- Copyright © 2000 American Society of Plant Physiologists
The availability of a nearly complete genome sequence for Arabidopsis has created many novel opportunities to identify, by computational methods, the genes that encode enzymes, which have been difficult to characterize by conventional means. We have used this approach to identify a large family of genes of unknown function that show sequence similarity to cellulose synthase. Our working hypothesis is that these genes encode enzymes that catalyze the synthesis of non-cellulosic polysaccharides (Cutler and Somerville, 1997).
A recent breakthrough in research concerning the biogenesis of plant cell walls was the identification, by genomic methods, of genes encoding cellulose synthase in cotton fibers (Pear et al., 1996;Delmer, 1999). The cotton cellulose synthase genes, now termedCesA1 and CesA2, were identified in a collection of expressed sequence tag (EST) sequences on the basis of weak sequence similarity to genes for cellulose synthase from bacteria. In addition, the genes were expressed at high levels in cotton fibers at the onset of secondary wall synthesis and a purified fragment of one of the corresponding proteins was shown to bind UDP-Glc, the proposed substrate for cellulose biosynthesis. The conclusion that the cottonCesA genes are cellulose synthases is supported by results obtained with two cellulose-deficient Arabidopsis mutants,rsw1 (Arioli et al., 1998) and irx3 (Turner and Somerville, 1997; Taylor et al., 1999). The genes corresponding to the RSW1 and IRX3 loci exhibit a high degree of sequence similarity to the cotton CesA genes and are considered orthologs. Ten full-length CesA genes have been sequenced from Arabidopsis, and there is a genome survey sequence that may indicate one additional family member (Fig.1).
Unrooted, bootstrapped tree of theCesA superfamily. ClustalX (version 1.8) was used to create an alignment of the full-length, publicly available protein sequences that was then bootstrapped (n = 5,000 trials) to create the final tree. Subfamilies are boxed. At, Arabidopsis; Gh, cotton; Le, tomato; Mt, Medicago truncatula; Os, rice; Pt, Populus tremuloides; Pt/Pa, Populus tremula × Populus alba.
It is not known at this time whether other polypeptides are also required for cellulose synthase activity (i.e. the CesA polypeptides may be a component of a multisubunit enzyme complex). Until this matter is resolved we consider it expedient to simply refer to the CesA family members as cellulose synthase. The observation that IXR3(AtCesA7), which is required for secondary wall cellulose synthesis, is in a different branch of the CesA tree than RSW1(AtCesA1), which is required for primary wall synthesis (Fig. 1), may indicate that there is sequence divergence between the enzymes involved in primary and secondary wall synthesis.
Reiterative database searches using the Arabidopsis Rsw1 (AtCesA1) and the cotton CesA polypeptide sequences as the initial query sequences revealed a large superfamily of at least 41CesA-like genes in Arabidopsis. Based on predicted protein sequences, we have grouped these genes into seven clearly distinguishable families (Fig. 1): the CesA family, which includes RSW1 and IRX3 (AtCesA7), and six families of structurally related genes of unknown function designated as the “cellulose synthase-like” genes (CslA,CslB, CslC, CslD, CslE, andCslG). The nomenclature for these families is still under discussion (http://mbclserver.rutgers.edu/CPGN/CelluloseWeb/CesA.proposal.html), so the Csl designation for these genes should be considered temporary and may be revised as the enzymatic function of the members of each family is determined.
All of the members of the cellulose synthase superfamily appear to be integral membrane proteins, with three to six transmembrane domains in the carboxy terminal region of the protein and one or two transmembrane domains in the amino terminal region. It is thought that the CesA proteins are located in the plasma membrane (Delmer, 1999). If the Csl proteins participate in the synthesis of non-cellulosic polysaccharides, they would be expected to be located in the Golgi apparatus. Preliminary analysis of CslB, CslG, and CslE fusions to green fluorescent protein appear to localize to the Golgi (T. Richmond and C. Somerville, unpublished data). Also, immunolocalization studies with an antibody to the CslA protein indicates that this family is localized to the cytoplasm (i.e. the Golgi apparatus) rather than the plasma membrane (N. Sprenger and C. Somerville, unpublished data).
Intron-exon organization is conserved among the CesA,CslB, CslG, and CslE gene families, but not the CslA, CslC, or CslDfamilies (Fig. 2). However, the C-terminus of a subset of the CslD genes is congruent with this organization as well. The CslD gene family is the most similar of the Csl gene families to the CesAfamily (approximately 45% identical at the amino acid level). The gene structure for this family is unusual in that the seven genes for which complete genomic sequence information is available have four different patterns of intron-exon organization. Based on recent thinking about the evolution of intron/exon structure (de Souza et al., 1998), the small number of introns in this family, and their divergent nature, would seem to suggest that this gene family is the oldest in the cellulose synthase superfamily and may predate the CesAfamily.
Comparison of the gene structure of representative genes of the Arabidopsis CesA superfamily. Colored boxes represent exons and the lines connecting them denote introns. Thick vertical black bars indicate predicted transmembrane domains as predicted by HMMTOP (http://www.enzim.hu/hmmtop/). Thin blue bars represent conserved Asp residues, and the thicker gray bar represents the QxxRW domain. Thin lines connecting different genes indicate conserved intron-exon junctions.
All members of the CesA family contain a putative LIM-like Zn-binding domain/RING finger domain in the N-terminal region, which is similar to several putative plant Leu zipper transcription factors (Kawagoe and Delmer, 1997a, 1997b; Arioli et al., 1998). LIM domains are known to mediate protein-to-protein interactions (Bach, 2000), whereas RING finger domains are thought to play a role in ubiquitin-mediated proteolysis (Freemont, 2000). These domains may play a role in mediating CesA function via protein partners or targeted degradation. All of the Csl proteins lack this amino terminus extension, including the CslD family, which contains proteins similar in size to the CesAs.
Although the various CesA and Csl proteins vary in their degree of sequence similarity to one another (TableI), they share several features that have been proposed to be indicative of processive glycosyltransferases (Saxena et al., 1995). All of the CesA and Cslgene products contain a D,D,D,QxxRW motif (Fig. 2), which has been proposed to define the nucleotide sugar-binding domain and the catalytic site of these enzymes. Based on this motif, the proposed topology of these proteins (discussed above), and sequence-based classification, the various members of the Arabidopsis cellulose synthase superfamily appear to belong to family 2 of the inverting nucleotide-diphospho-sugar glycosyltransferases (Campbell et al., 1997) that synthesize repeating β-glycosyl unit structures. To date, this family includes over 500 putative members, including cellulose synthase, chitin synthase, hyaluronan synthase, β-1,3-glucan synthase, and a number of uncharacterized genes from many organisms (Campbell et al., 1997;http://afmb.cnrs-mrs.fr/∼pedro/CAZY/gtf_2.html). The function of the various Csl families is not known, but speculation is that they are responsible for producing some of the other polysaccharides found in plant cell walls and in secretions such as root cap or stylar mucilage (Cutler and Somerville, 1997). Although the D,D,D,QxxRW motif is thought to be indicative of processive β-glycosyltransferases, there is no comparative sequence data available on processive α-glycosyltransferases. Therefore we cannot rule out the possibility that some of these enzymes produce polysaccharides with α-linkages, such as rhamnogalacturonan I or rhamnogalacturonan II. It is possible that linkage specificity is determined by subtle features in the active site of the proteins (Stasinopoulos et al., 1999) and that members of the Arabidopsis cellulose synthase superfamily make polysaccharides with both β- and α-linkages.
Identity/similarity matrix for selected members of the CesA superfamily
DISCUSSION
With six families of Csl genes and six major non-cellulosic polysaccharides in Arabidopsis (i.e. callose, xyloglucan, glucuronoarabinoxylan, homogalacturonan, rhamnogalacturonan I, and rhamnogalacturonan II), it is tempting to speculate that each family is responsible for the biosynthesis of one of the principal polysaccharides of the cell wall. Although we consider it possible that the gene superfamily described here encodes enzymes that catalyze the synthesis of different polymers, there is at present no evidence for this other than the observation that sequence divergence is frequently associated with functional divergence. It is also possible that there are additional functional divisions within the gene families that are not evident from our analysis. Recent results concerning the relationship between enzyme structure and function, such as experiments showing that as few as four amino acid changes can alter the catalytic outcome of an enzymatic reaction from desaturation to hydroxylation (Broun et al., 1998), emphasize the need for caution in using sequence similarity to infer function based on sequence.
The amount of plant genome sequence and EST information in the public sequence databases is expanding rapidly. At present there are more than 900,000 plant ESTs and genome survey sequences in GenBank, most of which are from 35 species. In the first 8 months of the year 2000, more than 516,000 new ESTs and genome survey sequences from 16 plant species were deposited. Thus except for species such as Arabidopsis, which will soon be completely sequenced, any attempt at a comprehensive compilation of CesA-related sequence information represents a continuing challenge. To facilitate research on these genes, we have established a website (http://cellwall.stanford.edu) that summarizes the ever-increasing number of cellulose synthase and cellulose synthase-like genes. At present, there are more than 1,250CesA and Csl sequences, from 29 different plant species in GenBank. Although the most extensive information available is for Arabidopsis where there are more than 330 partial or complete gene sequences, there is also a significant amount of information available for several other species, especially rice, maize, soybean, and tomato. A crude estimate of the relative abundance of mRNA for the various family members can be calculated from the frequency with which each gene family is represented by EST sequences in the public databases (Fig. 3).
Relative abundance of EST sequences for members of the CesA and Csl families in GenBank.
Polysaccharides found in other plant species, but not in Arabidopsis (Zablackis et al., 1995), such as mixed linkage xylans, mannans, or arabinans, may be synthesized by genes that are not represented by orthologs in Arabidopsis. A number of gene sequences from plants in GenBank show limited similarity (<50% identity) to the members of the various Csl families in Arabidopsis. This and other issues will undoubtedly become more transparent when the function of theCsl genes in Arabidopsis is known from direct experimental evidence. Our laboratory, along with others, is examining the patterns of gene expression and protein localization of the ArabidopsisCsl genes, and attempting to characterize their enzymatic function using reverse genetics. We are confident that in the next several years the function of these genes will be understood and it will then be possible to begin to unravel the challenge of understanding how cell wall composition and deposition is controlled.
Footnotes
- Received May 25, 2000.
- Accepted July 7, 2000.
LITERATURE CITED
