This Article Abstract Full Text (PDF) Supplemental Data All Versions of this Article: 143/1/339    most recent pp.106.086405v1 Alert me when this article is cited Alert me if a correction is posted Services Email this article to a friend Similar articles in this journal Similar articles in ISI Web of Science Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via CrossRef Citing Articles via ISI Web of Science (4) Citing Articles via Google Scholar Google Scholar Articles by Guillaumie, S. Articles by Goffner, D. Search for Related Content PubMed PubMed Citation Articles by Guillaumie, S. Articles by Goffner, D. Agricola Articles by Guillaumie, S. Articles by Goffner, D.

GENETICS, GENOMICS, AND MOLECULAR EVOLUTION

MAIZEWALL. Database and Developmental Gene Expression Profiling of Cell Wall Biosynthesis and Assembly in Maize ^1,[W]

Université Paul Sabatier, Centre National de la Recherche Scientifique, Unité Mixte de Recherche 5546, 31326 Castanet-Tolosan, France (S.G., H.S.-C., C.D., Y.M., M.P., D.G.); Institut National de la Recherche Agronomique (INRA), Unité de Génétique et d'Amélioration des Plantes Fourragères, 86600 Lusignan, France (S.G., Y.B.); INRA-Institut National Agronomique Paris-Grignon, Unité de Chimie Biologique, 78850 Thiverval-Grignon, France (C.L.); and Biogemma, Campus Universitaire des Cézeaux, 63170 Aubière, France (A.M.)

	ABSTRACT

TOP ABSTRACT RESULTS DISCUSSION MATERIALS AND METHODS LITERATURE CITED

 
An extensive search for maize (Zea mays) genes involved in cell wall biosynthesis and assembly has been performed and 735 sequences have been centralized in a database, MAIZEWALL (http://www.polebio.scsv.ups-tlse.fr/MAIZEWALL). MAIZEWALL contains a bioinformatic analysis for each entry and gene expression data that are accessible via a user-friendly interface. A maize cell wall macroarray composed of a gene-specific tag for each entry was also constructed to monitor global cell wall-related gene expression in different organs and during internode development. By using this macroarray, we identified sets of genes that exhibit organ and internode-stage preferential expression profiles. These data provide a comprehensive fingerprint of cell wall-related gene expression throughout the maize plant. Moreover, an in-depth examination of genes involved in lignin biosynthesis coupled to biochemical and cytological data from different organs and stages of internode development has also been undertaken. These results allow us to trace spatially and developmentally regulated, putative preferential routes of monolignol biosynthesis involving specific gene family members and suggest that, although all of the gene families of the currently accepted monolignol biosynthetic pathway are conserved in maize, there are subtle differences in family size and a high degree of complexity in spatial expression patterns. These differences are in keeping with the diversity of lignified cell types throughout the maize plant.

As a first step in obtaining information about individual cell wall-related gene products and their role in plant growth and development, many genomic approaches have been undertaken. An extensive genomic cell wall database has recently become available in Arabidopsis (Arabidopsis thaliana; Girke et al., 2004). In this Cell Wall Navigator database, more than 5,000 putative cell wall genes coding for enzymes and structural proteins known to be involved in primary wall metabolism have been assembled. It has been predicted that over 100 protein families with over 1,000 members are involved in cell wall-related processes (Henrissat et al., 2001). As for genomic initiatives of secondary walls, large expressed sequence tag (EST) and microarray datasets have been generated for agronomic and model species, including cotton (Gossypium hirsutum) fibers (Arpat et al., 2004), poplar (Populus tremuloides; Hertzberg et al., 2001) and pine (Pinus sylvestris; Pavy et al., 2005) wood, zinnia (Zinnia elegans) TEs (Demura et al., 2002; Milioni et al., 2002; Pesquet et al., 2005), and Arabidopsis (Zhao et al., 2005). Despite the abundance of genomic information now available, the function of the large majority of these genes is still unknown.

As compared to Arabidopsis and dicotyledonous plants in general, cell wall research in monocots and, in particular, maize (Zea mays) has been much less explored. In the case of primary walls, some isolated examples of the role of cell wall enzymes in relation to specific physiological processes have been reported. For example, the role of cell wall-loosening proteins, including expansins and xyloglucan endotransglycosylase/hydrolases (XTHs), in promoting cell elongation during water deficit has been investigated in maize root tips (Wu et al., 1994, 1996). The inventory of cell wall proteins in the elongation zone of maize roots has recently been enlarged through a systematic proteomic approach (Zhu et al., 2006). Interestingly, although many identified proteins are similar to those previously found in dicot primary walls, many maize wall proteins appear to be specific to monocot walls. It should be possible, at least to some extent, to extrapolate information from cell wall genomics in dicot species, but, considering the profound differences in cell wall structure and composition of monocot and dicot cell walls, it is clear that many biosynthetic and remodeling regulatory mechanisms will not be conserved among these phylogenetically distant taxa. The genomic sequence of rice (Oryza sativa) has recently been completed and a comparative approach between cell wall-related gene families in Arabidopsis and rice was undertaken (Yokoyama and Nishitani, 2004). Among the 32 rice cell wall gene families examined, there was a great deal of sequence overlap with cell wall genes in Arabidopsis, but the structural differences in type I (Arabidopsis) versus type II (rice) walls were indeed reflected in certain families. For example, in Arabidopsis pectin is an abundant wall component, whereas in rice it is not. This structural difference was mirrored in the genomic data in that pectin-related gene families were much larger in Arabidopsis than in rice.

At the time of this writing, more than 700,000 ESTs from maize were available. A comparative genomic survey between maize and Arabidopsis revealed that only 60% to 70% of the maize sequences matched with Arabidopsis sequences, indicating a significant proportion of highly diverged or putative maize-specific genes in the maize genome (Brendel et al., 2002). Although global transcriptomic analysis has been undertaken in different physiological contexts, including roots (Poroyko et al., 2005), kernels in response to water stress (Andjelkovic and Thompson, 2006), or organ response to UV light (Casati and Walbot, 2004), an in-depth, systematic characterization of the maize cell wall transcriptome has not been examined (to our knowledge). It should be noted that it is impossible to speak of a maize cell wall per se because the spatial and temporal distribution of wall components is organ, tissue, and cell specific. In general, as stated above, there are several major chemical differences between maize primary cell walls as compared to dicot species (for review, see Carpita and Gibeaut, 1993; Carpita, 1996). Briefly, as is the case in dicots, maize walls contain cellulose microfibrils, but instead of xyloglucan as the main cellulose-tethering molecule, maize is rich in glucuronoarabinoxylans. Maize primary walls are relatively low in pectins, and, based on immunolabeling experiments with antibodies raised against different pectin structures, they appear to be cell type-specific with esterified polygalacturonic acids preferentially localized in vascular tissues, whereas unesterified polygalacturonic acids are mainly found in cortical and parenchyma cells (Knox et al., 1990). The primary walls of maize also contain relatively large amounts of phenolic compounds and little protein in comparison to dicots.

In relation to silage maize digestibility, the lignified secondary cell wall has been extensively studied at the biochemical level (Barrière et al., 2003). However, despite the importance of lignin content and structure in determining forage digestibility, biochemical and molecular regulation of the lignin biosynthetic pathway has been virtually unexplored in maize. That said, expression of a few key genes, including cinnamoyl-CoA reductase (CCR; Pichon et al., 1998), cinnamyl alcohol dehydrogenase (CAD; Halpin et al., 1998), and caffeic acid O-methyltransferase (COMT; Capellades et al., 1996), have been spatially and temporally correlated to lignifying tissues. Beyond correlative gene expression data, the function of only two genes, COMT and CAD, has been unambiguously demonstrated (Vignols et al., 1995; Halpin et al., 1998). A well-characterized, naturally occurring mutant, brown midrib 3 (bm3), is mutated in the COMT gene, resulting in less lignin, a decrease in syringyl (S) units, and improved digestibility (Vignols et al., 1995). COMT maize antisense lines were also generated and have a similar, yet less severe, phenotype (Piquemal et al., 2002; He et al., 2003). Similarly, the bm1 mutant, characterized by altered lignin content and composition, is severely affected in CAD expression (Halpin et al., 1998).

Two specific facts concerning lignification in maize should also be pointed out. First, maize lignin contains relatively high amounts of hydroxycinnamyl (H) units in addition to the guaiacyl (G) and S units typically found in dicotyledonous angiosperms (Lapierre, 1993). This is also the case in most cereals (Lapierre, 1993). Second, maize organs have a wide diversity of lignified cell types. For example, maize internodes are characterized not only by sclerenchyma fibers (located both in the subepidermal cell layers and in close association with vascular bundles) and xylem vessel elements, but also lignified parenchyma cells that represent a very high proportion of the total lignified surface area in older tissues (M. Pichon, unpublished data). Until now, to our knowledge, data have not been made available to integrate knowledge of the specifics of cell wall composition in relation to gene expression in maize.

In this article, we provide a user-friendly, maize cell wall database, MAIZEWALL, containing 735 accessions associated directly or indirectly with primary and secondary wall metabolism. MAIZEWALL is composed of maize homologs resulting from (1) an extensive cell wall-related keyword and BLAST search based on existing knowledge of cell walls in other species; and (2) a BLAST search with ESTs derived from secondary wall-forming in vitro TEs from zinnia (Pesquet et al., 2005). A complete bioinformatic analysis of each gene is provided. A maize cell wall macroarray consisting of gene-specific tags (GSTs), each corresponding to the 3'-untranslated region (UTR) per gene, was constructed. This study provides an organ-specific fingerprint of cell wall-related gene expression in maize. Finally, an in-depth transcriptome analysis of the gene families encoding enzymes of the lignin biosynthetic pathway allowed us to identify putative preferential routes for lignin biosynthesis in different organs and throughout internode development in maize.

	RESULTS

TOP ABSTRACT RESULTS DISCUSSION MATERIALS AND METHODS LITERATURE CITED

 

MAIZEWALL: A Bioinformatic and Gene Expression Database of Cell Wall Genes in Maize

An overview of the strategy used to construct the cell wall gene catalog found in MAIZEWALL is illustrated in Figure 1 . First, a cell wall-related keyword list of nearly 100 words was established based on current knowledge of cell wall synthesis and assembly genes in plants. When available, maize sequences with the appropriate keyword annotation were retrieved from public databases, or, if not, sequences from other plant species were subsequently used as bait to identify the most closely related maize sequences. In this search, we also included genes involved in closely related metabolism (i.e. general phenylpropanoid and shikimic acid pathways) and those controlling vascular patterning that have been identified by the characterization of Arabidopsis mutants (for review, see Scarpella and Meijer, 2004). Second, maize sequences were retrieved based on sequence similarities with zinnia genes expressed during in vitro secondary wall formation (Pesquet et al., 2005). All sequences were then BLASTed against the unannotated maize GénoPlanteInfo (GPI) contig database (Samson et al., 2003) to obtain the corresponding maize contigs. Only contigs with the expected keyword annotation when BLASTed against the public protein databases (SWALL and nonredundant [NR]) were retained. Based on these criteria, 735 contigs were selected as entries for MAIZEWALL (http://www.polebio.scsv.ups-tlse.fr/MAIZEWALL). The 735 contigs belong to 174 putative gene functions, which were further classified into 19 functional categories. The complete cell wall catalog, along with the number of contigs identified for each putative gene function, is found in Supplemental Table S1.

View larger version (10K): [in this window] [in a new window] [as a PowerPoint slide]   Figure 1. Schematic view of the strategy and content of the cell wall gene catalog found in MAIZEWALL.

View larger version (19K): [in this window] [in a new window] [as a PowerPoint slide]   Figure 2. Structure of the MAIZEWALL database. Each box represents a different Web page. Arrows indicate links between pages.

To identify a transcriptional fingerprint of cell wall gene expression in different organs, we compared the global transcript profiles in roots, leaves, and young stems of plants at the four- to five-leaf stage. We first examined the most highly expressed genes, as judged by hybridization signal intensity in each organ (Supplemental Table S2). Data expressed in this manner provide a snapshot of the most highly active metabolic pathways for each organ. Interestingly, when comparing the 30 most highly expressed genes, there is a high degree of overlap among organs (more than one-half), suggesting that the metabolic demands are similar throughout the young maize plant. Not surprisingly, many fall into functional categories related to polysaccharide synthesis. For example, among the 14 cellulose synthase genes spotted on the array, only one contig (QBS7b05.x.g.2.1) is highly expressed in all organs. Along with this cellulose synthase gene, two of the 11 spotted Suc synthase genes are also among the most highly expressed genes, regardless of organ location. These results suggest that the same actors are likely to be important in cellulose synthesis during early maize development. As for hemicelluloses, among the four UDP-Glc-6-dehydrogenase genes, again the same two genes are highly expressed in all organs of young maize plants. These genes encode central enzymes of hemicellulose biosynthesis and appear to be essential for cell wall formation in young organs. Interestingly, some conspicuous differences were observed among the sugar nucleotide-converting enzymes. GDP-Man-4,6-dehydratase (mur4) and UDP-D-Xyl-4-epimerase (mur1) were among the most highly expressed genes uniquely in the aerial portions of the plant (leaves and young stems).

As a further step toward understanding cell wall gene function, we then searched for genes that exhibited differential expression profiles among organs at the four- to five-leaf stage. A gene was considered differentially expressed when its signal intensity was twice that of one or both of the other two organs examined. Of the 651 GSTs spotted on the array, 180 were not expressed in any organ at this stage of development under our hybridization conditions. Among the remaining 471, 43 were differentially expressed (Table I ). Seven genes were preferentially expressed uniquely in young stems and three genes only in roots. Interestingly, there were no cell wall genes exhibiting exclusively leaf-preferential expression. Among the seven genes exhibiting young stem-preferential expression profiles, two of them encoded XTH and three cell wall proteins: a Gly-rich protein (GRP) and two Pro-rich proteins (PRPs). Fourteen were expressed preferentially in both young stems and roots. Among them were four genes encoding enzymes of the phenylpropanoid pathway: two Phe ammonia lyases (PALs), a 4-coumarate:coenzyme A ligase (4CL), and a COMT (Table I). Eight genes were preferentially expressed in the aerial portion of the plant—young stems and leaves. Among them are a mur1 equivalent and two genes of the phenylpropanoid pathway—a ferulate 5-hydroxylase (F5H) and a CAD.

We then searched for genes that were differentially expressed at a given moment during internode development. Among the 651 spotted genes, 133 were differentially expressed in at least one stage (Table II ). As compared to IN6, very few were preferentially expressed uniquely in young stems or IN1. Of the seven genes expressed in young stems, we detected a pectinesterase and a UDP-D-Gal-4-epimerase, suggesting the importance of pectin modification in the early stages of development. In IN6, many genes involved in phenylpropanoid metabolism are preferentially expressed: two PALs (the two that were among the 30 most highly expressed genes), one 4CL, two CCoAOMTs, one COMT, one CCR, and one CAD). There are also two genes of unknown function (contig nos. 3829406.2.1 and 3071483.2.1) with an extremely high degree of specificity in IN6 as indicated by the relative signal values for the three developmental stages in Table II. When comparing IN6 with young stems, it is interesting to note that different classes of transcription factors (three Myb factors and three monopteros genes) and several members of functionally ill-defined families, such as five callose synthases and three chitinase-like genes, are preferentially expressed. In IN1, lignification has presumably slowed down considerably because there are no phenylpropanoid genes that exhibit preferential expression exclusively at this stage.

Despite the economic importance of lignin quantity and quality in dictating certain agronomic traits (Barrière et al., 2003; Méchin et al., 2005), very little is known about the monolignol biosynthetic pathway in maize. By mining the maize databases, we have determined that, as is the case in dicots, most of the monolignol biosynthetic enzymes are encoded by multigene families. The first step in assigning a function to each gene family member is to precisely map out expression patterns for each in developmental fashion. In this way, we provide insight into putative preferred routes of monolignol biosynthesis throughout the maize plant. For each of the monolignol biosynthetic pathway genes, a complete enzyme-by-enzyme survey of developmental expression patterns in different organs and throughout internode development is described below.

PAL/Tyr Ammonia Lyase

PAL catalyzes the first step in the phenylpropanoid pathway by removing ammonia from L-Phe to produce p-coumaric acid. In maize, PAL also has Tyr ammonia lyase (TAL) activity in that the enzyme utilizes Tyr in addition to Phe as substrate (Rösler et al., 1997). In all plants analyzed thus far, PAL is encoded by multigene families. In the complete genome of Arabidopsis, four genes encoding PAL proteins have been identified (Raes et al., 2003). In maize, one PAL/TAL cDNA has been previously reported, but no data on gene number or expression patterns are available (Rösler et al., 1997).

Using the sequence previously described by Rösler et al. (1997), a total of five contigs annotated PAL/TAL were identified with nucleic acid sequence identities ranging from 57% to 84% among them (Supplemental Fig. S4A). Phylogenetic analysis allowed us to define three classes: classes I, II, and III (Fig. 3A ). Class I contains two sequences: the one originally described by Rösler et al. (1997) (contig no. 2161072.2.1) and another with 84% identity at the nucleic acid level (contig no. 2161072.2.3). Class II contains only one member, and class III contains two members that exhibit 77% identity between them. Based uniquely on sequence identity, we were unable to establish an unambiguous relationship with other previously described PAL genes from other species. For example, the class I maize PAL (contig no. 2161072.2.3) exhibited 64% and 66% sequence identity at the protein level with Arabidopsis PAL1 and PAL2, respectively. On the other hand, the same maize contig exhibited 89% and 71% identity at the protein level with rice PAL1 and PAL2, respectively, suggesting that this maize gene is most likely the rice PAL1 putative ortholog (Zhu et al., 1995).

View larger version (24K): [in this window] [in a new window] [as a PowerPoint slide]   Figure 3. Characterization of the PAL gene family in maize. A, Phylogenetic analysis. Classes were defined according to evolutionary distances defined by Phylip 3.5 software. B, Gene expression in organs at the four- to five-leaf stage and IN1 and IN6 at silking. Values indicate the normalized signal hybridization intensity for each gene and (–) signifies a below-background signal intensity of 6,000. C, Relative contribution of each gene to total PAL gene expression. R, Roots; L, leaves, YS, young stems, which results from piled-up internodes at the four- to five-leaf stage. IN6, Internode just below the node bearing the ear at silking; IN1, basal internode at silking. Note that equivalent datasets as found in B and C are available for all gene entries in MAIZEWALL.

Cinnamate 4-Hydroxylase

In conjunction with two other key enzymes of the core phenylpropanoid pathway, PAL and 4CL, cinnamate 4-hydroxylase (C4H) directs carbon flux into phenylpropanoid metabolism. C4H belongs to the CYP73A group of the cytochrome P450 family and catalyzes the first oxidative reaction in phenylpropanoid metabolism, namely, the conversion of trans-cinnamic to p-coumaric acid. This reaction consumes molecular oxygen and a reducing equivalent from NADPH delivered via cytochrome P450 reductase (Meijer et al., 1993). In Arabidopsis, C4H is encoded by one gene, which is expressed in all tissues (Mizutani et al., 1997). In maize, we identified two sequences, C4H1 corresponding to contig number 2521589.2.1 and C4H2 cloned by reverse transcription (RT)-PCR in our laboratory. They exhibit 72% identity between them at the nucleotide level. At the protein level, both C4H1 and C4H2 from maize exhibited 75% identity/86% similarity and 81% identity/91% similarity, respectively, with Arabidopsis C4H. Both genes are expressed in all organs and at all stages of internode development (Table III ). C4H1 (contig no. 251589.2.1) is by far the predominant gene in IN6. C4H2, on the other hand, has a tendency to be more highly expressed in all organs of young plants.

4CL, which catalyzes the formation of CoA esters of p-coumaric acid, caffeic acid, ferulic acid, 5-hydroxyferulic acid, and sinapic acid, plays a pivotal role in channeling phenylpropanoid precursors into different downstream pathways, each leading to a variety of functionally distinct end products (Harding et al., 2002). This probably explains why 4CL is encoded by relatively large multigene families. 4CL isoforms in Arabidopsis have been extensively characterized at the biochemical level (Hamberger and Hahlbrock, 2004; Costa et al., 2005) and their gene expression profiles analyzed (Raes et al., 2003).

To date, only two sequences have been reported in maize (Puigdomenech et al., 2001; S. Sivasankar, D. Sapienza, and T. Helentjaris, unpublished data). Using these maize sequences as bait, we retrieved a total of seven contigs (Fig. 4A ). The percentage of nucleic acid identity among the sequences ranges from 43% to 82% (Supplemental Fig. S4B). Among them, a class I member (contig no. 3071761.2.1) corresponding to the maize sequence isolated by S. Sivasankar, D. Sapienza, and T. Helentjaris (unpublished data) is a putative ortholog to 4CL2 from Arabidopsis (Hamberger and Hahlbrock, 2004) and poplar (Hu et al., 1998).

View larger version (25K): [in this window] [in a new window] [as a PowerPoint slide]   Figure 4. Characterization of the 4CL gene family in maize. A, Phylogenetic analysis. Classes were defined according to evolutionary distances defined by Phylip 3.5 software. Contig numbers 2448387.2.1 and 17163.2.2 annotated as 4CL-like; contig numbers 3106166.2.1 and 1716323.2.1 annotated as putative 4CL. For 4CL2, a contig was not found in the GPI databases. B, Gene expression in organs at the four- to five-leaf stage and internodes at different developmental stages. Values indicate the normalized signal hybridization intensity for each gene and (–) signifies a below-background signal intensity of 6,000. Note that contig numbers 1716323.2.2 and 2448387.2.1 were not included in expression studies because we were unable to amplify corresponding GSTs. C, Relative contribution of each gene to total 4CL gene expression. R, Roots; L, leaves; YS, young stems, which results from piled-up internodes at the four- to five-leaf stage. IN6, Internode just below the node bearing the ear at silking; IN1, basal internode at silking. Note that equivalent datasets as found in B and C are available for all gene entries in MAIZEWALL.

Hydroxycinnamoyl-CoA Transferase

Hydroxycinnamoyl-CoA transferase (HCT) is the most recently identified actor in monolignol biosynthesis and belongs to a large family of acyltransferases (Hoffmann et al., 2003). It catalyzes the conversion of p-coumaroyl-CoA and caffeoyl-CoA to their corresponding shikimate or quinate esters just up and downstream of p-coumarate 3-hydroxylase (C3H). In Arabidopsis, only one gene has been detected in the genome and it is expressed in all tissues investigated, with the strongest expression in inflorescence stems (Raes et al., 2003). In maize, we identified two HCT contigs with 64% identity between them at the nucleotide level. HCT1 (contig no. 2478084.2.1) exhibited 52% identity/63% similarity at the protein level with HCT originally isolated from tobacco (Nicotiana tabacum; Hoffmann et al., 2003). HCT2 (contig no. 2619423.2.1) exhibited 58% identity/68% similarity with the tobacco protein. HCT2 (contig no. 2619423.2.1) is expressed at relatively low levels in all young organs and throughout internode development, whereas HCT1 (contig no. 2478084.2.1) is only expressed in IN6 (Table III).

C3H

It was originally postulated that this enzyme catalyzed the C3 hydroxylation step from p-coumaric to caffeic acid. More recently, it has been shown that C3H preferentially converts the shikimate and quinate esters of p-coumaric acid into their corresponding caffeic acid conjugates (Schoch et al., 2001; Franke et al., 2002). C3H belongs to the CYP98 cytochrome P450 family. In Arabidopsis, three C3H genes have been identified in the genome (Raes et al., 2003). One of them, C3H1, clusters with all known C3Hs from other species, whereas C3H2 and C3H3 are more divergent and constitute a separate class and do not appear to hydroxylate shikimate and quinate esters of p-coumaric acid (Schoch et al., 2001). In maize, we found only one contig (no. 2643622.2.1) that annotated C3H (Table III). The sequence exhibited 65% identity with the C3H1 Arabidopsis gene, with much lower homology to C3H2 and C3H3 (49% identity with each). In maize, C3H exhibited relatively low levels of expression in all organs studied, with slightly higher levels in IN6.

CCoAOMT

CCoAOMT, by catalyzing the methylation of caffeoyl-CoA to feruloyl-CoA and 5-hydroxyferuloyl-CoA to sinapoyl-CoA, plays a pivotal role in determining lignin composition (Zhong et al., 1998; Pinçon et al., 2001). Seven CCoAOMT gene family members with different spatiotemporal gene expression patterns were identified in the Arabidopsis genome (Raes et al., 2003). In maize, two CCoAOMT sequences, CCoAOMT1 and CCoAOMT2, have been previously deposited in public databases (Civardi et al., 1999), but to our knowledge no expression data for these genes have ever been reported. By BLASTing these two sequences in GPI databases, we were able to identify a total of five contigs that fell into three classes (Fig. 5A ). Class I contains two members (contig nos. 2455940.2.1 and 2455940.2.2) that correspond to the previously identified CCoAOMT1 and CCoAOMT2, respectively. They exhibited 83% identity at the nucleotide level (Supplemental Fig. S4C). Class II contains one member (contig no. 2591258.2.1) and class III contains two members (contig nos. 2943966.2.1 and 2430769.2.1). In general, global CCoAOMT expression was similar in all organs studied, except in IN1, where expression was significantly lower (Fig. 5B). In all young organs, although all five genes are constitutively expressed, the class II member (2591258.2.1) is the most highly expressed (Fig. 5, B and C). At silking, although all five genes are still expressed in IN6 and IN1, contig number 2455940.2.2 (CCoAOMT2) and contig number 2430769.2.1 account for a large majority of CCoAOMT expression in IN6. It should be noted that CCoAOMT1 (contig no. 2455940.2.1) is expressed at low levels in all organs examined, suggesting that it is not the major actor in lignin biosynthesis in maize. In Arabidopsis, among the seven putative CCoAOMT genes, based on expression patterns, CCoAOMT1 was considered to be the gene most likely involved in constitutive lignification (Raes et al., 2003). Based on sequence homology alone, it is impossible to determine whether CCoAOMT1 or CCoAOMT2 is the functional maize equivalent of this Arabidopsis gene. In any case, based on expression data, CCoAOMT2 appears to be more important for constitutive lignification in maize as compared to CCoAOMT1.

View larger version (24K): [in this window] [in a new window] [as a PowerPoint slide]   Figure 5. Characterization of the CCoAOMT gene family in maize. A, Phylogenetic analysis. Classes were defined according to evolutionary distances defined by Phylip 3.5 software. Contig numbers 2591258.2.1, 2430769.2.1, and 2943966.2.1 annotated as putative CCoAOMT. B, Gene expression in organs at the four- to five-leaf stage and internodes at different developmental stages. Values indicate the normalized signal hybridization intensity for each gene. C, Relative contribution of each gene to total CCoAOMT gene expression. R, Roots; L, leaves; YS, young stems, which results from piled-up internodes at the four- to five-leaf stage. IN6, Internode just below the node bearing the ear at silking; IN1, basal internode at silking. Note that equivalent datasets as found in B and C are available for all gene entries in MAIZEWALL.

CCR catalyzes the conversion of hydroxycinnamoyl-CoA esters (p-coumaroyl-CoA, feruloyl-CoA, sinapoyl-CoA) into their corresponding cinnamyl aldehydes and is therefore the first committed enzyme of the monolignol pathway. Two full-length cDNAs, ZmCCR1 and ZmCCR2, have previously been cloned and their expression pattern in different organs characterized (Pichon et al., 1998). ZmCCR1 was highly expressed along the stalk, suggesting that the corresponding enzyme was probably involved in constitutive lignification. Using ZmCCR1 to BLAST the GPI database, in addition to ZmCCR1 (contig no. 3012873.2.2), six other contigs annotated as CCR/putative CCR were identified (Fig. 6 ). According to phylogenetic analysis, the seven CCR/putative CCRs formed three classes (Fig. 6A). CCRs exhibited 39% to 59% identity with ZmCCR1 (contig no. 3012873.2.2) and 38% to 83% among all family members (Supplemental Fig. S4D).

View larger version (24K): [in this window] [in a new window] [as a PowerPoint slide]   Figure 6. Characterization of the CCR gene family in maize. A, Phylogenetic analysis. Classes were defined according to evolutionary distances defined by Phylip 3.5 software. Contig numbers 131555.2.2, 4695478.2.1, 2969912.2.1, 3230260.2.1, and 131555.2.3 annotated as putative CCRs. For CCR2, a contig was not found in the GPI database. B, Gene expression in organs at the four- to five-leaf stage and internodes at different developmental stages. Values indicate the normalized signal hybridization intensity for each gene and (–) signifies a below-background signal intensity of 6,000. C, Relative contribution of each gene to total CCR gene expression. R, Roots; L, leaves; YS, young stems, which results from piled-up internodes at the four- to five-leaf stage. IN6, Internode just below the node bearing the ear at silking; IN1, basal internode at silking. Note that equivalent datasets as found in B and C are available for all gene entries in MAIZEWALL.

F5H

F5H, or coniferaldehyde 5-hydroxylase, is a cytochrome P450-dependent monooxygenase (CYP84) and a key enzyme for the production of S-unit lignin in dicotyledonous angiosperms (Humphreys et al., 1999). It catalyzes the 5-hydroxylation of coniferaldehyde and/or coniferyl alcohol. In Arabidopsis, two genes encoding F5H have been identified (Raes et al., 2003). F5H1 is expressed in all tissues, with maximal expression in developing stems, whereas F5H2 appears to be more closely associated with earlier stages of plant development. In maize, there are two F5H genes: F5H1 was previously described by Puigdomenech et al. (2001) and F5H2 was identified in this study. They exhibited 92% identity between them at the nucleotide level. Maize F5H1 and F5H2 exhibit 60% identity/72% similarity and 51% identity/65% similarity, respectively, with the Arabidopsis F5H1 gene at the protein level. Expression analysis indicated that F5H1 was highly expressed in young stems and with the highest expression in leaves. Interestingly, F5H1 transcripts could not be detected in roots (Table III). On the contrary, F5H2 was predominantly expressed in roots and, to a lesser extent, in other organs with the lowest levels in young stems and IN6 (data not shown).

COMT

COMT was originally thought to be a bifunctional enzyme that sequentially methylated caffeic and 5-hydroxyferulic acids. More recently, it has been shown that COMT acts downstream in monolignol biosynthesis by methylating the aldehyde and alcohol backbones (Osakabe et al., 1999; Parvathi et al., 2001). In maize, a single gene encoding COMT has previously been identified (Collazo et al., 1992). Down-regulation of this gene, both in the bm3 mutant (Vignols et al., 1995) and in antisense transgenic maize (Piquemal et al., 2002), led to a drastic decrease in lignin content and S-unit lignin. In the GPI database, we detected one contig (no. 2192909.2.3) that annotated COMT. This contig corresponded to the previously described COMT. In agreement with published expression data (Collazo et al., 1992), this gene was expressed in all organs, with the highest levels in IN6 and the lowest in leaves (Table III).

CAD

CAD catalyzes the reduction of p-hydroxycinnamaldehydes into their corresponding alcohols and is the last enzyme in monolignol biosynthesis. In the Arabidopsis genome, nine putative CAD genes have been identified (Raes et al., 2003). Whether different CAD genes possess different substrate specificity toward coniferyl and sinapyl alcohol and hence dictate, at least to some extent, lignin composition in planta has been the subject to debate (Li et al., 2001; Sibout et al., 2005). Because this issue has not been addressed in monocots (to our knowledge), we chose to classify all contigs annotating CAD, CAD-like, sinapyl alcohol dehydrogenase (SAD), and SAD-like into one CAD family.

In maize, one CAD cDNA had previously been characterized (Halpin et al., 1998). Its expression was also shown to be severely affected in the bm1 mutant, resulting in modified lignin content and structure. Beyond the identification of the previously described CAD (Y13733), searching the GPI databases enabled us to identify six other CAD family contigs (Fig. 7A ). One contig (no. 2405118.2.1) exhibited 92% identity at the nucleotide level with the original CAD gene (Supplemental Fig. S4E). Sequence identities ranged from 39% to 92% among the different family members. Contig numbers 3071507.2.1 and 2949673.2.1 were most similar to the SAD gene originally characterized in poplar and annotated as such (Li et al., 2001). They exhibit, respectively, 51% and 49% identity at the nucleotide level with the poplar SAD gene.

View larger version (24K): [in this window] [in a new window] [as a PowerPoint slide]   Figure 7. Characterization of the CAD gene family in maize. A, Phylogenetic analysis. Classes were defined according to evolutionary distances defined by Phylip 3.5 software. Note that contig number 2485944.3.1 was annotated putative CAD, contigs numbers 3203838.2.1 and 4424417.2.1 were annotated SAD-like, and contig numbers 3071507.2.1 and 2949673.2.1 were annotated SAD. B, Gene expression in organs at the four- to five-leaf stage and internodes at different developmental stages. The values indicate the signal-normalized hybridization intensity for each gene and (–) signifies a below-background signal intensity of 6,000. Note that contig number 2949673.2.1 was not included in CAD expression studies because we were unable to amplify a corresponding GST. C, Relative contribution of each gene to total CAD gene expression. R, Roots; L, leaves; YS, young stems, which results from piled-up internodes at the four- to five-leaf stage. IN6, Internode just below the node bearing the ear at silking; IN1; basal internode at silking. Note that equivalent datasets as found in B and C are available for all gene entries in MAIZEWALL.

Histochemical Analysis Reveals a Large Diversity of Spatially Regulated, Lignified Cell Types in Maize

To associate global cell wall gene expression with cell wall type and composition, cross-sections of the same organs analyzed by transcriptomics were observed (Fig. 8 ). Sections were stained with Maüle reagent, which is classically used to distinguish S-unit (red coloration) and G-unit (brown coloration) lignins (Nakano and Meshitsuka, 1992). In primary roots, three to four cell layers of hypodermal cells just beneath the epidermis exhibited strong staining (Fig. 8, A and B). Interestingly, a closer examination of this region revealed that the outermost cell layer had thin walls that stained brown (G units; see Fig. 8B, thick arrow), whereas the three innermost layers had thick walls that stained red-purple (S units). In the central portion of the root (Fig. 8, A and C), the vascular cylinder is delimited from the cortex by the endodermis, which possesses a thickened cell wall containing suberin and lignin known as the Casparian strip (Zeier et al., 1999). At the four- to five-leaf stage, the Casparian strip is weakly stained with Maüle reagent (Fig. 8C). Two types of xylem vessels can be distinguished based on their diameter and location. Xylem vessels of narrow diameter are intercalated with the phloem strands at the periphery and large diameter vessels that alternate with lignified parenchyma cells, forming an inner ring. Both xylem and lignified parenchyma exhibited red-purple coloration, indicating the presence of substantial amounts of S units. Roots are therefore characterized by a wide variety of lignified, S-rich cells with the exception of a single layer of hypodermal cells just below the epidermis.

View larger version (101K): [in this window] [in a new window] [as a PowerPoint slide]   Figure 8. Histochemical localization of lignin in maize organs at the four- to five-leaf stage and internodes at different stages of development at silking. Sections were stained with Maüle reagent. A to C, Transverse sections of primary roots. A, Overall view of a root section. Red-purple staining in the cells of the hypodermal layers and vascular cylinder indicates the presence of S-unit lignin. B, Close-up of hypodermal layers. The three innermost cell layers stained red-purple and the outermost layer just below the epidermis is stained brown, indicating G-unit lignin (large arrow). C, Close-up of the vascular cylinder. The Casparian strip of the endodermis and two rows of xylem vessels embedded in parenchyma cells are all stained red-purple. D to G, Transverse sections in leaves. D, Leaf blade. E, Midrib. F, Close-up of a vascular bundle from a large vein in the blade. G, Close-up of a vascular bundle from a large vein in the midrib. In all cases, note the yellow-brown coloration of the sclerenchyma cells above the vascular bundle, indicating the absence of S-unit lignin. H, Longitudinal section through the young stem, resulting from piled-up internodes of plants at the four- to five-leaf stage. Black lines indicate the positions of transverse sections in internodal and nodal regions shown in I and J, respectively. I, Transverse section in the first internode. J, Transverse section in the node. Only walls of the protoxylem are stained red. K, General view of a maize plant at the silking stage grown under greenhouse conditions. White lines indicate the positions of transverse sections of IN1 and IN6 shown in L and M, respectively. L, Transverse sections of IN1 showing red staining of all lignified tissues, including sclerenchyma, parenchyma cells, and xylem. M, Transverse sections in IN6. Note that sclerenchyma cells are very weakly stained and parenchyma cells are not at all. en, Endodermis; ph, phloem; p, pith; ep, epidermal cells; hp, hypodermal cell; c, cortex; s, sclerenchyma; x, xylem; lv, large veins; sv, small veins; vb, vascular bundle; lp, lignified parenchyma; px, protoxylem; mx, metaxylem. Magnification bars: B, C, and F, 100 µm; A, D, E, G, L, and M, 200 µm; I and J, 400 µm.

At the four- to five-leaf stage, young stems are formed by the piling up of nodes and internodes that will subsequently elongate during development (Fig. 8, H–J). Staining with Maüle revealed very few lignified cells at this stage. Indeed, only the protoxylem elements stained red-purple in both internodes and nodes (Fig. 8, I and J, respectively). These results suggest that the young stem is an excellent indicator of gene expression associated with the onset of lignification in maize. At silking, internode elongation has ceased and a gradient of lignification exists between the young, upper internodes (IN6) and old, basal internodes (IN1) of the plant (Scobbie et al., 1993). To examine this gradient at the cellular level, IN1 and IN6 were histochemically analyzed (Fig. 8, K–M). In IN6, lignifying cells included the cell layer just below the epidermis and sclerenchyma cells surrounding xylem elements that exhibited a centrifugal red-to-pink color gradient (Fig. 8M). On the contrary, in IN1, a large proportion of the total cell surface area was already lignified. Only a few cell layers of parenchyma cells located just inside the peripheral sclerenchyma did not possess a lignified secondary cell wall. Lignified cell types included xylem vessel elements, sclerenchyma just below the epidermis and surrounding the vascular bundles, and a high proportion of lignified parenchyma between vascular bundles (Fig. 8L). All the lignified cells stained red, indicating the presence of S units. Complementary staining techniques (phloroglucinol and Mirande's reagent) were also used to examine lignified tissues in all organs. These results indicated the same spatial distribution and diversity of lignified cell types throughout the maize plant (data not shown). Together, these results indicate that (1) vascular tissue contributes only a minor fraction to total lignin content throughout the maize plant and (2) IN6 constitutes the ideal organ to study highly active constitutive lignification in adult maize plants. These observations are in perfect agreement with transcriptomic data pointing to the importance of phenylpropanoid metabolism genes at this stage of internode development.

Lignin Structure Varies Greatly throughout the Maize Plant

Plant material used for transcriptomics and histochemistry was also subjected to lignin analysis. Lignin structure was investigated by thioacidolysis, which is an analytical degradation method that proceeds by cleavage of labile -O-4 bonds (Lapierre et al., 1986). The total amount and relative frequency of the H, G, and S units cleaved by thioacidolysis provides an estimate of the amount and composition of these units that are uniquely -O-4 linked. Comparison of H, G, and S distribution among organs at the four- to five-leaf stage revealed striking differences (Table IV ). Root lignin was particularly rich in S units (62.5%), whereas, on the contrary, leaf lignin was particularly rich in G units (72.9%). These results are in good agreement with Maüle histochemical staining. When examining internode development, young stems contain the lowest proportion of S, with slightly more in IN6, followed by IN1. These results suggest that young tissues preferentially accumulate G-unit lignin, whereas S-unit lignin content increases with maturity. This has already been observed in other species (Meyer et al., 1998). Interestingly, H units appear to be more abundant in younger tissues (7.2% in roots, 5.3% in young stems, and 5.4% in IN6) as compared to older tissues (2.2% in IN1). Finally, if one uses thioacidolysis yield (Table IV, H + G + S) as an indicator of lignin content, it may be deduced that IN1 and roots are the most lignified tissues (83 and 53 µmol, respectively), whereas leaves are the least lignified (34 µmol). This is also in agreement with histochemical data. In conclusion, lignin deposition and composition are highly regulated at the spatial level and reflect the extreme diversity of lignified cell types throughout the maize plant.

	DISCUSSION

TOP ABSTRACT RESULTS DISCUSSION MATERIALS AND METHODS LITERATURE CITED

At present, our knowledge of cell wall biosynthesis and assembly, especially in monocotyledons, is fragmentary. When taking into consideration the extent of wall polymer diversity, the multitude of ways they assemble in muro, the high degree of wall specificity in different cell types, and the dynamic changes that take place during development within a given cell, it is predicted that several hundreds or even thousands of genes are required for proper, coordinated wall formation. Cell wall-related genomic and proteomic datasets are now available in herbaceous model species, such as Arabidopsis (Zhao et al., 2005; Jamet et al., 2006) and zinnia (Demura et al., 2002; Milioni et al., 2002; Pesquet et al., 2005), and in woody species, including poplar (Hertzberg et al., 2001) and pine (Pavy et al., 2005). Some Web sites are also dedicated to specific protein families, including expansins (http://www.bio.psu.edu/expansins/index.htm), cellulose synthases and cellulose synthase-like genes (http://cellwall.stanford.edu/cesa/index.shtml), and XTHs (http://labs.plantbio.cornell.edu/XTH/overview.htm). In this respect, a large gap exists for monocot species in that genomic inventories specifically devoted to cell walls are lacking. That said, a high-throughput, Fourier transform infrared-based screen has been set up to identify maize plants with modifications in wall assembly or architecture from the UniformMu maize mutant population developed at the University of Florida (http://www.cellwall.genomics.purdue.edu; Bleecker et al., 2004). In this article, we describe the construction of MAIZEWALL, a repertory of 735 genes complete with in-depth bioinformatic analysis and a transcriptional fingerprint of cell wall-related gene expression throughout the maize plant. We consider MAIZEWALL to be a significant leap forward for maize cell wall research while awaiting the full genome sequence in public databases.

The 735 sequences of MAIZEWALL are divided into 174 families based on known gene annotations. As our aim was to subsequently use this database and the corresponding macroarray for a wide range of applications, including mutant analysis, we deliberately chose to widen the scope of entry annotations to include not only primary and secondary cell wall biosynthetic and assembly genes, but also genes involved in closely related metabolism. To provide the most comprehensive and original overview of gene families in secondary wall formation, we also included maize homologs of cDNAs derived from our zinnia TE differentiation library (Pesquet et al., 2005). The sequences were organized into functional categories, including cellulose synthesis, noncellulosic polysaccharide synthesis, general phenylpropanoid, lignin/lignan, transcription factors, etc. This classification was temporary because an actual proof-of-function for any of the MAIZEWALL entries is extremely rare. In our study, as part of the bioinformatic analysis, the most homologous genes in other species have been provided in the database, but these data are based solely on sequence homology and should not be taken as an indication of equivalent gene function. When the maize genome becomes available, it will be necessary to update this search.

Transcriptome-Based Identification of Developmental Marker Genes of Cell Wall Formation in Maize

Global gene expression analysis in different organs and developmental stages of internode development provided us with a fingerprint of the cell wall transcriptome throughout the maize plant. In this way, we identified (1) individual gene family members that are preferentially expressed in a particular organ or developmental stage, and (2) genes that are spatially and temporally coregulated. When considering young plants, the observation that one-third of the genes spotted on the macroarray are not expressed at all is, in itself, of interest. This implies that the transcriptional needs in the early stages of plant development revolve around a restricted set of metabolic functions. When examining the 30 most abundantly expressed genes in all young organs, it is clear that genes involved in polysaccharide biosynthesis and modification genes are predominant. Regarding cellulose synthesis, a cellulose synthase (QBS7b05.xg.2.1), two isoforms of Suc synthase (contig nos. 3748394.2.2 and 131537.2.202), and three endoglucanases (contig nos. 2493751.2.1, 3713002.2.1, and 2922138.2.1) were identified. Among the 14 cellulose synthase genes spotted on the macroarray, only one gene (QBS7b05.xg.2.1) exhibited consistently high levels of expression, not only throughout young plants but also during later stages of internode development. The expression patterns of 12 cellulose synthase genes (ZmCesA1–12) in maize have been previously reported (Holland et al., 2000; Appenzeller et al., 2004). ZmCesA1 to 9 were considered more closely associated with cellulose synthesis in primary walls, whereas ZmCesA10 to 12 were more closely associated with secondary wall cellulose synthesis. QBS7b05.xg.2.1 shares 95% identity with ZmCesA12. The most homologous Arabidopsis gene is irx3, which has been clearly shown to be associated with secondary wall cellulose biosynthesis (Taylor et al., 1999). Detailed functional analysis of QBS7b05.x.g.2.1 in maize would enable us to determine its role in either primary or secondary wall cellulose metabolism. Similarly, two of the four spotted contigs annotated UDP-Glc dehydrogenase were among the most highly expressed genes in all young tissues (but not in IN6 or IN1). UDP-Glc dehydrogenase is a key enzyme that oxidizes UDP-Glc to UDP-GlcUA and thus directs carbohydrates irreversibly into the cell wall-specific pool of nucleotide sugars. These transcriptomic data allowed us to identify major actors of cell wall metabolism in young maize tissues, some expressed organ preferentially and others expressed ubiquitously.

A completely different transcriptome fingerprint was obtained when examining the 30 most highly and preferentially expressed genes at midstage internode development (IN6). Clearly, the most highly abundant and/or preferentially expressed transcripts in IN6 include genes in secondary cell wall synthesis and, more particularly, certain gene family members of the lignin biosynthetic pathway (three PALs, one C4H, one 4CL, two CCoAOMTs, one CCR, one COMT, one F5H1, and one CAD). These transcriptomic data are in perfect agreement with cytological observations indicating that IN6 is clearly at the onset of lignification. Moreover, these results suggest highly coordinated transcriptional coregulation among many genes along the monolignol biosynthetic pathway. It is interesting to note that Myb and monopteros transcription factors are also preferentially expressed in IN6, making them excellent candidates for transcriptional regulation of lignin biosynthesis. Myb factors have already been shown to play a role in the coordinated regulation of lignin genes in several different plant species (Karpinska et al., 2004; Goicoechea et al., 2005).

Toward a Division of Labor among Family Members of Lignin Biosynthetic Genes in Different Organs

Until now, the known actors involved in dicot lignin biosynthesis have not been studied in any detail in maize, with the exception of a few isolated gene expression studies for COMT (Capellades et al., 1996), CCR (Pichon et al., 1998), and CAD (Halpin et al., 1998). Allelic variations have also been evaluated for two CCoAOMT and COMT genes in 34 lines of maize selected for their varying digestibility (Guillet-Claude et al., 2004). Herein, we combine cytological, chemical, and transcriptomic approaches as a first step in determining the gene family member for each of the 10 enzymatic steps of the monolignol biosynthetic pathway that is most likely involved in lignin formation in different organs and stages of internode development. As a general rule, monolignol biosynthetic enzymes that are encoded by multigene families in Arabidopsis are also encoded by multigene families in maize, although the numbers may vary. A similar observation has already been made between rice and Arabidopsis (Yokoyama and Nishitani, 2004). Although these differences in gene number certainly exist among species, the predicted number of genes for a given function also depends on the bioinformatic parameters used for gene annotation in the different studies. That said, we did find some essential differences in gene number between Arabidopsis and maize. First, C4H and HCT are encoded by single genes in Arabidopsis, whereas we identified two genes for each of these functions in maize. Second, in Arabidopsis, C3H is encoded by three genes, whereas only one was identified in maize. It is possible that other maize C3Hs may be identified when the complete genome sequence becomes available.

In Supplemental Figure S5, we have compiled gene expression data for each of the 10 gene families (based on Figures 3–7 and Table III) to propose the most likely routes of monolignol biosynthesis in young organs and during internode development in maize. This scheme is based exclusively on relative gene expression levels for each family, and, for this reason, we cannot exclude that the most highly expressed genes may be involved in the synthesis of other phenylpropanoid metabolites and not necessarily monolignols.

Is There a Sclerenchyma-Specific Route for G-Unit Lignin Synthesis in Leaves?

Lignin analysis of maize plants at the four- to five-leaf stage indicated that leaves were characterized by a high proportion of G units. Similar results have already been reported for leaves at the flowering stage (Piquemal et al., 2002). Our cytological observations revealed that leaf veins are surrounded by patches of heavily lignified sclerenchyma cells that account for a large majority of lignified cells in leaves. These cells stain yellow-brown with Maüle, thereby explaining the high G-unit lignin content in leaves. To our knowledge, G-rich sclerenchyma fibers have not been described in higher plants. In dicots, it is commonly accepted that xylem vessels contain G units, whereas fibers and parenchyma contain both G- and S-unit lignin, with the latter predominating in fibers (Donaldson, 2001). These results suggest that a novel genetic program leading to G units in sclerenchyma fiber cells may be functional in maize leaves.

When analyzing gene expression data, we noted that some members of the multigene families of the monolignol pathway exhibited proportionally high levels of expression in the leaves. In the case of PAL, among the five contigs, three were expressed in leaves, one of which (contig no. 3858663.2.1) contributed to the large majority of total PAL transcriptional activity in leaves. This class II maize PAL is most homologous to Arabidopsis PAL2. In Arabidopsis, PAL2 is expressed in all organs, although most abundantly in roots and stems (Raes et al., 2003). Based on gene expression data, Arabidopsis PAL2, along with PAL1, was considered to be most likely involved in constitutive lignin biosynthesis in Arabidopsis. Detailed sequ