A Comprehensive Toolkit of Plant Cell Wall Glycan-Directed Monoclonal Antibodies

A collection of 130 new plant cell wall glycan-directed monoclonal antibodies (mAbs) was generated with the aim of facilitating in-depth analysis of cell wall glycans. An ELISA-based screen against a diverse panel of 54 plant polysaccharides was used to characterize the binding patterns of these new mAbs, together with 50 other previously generated mAbs, against plant cell wall glycans. Hierarchical clustering analysis was used to group these mAbs based on the polysaccharide recognition patterns observed. The mAb groupings in the resulting cladogram were further verified by immunolocalization studies in Arabidopsis thaliana stems. The mAbs could be resolved into 19 clades of antibodies that recognize distinct epitopes present on all major classes of plant cell wall glycans, including arabinogalactans (both protein and polysaccharide-linked), pectins (homogalacturonan, rhamnogalacturonan I), xyloglucans, xylans, mannans, and glucans. In most cases, multiple sub-clades of antibodies were observed to bind to each glycan class, suggesting that the mAb in these sub-groups recognize distinct epitopes present on the cell wall glycans. The epitopes recognized by many of the mAbs in the toolkit, particularly those recognizing arabinose- and/or galactose-containing structures, are present on more than one glycan class, consistent with the known structural diversity and complexity of plant cell wall glycans. Thus, these cell wall glycan-directed mAbs should be viewed and utilized as epitope-specific, rather than polymer-specific probes. The current world-wide toolkit of ~180 glycan-directed antibodies from various laboratories provides a large and diverse set of probes for studies of plant cell wall structure, function, dynamics, and biosynthesis.


INTRODUCTION
Cell walls play important roles in the structure, physiology, growth and development of plants (Carpita and Gibeaut, 1993). Plant cell wall materials are also important sources of human and animal nutrition, natural textile fibers, paper and wood products, and raw materials for biofuel production (Somerville, 2007). Many genes thought to be responsible for plant wall biosynthesis and modification have been identified (Burton et al., 2005;Lerouxel et al., 2006;Mohnen et al., 2008) and 15% of the Arabidopsis thaliana genome is likely devoted to these functions (Carpita et al., 2001). However, phenotypic analysis in plants carrying cell wallrelated mutations has proven particularly difficult. First, cell wall-related genes are often expressed differentially and at low levels between cells of different tissues (Sarria et al., 2001).
Also, plants have compensatory mechanisms to maintain wall function in the absence of a particular gene (Somerville et al., 2004). Thus, novel tools and approaches are needed to characterize wall structures and the genes responsible for their synthesis and modification.
Here we report the generation of 130 new mAbs that bind to diverse epitopes present on a broad spectrum of plant cell wall glycans. In addition, ~50 previously reported or generated by aspiration and 50 μ L of undiluted hybridoma supernatant were added to each well and incubated for one hour at room temperature. Supernatant was removed and wells were washed three times with 300 μ L of 0.1% (w/v) instant non-fat dry milk in TBS (wash buffer).
Peroxidase-conjugated goat anti-mouse IgG or goat anti-rat IgG antibodies (Sigma-Aldrich, St. Louis, MO), depending on the primary antibody used, were diluted 1:5000 in wash buffer and 50 μ L was added to each well and incubated for one hour. Note that the secondary antibodies used in this study are generated against whole immunoglobin molecules and thus bind to several isotypes of primary antibodies, including IgGs, IgMs, and IgAs, according to the manufacturers.

Immobilization of Plant Polysaccharides for ELISA
Eight commercially available 96-well plates were tested to determine their suitability for plant cell wall polysaccharide-based ELISA assays. These plates were tested simultaneously using a standard ELISA protocol and twelve mAb/polysaccharide pairs: CCRC-M1/sycamore xyloglucan, CCRC-M2/gum karaya, CCRC-M7/sycamore pectic polysaccharides, CCRC-

Data Correlation Analysis Emphasizes Reproducibility of ELISAs
ELISA analyses were used to test monoclonal antibody binding specificities against a diverse panel of plant polysaccharide preparations. The reproducibility of the ELISA data pattern for each antibody was examined by generating a correlation heat map (see Supplementary Materials). This correlation heat map analysis was done using the data obtained from six replicate experiments involving ELISA screening of 41 antibodies against a panel of diverse polysaccharides. In the correlation heat map, the value (color) of each square corresponds to the correlation of the ELISA response vector for one mAb in one experiment to the ELISA response vector for each mAb in another experiment or group of experiments.
Perfect reproducibility corresponds to a heat map with all diagonal elements equal to 1.0 (brightest yellow) and perfect symmetry about the diagonal. (This would be the result if a data set were compared to itself as shown in Figure 1A). A second heat map, shown in Figure 1B, depicts the correlation of a randomly selected ELISA panel test replicate with the average of 6 replicates. In this case, each diagonal heat map element correlates the response pattern of a specific mAb in the selected experiment with the average ELISA response pattern for that mAb.
Almost all of the correlation coefficients were greater than 0.98 (Table S3; Supplementary Materials). Each off-diagonal heat map element in Figure 1B shows the correlation of the ELISA response pattern of a mAb in the selected experiment to the mean ELISA response pattern of a different mAb. The presence of very few deviations from symmetry about the diagonal in the correlation heat map indicates that the ELISAs are highly reproducible. The reproducibility of ELISAs is also emphasized by the significant resemblance of the experimental correlation heat map (shown in Figure 1B) to the autocorrelation heat map (shown in Figure 1A).
The correlation heat maps ( Figure 1) also group antibodies that show similar binding patterns to the panel of polysaccharides. These groups (clades) of antibodies are highlighted by the coloring of the correlation heat map (from black to bright yellow) ( Figure 1). For example, the relatively brighter yellow "correlation square" of the CCRC-M79 group (red-highlighted antibodies, red outline) indicates that the binding patterns of mAbs in this clade are very similar.
In contrast, the relatively dark coloring of the "correlation square" in the lower left corner of the map [corresponding to the clade containing JIM3 and five other antibodies (blue-highlighted antibodies, blue outline); Figure 1B] indicates that the binding patterns of the mAbs in this clade are not closely related; such outliers fall into a single cluster only because they do not fit into any well-defined cluster within the scope of this analysis.

Polysaccharide Panel Screening and Hierarchical Clustering of Monoclonal Antibodies
The current toolkit of mAbs studied here, when screened against 54 diverse plant polysaccharide preparations whose detailed chemical compositions are either previously known or determined during the current study (Table S1; Supplemental Materials), yielded diverse polymer binding patterns when viewed as a whole. However, subsets of the collection shared similar polymer binding patterns, suggesting that hierarchical relationships exist among these mAbs. To study these relationships in greater detail and to group mAbs based on their polymer binding fingerprints, hierarchical clustering analysis of the ELISA data was performed using a modification of previously described methods (Ferguson et al., 1988 values, one for each polysaccharide). A matrix, in which each row consisted of the ELISA response vector for a particular mAb, was generated. The rows and columns in the matrix were clustered (see Supplemental Materials) to generate dendrograms that allowed similarities in the ELISA response patterns of mAbs (rows) and polysaccharides (columns) to be visualized. The clustered ELISA data are represented as a heat map ( Figure 2) along with the dendrograms that were used to order the data. The color of each cell in the heat map represents the ELISA response of a particular mAb when tested against a particular polysaccharide.
Dendrograms generated by our initial clustering experiments, performed essentially as described previously (Ferguson et al., 1988), were often in disagreement with groupings obtained by manual comparison of the ELISA responses. We showed that this initial approach, which is based on using the Pearson correlation coefficient as the distance metric for clustering, can lead to dendrograms that imply close associations between dissimilar patterns. Therefore, we used a different approach in which the inverse cosine of the dot product of each pair of ELISA response vectors was used as the distance metric for clustering. This is similar to the use of Pearson correlation coefficients in that it builds dendrograms using response patterns rather than absolute responses. When applied to our ELISA data, this new approach produced dendrograms that were more consistent with the mAb groupings obtained by manual comparison of the ELISA responses.
Encouraged by these results, we used the R language (R Development Core Team, 2006) to develop a software application that uses the alternative approach. Given the ELISA response data for a collection of mAbs against the panel of polysaccharides, the software provides dendrograms for the mAbs and the polysaccharides, and a heat map ordered using the dendrograms. The software also allows the sub-trees of a selected vertex in either dendrogram to be reversed, which does not formally or materially alter the dendrogram, but can provide images that are easier to interpret. This software also can produce a heat map that illustrates the correlation of one data set to another, providing a rapid method of assessing reproducibility between data sets and identifying those data points and ELISA response patterns that differ significantly between the two data sets. We are making this software freely available for use by others (http://glycomics.ccrc.uga.edu/cluster/).
The hierarchical clustering analysis grouped the mAbs into well-resolved clades that are characterized by commonalities in polymer recognition (Figure 2 analysis we identified 19 groups of mAbs that recognize a range of glyco-structures covering most major cell wall polysaccharides (circled in white in Figure 2). Some examples include a non-fucosylated xyloglucan-directed clade of mAbs, a fucosylated xyloglucan-directed calde, the pectic backbone-directed clade, the RG-1/AG clade, four distinct xylan-directed clades of mAbs (Xylan 1-4) and several arabinogalactan-directed clades (AG-1-4). The mAbs that are grouped within each clade are identified in Table S2 ( among three distinct clades of mAbs, RG-I/AG, AG-3, and AG-4. The mAbs in the RG-1/AG clade (which includes JIM1, JIM16, JIM131 and JIM132) bind to RG-I preparations from a broad range of plants, but do not bind to Gum Arabic. The mAbs in the AG-3 clade (which includes JIM4, JIM17, JIM8, and JIM15) bind strongly to Gum Ghatti and Gum Arabic and also to pectic polysaccharide preparations from tomato and sycamore maple. The mAbs in the AG-4 clade (with JIM13 and JIM133) bind to RG-I preparations from a broad range of plants and also binds to Gum Arabic.

Immunolabeling
Immunolabeling studies were carried out on Arabidopsis influorescence stems (Figures 3,   4 and 5) to obtain independent verification of the clades or sub-clades resulting from the hierarchical clustering of ELISA data. These studies were done using three sets of mAbs that resulted from the hierarchical clustering analyses (Figure 2; Table S2), two distinct sets of xylandirected mAbs (Xylan-3 and Xylan-4) and another set of mAbs directed against the arabinogalactan side chains of rhamnogalacturonan I (RG-I/AG).
All of the mAbs in the Xylan-4 clade labeled xylem tissues in Arabidopsis stems in a similar fashion, though the labeling intensity differed amongst some of the mAbs (Figure 3). These intensity differences likely result from differences in the epitope structures recognized by the mAbs and/or the different epitope density distribution patterns within the xylans synthesized in different cells. The similar localization patterns among these xylan-directed mAbs support the results of the hierarchical clustering that grouped these mAbs into a tight cluster ( Figure 2).
Interestingly, another set of xylan-directed mAbs, which form a distinct sub-set within the Xylan-3 clade, also label xylem tissue in a similar pattern as observed with the Xylan-4 mAbs ( Figure 4A). This suggests that the xylans made in Arabidopsis stem xylem tissues carry at least two distinct epitopes recognized by these two groups of xylan-directed antibodies, which is M145, CCRC-M146, and CCRC-M155) all show a labeling pattern that is very distinct from the first sub-set of Xylan-3 mAbs, labeling only specific cells in Arabidopsis stems ( Figure 4B).
The RG-I/AG clade of mAbs was grouped into three sub-clades by the hierarchical clustering [ Figure 2; RG-I/AG clade (top group of 15 mAbs, middle 9 mAbs and the bottom 7 mAbs)]; two of those sub-clades appeared to be distinguished only by their ability to bind to larch and soybean arabinans/galactans, while the third sub-clade showed more diverse binding patterns against the pectic arabinogalactans tested here. The first two sub-clades showed very similar immunolabeling patterns in Arabidopsis stems, although differences in labeling intensities and subtle differences in labeling patterns were observed (Figures 5A and B).
Antibodies in the third RG-I/AG sub-clade showed disparate labeling patterns from each other and from the other two sub-clades ( Figure 5C). Thus, the observed immunolabeling patterns support the sub-clade structure within the RG-I/AG clade of mAbs that had been defined by the hierarchical clustering of the ELISA data.  (Table S2; Supplementary Materials). The world-wide collection of cell wall glycan-directed mAbs is now sufficiently large and diverse to constitute a comprehensive resource that will prove invaluable for detailed studies of the structure, dynamics, function, and biosynthesis of plant cell walls.

DISCUSSION
The ELISA used in this study for determination of antibody binding specificities is a reliable and a highly reproducible assay for quantifying the binding of mAbs to polysaccharide antigens. One key aspect of this method that is different from many ELISA applications is the drying down of the polysaccharides to the bottom of the ELISA plate wells, rather than allowing adsorption to take place only in the liquid phase as is commonly done. Adsorbtion of polysaccharides in the liquid phase yielded greater variability in the ELISA data, probably due to less consistent immobilization of diverse polysaccharides to the ELISA plates (data not shown).
The ability of a polysaccharide to bind to the plate under our assay conditions does not depend on its glycosyl composition or charge; both neutral and charged polysaccharides of diverse structures bind to the plates. However, the molecular size of a polysaccharide does affect its ability to adhere to the ELISA plates.  Table S2). These mAb clusters are also supported by immunolocalization studies (Figures 3, 4, 5). Comparisons of the glycosyl compositions of the polysaccharide preparations (Table S1; Supplemental Materials) recognized by each group of antibodies reveal compositional commonalities. The polysaccharide preparations used in the panel can each be viewed as a collection of epitopes, each of which typically ranges in size from one to eight glycosyl residues (Clausen et al., 2003;Kabat, 1966;Puhlmann et al., 1994;Reimer et al., 1992;Steffan et al., 1995). Different polysaccharide preparations will vary in both the types of epitopes (structural features) present and the amounts of each epitope. A minor polysaccharide component will contribute a correspondingly low proportion of epitopes to the overall epitope composition of a given polysaccharide preparation. The ELISAs provide both qualitative and quantitative measures of the epitope compositions of each of the polysaccharide preparations.
Hierarchical clustering as applied to the ELISA data will then group the antibodies based both on which polysaccharide preparations are recognized and on the strength of the ELISA signal for each polysaccharide. Minor contaminants in an individual polysaccharide preparation will not www.plantphysiol.org on August 19, 2017 -Published by Downloaded from Copyright © 2010 American Society of Plant Biologists. All rights reserved.
significantly affect the outcome of the clustering. This can be most clearly seen in the heat map ( Figure 2) in the case of the two xyloglucan clusters. These antibodies show some crossreactivity with a sycamore pectic polysaccharide preparation, but the hierarchical clustering still groups these antibodies on the basis of the strong signals to the xyloglucans and clearly distinguishes these two groups of mAbs from each other and from other groups of mAbs that bind to RG-I epitopes. Thus, the power of the hierarchical clustering approach is that it can identify the commonalities in epitope recognition patterns across the collection of antibodies and polysaccharides, and group the antibodies accordingly.
The current study included a larger number of mAbs and utilized a greater diversity of polysaccharides in the binding studies used as the foundation for the hierarchical clustering than used in a previous study that yielded a dendrogram of cell wall-reactive mAbs (Moller et al., 2008), resulting in differences in the clade compositions between the two studies. For example, JIM5 and JIM7, which fell into the same cluster previously (Moller et al., 2008), are clearly resolved into distinct, but related, sub-clusters in the current analysis ( Figure 2, Table S2). The separate clustering of JIM5 and JIM7 noted here is consistent with the reported differences in epitopes recognized by these mAbs (Clausen et al., 2003). Likewise, several JIM and MAC mAbs that recognize AGP and/or extensin (HRGP) epitopes and had been grouped largely into two clades in previous analyses (Moller et al., 2008;Yates and Knox, 1994) are now distributed among several clades, each of which shows distinct arabinogalactan-glycan binding patterns.
These new clusterings suggest that the set of cell wall glycoprotein-binding JIM and MAC mAbs bind to a greater diversity of glycans, specifically pectic arabinogalactans, than had previously been recognized, and cannot be viewed as being specific to a particular class of cell wall glycoproteins.
The observed complexities in the mAb binding patterns to plant cell wall glycans reflect the known structural complexities of plant cell wall polysaccharides (O'Neill and York, 2003;Ridley et al., 2001). Some of the observed cross-reactivities can be readily explained by the covalent association of different glycans with one another in the wall and in glycan preparations.
Thus, antibodies that bind to homogalacturonan epitopes (e.g., CCRC-M38, JIM5, CCRC-M34, JIM7) also bind to a broad range of RG-I preparations due to the covalent association between homogalacturonan and RG-I (Mohnen, 2008). In other cases, the cross-reactivities can be explained by the presence of similar/identical epitopes on a given polysaccharide isolated from  In still other cases, the cross-reactivities can be explained by the fact that some structural features (epitopes) are present on multiple glycans that occur in plant cell walls. This is particularly the case with antibodies that bind to epitopes containing arabinosyl and/or galactosyl residues, which are present in multiple structural contexts within diverse plant cell wall glycans.
For example, mAbs that bind to arabinogalactan side-chains of RG-I frequently, but not always, also bind to free and/or protein-linked arabinogalactans, which contain similar structural features (Ridley et al., 2001;Seifert and Roberts, 2007). The data presented here emphasizes that glycandirected antibodies should be utilized as epitope-directed reagents, and frequently are not polymer-selective.
Some observed cross-reactivities are not as readily explained based on current knowledge of cell wall glycan structures. For example, mAbs that bind to fucosylated xyloglucans (e.g., CCRC-M1) also bind strongly to sycamore RG-I (but not to other RG-I's included in the current study). This cross-reactivity is not due contamination of the sycamore RG-I preparation with xyloglucan, since treatment of sycamore RG-I with a xyloglucan-specific endoglucanase did not affect binding of CCRC-M1 (data not shown). The cross-reactivity of the mAbs in the Xylan-1 clade with xyloglucans included in this study is also not readily explained. The epitope(s) recognized by the Xylan-1 mAbs appear not to be present on all xyloglucans, as these mAbs do not label any cells in Arabidopsis tissues (data not shown). Interestingly, studies of carbohydrate binding modules that recognize xylan have been reported to also bind to xyloglucans (Boraston et al., 2001;Gunnarsson et al., 2006). Xyloglucan and xylan are not known to be covalently linked or to share common structural features (except that both have a β-1,4 linked backbone composed of pyranosyl residues in which all exocyclic oxygens are equatorial). Resolution of these cross-reactivities must await detailed characterizations of the epitopes recognized by these mAbs.
Most of the mAb clades identified through hierarchical clustering are divided further into sub-clades. Some of these sub-divisions are informative with respect to possible epitopes recognized by newly generated mAbs due to tight clustering with previously characterized mAbs. For example, several new mAbs (CCRC-M39, CCRC-M84, CCRC-M102, CCRC-M106) cluster with CCRC-M1, suggesting that the newly generated mAbs bind to the same or similar fucosylated xyloglucan epitope recognized by CCRC-M1 (Puhlmann et al., 1994). Other newly reported mAbs cluster in distinct sub-clades with previously characterized mAbs directed against homogalacturonans (Figure 2, Table S2). CCRC-M34, CCRC-M130 and JIM136 cluster closely in a sub-clade with JIM7 and LM7, mAbs that bind to densely methyl-esterified homogalacturonan epitopes (Clausen et al., 2003), suggesting that these three mAbs bind to methylated homogalacturonan epitopes. In contrast, CCRC-M38, CCRC-M131 and CCRC-M132 cluster tightly in a sub-clade with JIM5, a mAb that binds to a homogalacturonan epitope having a low density of methylesterification (Clausen et al., 2003), suggesting that these three mAbs bind to a largely or completely de-esterified homogalacturonan epitope. Lastly, about a dozen newly generated mAbs cluster tightly in a sub-clade of the RG-I/AG clade (Figure 2, Table S2) that includes CCRC-M7, suggesting that these mAbs recognize a β -1,6 galactan epitope similar or identical to that recognized by CCRC-M7 (Steffan et al., 1995). Verification of these tentative epitope assignments awaits further more detailed studies, which are currently underway in our laboratory. analysis. Well-defined clades correspond to bright "correlation squares" (e.g., the square outlined in red). Poorly-defined clades correspond to darker correlation squares (e.g., the square outlined in blue).    Tranverse sections (250 nm) were taken from the base of inflorescence stems of 40 day old Arabidopsis thaliana plants. Immunolabeling was carried out using two sub-groups (A and B) of xylan-binding antibodies within the Xylan-3 clade identified by hierarchical clustering of ELISA data obtained from the polysaccharide panel screens (Figure 2; Table S2). The toluidine blue section shown in A and B is identical to that shown in Figure 3, and is included for orientation of the immunofluorscent panels. Arrows identify the labeled cell(s) in B. Scale bar = 50 μm. Tranverse sections (250 nm) were taken from inflorescence stems of 40 day old Arabidopsis thaliana plants. Immunolabeling was carried out using a group of pectic arabinogalactanbinding mAbs (RG-I/AG) identified by hierarchical clustering of ELISA data obtained from the polysaccharide panel screens (Figure 2; Table S2). One toluidine blue-stained section is included in C for orientation of all immunofluorescence panels. Scale bar = 50 μm.