Molecular analysis of a family of Arabidopsis genes related to galacturonosyltransferases.

We are studying a Galacturonosyltransferase-Like (GATL) gene family in Arabidopsis (Arabidopsis thaliana) that was identified bioinformatically as being closely related to a group of 15 genes (Galacturonosyltransferase1 [GAUT1] to -15), one of which (GAUT1) has been shown to encode a functional galacturonosyltransferase. Here, we describe the phylogeny, gene structure, evolutionary history, genomic organization, protein topology, and expression pattern of this gene family in Arabidopsis. Expression studies (reverse transcription-polymerase chain reaction) demonstrate that all 10 AtGATL genes are transcribed, albeit to varying degrees, in Arabidopsis tissues. Promoter::β-glucuronidase expression studies show that individual AtGATL gene family members have both overlapping and unique expression patterns. Nine of the 10 AtGATL genes are expressed in all major plant organs, although not always in all cell types of those organs. AtGATL4 expression appears to be confined to pollen grains. Most of the AtGATL genes are expressed strongly in vascular tissue in both the stem and hypocotyl. Subcellular localization studies of several GATL proteins using yellow fluorescent protein tagging provide evidence supporting the Golgi localization of these proteins. Plants carrying T-DNA insertions in three AtGATL genes (atgatl3, atgatl6, and atgatl9) have reduced amounts of GalA in their stem cell walls. The xylose content increased in atgatl3 and atgatl6 stem walls. Glycome profiling of cell wall fractions from these mutants using a toolkit of diverse plant glycan-directed monoclonal antibodies showed that the mutations affect both pectins and hemicelluloses and alter overall wall structure, as indicated by altered epitope extractability patterns. The data presented suggest that the AtGATL genes encode proteins involved in cell wall biosynthesis, but their precise roles in wall biosynthesis remain to be substantiated.

Plant cell walls are composed mostly of networks of polysaccharides, primarily cellulose, pectins, and hemicelluloses. The synthesis of these polysaccharides requires a significant commitment of the plant's genomic resources; perhaps as many as 10% of genes in Arabidopsis (Arabidopsis thaliana) have been estimated to be involved in plant cell wall synthesis, maintenance, modification, and degradation (Carpita et al., 2001). Many of these genes belong to multigene families (Henrissat et al., 2001) whose individual members have distinct patterns of expression among plant cells and tissues (Taylor et al., 1999(Taylor et al., , 2000Fagard et al., 2000;Peng et al., 2000;Sarria et al., 2001;Williamson et al., 2001;Orfila et al., 2005;Burton et al., 2006;Harholt et al., 2006;Peñ a et al., 2007;Persson et al., 2007). At least 50 glycosyltransferases (GTs) are predicted to be required for pectin synthesis (Ridley et al., 2001). Recently, several putative GTs that may be involved in the biosynthesis of different pectins have been identified using mutational and/or biochemical approaches. These include three members from the CAZy (Cantarel et al., 2009) family GT47, two members from family GT77, and two members from family GT8. Specifically, these genes are: ARAD1 (GT47), which is believed to encode an a-L-arabinosyltransferase involved in the synthesis of arabinan side chains of rhamnogalacturonan I (RG-I; Harholt et al., 2006); XGD1 (GT47), which encodes a b-(1,3)-xylosyltransferase possibly involved in xylogalacturonan synthesis (Jensen et al., 2008); NpGUT1 (GT47), involved in RG-II side chain synthesis in tobacco (Nicotiana plumbaginifolia) as a putative glucuronosyltransferase (Iwai et al., 2002), although mutants in orthologous genes in Arabidopsis evidence xylan defects (Brown et al., 2009;Séveno et al., 2009;Wu et al., 2009); RGXT1 and RGXT2 genes (GT77), which may participate in the synthesis of side chain A of RG-II, as (1,3)-a-D-xylosyltransferases (Egelund et al., 2006); QUA1 (GT8), the mutation of which results in a dwarf phenotype, reduced cell adhesion, a 25% reduction in the amounts of GalA in the leaves, and slightly lower levels of Xyl and xylosyltransferase activity (Bouton et al., 2002;Orfila et al., 2005); and Galacturonosyltransferase1 (GAUT1 [GT8]), which is involved in homogalacturonan (HG) synthesis as a (1,4)-a-D-galacturonosyltransferase (Sterling et al., 2006). Other studies have linked at least two other members of the GAUT1-related gene family (IRX8/GAUT12 and PARVUS/AtGATL1) to the synthesis of a specific subfraction of pectin (Persson et al., 2007) and/or to the synthesis of the GalA-containing tetrasaccharide sequence located at the reducing end of dicot and gymnosperm xylans (Brown et al., 2007(Brown et al., , 2009Peñ a et al., 2007;Persson et al., 2007).
Among these previously identified genes, however, GAUT1 is the only galacturonosyltransferase that has been functionally determined to be involved in pectin synthesis (Mohnen, 2008). Bioinformatic analysis of the Arabidopsis genome identified 24 other genes with high sequence similarity to GAUT1 (Sterling et al., 2006). Sequence alignment and phylogenetic analysis of the Arabidopsis GAUT1-related gene family separates them into the GAUT and Galacturonosyltransferase-Like (GATL) families. There are 15 GAUT genes in Arabidopsis that encode proteins predicted to be 61 to 78 kD, whereas the 10 Arabidopsis GATL genes encode proteins that have molecular masses between 39 and 44 kD (Sterling et al., 2006).
The absence of functional characterization of most members of the GAUT1-related gene family leaves open the question of the roles of these genes and their encoded proteins in cell wall synthesis. A recent analysis (Caffall et al., 2009) of cell walls isolated from homozygous mutants of 12 members of the GAUT family demonstrated that mutations in eight of these genes resulted in discernible changes in cell wall monosaccharide composition. Mutations in AtGAUT6 result in a reduction in GalA that coincides with higher levels of Xyl and Rha in the wall, and preliminary results suggested a role for AtGAUT6 in HG synthesis. Mutations in AtGAUT9, -10, -11, and -12 resulted in significant reductions in GalA content without decreases in Xyl content. Mutations in AtGAUT13 and AtGAUT14 resulted in increased GalA and Gal content coinciding with reduced Xyl and Rha content compared with the wild type. The results of this study reinforce the hypothesis that the proteins encoded by the AtGAUT genes are involved in pectin and/or xylan synthesis, although further work is necessary to validate this hypothesis. We report here on a study of the GATL family in Arabidopsis to lay the foundation for functional characterization of these putative GTs and identification of their role(s) in plant cell wall biosynthesis.

Gene Structure and Phylogenetic Relationship of the GATL Family
A previous study had identified 10 GATL genes within the Arabidopsis GAUT1-related gene family that have high sequence similarity (Sterling et al., 2006). Pairwise comparisons of the amino acid sequences of the whole coding regions of these proteins showed between 56% and 84% identity and between 71% and 89% similarity ( Table I). The initial analysis of the GATL phylogeny in Arabidopsis did not yield a statistically robust subclade structure for this GT8 subfamily (Sterling et al., 2006). Subsequent inclusion of additional GATL protein sequences from eight additional fully sequenced plant genomes resolved the GATL subfamily into six subclades with robust statistical support (Yin et al., 2010). This substructure of the GATL tree remains largely unchanged by the addition of sequences from three additional sequenced plant genomes (maize [Zea mays], Medicago truncatula, and papaya [Carica papaya]). In this analysis, an additional clade (GATL-c) that contains only monocot genes was resolved with high statistical support (Fig.  1A). The phylogram shows a basal clade (GATL-g) of GATLs from Physcomitrella patens and Selaginella moellendorfii and six clades of angiosperm GATLs. The functional significance of the GATL subclades remains to be determined. All but one of the angiosperm GATL subclades have both monocot and dicot representatives, suggesting that the divergence of the GATL  Table  S1). The phylogenetic reconstruction was carried out using PhyML as described in "Materials and Methods." The six genes in GATL-g were selected as the outgroup to root the phylogeny. Selected supporting values for nodes greater than 70% are shown. The phylogeny is displayed using the Interactive Tree of Life Web server (Letunic and Bork, 2007 family occurred before the evolutionary split between monocots and dicots. As several of the AtGATL proteins appeared as pairs of paralogs in the phylogenetic tree (Fig. 1A), we investigated whether traceable genome duplication events contributed to the complexity of the GATL family in Arabidopsis. The chromosomal locations of AtGATL genes in relation to the segmental duplication history of these regions were analyzed (Blanc and Wolfe, 2004). The resulting chromosome map (Fig. 1B) shows that AtGATL genes are only present on chromosomes I, III, and IV. According to the map, AtGATL8/ AtGATL9 and AtGATL5/AtGATL6 constitute pairs of paralogous genes evolved from recent segmental duplication events. The AtGATL1/AtGATL2, AtGATL9/ AtGATL10, and AtGATL5/AtGATL7 pairs evolved from older segmental duplication events with good statistical significance. The duplication history of AtGATL genes supports the observed close phylogenetic associations between some members of the AtGATL family. For AtGATL3 and AtGATL4, no traceable duplication history could be found, even though their genome positions are close to duplicated blocks.
The AtGATL genes are similar not only in terms of their primary sequence and the size of their encoded proteins but also in their intron/exon organization (Fig. 1C). Of the 10 AtGATL genes, only AtGATL5 and AtGATL6, which appear as paralogs, contain one intron at the same position in their 3# untranslated region. No introns are present in any of the other AtGATLs. The common intron/exon organization shared by these AtGATL genes supports the results from the phylogenetic analysis and the duplication history.

Expression Profiles of the AtGATL Gene Family by Reverse Transcription-PCR
To understand the roles of the AtGATL genes in plant development, we conducted semiquantitative reverse transcription (RT)-PCR analysis to determine the tissue-specific expression patterns for all AtGATL genes in different tissues taken from 7-week-old plants (Fig. 2). Analyzing different parts of the inflorescence stem makes it possible to monitor primary cell wall and secondary cell wall synthesis in a single stem simply by sectioning at different distances from the apical meristem, because secondary cell wall formation increases from the top to the base of the stem (Turner and Somerville, 1997). An ACTIN gene served as a positive control because it is expressed ubiquitously in all organs.
All 10 AtGATL genes are expressed, but at different levels and with different tissue distribution patterns (Fig. 2). The majority of the AtGATL genes are expressed at various levels in all of the different organs analyzed. However, AtGATL4 is expressed only in flowers. AtGATL2 expression is primarily limited to leaves, upper and lower stems, and, at a low level, in roots. AtGATL6 is highly expressed in upper and middle stems but not in lower stems, suggesting that this gene may be involved in primary cell wall synthesis or the initiation of secondary cell wall synthesis.

Expression Profiles of the AtGATL Proteins
To extend the observations made with the RT-PCR analyses, we generated promoter::GUS reporter lines in a wild-type Columbia background for each of the AtGATL genes by fusing the GUS gene (Jefferson et al., 1986;Jefferson, 1987) to upstream (promoter and 5# untranslated regions) and downstream (3# untranslated regions) sequences of each of the AtGATL genes (AtGATL::GUS). Histochemical staining analyses of GUS activity in the transgenic plants revealed expression patterns that were consistent with the outcome of the RT-PCR analyses done at the whole-tissue level, but with higher resolution (Figs. 3 and 4). For Semiquantitative RT-PCR of total RNA isolated from siliques (S), flowers (F), leaves (L), roots (R), upper stems (Us), middle stems (Ms), and lower stems (Ls) was used to assess AtGATL transcript levels in tissues of 7-week-old Arabidopsis plants. The ACTIN2 gene (At3g18780) was used as a control. All AtGATL gene amplification reactions were carried out for 35 PCR cycles, and the ACTIN gene was amplified for 30 cycles. Whole seedlings were cleared and stained for GUS activity (blue coloring). The insets in the AtGATL1, -2, -7, -8, and -10 seedling panels show intense staining of trichomes, and the inset in the AtGATL4 single flower panel shows staining in pollen grains and elongating pollen tubes. A summary of the observed AtGATL expression patterns is provided in Supplemental Table S2.
example, expression of the GUS reporter driven by the AtGATL4 promoter was restricted to flowers, in agreement with the RT-PCR analysis, but was confined to pollen grains and elongating pollen tubes. The tissue/ organ-specific GUS expression profiles of all AtGATLs are summarized below.

10-d-Old Seedlings
Expression of the GUS reporter gene constructs in 10-d-old seedlings was detected in the cotyledons, primary leaves, shoot apices, and roots, depending on the AtGATL gene examined, although frequently only in specific cell types within those tissues. As shown in Figure 3, AtGATL1, -3, -7, -8, and -10 are expressed in cotyledons and primary leaves, whereas AtGATL5 and -9 are only detected in primary leaves. A close examination of primary leaves shows that AtGATL1 and -5 are only expressed in the vasculature of primary leaves. Most AtGATLs, except for AtGATL3, -4, -5, and -9, are also expressed in the trichomes, which form complex secondary walls (Marks et al., 2008). Interestingly, AtGATL1, -2, -3, -5, -6, and -8 are also expressed in the shoot meristem, where only primary wall synthesis would be expected. All AtGATL reporters, with the exception of AtGATL4, are expressed in roots, but with distinct expression patterns for each gene. AtGATL2 is expressed in root tips, developing lateral root meristems, and at the base of extended lateral roots. AtGATL1 and AtGATL3 are expressed in root tips and portions of older roots. The restricted expression pattern of these two GATLs might explain why no expression of these genes could be detected in RT-PCR analyses of whole roots. Both AtGATL6 and AtGATL9 are only expressed in root tips. In addition, AtGATL5, -7, -8, and -10 are strongly expressed throughout the root, but only AtGATL8 and AtGATL10 are expressed in root hairs.

Developing Flowers
Quite diverse patterns of expression were observed in developing flowers among the various AtGATL gene family members (Fig. 3). All AtGATL family members, except for AtGATL2, are expressed in a variety of floral organs, which is consistent with the RT-PCR analysis. AtGATL5 expression in flowers can be detected only in filaments, where AtGATL1 and At-GATL10 expression can also be detected. AtGATL1, -4, and -7 are expressed in the anthers. Expression of AtGATL4 was also observed in elongating pollen tubes (Fig. 3,inset). Carpel tissue expresses AtGATL3, -4, -6, and -10 in the stigma and AtGATL3 in the style. AtGATL3, -8, -9, and -10 are expressed in the sepals of the flower, whereas expression of the AtGATL3, -9, and -10 genes is detected in petals.

Stems and Hypocotyls
The RT-PCR data documented that all AtGATL gene family members, except for AtGATL4, are highly ex-  Table S3.
pressed in inflorescence stems (Fig. 2). To determine in which cells within the stems these genes are expressed, hypocotyls and upper, middle, and lower parts of inflorescence stems were sectioned, stained, and analyzed. Overall, AtGATLs show overlapping, but not identical, expression patterns in stem and hypocotyl. Figure 4 shows that expression of many of the AtGATL genes is restricted primarily to the vascular tissues, although the patterns of expression vary depending on the AtGATL gene. The simplest expression pattern was observed for AtGATL5 and AtGATL10, whose expression could be detected only in xylem and phloem, respectively, in different parts of stems and hypocotyls. AtGATL6 is expressed in the phloem of the upper stem and in the phloem and developing secondary xylem of the hypocotyl (Fig. 4). In contrast, AtGATL8 expression varies a lot in different parts of stems and is expressed in almost all vascular tissues except for fibers and secondary xylem. It is noteworthy that AtGATL1 was specifically expressed in cells undergoing secondary wall thickening, including interfascicular fibers and metaxylem in stems and secondary xylem in hypocotyls. The AtGATL1 expression pattern observed here is consistent with the finding that this gene is specifically associated with secondary wall thickening in fibers and vessels (Brown et al., 2007;Lee et al., 2007). Interestingly, AtGATL2 shows an overlapping expression pattern with AtGATL1 in all parts of the stem, except that there is no expression in fibers. In hypocotyls, AtGATL2 can only be detected in the developing secondary xylem. AtGATL3 is specifically expressed in the epidermis, cortex, and phloem of stems and hypocotyls. AtGATL7 is detected in the phloem of lower stems, the cortex of hypocotyls, upper and middle stems, and the epidermis of upper and middle stems. Finally, AtGATL9 is highly expressed in the protoxylem of all parts of the stem and the cortex of upper stems and hypocotyls. The overall expression patterns in young seedlings are summarized in Supplemental Table S3.

Subcellular Localization of the AtGATL Proteins
The topology of AtGATL proteins was analyzed using the plant membrane protein database, Aramemnon (http://aramemnon.botanik.uni-koeln.de/index. ep). The overall results of these analyses are shown in  AtGATL transcription levels in homozygous mutants of atgatl3, atgatl5, atgatl6, atgatl8, and atgatl9. RT-PCR of total RNAs isolated from stems of 7-week-old homozygous atgatl mutant lines was performed using gene-specific primers (Supplemental Table S6). The ACTIN2 gene (At3g18780) was used as a control. All AtGATL gene amplification reactions were carried out for 35 PCR cycles, and the ACTIN gene was amplified for 30 cycles. WT, Wild type.
Supplemental Table S4. All AtGATLs were predicted to be soluble proteins by this analysis. However, the Aramemnon database strongly predicts that all GATL proteins are targeted to the secretory pathway, with at least half of the 18 programs used in Aramemnon predicting an N-terminal hydrophobic signal peptide domain for all of the GATL proteins.
To establish the subcellular localization of the AtGATL proteins experimentally, we chose four AtGATL isoforms, AtGATL2, AtGATL3, AtGATL7, and AtGATL9, belonging to four different GATL subclades (GATL-d, GATL-e, GATL-b, and GATL-a; Fig.  1A), and generated fluorescently tagged fusion proteins by fusing enhanced yellow fluorescent protein (EYFP) to the C terminus of the full-length AtGATL proteins. The recombinant constructs were transiently coexpressed in Nicotiana benthamiana leaf epidermal cells with Gmct-ECFP (for enhanced cyan fluorescent protein, a Golgi marker; Saint-Jore- Dupas et al., 2006;Nelson et al., 2007) or ECFP-WAK2-HDEL (endoplasmic reticulum marker; Nelson et al., 2007) constructs, respectively. Confocal microscopy was used to determine the subcellular localization of the recombinant AtGATLs. EYFP-tagged AtGATL3 showed a punctate localization pattern when expressed in tobacco leaf epidermal cells (Fig. 5C). Colocalization experiments revealed that the localization pattern of AtGATL3-EYFP is identical to that of Gmct-ECFP (Fig. 5, B and D), which was previously shown to be localized in the Golgi (Saint-Jore- Dupas et al., 2006;Nelson et al., 2007). Together, these results demonstrate that At-GATL3 is a Golgi-localized protein. Similar subcellular localizations were also observed for AtGATL2, At-GATL7, and AtGATL9 proteins (data not shown). The Golgi localization of the AtGATL proteins is consistent with their possible role(s) in the biosynthesis of noncellulosic polysaccharides, which occurs in the Golgi (Nebenfü hr and Staehelin, 2001).

Identification of T-DNA-Tagged Mutants
To obtain homozygous mutant plants with disruptions in the AtGATL genes, we screened T-DNA insertion lines available from the Salk Institute (http://signal.salk. edu/cgi-bin/tdnaexpress) through the Arabidopsis Biological Resource Center (for details, see "Materials and Methods"). Homozygous lines with an insertion in the exon or within the 5# or 3# noncoding regions were obtained for AtGATL3, -5, -6, -8, and -9 (Fig. 6A). RT-PCR of total RNAs isolated from homozygous atgatl mutant lines allowed us to identify four knockout mutants (atgatl5, atgatl6, atgatl8, and atgatl9) and one knockdown mutant (atgatl3; Fig. 6B). No discernible phenotypic changes were observed in the growth or morphology of any of the mutant plants compared with the wild type under normal growth conditions.

Cell Wall Composition of Five atgatl Mutants
The amino acid sequences of the AtGATLs contain domains characteristic of GTs (Sterling et al., 2006;Yin et al., 2010), suggesting that these proteins may be involved in cell wall polysaccharide biosynthesis. To determine whether disruptions of the five AtGATL genes caused alterations in the monosaccharide composition of total cell wall material, the relative amounts of neutral monosaccharides and GalA were determined for wide-type and five atgatl mutant plants. Stems from 10-week-old plants were chosen for this analysis because this tissue can be harvested in large quantities and all four of the disrupted AtGATL genes are highly expressed in stems, based both on RT-PCR (Fig. 2) and promoter::GUS results (Fig. 4). Compared with wide-type plants, atgatl3, atgatl6, and atgatl9 mutants showed 23.7%, 16.4%, and 20.5% reduction in GalA content, respectively, which was counterbalanced by increases in Xyl in atgatl3 and atgatl6 (Table II). The decrease in GalA content of atgatl3, atgatl6, and atgatl9 supports the argument that, like GAUT1 and QUA1, two other members of the GT8 family, the three AtGATL genes function as putative galacturonosyl transferases involved in pectin synthesis. For atgatl5 and atgatl8, no significant changes in sugar content were observed relative to wild-type plants.
To obtain a more complete picture of possible changes in cell wall composition and structure resulting from mutations in the five GATL genes, we analyzed stem cell walls from the mutants using glycome profiling. This method (S. Pattathil and M.G. Hahn, unpublished data) involves ELISA-based screening of sequential extracts prepared from the plant cell walls Table II. Glycosyl residue compositions of cell walls isolated from stems of wild-type and atgatl plants The glycosyl residue compositions are for cell walls isolated from inflorescence stems of 10-week-old Arabidopsis plants as determined by GC-MS of trimethyl silyl derivatives as described in "Materials and Methods." Data are mol % 6 SE of three independent analyses. Glycosyl residues are abbreviated as Ara, Rha, Fuc, Xyl, Man, Gal, Glc, and GalUA (GalA). . Glycome profiling of sequential stem wall extracts prepared from five atgatl mutants and wild-type (WT) plants. Sequential stem cell wall extracts of 10-week-old atgatl and wild-type plants were generated as described in "Materials and Methods." The presence of cell wall glycan epitopes in each extract was determined by ELISAs using 150 glycan-directed monoclonal antibodies , and the data are presented as heat maps. Reagents used for extracting stem materials are indicated on the top of each column. The panel on the right lists the array of antibodies used (left-hand side) grouped according to the principal cell wall glycan (right-hand side) recognized by the antibodies. The colored outlines highlight changes in glycome profiles in the mutant walls compared with the wild type: green outlines, xyloglucan epitopes; blue outlines, pectin and arabinogalactan epitopes; red outlines, xylan epitopes. The yellow-black scale indicates the strength of the ELISA signal: bright yellow color depicts strongest binding, and black color indicates no binding. Col., Ecotype Columbia.
using a toolkit of approximately 150 plant cell wall glycan-directed monoclonal antibodies that recognize diverse epitopes present on most major classes of plant cell wall polysaccharides, including xyloglucans, xylans, pectins, and arabinogalactans . Glycome profiling gives information about changes both in the nature of the epitopes present in the cell walls and in the extractability of those epitopes from the walls (Zhu et al., 2010), the latter providing some information about larger scale changes in wall structure. Glycome profiles of the five atgatl mutant walls show subtle differences, primarily in epitope extractability patterns, when compared with the glycome profile of wild-type cell walls (Fig. 7). For example, the oxalate extracts of atgatl5 and atgatl6 stems contain easily detectable levels of xyloglucan epitopes, whereas the oxalate extract of wild-type walls show no detectable xyloglucan. There are also subtle differences in the levels of arabinogalactan epitopes in the 1 M KOH extracts of atgatl3, atgatl5, and atgatl9 stem walls compared with the equivalent extract of wildtype stem walls. Lastly, subtle differences in the extractability patterns of xylan epitopes are observable in the glycome profiles of all five atgatl mutant stem walls compared with the profile of wild-type stem walls.

DISCUSSION
The GAUT1-related gene family in Arabidopsis consists of 25 genes, all of which encode proteins belonging to CAZy GT8 family. Phylogenetically, this gene family splits into two clades of related genes, the GAUT clade of 15 genes, one of which, AtGAUT1, has been shown to encode a functional HG galacturonosyltransferase (Sterling et al., 2006), and the GATL clade of 10 genes. The proteins encoded by both clades contain several conserved amino acid sequences that are unique to these clades within the larger GT8 family (Yin et al., 2010). However, AtGATLs and AtGAUTs also differ from one another in a couple of ways. These differing characteristics raise questions about the possible role(s) of AtGATL proteins in cell wall glycan synthesis.
AtGATL proteins are smaller and lack an obvious transmembrane domain such as that present in the AtGAUTs. Nonetheless, AtGATL proteins do appear to contain a hydrophobic signal peptide at their N termini, which would be expected to target AtGATL proteins to the endomembrane system. Previous studies had shown that AtGATL1 is retained within the endomembrane system, specifically in the endoplasmic reticulum and Golgi compartments (Kong et al., 2009;Lee et al., 2009). We demonstrate here that four other AtGATL proteins, belonging to four different subclades within the GATL subfamily, colocalize with a Golgi marker (Fig. 5). Thus, the localization of these AtGATL proteins places them in a position within the cell to participate in plant cell wall matrix polysaccharide synthesis, which is thought to take place in the Golgi (Scheible and Pauly, 2004;Sandhu et al., 2009). However, AtGATL proteins contain no transmembrane domain, nor do they contain any other known signal sequences that would retain them within the Golgi. Thus, these AtGATL proteins must be retained within the Golgi by another mechanism. There is precedence for transmembrane domain-independent Golgi localization. For example, reversibly glycosylated polypeptide showed cytoplasmic and transient association with the Golgi, yet it does not possess any signal sequence (Sagi et al., 2005;Drakakaki et al., 2006). We hypothesize that the AtGATL proteins are synthesized in the cytosol and retained in the Golgi by virtue of their interaction with proteins that are membrane embedded or anchored and that carry Golgi retention signals. This hypothesis awaits further experimental investigation, particularly with respect to possible protein-protein interactions involving AtGAUT proteins, which have transmembrane domains and are localized to the Golgi (Sterling et al., 2001).
The AtGATL proteins show a significantly higher sequence identity/similarity (Table I) than do the AtGAUTs (Sterling et al., 2006;Caffall et al., 2009). For example, AtGATL5 and AtGATL6 share 77% identity to each other, and AtGATL8 and AtGATL9 are 84% identical in their coding regions. Such high sequence identity might be indicative of functional redundancy among these sets of AtGATL family members. The phylogenetic analysis of the GATL protein family also suggests such functional redundancy, particularly among the AtGATL5/AtGATL6 and AtGATL8/AtGATL9 pairs, which appear to have arisen by relatively recent partial genome duplication events in Arabidopsis.
However, the gene expression data presented here, as examined using transgenic pGATL::GUS expression, argues against the functional redundancy of these proteins. The constructs used for these studies included the 3# untranslated regions of the GATL genes to better ensure that any GUS expression observed accurately reflects AtGATL gene expression, in light of previous reports that DNA elements in the 3# regions of plant genes can regulate gene expression (Dean et al., 1989;Dietrich et al., 1992;Larkin et al., 1993;Fu et al., 1995aFu et al., , 1995bChen et al., 1998). As indicated by GUS expression, each of the AtGATL genes shows a unique expression pattern, although there is some overlap, particularly in the vasculature. For example, AtGATL5 is expressed in secondary xylem of stems and in filaments of flowers, while AtGATL6 is expressed in phloem of upper stems and in stigma and style of the flower. Different expression patterns were also observed for the AtGATL8 and AtGATL9 genes. Furthermore, recent research on two poplar (Populus deltoides) isoforms of AtGATL1 also suggests functional specialization of GATL proteins (Kong et al., 2009). The differing tissue-specific expression patterns between such duplicated genes suggests that subfunctionalization or neofunctionalization has happened, a process that occurs mainly through mutations in reg-ulatory sequences instead of via mutations in the coding sequence (Blanc and Wolfe, 2004;Haberer et al., 2004;Langham et al., 2004).
The GATL proteins also share high sequence identity/similarity among species. For example, PdGATL1.1 and PdGATL1.2 share about 80% identity at the amino acid level with AtGATL1. Transformation of either poplar GATL1 ortholog into the Arabidopsis atgatl1 mutant background complemented several phenotypes of the atgatl1 mutant, including decreased Xyl content, tissue morphology, and growth habit (Kong et al., 2009). Thus, the functions of GATL genes may be largely conserved between different species.
In order to gain insight into the biochemical functions of AtGATL family members, we analyzed the cell walls isolated from stems of T-DNA insertion mutants of AtGATL3, -5, -6, -8, and -9 using both glycosyl composition analysis and glycome profiling. Both analyses revealed only subtle changes in wall composition and structure and thus did not yield definitive information about the functions of these five AtGATL genes in cell wall biogenesis. For example, the glycosyl composition analyses showed significant decreases in the GalA content of stem walls in atgatl3, atgatl6, and atgatl9 but also showed changes in other glycosyl residues (Table II). Glycome profiling of the stem cell walls did not reveal the absence of any particular polysaccharide component recognized by the diverse suite of plant glycan-directed monoclonal antibodies used in any of the five atgatl mutants analyzed (Fig. 7) and therefore also did not provide direct evidence of AtGATL function in these plants. Nonetheless, glycome profiling did demonstrate changes in overall wall structure in each of the five mutants, as evidenced by altered patterns of epitope extractability. For example, some xyloglucan epitopes were more easily extracted from the walls of atgatl5 and atgatl6 stems compared with wild-type walls. Changes in the extractability patterns of pectic arabinogalactan epitopes were also observed. Our results suggest that cell wall analyses carried out at whole-plant or whole-tissue level resolution are unlikely to document dramatic changes in cell wall composition and structure in mutants when expression of the affected gene is highly localized to specific cell types in wild-type plants, as is the case for most of the GATL genes in Arabidopsis (Figs. 3 and 4). Thus, it will likely be more informative to examine specific cell types that express a GATL gene of interest for changes in wall structure in order to infer the function of that GATL protein in cell wall synthesis. For example, we have found that atgatl5 shows defects in seed coat mucilage production, suggesting that AtGATL5 plays a role in synthesizing the pectic polysaccharides that are the principal components of this mucilage (Y. Kong and M.G. Hahn, unpublished data). This phenotype would not be observed in analyses carried out at the whole-plant or whole-tissue level.
Several pieces of data, in addition to the decreased GalA content mentioned above, implicate AtGATL3, AtGATL6, and AtGATL9 in pectic polysaccharide synthesis, which occurs in primary wall synthesis. The AtGATL3 promoter::GUS expression results, which show that AtGATL3 expression in stem is mainly localized in primary cell wall-rich cells, such as epidermis and cortex, support the hypothesis that AtGATL3 is a primary cell wall-associated gene. The RT-PCR data showed that AtGATL6 expression is lower at the base of the stem compared with the top, indicating that AtGATL6 is preferentially expressed in younger stems, where primary cell wall synthesis predominates. Given the fact that AtGATL6 promoter::GUS is expressed only in the phloem of upper stems, we suggest that AtGATL6 may be involved in primary cell wall synthesis in developing phloem. Transcriptional profiling of genes differentially expressed during in vitro xylem differentiation in Arabidopsis suspension cells showed that AtGATL6 and AtGATL9 expression levels decreased rapidly before xylem vessel element formation (Kubo et al., 2005), further supporting a role for these two genes in primary cell wall synthesis. However, the hypothesis that AtGATL3, -6, and -9 are involved in pectin synthesis will require further experimental substantiation.
It is noteworthy that all AtGATL genes except AtGATL4 are expressed in vascular tissue, with several family members showing overlapping expression patterns in different vascular cell types. This is true for the phloem (AtGATL3, -6, -7, -8, and -10), protoxylem (AtGATL8 and -9), metaxylem (AtGATL1, -2, and -5), secondary xylem (AtGATL1, -2, -5, -6, and -8), and cortex (AtGATL3, -7, -8, and -9). Such overlapping expression patterns for many family members during vascular development suggest a potential for combinational AtGATL action in vascular cell wall synthesis or, alternatively, that some functional redundancy may occur among GATL proteins, at least in some tissues. The latter could explain the absence of dramatic phenotypes in the five mutant lines that were examined in this study.
The involvement of AtGATL1 in xylan synthesis during secondary wall formation has been reported previously (Brown et al., 2007;Lee et al., 2007), and a close correlation between AtGATL1 expression and the expression of AtCESA (for cellulose synthase) genes associated with secondary cell wall formation has also been reported (Mutwil et al., 2009). Although AtGATL1 has been linked to xylan synthesis, it is not clear whether this GATL is directly involved in xylan biosynthesis, possibly in connection with the synthesis of the GalA-containing reducing end oligosaccharide (Peñ a et al., 2007) or indirectly through the synthesis of another polysaccharide that establishes a foundation for xylan synthesis (Mohnen, 2008). In this context, the very strong AtGATL1 expression observed in the root apical meristem and elongation zones is interesting (Fig. 3), tissues where xylan synthesis is not known to occur in Arabidopsis. Furthermore, four AtGATLs, AtGATL3, -6, -7, and -8, have been shown to be coexpressed with AtCESA genes that are involved in primary cell wall synthesis (Mutwil et al., 2009). These data suggest that the majority of AtGATL genes expressed in vascular tissues are involved in primary cell wall synthesis in these tissues. It is also notable that five of the AtGATL members are expressed in the abscission zone of silique and six are expressed in trichomes. Both trichomes and abscission zones are rich in pectins (Henderson et al., 2001;Marks et al., 2008), polysaccharides characteristic of primary cell walls, further suggesting that these AtGATL genes are involved in pectin synthesis in these tissues.
AtGATL4 stands out among the AtGATL family because it is expressed exclusively in the flower. The GUS reporter-based expression analyses localized AtGATL4 expression specifically to the pollen grain and elongating pollen tube. Pectic polysaccharides are a major component of pollen grains and pollen tube walls, where they help maintain the cylindrical shape of the pollen tube and act as adhesion molecules during the fertilization process (Lord and Russell, 2002;Bosch et al., 2005). AtGATL1 and -7 are also expressed in pollen, although not specifically, implicating a potentially specialized function of some AtGATL family members in the deposition of pollen cell wall pectin components.
In summary, the unique and overlapping cell typespecific expression of each AtGATL family member provides useful information and a platform for understanding their functions. The Golgi localization of the AtGATLs, their expression patterns, the available microarray data and coexpression analyses, as well as the cell wall compositional analyses strongly implicate at least some of the AtGATL family members, like AtGATL3, -6, and -9, in the biosynthesis of primary cell walls in diverse organs and tissues of Arabidopsis by contributing to pectin synthesis. However, given the partial overlapping gene expression patterns for several GATLs in some Arabidopsis tissues, more extensive gene knockout analyses, either simultaneously using RNA interference or by piling up multiple insertional mutations, and biochemical studies in specific cell and tissue types will be required to address the biological functions of the AtGATL genes.

Bioinformatic Analyses of GATL Protein Sequences
The 12 draft plant genomes, predicted genes, and translated protein data used for bioinformatic analysis of the GATL family were downloaded from various sources, as specified in Supplemental Table S1. GATL proteins were identified from these genomes using an HMMER search (Eddy, 1998) for the Pfam (Finn et al., 2006) Glyco_transf_8 (PF01501; 345 amino acids long) domain as the query. This HMMER search, using an E-value cutoff of 1e-2 or less, identified 99 GATL proteins from the 12 genomes. Four protein sequences were removed due to the fact that they were truncated or lacked key amino acid motifs characteristic of GT8 (Yin et al., 2010). A maximum likelihood phylogenetic tree was constructed using PhyML version 2.4.4 (Guindon and Gascuel, 2003) for the 95 full-length proteins based on their multiple sequence alignment generated using MAFFT version 6.603 (Katoh et al., 2005) using the L-INS-I option as recommended to give the most accurate alignment. Specifically, PhyML analyses were conducted with the JTT model, 100 replicates of bootstrap analyses, estimated proportion of invariable sites, four rate categories, estimated g-distribution parameter, and an optimized starting BIONJ tree. A clade of six genes (five from Physcomitrella patens and one from Selaginella moellendorfii) that are basal to all other plant GATL proteins examined was selected as the outgroup to root the phylogeny. A rectangular phylogram of the GATL protein sequences was generated using the Interactive Tree of Life Web server (Letunic and Bork, 2007).

Plant Material and Growth Conditions
Arabidopsis (Arabidopsis thaliana) plants were grown on soil in a controlled-environment chamber under a 14-h-light/10-h-dark cycle at 19°C during the light period and 15°C during the dark period. The light intensity was 150 mE m 22 s 21 , and the relative humidity was maintained at 60% to 70%. Arabidopsis plants of the Columbia ecotype were used for transformation and isolation of DNA and RNA. T-DNA-mutagenized seeds were obtained from the Salk Institute (http://signal.salk.edu/cgi-bin/tdnaexpress) through the Arabidopsis Biological Resource Center (Supplemental Table S5). T-DNA insertion mutants for gatl3, -5, -6, -8, and -9 were identified using the flanking primers (LP and RP) generated by the SIGnal T-DNA verification primer design Web site (http://signal.salk.edu/tdnaprimers.html) and primers from the T-DNA left border, LBa1 (5#-GCGTGGACCGCTTGCTGCAACT-3#) and LBb1 (5#-TCAAACAGGATTTTCGCCTGCT-3#). The sequences of the flanking primers for AtGATL genes are provided in Supplemental Table S6.

RNA Extraction and RT-PCR
For semiquantitative RT-PCR analysis of AtGATL gene expression, siliques, flowers, leaves, upper stems, middle stems, and lower stems were harvested from 7-week-old Arabidopsis plants and frozen immediately in liquid nitrogen. Roots for RT-PCR were obtained from plants grown hydroponically for 14 d under sterile conditions in B5 vitamin, with 1% (w/v) Suc, pH 5.8, at 22°C in constant white light. Approximately 100 mg of tissue samples was ground in liquid nitrogen, and total RNAs were extracted with the RNeasy plant mini kit (Qiagen) and treated with RNase-free DNase (Qiagen) to remove contaminating genomic DNA. One microgram of total RNA was reverse transcribed using SuperScript III reverse transcriptase (Invitrogen) in a 20-mL RT firststrand synthesis reaction that contained oligo(dT) primers. RT-PCR products were generated using primer sequences unique to each of the 10 AtGATL genes (Supplemental Table S7). Semiquantitative RT-PCR was performed using the following program: 95°C for 3 min; 30 to 35 cycles of 30 s at 95°C, 30 s at 55°C, and 1 min at 72°C; then hold at 72°C for 5 min. To determine whether comparable amounts of RNA had been used for RT-PCR from the different tissues, the ACTIN2 gene (At3g18780) was used as a control.
To determine AtGATL transcript levels in their respective homozygous mutant lines, total RNAs were isolated from stems of 7-week-old wild-type and homozygous atgatl mutant lines, and the gene transcript levels were analyzed according to the method described above. Knockouts were defined as mutants with RT-PCR reactions that yielded no detectable PCR product using genespecific primers. Knockdown mutants were those that yielded significantly less, but detectable, PCR product compared with the wild-type plants.

Plasmid Construction and Plant Transformation
The cell-specific expression pattern of the AtGATL genes was studied using the GUS reporter gene. For each AtGATL promoter::GUS reporter gene construct, approximately 2.5 kb upstream of the predicted ATG start codons (including 5# untranslated regions) and 1 to 1.5 kb downstream of the predicted stop codons (including 3# untranslated regions) were PCR amplified using Pfx50 Taq polymerase (Invitrogen) with gene-specific primers containing appropriate restriction sites. Then, the two amplified genomic DNA fragments were fused with the GUS gene in the binary vector pBI101 to create the AtGATL promoter::GUS reporter gene construct. Sequences of the individual primers used are listed in Supplemental Table S8. Each of the AtGATL promoter::GUS reporter gene constructs was sequenced to verify its construction. The gene fusions were first electroporated into Agrobacterium tumefaciens strain GV3101 and then introduced into Arabidopsis wild-type plants (Columbia ecotype) via the floral dip method (Clough and Bent, 1998). Transgenic plants were selected on plates containing kanamycin (50 mg L 21 ).

Histochemical GUS Assays
Expression of AtGATL::GUS transgenes was visualized by staining for GUS activity as described . Briefly, transgenic plants or excised tissues were stained in GUS staining solution (100 mM sodium phosphate, pH 7.0, 10 mM EDTA, 1 mM ferricyanide, 1 mM ferrocyanide, 0.1% [v/v] Triton X-100, and 1 mM 5-bromo-4-chloro-3-indolyl b-D-GlcA) at 37°C. The staining buffer was removed, and the samples were cleared in 70% (v/v) ethanol until the blue color became visible. For each construct, plants from five to 10 independent transgenic lines were examined. Patterns of gene expression for each construct were consistent across multiple transgenic lines, and representative plants were photographed with a stereoscopic microscope (Olympus SZH-ILLD) equipped with a Nikon DS-Ril camera head using NIS-Elements Basic Research software.

Subcellular Localization of AtGATL Proteins
The coding regions for selected AtGATL genes were cloned in frame with an EYFP gene under the control of the 35S promoter in a pCAMBIA-based binary vector (Kong et al., 2009) to generate the fusion constructs (35S-AtGATL-EYFP). Primers used for creating the EYFP fusions are listed in Supplemental Table S9. The AtGATL constructs were sequenced to verify their construction and then transformed individually into A. tumefaciens strain GV3101. The constructs were individually cotransformed into fully expanded leaves of Nicotiana benthamiana plants together with the ECFP-tagged Golgi marker Gmct-ECFP (Saint-Jore- Dupas et al., 2006;Nelson et al., 2007). Cotransformation and signal observation were done as described previously (Kong et al., 2009).

Cell Wall Extraction and Sugar Analysis
Cell walls were prepared as alcohol-insoluble residues (AIR) as described previously (Persson et al., 2007). In brief, stems from 10-week-old Arabidopsis wild-type and mutant plants were harvested on ice, flash frozen in liquid nitrogen, and ground to a fine powder with a mortar and pestle. The ground materials were consecutively extracted with 2 volumes of 100 mL of ice-cold 80% (v/v) ethanol, 100% ethanol, chloroform:methanol (1:1, v/v), and 100% acetone. Starch was removed from the walls by treatment with type I porcine a-amylase (Sigma-Aldrich; 47 units per 100-mg cell wall) in 100 mM ammonium formate (pH 6.0) for 48 h at room temperature with constant rotation. Destarched walls were centrifuged, washed three times with sterile water and twice with 100% acetone, and air dried. Sugar composition analyses were carried out on three independently prepared cell wall preparations using trimethyl silyl ethers of methyl glycosides as described (Caffall et al., 2009).

Cell Wall Fractionation
Sequential extraction of cell walls (AIR) isolated from wild-type and atgatl mutant plants was done on 10 mg mL 21 suspensions. First, the AIR samples were suspended in 50 mM ammonium oxalate (pH 5.0). The suspension was incubated overnight at room temperature with constant mixing. After the incubation, the suspension was centrifuged at 3,400g, the supernatant was decanted and saved, and the pellet was washed three times with deionized water before subsequent extraction steps. The washed pellet was then sequentially extracted in the same manner using 50 mM sodium carbonate (pH 10), 1 M KOH, and 4 M KOH. In each step, the supernatants were individually decanted and saved. The 1 M and 4 M KOH extracts were neutralized with glacial acetic acid. All cell wall extracts were dialyzed against four changes of 20 L of deionized water (sample:water, approximately 1:60) at room temperature for a total of 48 h and then lyophilized.

Total Sugar Estimation and ELISA
Cell wall extracts were dissolved in deionized water (0.2 mg mL 21 ), and total sugar contents of cell wall extracts were estimated using the phenolsulfuric acid method (Masuko et al., 2005). Cell wall extracts (60 mg sugar mL 21 ) were applied to the wells of ELISA plates (Costar 3598) at 50 mL per well and allowed to evaporate to dryness overnight at 37°C. A Biotek robotic system was used to perform fully automated ELISA using a series of 150 monoclonal antibodies directed against plant cell wall carbohydrate epitopes . ELISA data are presented as a heat map in which the antibody order is based on a hierarchical clustering analysis of the antibody collection that groups the antibodies according to their binding patterns to a panel of diverse plant glycans .

Monoclonal Antibodies
Monoclonal antibodies were obtained as hybridoma cell culture supernatants either from laboratory stocks at the Complex Carbohydrate Research Center (JIM and MAC series; available from CarboSource Services [http:// www.carbosource.net]) or from Plant Probes (LM series, PAM1 [http://www. plantprobes.net]).

Supplemental Data
The following materials are available in the online version of this article.
Supplemental Table S1. Sources for the 12 genomes used for bioinformatic analysis of the GATL family.
Supplemental Table S2. Summary of whole-plant AtGATL expression patterns based on Figure 3.
Supplemental Table S3. Summary of whole-plant AtGATL expression patterns based on Figure 4.
Supplemental Table S4. Analysis of the topology of AtGATLs using the plant membrane protein database, Aramemnon.
Supplemental Table S5. T-DNA insertion lines used in this study.
Supplemental Table S6. Flanking primer sequences for insertions in AtGATL genes.
Supplemental Table S7. Primers used for RT-PCR.
Supplemental Table S8. Primers used for construction of AtGATL promoter::GUS fusions.