manner. Two groups have made use of transcriptomics in order to identify further components necessary for cellulose synthesis. Brown et al. (2005) generated a new transcriptomic data set from different stages of plant development, in which the expression of the three secondary cell wall cellulose synthases showed a very clear pattern. Further genes showing a similar expression pattern were identified and publicly available data were cross-referenced. This led to the identification of seven genes that, when disrupted by an inserted T-DNA, showed an irregular xylem phenotype similar to mutations in the secondary cell wall CesAs. Surprisingly, only one of these insertion mutants, in the AtCOBL4 gene had a significant decrease in cellulose content. The other genes are likely to define novel processes that are required for secondary cell wall formation (Brown et al., 2005).

A second study using a different statistical approach was carried out on publicly available microarray data, and identified a list of genes coexpressing with both primary cell wall CesAs (including COBRA) and secondary cell wall CesAs. This second list contained many of the same candidates as the study specifically on secondary cell walls including COBL4, although in contrast to the study by Brown et al. (2005), no data on the cellulose content of mutants in this gene were shown (Persson et al., 2005).

While this approach has yielded new information on genes involved specifically in primary and secondary cell wall cellulose synthesis, it may not identify components that are required for both types of cell wall. For example, KORRIGAN, which has been shown by mutational analysis to be involved in both primary and secondary cell wall synthesis, shows a low level of correlation with the expression of primary cell wall CesAs but does not have a similar enough expression pattern to the secondary cell wall CesAs to be identified as being coexpressed (Brown et al., 2005; Persson et al., 2005). It is clear, then, that a combination of complementary approaches is required for identification of the full complement of components required for cellulose synthesis.

VII. Purification of the cellulose synthase complex

One way to definitively identify the essential components required for cellulose synthesis would be to purify the active cellulose synthase complex for subsequent proteomic analysis. Assaying cellulose synthase activity has been problematic, however, with only low levels of cellulose synthase activity in preparations that also contain very high levels of callose synthase activity (Kudlicka et al., 1995; Kudlicka & Brown, 1997). Attempts have been made over many years to purify an active cellulose synthase complex with little success. These problems may be caused by the predicted large size of the protein complex (if the complex contained only 36 CesA proteins, it would be approx. 4 MDa) and/or the presence of labile cofactors. High levels of cellulose synthase activity have been reported in vitro from solubilized microsome preparations from blackberry cell lines, but this method does not involve purification of the protein complex, and the assay conditions are not suitable for assaying cellulose synthesis in Arabidopsis, where a proteomics approach would be most applicable (Lai-Kee-Him et al., 2002). Attempts to purify the cellulose synthase complex from Arabidopsis plants expressing a functional epitope-tagged CesA7 were unsuccessful, mainly because of problems in solubilizing CesA7 from a microsomal fraction (Taylor et al., 2004).

VIII. Regulation of cellulose synthesis

The deposition of cellulose is clearly controlled in a very specific manner, both temporally and spatially. There is significant interest currently in identification of the mechanisms that control cellulose synthesis in the secondary cell wall as an approach to modify cell wall composition in order to improve its feasibility as a source of bioenergy. The identification of transcription factors that regulate overall secondary cell wall synthesis is a major step in this direction. NAC and MYB family transcription factors have recently been shown to be key players in regulating secondary cell wall biosynthesis. SND1 (secondary wall-associated NAC domain protein 1) is specifically expressed in interfascicular and xylem fibres, and dominant repression of SND1 drastically reduces the secondary wall thickening in fibres (Zhong et al., 2006). Overexpression of SND1 leads to activation of secondary cell wall biosynthetic pathways and causes ectopic deposition of secondary cell walls. It was recently found that SND1 acts redundantly with NST1 (NAC secondary wall-thickening-promoting factor 1) in the regulation of secondary wall synthesis in fibres. Simultaneous knockout of both genes results in severe reduction in the expression of secondary cell wall biosynthetic genes and the loss of all three major secondary wall components (cellulose, lignin and xylan) (Mitsuda et al., 2007; Zhong et al., 2007a). NST1 and another gene, SND2, are redundant in regulating secondary cell wall formation in the endothecium cell layer of anthers, where a secondary wall is required so that rupture can occur to release pollen grains (Mitsuda et al., 2005). Mutation of both genes simultaneously results in a lack of secondary wall thickening in the endothecium, and overexpression of either transcription factor results in ectopic secondary cell wall deposition (Mitsuda et al., 2005). A MYB transcription factor, MYB26, is also essential for secondary wall thickening in the endothecium (Steiner-Lange et al., 2003), and its overexpression also causes ectopic secondary wall formation (Yang et al., 2007). It has been shown recently that another MYB transcription factor, MYB46, is a direct target of SND1, and dominant repression of MYB46 resulted in a lack of secondary cell wall thickening in fibre and xylem vessels in stems (Zhong et al., 2007b). These data all suggest that all three major pathways for secondary cell wall synthesis (cellulose, lignin and xylan) are coordinately regulated at the transcriptional level. It is likely, however, that there is a cascade of transcription factors regulating secondary cell wall synthesis, and that further dissection of these networks may identify transcription factors involved in activation of the individual pathways for the three major secondary cell wall polymers, including cellulose.

While the transcriptional control of cellulose synthesis is important as a general regulatory step, the assembly and transport to the plasma membrane of the cellulose synthase complex is a tightly regulated process, likely to be regulated by post-translational modification. There is also a requirement for some feedback communication between the cell wall, where the cellulose is deposited, and the cytosol, which contains the catalytic site of cellulose synthase. One way in which this signalling across the plasma membrane could occur is through the action of membrane-spanning receptor-like kinases (RLKs) that trigger a kinase cascade. Phosphorylation sites have been identified on a number of Arabidopsis CesA proteins in vivo. Most of these phosphorylation sites are found in the variable region in the amino-terminal cytosolic domain (Nuhse et al., 2004; Taylor, 2007). In vitro, the region of AtCesA7 that is phosphorylated in vivo can be phosphorylated by a detergent-solubilized stem extract, and this phosphorylation is accompanied by rapid protein degradation via a 26s proteasome-dependent pathway. Full-length CesA7 was also degraded by a 26s proteasome-dependent pathway in solubilized stem extracts, suggesting that phosphorylation of CesA proteins may target them for degradation (Taylor, 2007). This is one way in which plants could regulate the relative levels of the three CesAs that interact within the same protein complex. The precise role of phosphorylation of CesAs in vivo is unclear, however, and is still under investigation. Recent work has shown the half-life of a CesA protein in cotton fibres that produce large amounts of secondary cell wall cellulose to be < 30 min, much lower than the average membrane protein (Jacob-Wilk et al., 2006). This is consistent with the visualization of cellulose synthase complexes moving at a constant rate in the plasma membrane for at least 15 min (Paradez et al., 2006). It has also been shown that monomeric zinc-binding domains are far more susceptible to degradation than dimeric forms (Jacob-Wilk et al., 2006), although it should be noted that this study used a different portion of the protein (the N-terminal 66 amino acids) compared with the region of the protein that is phosphorylated (amino acids 114–203) (Taylor, 2007). In addition, cotton fibres show a high potential for proteasome activity when they are at the peak of their cellulose synthesis (Jacob-Wilk et al., 2006), suggesting that the proteasome may be involved in specific degradation of monomeric proteins that are not associated into a rosette. Phosphorylation sites have also been identified on the intracellular domain of the glucanase KORRIGAN (Nuhse et al., 2004). At present, no kinases involved in phosphorylating these proteins have been definitively identified.

IX. Feedback responses to cell wall defects

The cell wall is a dynamic structure that is central to plant development. Responding to defects in cell wall integrity is an important way in which the plant can maintain cell wall functionality by either altering cell wall composition or activating stress-response pathways. Understanding of how plants sense and respond to alterations in the extracellular environment is limited, but an excellent review of this area has been published recently (Humphrey et al., 2007). Inhibition of cellulose synthesis by chemical inhibitors or in mutant backgrounds can lead to the accumulation of ectopic lignin and changes in the composition of matrix polysaccharides (His et al., 2001; Desprez et al., 2002; Cano-Delgado et al., 2003). Jasmonate and ethylene pathways and stress-response gene expression are also activated in AtCesA3eli1 and AtCesA3cev1 (Ellis & Turner, 2001; Cano-Delgado et al., 2003). The way in which cellulose deficiency is sensed is not currently well understood. Recent work has, for the first time, identified a possible mechanism for sensing cell wall defects. A receptor-like kinase was identified from a screen for suppressors of AtCesA6prc1-1. Mutation of this RLK, THESEUS1 (THE1), attenuated the growth defect and ectopic lignification of AtCesA6prc1-1, without rescuing the cellulose deficiency (Hematy et al., 2007). This effect was also seen in other cellulose-deficient mutants (including AtCesA3eli1-1 and AtCesA1rsw1-10). Overexpression of THE1 in AtCesA6prc1-1 and AtCesA3eli1-1 mutants increased the ectopic lignification phenotype. Interestingly, there was no effect of either mutation or overexpression of THE1 in wild-type plants, suggesting that there is some signal in particular cellulose deficient mutants required for THE1 activity. The nature of this signal is not currently known (Hematy et al., 2007).

Analysis of mutant phenotypes in different genetic backgrounds suggests that responses to cell wall defects differ between ecotypes of Arabidopsis. In the Columbia ecotype, knockout mutants of AtCesA2 are indistinguishable from wild-type in light-grown plants, and have only a slightly reduced hypocotyl length in dark-grown plants (Desprez et al., 2007). In a Landsberg erecta background, however, knockout of CesA2 results in severe stunting and a significant reduction in cellulose content in leaves not seen in knockout mutations in a Columbia background in the same study (Chu et al., 2007). This suggests that there is a significant difference either in the redundancy between different CesAs, which is unlikely given the apparent similarity between CesA function between species, or more likely a difference in the response of the different ecotypes to alterations in cell wall integrity. Mutations in secondary cell wall CesAs also show a different severity of phenotype between ecotypes. In Landsberg erecta, null mutations in CesA7 cause a lack of cellulose in the secondary cell wall, but the plants remain relatively healthy and fertile (Turner & Somerville, 1997). Null mutations in CesA7 in a Columbia background, however, are very small and almost completely sterile (Brown et al., 2005). Another allele, AtCesA7mur10 (Bosca et al., 2006) in the Columbia background, was identified from a screen for mutants showing altered primary cell wall composition (Reiter et al., 1997). CesA7 is clearly expressed only in secondary cell wall-forming cells (Taylor et al., 2003; Bosca et al., 2006), demonstrating that in this genetic background, an alteration in secondary cell wall cellulose synthesis fed back into a signalling pathway, resulting in altered primary cell wall synthesis. It is tempting to speculate that these differences may be caused by the erecta mutation in the Landsberg erecta background. ERECTA is a receptor-like kinase that has been shown to be involved in inflorescence architecture because of the altered organ shape and internodal elongation pattern in erecta mutants (Torii et al., 1996; Komeda et al., 1998). It has also been shown to regulate transpiration efficiency through a number of mechanisms, including stomatal density, epidermal cell expansion, mesophyll cell proliferation and cell-to-cell contact (Masle et al., 2005). The pleiotropic functions of ERECTA make it possible that it may be involved in general signalling from the cell wall, and that mutations in ERECTA result in a difference in the responses of plants to perturbations in cell wall structure or composition.

X. Regulation of orientation of cellulose deposition

Because of their physical properties, cellulose microfibrils must be deposited in a very specific orientation in order for them to provide the strength required in cell walls. In growing cells, cellulose microfibrils are usually deposited in parallel arrays transverse to the direction of growth. It has long been hypothesized that the orientation of deposition of cellulose microfibrils is associated with underlying cortical microtubules (Heath, 1974), which in growing cells also form a characteristic transverse array (Wick et al., 1981; Gunning & Hardham, 1982). This hypothesis has since been called the alignment hypothesis (Baskin, 2001), and it has been suggested that microtubules could influence the deposition of cellulose microfibrils either by directly guiding the cellulose synthase rosettes as they synthesize cellulose (Heath, 1974), or by acting as buffers that constrain the lateral movement of rosettes, leading to a deposition of cellulose microfibrils parallel to the underlying cortical microtubule array (Giddings & Staehelin, 1991).

There are, however, many inconsistencies in the model of microtubules guiding microfibril deposition. Short treatments of Arabidopsis with the microtubule-destabilizing drug oryzalin or the microtubule-stabilizing drug taxol caused no apparent change in the orientation of cellulose microfibrils in cells that expanded during treatment (Sugimoto et al., 2003; Baskin et al., 2004). Impairment of microtubule organization at the restrictive temperature of the temperature-sensitive microtubule mutant mor1-1 also did not alter the pattern of cellulose deposition (Sugimoto et al., 2000; Himmelspach et al., 2003). These studies relied on visualizing the most recently laid-down layer of cellulose microfibrils, but did not address cellulose synthesis directly.

Paradez et al. (2006) have recently visualized cellulose synthase movement using a functional YFP–CesA6 fusion protein in expanding hypocotyls cells. Roughly linear trajectories of this labelled protein complex were observed, and this movement was along tracks also labelled by a fluorescently labelled alpha tubulin marker. Reorientation of the cortical microtubule array by light or low concentrations of oryzalin led to a corresponding change in the direction of cellulose synthase movement. Partial disorganization of microtubules with oryzalin led to an aggregation of CesA complexes in the plasma membrane that appeared to follow the few remaining microtubules. Cellulose synthase complex movement was observed beyond the ends of microtubules (Paradez et al., 2006), demonstrating that microtubules are not responsible for the motile force behind cellulose synthesis. This supports the idea that the force for rosette movement is provided by the polymerization of cellulose (Robinson & Quader, 1981; Lloyd, 1984). Complete disassembly of the microtubules, however, resulted in cellulose synthase complexes becoming reorganized and more uniformly dispersed compared with cells with partial disassembly of the microtubules (Paradez et al., 2006). This suggests that, while microtubules do influence the orientation of cellulose synthesis, either there is some capacity for the complexes to self-organize in the complete absence of a microtubule network (Emons & Mulder, 2000), or there is a second organizational mechanism (Sugimoto et al., 2003; Wasteneys, 2004).

Microtubules are, however, required for the correct distribution of cellulose synthase complexes in the secondary cell wall of developing root xylem cells, where thick bands of secondary cell wall thickenings are laid down at specific sites. Depolymerizing microtubules with oryzalin rapidly alters the pattern of cellulose synthase complexes from a characteristic banded pattern to a diffuse pattern (Gardiner et al., 2003). Work with cellulose synthesis inhibitors also suggested that there might be some feedback between primary cell wall cellulose microfibrils and microtubules, with cellulose synthase activity required for microtubule orientation (Fisher & Cyr, 1998; Himmelspach et al., 2003; Lazzaro et al., 2003), but in the secondary cell wall, at least, this is not the case, as in the absence of one of the secondary cell wall CesAs, and therefore the lack of an active cellulose synthase complex at the plasma membrane, microtubule orientation appeared normal (Gardiner et al., 2003). It is possible that the relationship between CesA proteins and microtubules differs in different cell types, or that only a subset of microtubules are involved in directing microfibril deposition. It is clear that there is much we do not understand about the relationship between microtubules and cellulose deposition in different cell types. The use of functional fluorescently labelled CesA proteins (Gardiner et al., 2003; Paradez et al., 2006), however, provides the tools to allow us to answer further questions about the mechanisms responsible for the orientation of the deposition of cellulose, which is crucial for its role in cell walls.

XI. Cellulose as a source of renewable energy

Given the abundance of cellulose, and the fact that it is a rich carbon source, it is obviously attractive as a source of renewable energy. There are a number of ways in which cellulosic biomass could be used as an energy source, including direct combustion for electricity generation, or the release of carbon from cellulose to produce cellulosic ethanol. While this review has focused on our growing understanding of cellulose synthesis in the model plant Arabidopsis, it is clear that this process is highly conserved among plants, meaning that approaches based on knowledge gained from Arabidopsis are likely to be successful in manipulating the cellulose content of biomass crops. It is clear that there are a number of components required for cellulose synthesis, so in order to increase cellulose synthesis it would be necessary to increase the expression of all these proteins. The recent discovery of transcription factors that globally regulate secondary cell wall synthesis in Arabidopsis (Zhong et al., 2006; Mitsuda et al., 2007; Yang et al., 2007) suggests that these may be of use in biomass crops to increase secondary cell wall formation. This would probably be sufficient to increase biomass for combustion to generate electricity, but further delineation of the individual pathways for cellulose, lignin and xylan production will be required before the relative amounts of these polymers in the cell wall can be manipulated. An alternative approach is the use of cellulosic feedstocks to produce bioethanol, which can be used as a fuel. A current major source of bioethanol is corn (maize) grain, but this results in the supply of corn for nutrition becoming scarcer and hence more expensive. The release of carbon from cellulose would overcome this problem, as well as allowing the use of a wide range of feedstocks for ethanol production, including purpose-grown biomass crops and agricultural and forestry waste. The very properties that make cellulose such an important structural polymer in plant cell walls, its insolubility and crystalline nature, along with its association in secondary cell walls with a lignin matrix, make it recalcitrant to carbon release. Manipulating the cellulose as it is deposited, however, may make it easier to hydrolyse. Plant endo-β-1,4-glucanases have until recently been thought to be unable to degrade crystalline cellulose. This is thought to be caused by the absence of a cellulose-binding module. Recently, however, it has been shown that a tomato endo-β-1,4-glucanase, SlCel9C1, does contain a cellulose-binding module and is capable of binding crystalline cellulose and degrading various cellulosic polymers in vitro (Urbanowicz et al., 2007). It will be interesting to know if manipulating the expression of this glucanase results in changes in cellulose synthesis and extractability. It has been shown previously that expression of a bacterial cellulose-binding module in tobacco resulted in plants that accumulated biomass more rapidly than wild-type plants, and this was attributed to the cellulose-binding module somehow interfering with microfibril biosynthesis and crystallization (Shoseyev et al., 2001). The use of cellulosic feedstocks for energy production is very attractive, given that the growth of biomass crops is not as energy-intensive as the growth of food crops. A recent study has suggested that biomass crops can also be produced from agriculturally marginal land with minimal energy input, thus having the potential to provide fuel supplies with greater environmental benefits than either fossil fuels or current food-based biofuels (Hill et al., 2006).

XII. Conclusions

The deposition of cellulose in the plant cell wall is central to plant growth and development. Advances in the past decade, particularly using the model plant Arabidopsis, have led to a great increase in our knowledge of the proteins involved in cellulose synthesis, particularly the cellulose synthase proteins themselves. Despite this, there is still much to learn about cellulose synthesis, especially about how it is regulated at both transcriptional and post-transcriptional levels. The structure of cellulose itself dictates that the orientation of its deposition must be closely regulated for its physical properties to be utilized properly. Recent approaches visualizing moving fluorescently tagged cellulose synthase complexes are starting to answer some of the questions about how the orientation of deposition of cellulose microfibrils is controlled. Resolving some of the unanswered questions about cellulose synthesis is essential if we are to understand fundamental processes in plant development as well as to utilize the vast quantities of sugars contained in plant cell walls as a source of bioenergy in order to combat global climate change. This apparently ‘simple’ polymer has proved a frustrating subject for research, but recent developments have given us hope that one day we may truly understand its synthesis.


Neil Taylor is supported by a Royal Society University Research Fellowship. I would like to thank the anonymous reviewers for their constructive comments.